Compare commits

...
Author SHA1 Message Date
poli0981andClaude Fable 5.2 ffd5501811 docs(guide): the settings page listed Controllers and input twice
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-15 04:56:30 +07:00
poli0981 a7ee7fd8ed Merge main (#467 squashed) into feat/input-agent: the trees are the same below this branch's own commits 2026-11-15 04:35:33 +07:00
poli0981andClaude Fable 5.2 e3dab94834 feat(input): the input reader, which the Agent does not start yet - raw input for controllers, XInput, the time of the last input (D155, D158, D159) (#467)
* feat(input): the arithmetic of controller statistics - classes, batteries, makers, one pad seen twice (D155-D163)

Nothing reads a device yet. Domain.Input says whether an XInput state or a DualSense report is a pad in use and what
its battery is, adds a battery's readings up, counts switches and decides when two entries are one pad;
Application.Input holds one session's slices and turns them into a reading a second; InputSessionBuilder turns a
session's readings into its totals, its controllers as JSON and its timeline. TelemetrySample can carry a reading.

Every byte and reading under test was written by hand: no report from a real controller has passed through this code.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(input): a session can store what it was played with - schema 0024, the .partial file's chunk 11, both tiers (D160)

Nine columns on sessions (the time in front and its four classes, the switches of method, the sources that answered,
the controllers as JSON) and three on games (what a store lists of controller support), ADD COLUMN only. The .partial
file gains chunk 11, so a session recovered after the Agent died keeps its readings. SessionFinalizer.Build stores
them before the tier branch: hooked, unhooked and recovered sessions alike. The input series are sensor_blobs rows
named input.*, which the sensors' read now leaves out and FindInputSensorsAsync returns.

Nothing produces a reading yet. A ledger this build has opened cannot be opened by 0.1.0-beta.49 or earlier.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(input): the input reader, which the Agent does not start yet - raw input for controllers, XInput, the time of the last input (D155, D158, D159)

Application.Input.InputReader takes a slice about ten times a second from Infrastructure.Input.Win32InputBackend, on
the thread of InputDeviceHub, which exists only while a session has a view of it. It opens no device, sends nothing
to one and registers for no keyboard or mouse. tools/input-readonly-check.ps1 holds its Windows calls to a list.

With no controller at hand, what the reader does with one connected is not established.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(input): an arm of XInputPad.Battery's switch said what the default says (Qodana, on #464 after it merged)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(input): no using that makes the file's own qualified names redundant (InspectCode)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(input): what InspectCode reported in the reader's tests - redundant casts, a hub disposed twice, a suppression not needed

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 16:20:55 +00:00
poli0981 9afbeb6801 Merge feat/input-reader (main with #466) into feat/input-agent 2026-11-14 23:04:48 +07:00
poli0981 0378b7dfa2 Merge main (#466 squashed) into feat/input-reader: the trees are the same below this branch's own commits 2026-11-14 23:04:44 +07:00
poli0981andClaude Fable 5.2 5d6434c360 feat(input): a session can store what it was played with - schema 0024, the .partial file's chunk 11, both tiers (D160) (#466)
* feat(input): the arithmetic of controller statistics - classes, batteries, makers, one pad seen twice (D155-D163)

Nothing reads a device yet. Domain.Input says whether an XInput state or a DualSense report is a pad in use and what
its battery is, adds a battery's readings up, counts switches and decides when two entries are one pad;
Application.Input holds one session's slices and turns them into a reading a second; InputSessionBuilder turns a
session's readings into its totals, its controllers as JSON and its timeline. TelemetrySample can carry a reading.

Every byte and reading under test was written by hand: no report from a real controller has passed through this code.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(input): a session can store what it was played with - schema 0024, the .partial file's chunk 11, both tiers (D160)

Nine columns on sessions (the time in front and its four classes, the switches of method, the sources that answered,
the controllers as JSON) and three on games (what a store lists of controller support), ADD COLUMN only. The .partial
file gains chunk 11, so a session recovered after the Agent died keeps its readings. SessionFinalizer.Build stores
them before the tier branch: hooked, unhooked and recovered sessions alike. The input series are sensor_blobs rows
named input.*, which the sensors' read now leaves out and FindInputSensorsAsync returns.

Nothing produces a reading yet. A ledger this build has opened cannot be opened by 0.1.0-beta.49 or earlier.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(input): an arm of XInputPad.Battery's switch said what the default says (Qodana, on #464 after it merged)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(input): no using that makes the file's own qualified names redundant (InspectCode)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 16:02:29 +00:00
poli0981 7dbb3c573a Merge feat/input-reader (main at 1d1d4ff) into feat/input-agent 2026-11-14 22:53:00 +07:00
poli0981 102dcdd3f1 Merge feat/input-storage (main at 1d1d4ff) into feat/input-reader 2026-11-14 22:52:58 +07:00
poli0981andClaude Fable 5.2 ac360eae01 style(input): import order in the Agent's composition; no redundant qualifier in a test (InspectCode)
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:52:56 +07:00
poli0981 43e0c2b347 Merge branch 'main' into feat/input-storage 2026-11-14 22:48:46 +07:00
poli0981andClaude Fable 5.2 b579133128 wip(input): the Agent records what a session was played with - the switch, the terms, the legal texts
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:48:03 +07:00
poli0981andClaude Fable 5.2 1d1d4ff855 docs(roadmap): the owner's order of work after the controller train (#465)
* docs(roadmap): the owner's order of work after the controller train - time played, more from stored data, shader-compilation stutter, the in-game overlay, winget and the Microsoft Store

The owner was shown six directions on 2026-11-14 and chose four, and added distribution; the
Store upload and its listing are the owner's by hand. What must be verified before anything is
promised of the two channels is listed as leads, not findings. Documents only.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(roadmap): FrameLedger has no product in Partner Center yet - the one there is another program of the owner's

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 15:31:51 +00:00
poli0981andClaude Fable 5.2 f7129807e6 test(input): what InspectCode reported in the reader's tests - redundant casts, a hub disposed twice, a suppression not needed
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:20:52 +07:00
poli0981 be96564a8f Merge feat/input-storage (the inspection fix) into feat/input-reader 2026-11-14 22:19:55 +07:00
poli0981andClaude Fable 5.2 e24f09ae8a test(input): no using that makes the file's own qualified names redundant (InspectCode)
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:19:52 +07:00
poli0981 80155d7638 Merge feat/input-storage (main at 8f24854, the Qodana fix) into feat/input-reader 2026-11-14 22:14:22 +07:00
poli0981andClaude Fable 5.2 8b4cccbb25 fix(input): an arm of XInputPad.Battery's switch said what the default says (Qodana, on #464 after it merged)
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:14:11 +07:00
poli0981 2dd0e4f271 Merge main (8f24854, #464 squashed) into feat/input-storage: the trees are the same below this branch's own commit 2026-11-14 22:13:56 +07:00
poli0981andClaude Fable 5.2 e2675cd91c feat(input): the input reader, which the Agent does not start yet - raw input for controllers, XInput, the time of the last input (D155, D158, D159)
Application.Input.InputReader takes a slice about ten times a second from Infrastructure.Input.Win32InputBackend, on
the thread of InputDeviceHub, which exists only while a session has a view of it. It opens no device, sends nothing
to one and registers for no keyboard or mouse. tools/input-readonly-check.ps1 holds its Windows calls to a list.

With no controller at hand, what the reader does with one connected is not established.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 22:12:33 +07:00
poli0981andClaude Fable 5.2 8f24854aa5 feat(input): the arithmetic of controller statistics - classes, batteries, makers, one pad seen twice (D155-D163) (#464)
Nothing reads a device yet. Domain.Input says whether an XInput state or a DualSense report is a pad in use and what
its battery is, adds a battery's readings up, counts switches and decides when two entries are one pad;
Application.Input holds one session's slices and turns them into a reading a second; InputSessionBuilder turns a
session's readings into its totals, its controllers as JSON and its timeline. TelemetrySample can carry a reading.

Every byte and reading under test was written by hand: no report from a real controller has passed through this code.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 15:02:57 +00:00
poli0981andClaude Fable 5.2 dca740e94b feat(input): a session can store what it was played with - schema 0024, the .partial file's chunk 11, both tiers (D160)
Nine columns on sessions (the time in front and its four classes, the switches of method, the sources that answered,
the controllers as JSON) and three on games (what a store lists of controller support), ADD COLUMN only. The .partial
file gains chunk 11, so a session recovered after the Agent died keeps its readings. SessionFinalizer.Build stores
them before the tier branch: hooked, unhooked and recovered sessions alike. The input series are sensor_blobs rows
named input.*, which the sensors' read now leaves out and FindInputSensorsAsync returns.

Nothing produces a reading yet. A ledger this build has opened cannot be opened by 0.1.0-beta.49 or earlier.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 21:54:13 +07:00
poli0981andClaude Fable 5.2 b29993ba82 feat(input): the arithmetic of controller statistics - classes, batteries, makers, one pad seen twice (D155-D163)
Nothing reads a device yet. Domain.Input says whether an XInput state or a DualSense report is a pad in use and what
its battery is, adds a battery's readings up, counts switches and decides when two entries are one pad;
Application.Input holds one session's slices and turns them into a reading a second; InputSessionBuilder turns a
session's readings into its totals, its controllers as JSON and its timeline. TelemetrySample can carry a reading.

Every byte and reading under test was written by hand: no report from a real controller has passed through this code.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 21:44:42 +07:00
poli0981andClaude Fable 5.2 e20bf9abc0 feat(advanced): Advanced metrics is closed - the ten Streamline readings that waited for a spike are not read, with the reason (D154) (#463)
The owner was shown the catalog (140 entries, 110 read) and what the ten entries of reach
Spike hold, read from Streamline 2.14.16's own headers, with what reading them takes: none
is an export, a game gets each function through slGetFeatureFunction and keeps the pointer,
so seeing its calls means handing the game a function of ours in place of Streamline's
(rule 4) - and a game that fetched its functions before the attach never calls it. The
owner closed it.

The ten are RefusedRoute; that reach's sentence names the third refused route in English,
Vietnamese and Japanese; the guide no longer says a later release reads more; 18_GPU's
inventory, 20_OPEN_QUESTIONS M15 and HANDOFF say train D is not built. Nothing that was
read stops being read; no hook, layout, schema or stored value changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 14:04:54 +00:00
poli0981andClaude Fable 5.2 5440f8a34a fix(app): the two GPU labels say what each figure is, not which holds the other (D152) (#462)
* docs: owner-only item 47 read (the Cyberpunk difference is the driver's; the holes are gone from steady play); Streamline's Reflex and DLSS-RR guides read against the code

Item 47, run by the owner on 0.1.0-beta.49 (sessions 236-245), read from a copy of the
ledger and the Overlay's logs: in Cyberpunk 2077 without frame generation the driver's
active GPU time is longer than its two marks are apart on 4 917 of 5 394 frames, and on
none of 129 900 frames in five other games; samples without a report are confined to a
game's start; off means off; D141's stop fired in a real game for the first time.

NVIDIA's Streamline guides for Reflex and DLSS Ray Reconstruction (2.14.16), read whole:
the guide's page on Advanced metrics now says what bUseGameSleep and bLowLatencyMode
mean; two design tables no longer promise PC latency (D147); spike-notes 30 has what the
guides add. Two comments (fl_shm.h, dllmain.cpp) said super resolution and ray
reconstruction run together - they do not; comments only, no object code changes.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): item 47 (f) and (g) in the owner's words - the NVIDIA App's Low Latency Mode is off, no hitch

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): item 47 (f) as the owner showed it - the NVIDIA App's Low Latency Mode is off for Cyberpunk 2077, the game's own Reflex is on

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(app): the two GPU labels say what each figure is, not which holds the other (D152)

"GPU active render (idle time excluded)" and "GPU render, start to end (idle time included)"
followed NVIDIA's header, which defines the active time as the span between the driver's two
GPU marks less its idle time. Item 47 counted it on 0.1.0-beta.49: in Cyberpunk 2077 without
frame generation the active time is the longer of the two on 4 917 of 5 394 frames. The owner
chose to reword the labels. They now read "GPU active render (the driver's own figure)" and
"GPU render (between the driver's start and end marks)", in English, Vietnamese and Japanese;
the guide says why. Strings only: the numbers and the means are unchanged.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): D153 - of Reflex it is enough to tell whether a game has it: Reflex and PCL leave train D, D141's stop is unchanged

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 13:13:15 +00:00
poli0981andClaude Fable 5.2 dc329c09b4 docs: owner-only item 47 read; Streamline's Reflex and DLSS-RR guides read against the code (#461)
* docs: owner-only item 47 read (the Cyberpunk difference is the driver's; the holes are gone from steady play); Streamline's Reflex and DLSS-RR guides read against the code

Item 47, run by the owner on 0.1.0-beta.49 (sessions 236-245), read from a copy of the
ledger and the Overlay's logs: in Cyberpunk 2077 without frame generation the driver's
active GPU time is longer than its two marks are apart on 4 917 of 5 394 frames, and on
none of 129 900 frames in five other games; samples without a report are confined to a
game's start; off means off; D141's stop fired in a real game for the first time.

NVIDIA's Streamline guides for Reflex and DLSS Ray Reconstruction (2.14.16), read whole:
the guide's page on Advanced metrics now says what bUseGameSleep and bLowLatencyMode
mean; two design tables no longer promise PC latency (D147); spike-notes 30 has what the
guides add. Two comments (fl_shm.h, dllmain.cpp) said super resolution and ray
reconstruction run together - they do not; comments only, no object code changes.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): item 47 (f) and (g) in the owner's words - the NVIDIA App's Low Latency Mode is off, no hitch

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): item 47 (f) as the owner showed it - the NVIDIA App's Low Latency Mode is off for Cyberpunk 2077, the game's own Reflex is on

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 12:55:35 +00:00
poli0981andClaude Fable 5.2 1a8a3cb713 docs(handoff): 0.1.0-beta.49's release facts - the tag's run, the digests, the delta, the wiki, the two rehearsals, what SonarCloud and Qodana said on the release's commit (#460)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 11:32:35 +00:00
poli0981andClaude Fable 5.2 afe498452b release: 0.1.0-beta.49 - the latency report's frames counted for the log; a read a second on a grid; the Agent keeps the reading that waits (D148-D151); owner-only item 47 (#459)
* feat(overlay): the frames of NVIDIA's latency report are counted for the native log - each by its GPU span and the driver's active time, in batches that belong to one state of the sleep status (D148, D151; nothing published, no new call, no further member; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the log case's expected words are widened before they are multiplied (Qodana native: bugprone-implicit-widening-of-multiplication-result, two)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay): the watchdog's periodic tick is on a grid of one second, and a tick the loader woke no longer takes the driver read or publishes region 5 - a read a second whatever a tick's work took (D150, the Overlay's half; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(agent): a reading that waits for a sample is no longer overwritten by the next one - two wait, the older is carried first; the guide says whose switches the latency table shows (D150, the Agent's half; D151)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* release: 0.1.0-beta.49 - item 46 read; the latency report's frames counted for the log, Reflex and frame generation together; a read a second on a grid; the Agent keeps the reading that waits (D148-D151); owner-only item 47

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the measurement case prints the numbers it parsed, never a line of the log as it stands (SonarCloud cppsecurity:S5145)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 17:42:58 +07:00
poli0981andClaude Fable 5.2 3838e26422 fix(agent): a reading that waits for a sample is no longer overwritten; the guide on whose switches the latency table shows (D150, D151) (#458)
* feat(overlay): the frames of NVIDIA's latency report are counted for the native log - each by its GPU span and the driver's active time, in batches that belong to one state of the sleep status (D148, D151; nothing published, no new call, no further member; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the log case's expected words are widened before they are multiplied (Qodana native: bugprone-implicit-widening-of-multiplication-result, two)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay): the watchdog's periodic tick is on a grid of one second, and a tick the loader woke no longer takes the driver read or publishes region 5 - a read a second whatever a tick's work took (D150, the Overlay's half; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(agent): a reading that waits for a sample is no longer overwritten by the next one - two wait, the older is carried first; the guide says whose switches the latency table shows (D150, the Agent's half; D151)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the measurement case prints the numbers it parsed, never a line of the log as it stands (SonarCloud cppsecurity:S5145)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 17:28:30 +07:00
poli0981andClaude Fable 5.2 f70a9153fa fix(overlay): the driver read is on a grid of one second; a tick the loader woke no longer takes it or publishes (D150, the Overlay's half) (#457)
* feat(overlay): the frames of NVIDIA's latency report are counted for the native log - each by its GPU span and the driver's active time, in batches that belong to one state of the sleep status (D148, D151; nothing published, no new call, no further member; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the log case's expected words are widened before they are multiplied (Qodana native: bugprone-implicit-widening-of-multiplication-result, two)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay): the watchdog's periodic tick is on a grid of one second, and a tick the loader woke no longer takes the driver read or publishes region 5 - a read a second whatever a tick's work took (D150, the Overlay's half; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the measurement case prints the numbers it parsed, never a line of the log as it stands (SonarCloud cppsecurity:S5145)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 17:03:08 +07:00
poli0981andClaude Fable 5.2 4f02a17380 feat(overlay): the latency report's frames counted for the native log, by GPU span and active time, a batch per sleep-status state (D148, D151) (#456)
* feat(overlay): the frames of NVIDIA's latency report are counted for the native log - each by its GPU span and the driver's active time, in batches that belong to one state of the sleep status (D148, D151; nothing published, no new call, no further member; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the log case's expected words are widened before they are multiplied (Qodana native: bugprone-implicit-widening-of-multiplication-result, two)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 16:14:48 +07:00
poli0981andClaude Fable 5.2 c97bf8bd98 docs(handoff): owner-only item 46 read - beta.48 on ten real sessions (#455)
* docs(handoff): owner-only item 46 read - ten sessions on beta.48: samples a second apart, both games gave the driver's own GPU times, the multipliers listed, off means off; active render longer than the span in Cyberpunk 2077 without frame generation, not explained

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): item 46 (g) - the owner's answer: no hitch with Advanced metrics on

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 07:36:55 +00:00
poli0981andClaude Fable 5.2 59be004be1 docs(handoff): 0.1.0-beta.48's release facts - the tag's run, the digests, the delta, the wiki, what SonarCloud and Qodana said on the release's commit (#454)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 05:10:45 +00:00
poli0981andClaude Fable 5.2 aec1736a1f release: 0.1.0-beta.48 - two real games read; the driver's own GPU times, every multiplier seen, samples a second apart (D142-D147); owner-only item 46 (#453)
* feat(overlay): two more members of NVIDIA's latency report - the driver's active GPU time and its GPU frame time - and every frame-generation multiplier a session saw; four reserved words, layout still 5; a read left by the session before is dropped (D142, D144; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(advanced): the driver's active GPU time and GPU frame time as two lines of their own; the GPU render label says what it includes; the multiplier row lists every value seen; the texts say two real games were read - Privacy Policy 2.21, Disclaimer 2.17 (D142-D145; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(telemetry): the poller keeps its period - it waited the whole interval after a poll's work, and on the owner's PC a second was 1.32 s; sensor_blobs.hz is the rate the samples had, not the constant 1 (D146; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* release: 0.1.0-beta.48 - two real games read; the driver's own GPU times beside a span that includes idle (D142, D143); every frame-generation multiplier a session saw (D144); samples a second apart (D146); Privacy Policy 2.21, Disclaimer 2.17 (D145); owner-only item 46

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 11:29:09 +07:00
poli0981andClaude Fable 5.2 5694ec91f7 fix(telemetry): the poller keeps its period, and sensor_blobs.hz is the rate the samples had (D146) (#452)
* feat(overlay): two more members of NVIDIA's latency report - the driver's active GPU time and its GPU frame time - and every frame-generation multiplier a session saw; four reserved words, layout still 5; a read left by the session before is dropped (D142, D144; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(advanced): the driver's active GPU time and GPU frame time as two lines of their own; the GPU render label says what it includes; the multiplier row lists every value seen; the texts say two real games were read - Privacy Policy 2.21, Disclaimer 2.17 (D142-D145; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(telemetry): the poller keeps its period - it waited the whole interval after a poll's work, and on the owner's PC a second was 1.32 s; sensor_blobs.hz is the rate the samples had, not the constant 1 (D146; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 11:06:21 +07:00
poli0981andClaude Fable 5.2 b693207387 feat(advanced): two lines for the graphics card's own time, an honest GPU render label, every multiplier seen; Privacy Policy 2.21, Disclaimer 2.17 (D142-D145) (#451)
* feat(overlay): two more members of NVIDIA's latency report - the driver's active GPU time and its GPU frame time - and every frame-generation multiplier a session saw; four reserved words, layout still 5; a read left by the session before is dropped (D142, D144; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(advanced): the driver's active GPU time and GPU frame time as two lines of their own; the GPU render label says what it includes; the multiplier row lists every value seen; the texts say two real games were read - Privacy Policy 2.21, Disclaimer 2.17 (D142-D145; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 10:41:29 +07:00
poli0981andClaude Fable 5.2 33a5da1373 feat(overlay): two more members of NVIDIA's latency report - the driver's active GPU time and its GPU frame time - and every frame-generation multiplier a session saw; four reserved words, layout still 5; a read left by the session before is dropped (D142, D144; in no release yet) (#450)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 10:16:39 +07:00
poli0981andClaude Fable 5.2 3df56f0ca1 docs: item 45 on Cyberpunk 2077 and three games the read leaves alone - read twice, 65 us at the slowest, GPU render sane there; the owner's answer: read the driver's active time (#449)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 02:31:32 +00:00
poli0981andClaude Fable 5.2 80b4437169 docs: the first real game read on 0.1.0-beta.47 - Phantom Blade Zero, three sessions: the driver found and read every second, the slowest call 45 us, no latch; its GPU render span is longer than its total there (item 45) (#448)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 02:09:38 +00:00
poli0981andClaude Fable 5.2 268ad75585 docs(handoff): 0.1.0-beta.47's release facts - the tag's run, the digests, the delta, the wiki, what SonarCloud and Qodana said on the release's commit; and a local gate that a vanished test program made red, not falsely green (#447)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 01:39:20 +00:00
poli0981andClaude Fable 5.2 3d7c273e3b release: 0.1.0-beta.47 - with Advanced metrics on, the part inside a game asks NVIDIA's driver for its latency report (D132, D136, D138-D141); the batch log says how a batch ended (D137); item 44 (d) run by the owner, twice; owner-only item 45 (#446)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 07:56:00 +07:00
poli0981andClaude Fable 5.2 29b84832ee docs(handoff): item 44 (d) was tried on 2026-11-14 and not exercised - the switch was on before the games started (two sessions refused, nothing injected), and the hooked session ran to its exit; how (d) is run (#445)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 00:15:09 +00:00
poli0981andClaude Fable 5.2 cd4f69af1a feat(advanced): the driver's latency report in the Agent and the App - seven series and one table under Latency; Privacy Policy 2.20 and Disclaimer 2.16; owner-only item 45 (D132, D136, D138-D141; in no release yet) (#444)
* feat(advanced): the driver's latency report in the Agent and the App - seven series and one table under Latency, what became of the read said in words; reaches GameDriver and RefusedRoute; Privacy Policy 2.20 and Disclaimer 2.16; owner-only item 45 (D132, D136, D138-D141; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: the owner kept the 2 ms latch at one call after the busy-PC run (D141)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(app): no redundant qualifier and no redundant null suppression in the driver-read tests (Qodana)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 06:30:00 +07:00
poli0981andClaude Fable 5.2 80d1048235 feat(overlay): the driver's latency report, read while a session asks - two read-only getters, once a second, on the presenting thread; the gates that hold it (D132, D136, D138-D141; in no release yet) (#443)
* feat(overlay): the driver's latency report, read while a session asks - two getters of an nvapi64.dll the game loaded, about once a second on the presenting thread, behind a gate of four states; a module beside the executable refused, a fault or a call of 2 ms the last one; the gates that hold it (D132, D136, D138-D141; in no release yet)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay): the latency log line measures its path with a bound (cpp:S5813)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay): the resolve switch names every outcome (Qodana native: clang-diagnostic-switch-enum)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-14 00:29:23 +07:00
poli0981andClaude Fable 5.2 70a81ed765 feat(shm): the Reflex latency words of region 5 - ten reserved words named, written by nobody; the layout stays 5 (D136, D138) (#442)
* spike(reflex): the leg with a sleep - the decision table, written before the run (D139)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* spike(reflex): the getters while the game sleeps - a leg of the probe with a sleep mode, three runs, no stop condition (D139); the plan's four decisions (D138-D141) and the train's design in HANDOFF

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(overlay): the PATCH_BATCH line says how a batch ended - install / restore for one apply, install singly / restore singly with the count in the high half of b (D137)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(shm): the Reflex latency words of region 5 - ten reserved words named, written by nobody; the layout stays 5 (D136, D138)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 23:43:12 +07:00
poli0981andClaude Fable 5.2 c02ee63a4e feat(overlay): the PATCH_BATCH line says how a batch ended - one apply, or finished one patch at a time (D137) (#441)
* spike(reflex): the leg with a sleep - the decision table, written before the run (D139)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* spike(reflex): the getters while the game sleeps - a leg of the probe with a sleep mode, three runs, no stop condition (D139); the plan's four decisions (D138-D141) and the train's design in HANDOFF

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* feat(overlay): the PATCH_BATCH line says how a batch ended - install / restore for one apply, install singly / restore singly with the count in the high half of b (D137)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 23:16:30 +07:00
poli0981andClaude Fable 5.2 f8d4945e6d spike(reflex): the getters while the game sleeps - the probe's leg with a sleep mode (D139), and the decisions and design of the train that follows (D138-D141) (#440)
* spike(reflex): the leg with a sleep - the decision table, written before the run (D139)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* spike(reflex): the getters while the game sleeps - a leg of the probe with a sleep mode, three runs, no stop condition (D139); the plan's four decisions (D138-D141) and the train's design in HANDOFF

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 22:46:16 +07:00
poli0981andClaude Fable 5.2 fc8072019e docs(handoff): what the owner's five sessions on 0.1.0-beta.46 showed (item 44) - the batches went in whole in five games; two long ones not explained; the stop not exercised (#439)
* docs(handoff): what the owner's five sessions on 0.1.0-beta.46 showed - eighteen install batches in five games, sixteen at 81-262 ms whatever they held, two not explained (Hogwarts Legacy 1.26 s, Cyberpunk 2077 0.61 s); the stop while a game runs was not exercised; item 44's sentence about a fallback corrected

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs: the owner's answers of 2026-11-13 - Reflex route D goes to a plan for one real title, Cyberpunk 2077 (D136); the PATCH_BATCH line says how a batch ended, with the next release (D137)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 14:27:34 +00:00
poli0981andClaude Fable 5.2 2c9bc3e17c test(native): a test that is cut short is not a pass - every native test names the line it ends on and ctest requires it, a test registered without one fails the configure, and four canaries prove the rule red and green (#438)
On 2026-11-13 security software removed a freshly linked fl_guard_test.exe while a local gate ran: the process
ended early with exit code 0 and ctest printed "fl_guard ... Passed 0.66 sec" for a suite that takes three
minutes. Only the next test, which runs the same executable, made the gate red.

- src/native/CMakeLists.txt: fl_test_completes(<test> <regex>...) and fl_catch2_test_completes(<test>...) set the
  line a test must print (Catch2: "All tests passed (", or its summary row with a skipped case, never a row that
  names a failure); fl_require_every_test_completes() fails the configure when a test in tests/ or tools/ was
  registered without one.
- All 35 tests go through one of the two, each on the last line it already printed; no program changed.
- Four canaries print what a Catch2 run would - nothing, its last line, a row with a skipped case, a row with a
  failure beside "passed" - through the same helper; the two that must be red are WILL_FAIL.
- fl_nvapi_bridge, fl_process_stats, fl_nvapi_probe and fl_signer_probe passed on a line printed in the middle of
  their run; they now need it and, after it, their last line. fl_nvapi_probe also passed whatever followed its
  branch line - a pass expression makes ctest ignore the exit code.

Docs: 14_TESTING, HANDOFF, the changelog. No shipped code.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 20:36:41 +07:00
poli0981andClaude Fable 5.2 aa30080e06 docs(handoff): 0.1.0-beta.46's release facts - the tag's run, the digests, the delta, the wiki, what SonarCloud and Qodana said on the release's commit; and a local gate that counted a test cut short as passed (#437)
- Released 2026-11-13T11:10:28Z from the signed tag on 47f010a by run 43412434074; eight assets, seven digests
  equal, the build id in four natives, the delta 634,023 bytes, the wiki run. SonarCloud's gate passed on the
  release's commit with no issue open; Qodana's two reports are green on it.
- Security software on the development PC removed the freshly linked fl_guard_test.exe while a local gate ran, and
  ctest reported the run that was cut short as Passed after 0.66 s (the suite takes 181 s); only the next test,
  which runs the same executable, made the gate red. Recorded with a proposal; nothing changed.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 19:27:10 +07:00
poli0981andClaude Fable 5.2 39a747f2a3 spike: Reflex route D on the harness - the driver's two getters from an unshipped probe, against a stand-in module and the installed driver; no game touched (spike-notes §31; D132) (#436)
* spike: Reflex route D on the harness - the driver's two getters called from an unshipped probe, against a stand-in nvapi64.dll and against the installed driver; a decision table written before the runs, and what they showed (spike-notes §31; D132)

No shipped code changes and no game was touched.

- tools/fl-probe-reflex: fl_reflex_candidate.h, the call sequence that would move into the Overlay (find
  nvapi64.dll only if the process already holds it, resolve NvAPI_D3D_GetLatency and NvAPI_D3D_GetSleepStatus by
  interface id, call them SEH-guarded); --stub against tools/vendor-stubs/stub_nvapi64.cpp (ctest
  fl_reflex_probe_stub); --real against the installed driver with the probe playing the game's part - it sets
  latency markers on its own device, in its own process.
- docs/spike-notes/31-reflex-route-d-harness.md: nine pre-committed measurements and their results. No fault
  before initialisation, without markers, or beside a marker-setting thread (38 million reads); a call costs half
  a microsecond; the Direct3D 12 device the Overlay holds is the object the driver takes; the report's unit is
  microseconds. The call belongs on the presenting thread once a second, not on the watchdog. The newest frames
  of a report are not final. Ends in the owner's go or no for one real title.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(spike): §31 in the index, §M14 and §30 point at what the harness spike showed; the probe and its stub in the build document and CLAUDE.md's tools list

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(spike): HANDOFF says the route D spike is done and what a plan must keep from it; the changelog names the probe as a developer's tool

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(probe): the Reflex probe's Measure takes its callable by const reference - a forwarding reference it never forwarded (Qodana native, cppcoreguidelines-missing-std-forward)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(probe): what SonarCloud reported on the Reflex probe - the counter read through the union member it was initialised with, a guard for an empty measurement, NvU32 printed as what it is, a variable no longer named module; and the note says what a fourth run measured (1.5 us a call, 70 us at worst)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 19:05:59 +07:00
poli0981andClaude Fable 5.2 47f010a7b7 release: 0.1.0-beta.46 - hooks go into a game together and come out together: one stop of the game's threads for a batch of patches, not one a patch (D134); the pages brought up to date with the first real sessions on 0.1.0-beta.45; owner-only item 44 (#435)
- docs/changelog/0.1.0-beta.46.md takes [Unreleased] as it stood, under an introduction that says what the
  release is and is not; CHANGELOG.md keeps an empty [Unreleased] and gains the table row.
- docs/HANDOFF.md: the beta.45 section moves verbatim to docs/handoff/history/0.1.0-beta.45.md; a new START HERE -
  what the release decided beyond D132-D135 (the refused first design and the rules that replaced it, the gate, the
  engine's verdict, the fallback), the traps of the train, NEXT, what is left open.
- Owner-only item 44, with what each outcome means written before the run.
- The release lines: ACCESSIBILITY.md, LIMITATIONS.md, README.md, legal/ACCURACY.md, legal/DISCLAIMER.md.

Documents only; nothing under src/ changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 17:51:54 +07:00
poli0981andClaude Fable 5.2 28c7760d28 feat(overlay): patches go in and come out under one stop of the game's threads - MinHook's queue, behind a gate (beta.46 PR-1; D134) (#434)
* feat(overlay): patches go in and come out under one stop of the game's threads - MinHook's queue for the rows a tick finds and for the stop, behind a gate, with only patches that are still ours ever marked for a restore (beta.46 PR-1; D134)

MinHook stops every thread of the process to write a jump, and enabling or disabling one hook did that for one
jump: thirteen Streamline patches took 0.74 s to 2.8 s of a game's start in the owner's sessions of 2026-11-13,
and a stop restored its patches the same way.

- fl_patch_registry.h: fl::patch::Batch (Add: slot and create, nothing marked; Commit: enable them all, then the
  re-read of the flag Install makes) and fl::patch::RestoreAll, the stop's walk. Install and Registry untouched.
- dllmain.cpp: InstallUpscalerHooks queues every row a tick finds something for and commits once, before the
  families published "when whole"; InstallRtHooks batches its two; StopObserving calls RestoreAll. A patch's
  HOOK_INSTALLED line, family bit and latch come in that order and only once it is in. A new log line an apply,
  PATCH_BATCH install|restore (a = microseconds, b = patches). No hook body changed.
- MH_ApplyQueued applies every hook whose mark differs from its state, whoever marked it. So no mark is left
  standing, only a patch that is still ours is marked for a restore and nothing is marked "all" (a draft that did
  could put a patch back in after a stop; refused on review against hook.c), and mark, apply and commit happen
  inside a gate - a slim lock on no hook body's path, never held across a symbol lookup, re-entered by its owner.
  The verdict on an install is the engine's, not the bytes'. An apply stops at its first failure: the hooks
  after it are enabled or restored one at a time.

Measured on hook-harness's Streamline target, three runs before and after: the nine Advanced-only patches,
0.60 / 0.70 / 0.60 s one by one, are one batch of 81 / 90 / 78 ms; the stop's eighteen, 1.37 / 1.56 / 1.37 s,
are one apply of 77 / 90 / 76 ms; unhook requested to UNHOOKED, 1.27-1.48 s, is 94-109 ms.

ctest fl_unhook_inline: six new cases, against real MinHook and against a model of its table. fl_guard's
Streamline [advanced] case asserts the batch lines. No real game has been run.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(overlay, tests): what SonarCloud reported on the batching pull request - the performance counter read through the union member it was initialised with, a test's strlen, and a model index that could be -1

- dllmain.cpp: QpcNow() names QuadPart in the LARGE_INTEGER's initialiser, so the member read back is the member
  the union holds (cpp:S6232, four times on MicrosSince); MicrosSince takes the ticks.
- guard_test.cpp: the log-field helper takes a string_view instead of calling strlen (cpp:S5813).
- unhook_inline_test.cpp: the model's Set refuses an index outside its table and the test marks a foreign patch
  through a method that looks the target up (cpp:S3519: enabled[-1] if Find had failed).

No behaviour changes. The batch timings on the fixture are as before (four in 70 ms, nine in 70 ms, eighteen
out in 76 ms).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 16:41:53 +07:00
poli0981andClaude Fable 5.2 92affcd11e docs: what the owner's seven sessions on 0.1.0-beta.45 showed (item 43), and the four decisions taken on them - Reflex route D to a harness spike with rule 4's note, route C refused and ETW left closed, installs and restores to be batched, slAllocateResources keeps its patch (D132-D135) (#433)
Read 2026-11-13 from a copy of the ledger (sessions 190-196), seven overlay logs, the Agent's log and Windows'
application log. Every session hooked, normal, no fault counted, no PATCH_FAILED. slIsFeatureSupported read in
every Streamline session; slIsFeatureLoaded in three games; slSetFeatureLoaded in two; slFreeResources and
slShutdown once in one game each; slAllocateResources in none. Hogwarts Legacy and Black Myth: Wukong fault
while closing at the offsets they had before. Thirteen interposer patches take 0.7 s to 2.8 s of a game's start.
Off means off and a stop while a game runs were not exercised. The pages that said "not confirmed" are corrected.

The owner's decisions the same day: D132 (route D goes to a spike on hook-harness and a stub; a driver's
read-only getter with the device of the hooked swap chain is inside rule 4 - CLAUDE.md carries the note, and
nothing shipped does it), D133 (route C refused, the ETW route stays closed), D134 (batch installs and restores
through MinHook's queue, released alone as 0.1.0-beta.46), D135 (slAllocateResources keeps its patch).

Documents only.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 15:48:15 +07:00
poli0981andClaude Fable 5.2 d3e8dd3609 docs(handoff): 0.1.0-beta.45's release facts - the tag's run, the digests, the delta, the wiki, Qodana; and SonarCloud's gate, failed on the release's commit by 39 duplicated lines of one enum mirrored by design - the Domain's mirror is left out of duplication detection (#432)
- docs/HANDOFF.md: released 2026-11-12T16:40:42Z from the signed tag on 550d830 by run 43257803723; eight assets,
  seven digests equal, the build id in four natives, the delta 1,762,465 bytes, the wiki run, what Qodana said.
- SonarCloud's quality gate failed on 550d830 on duplicated lines on new code (3.7 % against 3 %), with no issue
  open: the six new members of FlHookFamily and of its Domain mirror HookFamilies, which are the same on purpose
  and held together by MetricEnumMirrorTests. SonarQube.Analysis.xml leaves the mirror out of the detection, with
  the reason there and in docs/13_CI_CD.md.

Whether the gate is green again is for the analysis of this commit on main to say.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-13 00:10:24 +07:00
poli0981andClaude Fable 5.2 550d83089d release: 0.1.0-beta.45 - six more of Streamline's calls, hooked only where asked: whether each feature is supported and loaded, the game switching it, its resources allocated and freed, Streamline shut down; slInit stays out (D129-D131); owner-only item 43 (#431)
- docs/changelog/0.1.0-beta.45.md takes [Unreleased] as it stood, under an introduction that says what the
  release is and is not; CHANGELOG.md keeps an empty [Unreleased] and gains the table row.
- docs/HANDOFF.md: the beta.44 section moves verbatim to docs/handoff/history/0.1.0-beta.44.md; a new START
  HERE - what the release decided beyond D129-D131, the traps of the train, NEXT, what is left open, and one
  finding not acted on: a stop restores its patches one after another, about 160 ms each.
- Owner-only item 43, with what each outcome means written before the run.
- The release lines: ACCESSIBILITY.md, LIMITATIONS.md, README.md, legal/ACCURACY.md, legal/DISCLAIMER.md.

Documents only; nothing under src/ changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 23:21:23 +07:00
poli0981andClaude Fable 5.2 a2235d7075 docs(spike): Reflex latency per frame - the routes there are and what each would cost, a reading with nothing run; the owner's go or no comes before any code (spike-notes §30; D130, §M14) (#430)
Train C's spike as a document. Hooking NVAPI's latency functions is the refused class; hooking Streamline's
Reflex and PCL functions is train D's wrapper and would miss a game that fetched them before the attach;
the Overlay calling slReflexGetState breaks the vendor's stated thread contract. What remains is the
Overlay calling two driver getters with the game's device (no patch; the owner's ruling on rule 4 first)
and a no-injection ETW route that gives the game-side markers only. Every Streamline game the owner ran
asks about kFeatureReflex. nvapi64.dll exports two functions on the dev box, not one.

No shipped code changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 15:59:09 +00:00
poli0981andClaude Fable 5.2 b6d7a58ee1 feat(advanced): a session shows whether each Streamline feature is supported and loaded, and what the game did with it - two facts from six calls, one sentence for their states, Privacy Policy 2.19 (beta.45 PR-3; D129-D131) (#429)
The Agent and the App read what the six hooks of this release keep.

- InGameAdvancedSampler: sl.feature.support (<feature> · slIsFeatureSupported as NVIDIA's own sl::Result
  name, a refusal included; <feature> · slIsFeatureLoaded on or off) and sl.feature.lifecycle
  (<feature> · slSetFeatureLoaded on, off or both; slAllocateResources and slFreeResources as counts,
  ">= 15" where the four bits stopped; a row slShutdown). A feature with nothing to say has no row.
- AdvancedSources: one member a call (Read / NeverCalled / NotHooked), null in every row written before.
  The page says the six in one sentence, each call by its name under what became of it.
- sl.init.preferences has reach BeforeAttach (D129): the game makes the call once, before FrameLedger is
  inside it. No catalog entry waits for a hook any more.
- Privacy Policy 2.19 (wording approved with the plan): what is read for each feature, and two more
  things never read - what identifies the graphics card a game hands over with its question, and the
  content of a graphics resource. "Four further calls" is ten. The Disclaimer stays 2.15.
- Strings in en, vi, ja. LIMITATIONS, the guide, docs 06/08/18, the accuracy block (three copies),
  CLAUDE.md, the changelog.

No native change, no layout change (5), no schema change (23). The Agent's end-to-end test on the
harness's Streamline target stores both facts and all six calls as Read; the FSR target stores NotHooked.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 22:41:09 +07:00
poli0981andClaude Fable 5.2 eef4266152 feat(overlay): hook slAllocateResources, slFreeResources and slShutdown - a feature's resources allocated and freed and Streamline shut down, counted and nothing else, only once the Agent has asked (beta.45 PR-2; D130) (#428)
Three more patch sites on sl.interposer.dll (family bits 20-22), installed only in a process that was asked for
Advanced metrics. The feature's id is the only argument looked at; the command buffer and the viewport are forwarded
and never read. The two that sit on a game's way out note the call and then call the original, and run nothing after
it. Two four-bit counts in word 1 of the feature's slot, and slShutdowns in the first of three reserved words (layout
5 as it was) with FL_ADV_VALID_SL_SHUTDOWN; all three are a session's and a new request clears them.

Measured (the detours' bodies on the headers, the Overlay's flags, 20 M calls x 3): slAllocateResources +0.70/+0.72/
+0.80 ns not asked, +10.43/+10.53/+10.74 asked; slFreeResources +0.61/+0.84/+0.58, +10.33/+10.57/+10.44; slShutdown
within the bench's noise not asked, +3.39/+3.37/+3.52 asked.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 15:20:31 +00:00
poli0981andClaude Fable 5.2 a4baae26fc feat(overlay): hook slIsFeatureSupported, slIsFeatureLoaded and slSetFeatureLoaded - read only where asked, the adapter never read; slInit not hooked (beta.45 PR-1; D129-D131) (#427)
* feat(overlay): hook slIsFeatureSupported, slIsFeatureLoaded and slSetFeatureLoaded - whether Streamline says a feature is supported and loaded, and the game switching it, read only once the Agent has asked; the adapter a game asks about is never read; slInit is not hooked (beta.45 PR-1; D129-D131)

Three new patch sites on sl.interposer.dll, a row and a family bit each (17-19), installed only in a process that
was asked for Advanced metrics (D111). Bare arguments and results, no vendor struct: the feature's id, the sl::Result
of slIsFeatureSupported (a refusal is the answer), the bool slIsFeatureLoaded wrote when it said OK, the bool the
game passed to slSetFeatureLoaded. sl::AdapterInfo is forwarded and never looked at, and the fields gate refuses the
names of its members in code loaded into a game (D131). Packed into word 1 of the feature's slot (layout 5 as it
was); the switch bits are a session's, the answers the process's.

Also: a stop restores a patch in about 160 ms, and the watchdog's wait for the stop's log lines was two seconds
whatever the number of patches - with fifteen the log lost its last UNHOOK_RESTORED lines. The wait is now a quarter
of a second for every patch the registry can hold.

Measured (the detours' bodies on the headers, the Overlay's flags, 20 M calls x 3): slIsFeatureSupported +1.50/+1.48/
+1.50 ns not asked, +11.97/+11.98/+12.08 asked; slIsFeatureLoaded +1.79/+1.77/+1.89, +12.35/+12.31/+12.33;
slSetFeatureLoaded +1.44/+1.47/+1.80, +11.28/+11.54/+11.45.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(harness): compute two feature-slot indices in size_t in the unit probe (Qodana native, bugprone-implicit-widening-of-multiplication-result)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 22:04:47 +07:00
poli0981andClaude Fable 5.2 b04f9f564c docs: what the owner's eight sessions on 0.1.0-beta.44 showed (item 42) - Streamline's answers about a feature were read in every Streamline session, and the pages that said most sessions would show nothing are corrected (#426)
Read from a copy of the ledger, the Agent's log, eight overlay logs and Windows' application log. Both calls read
on five games; the degrees fix reads right on Cyberpunk 2077; a game with no Streamline says not hooked. Not
exercised: off means off. Not explained: Valor Mortis's crash at 51 s (also on beta.40), Black Myth: Wukong's fault
in Steam's crash handler at exit. HANDOFF's NEXT carries the owner's word on the next version (slInit,
slIsFeatureSupported and other parts, to be planned).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 14:07:40 +00:00
poli0981andClaude Fable 5.2 b47777e88d docs(handoff): 0.1.0-beta.44's release facts - the tag's run, the digests, the delta, the wiki, what SonarCloud and Qodana said on the release's commit, and the one flaky check on the way (#425)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 12:39:38 +00:00
poli0981andClaude Fable 5.2 f9e9fdb175 release: 0.1.0-beta.44 - what the first real games on beta.43 found, fixed (a field of view in degrees, the count of early abnormal ends, flags none), and Streamline's answers about a feature hooked only where asked (D126-D128): its changelog file, HANDOFF's beta.44 section, owner-only item 42, the release lines (#424)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 12:04:30 +00:00
poli0981andClaude Fable 5.2 99efc08423 feat(advanced): a session shows which version each Streamline feature is and what it requires; Privacy Policy 2.18 (beta.44 PR-3) (#423)
* feat(advanced): a session shows which version each Streamline feature is and what it requires - two catalog entries read inside the game, and for each of the two calls whether it was read, not hooked, or hooked and not asked again; Privacy Policy 2.18 (beta.44 PR-3)

The Agent turns region 5's feature slots into sl.feature.versions and sl.feature.requirements, rows named
<feature> - <member> in the vendor's words; AdvancedSources says for each call which of three things happened,
and the App words hooked-and-not-seen for a start-up call as what it is: the game did not ask again. The Agent's
end-to-end tests run a Streamline target for the first time. Privacy Policy 2.18 names what Streamline answers
about a feature and what of the answer is never read. No layout change (5), no schema change (23).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* style(recorder): drop two namespace qualifiers the new using made redundant (Qodana managed, RedundantNameQualifier)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 11:31:47 +00:00
poli0981andClaude Fable 5.2 91204ed8d3 feat(overlay): hook slGetFeatureVersion and slGetFeatureRequirements - Streamline's answers about a feature, read only where asked; never the lists the answer points to (beta.44 PR-2) (#422)
* feat(overlay): hook slGetFeatureVersion and slGetFeatureRequirements - what Streamline answers when a game asks which version a feature is and what it requires, read after the original call and only once the Agent has asked; never the lists the answer points to (beta.44 PR-2)

Two new patch sites on sl.interposer.dll, a row and a family bit each (15, 16), installed only in a process that
was asked for Advanced metrics (D111) and the first two rows the watchdog patches. The answer is read only when
the original returned eOk and the struct carries its own GUID: two versions; the flags, three counts and four
versions of the requirements - every one a member of struct version 1 at each of sixteen upstream tags (D115).
requiredTags and the Vulkan members are never read; no pointer of the game's is followed. Twelve slots in region 5
(layout 5 as it was), kept across a new request.

Measured (the detours' bodies on the headers, the Overlay's flags, 20 M calls x 3): the version +0.58/+0.61/+0.58 ns
not asked, +14.82/+14.54/+14.47 ns asked; the requirements +0.58/+0.63/+0.61 ns not asked, +20.88/+21.01/+20.96 ns
asked.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix(tests): compute the index of a feature slot in size_t - three products of two 32-bit values used as a pointer offset, in the injected test and the harness (Qodana native, bugprone-implicit-widening-of-multiplication-result)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 10:47:32 +00:00
poli0981andClaude Fable 5.2 3f16235c73 fix(advanced): a field of view a game gives in degrees is no longer labelled radians, switching hooking back on starts the count of early abnormal ends again, and a frame-generation configure with no flag says none (beta.44 PR-1; D126-D128) (#421)
What the owner's nine sessions on 0.1.0-beta.43 found. The Overlay does not change.
- D128: a field of view over pi is the number as the game gave it, with a row that says it is not radians; never converted.
- D126: the consent grant sets hook_crash_count to 0 - it was never set back, so one early end was enough once a game had had two.
- D127: a fault while a game is closing, unseen by a hooked session, is written down and not changed.
- The accuracy statement and LIMITATIONS say what was read from real games and what was not.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 10:11:23 +00:00
poli0981andClaude Fable 5.2 c8d0731ed8 docs(handoff): what the owner's nine sessions on 0.1.0-beta.43 showed - both new hooks read real games, off meant off, and Cyberpunk 2077 gives Streamline its field of view in degrees (item 41) (#420)
Read from a copy of the ledger, the Agent's log, nine overlay logs and Windows' application log.
Not exercised: a configure through the FidelityFX 2.3 loader, a game with neither library.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 09:44:28 +00:00
poli0981andClaude Fable 5.2 e422a416a2 docs(handoff): 0.1.0-beta.43's release facts - the tag's run, the digests, the delta, the wiki, and what SonarCloud and Qodana said on the release's commit (#419)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 07:15:20 +00:00
poli0981andClaude Fable 5.2 6c7ba74eb0 release: 0.1.0-beta.43 - two hooks installed only where asked for (what a game tells Streamline each frame, how it configures FSR frame generation), the guard reads a moving module list again, an FSR UPSCALE counted once (D123-D125): its changelog file, HANDOFF's beta.43 section, owner-only item 41, the release lines (#418)
docs/changelog/0.1.0-beta.43.md takes [Unreleased]; HANDOFF's beta.42 section moves verbatim to docs/handoff/history/0.1.0-beta.42.md and a new START HERE says what the release decided beyond D123-D125, the traps of the train and what is left open; owner-only item 41 is the first real games with the two new hooks, with what each outcome means written before the run.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 06:38:31 +00:00
poli0981andClaude Fable 5.2 3558284a75 feat(advanced): a session shows what the game told Streamline each frame and how it configured FSR frame generation - six catalog entries read inside the game, and for each of the two calls whether it was read, never called or not hooked; Privacy Policy 2.17 (beta.43 PR-4) (#417)
The Agent and the App for the two hooks of #415 and #416. InGameAdvancedSampler makes sl.constants.jitter, .motion_vectors, .depth, .projection, .resets and ffx.framegen.config from members region 5 already had; the count is brought up to the last publication when the session ends. AdvancedSources gains StreamlineConstants and FsrFrameGeneration (AdvancedInGameCall), from the hook families the writer state published and what the publications carried; the session's sources line says it in en, vi, ja. Privacy Policy 2.17 names the two readings and says the two calls are intercepted only once a session with the option on has started in the game; the accuracy block, LIMITATIONS, the guide and docs 06, 08, 18 follow. No schema change (23), no layout change (5).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 06:22:01 +00:00
poli0981andClaude Fable 5.2 b2e90324c5 feat(overlay): hook ffxConfigure on the four ffx-api modules - how a game configures FSR frame generation, read only while the Agent asks and patched only once it has; never the swapchain, the callbacks, the HUD-less resource or the frame id (beta.43 PR-3) (#416)
Four inventory rows, family FL_HOOK_FFX_CONFIGURE (bit 14), installed only in a process asked for Advanced metrics at least once and published when every loaded ffx-api module is patched. fl_adv_fields.h reads frameGenerationEnabled, allowAsyncWorkloads, flags, onlyPresentGenerated and generationRect of a frame-generation configuration into region 5's ffxFg group, counted at the first module that saw one; the rectangle is kept in one word so it cannot tear, and one that does not fit is four -1. Every configure type is kept for the log (FFX_CONFIGURE). tools/overlay-fields-check.ps1 gains the parameter g and five list entries.

Measured on the headers with the Overlay's flags: +1.7 ns not asked, +15.2 ns on the counted module, +2.7 ns on the forwarded copy.

The test's 'is this export patched' compares the target's bytes with the file's own.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 06:05:06 +00:00
poli0981andClaude Fable 5.2 b4feea1b0b feat(overlay): hook slSetConstants - what a game tells Streamline about each frame, read only while the Agent asks and patched only once it has; never the matrices or the camera's position and axes (beta.43 PR-2) (#415)
A new inventory row, sl.interposer.dll!slSetConstants, family FL_HOOK_SL_CONSTANTS (bit 13). InstallByFamily returns before resolving the symbol in a process never asked for Advanced metrics (D111), so nothing is patched with the option off. The detour checks the request, MayObserve, the struct's GUID and version and the viewport handle's on every call; fl_adv_fields.h reads jitterOffset, mvecScale, cameraNear, cameraFar, cameraFOV, cameraAspectRatio and the seven Booleans into region 5's sl group, summarised for the first viewport seen. tools/overlay-fields-check.ps1 gains the parameter k and thirteen list entries; the camera's names are still refused everywhere.

Measured on the headers with the Overlay's flags: +0.6 ns not asked, +19.6 ns asked on the summarised viewport, +5.8 ns on another.

ctest fl_adv_fields and fl_guard [advanced]: never asked - the export's first 16 bytes in the target are the file's; asked - the first viewport's numbers and none of the second's; no longer asked - the region unchanged; stopped - the bytes restored.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 04:43:24 +00:00
poli0981andClaude Fable 5.2 31d79184eb fix(overlay): an FSR UPSCALE is counted at the first source that saw one - a loader that hands a dispatch on to the leaf's export brought it to two hooked modules, and the frame count would have doubled with frame generation off (beta.43 PR-1, D125; decisions D124-D125) (#414)
g_ffxUpscaleOwner gates the count and the source word through fl::adv::Owns; the FSR 3.0 host is a source like the four leaves; the render size and the quality byte are not gated. Measured on Cyberpunk 2077's loader 2.3.0 for frame-generation dispatches; no title run so far upscales through such a loader, so it is a precaution, held by the stub loader bound through the leaves' exports (hook-harness --ffx-topology 2x-export): 60 presents read as 120 frames before, 60 after.

Measured on the header's Owns with the Overlay's flags: +0.02 to +0.03 ns on the counted module.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 04:22:06 +00:00
poli0981andClaude Fable 5.2 ab8b4ac158 fix(guard): a module list that changed while it was read is read again before the guard refuses (D123) (#413)
* fix(guard): a module list that changed while it was read is read again, at most three times, before the guard refuses - launches were being lost as a safety unhook with no frame recorded (owner decision D123)

Collected::kUnsettled for a module that was in the list and could not be named (until now kIncomplete, an immediate refusal). CheckModules re-reads the whole scan set from a fresh snapshot after kScanRetryPauseMs, up to kMaxUnsettledRereads (3) times; a list still unsettled on the last read refuses kModuleScanFailed with its own signal. A finding - also among the modules that were read from a half-read list - a protected process and a list too long for the buffer still refuse at once; D48's re-read stays one. The launch poll's runtime probe asks again on an unsettled list.

ctest fl_guard [D123]: ten sections against fakes, and the real source against a process that loads and frees a module in a loop (hook-harness --churn-module): about one read in four unsettled, none anything else.

HANDOFF: D123; the Cyberpunk FSR 3 session cannot be run (the game's only FSR upscaler is 2.1), D121 closed.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(guard): the live D123 case counts its answers with an if-chain - Qodana's switch-enum on a switch that named two of five values

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 04:05:20 +00:00
poli0981andClaude Fable 5.2 6d4e799f68 docs(handoff): what the first four sessions on beta.42 showed - Cyberpunk's upscale was FSR 2.1, built into the game (D121) (#412)
* docs(handoff): what the owner's first four sessions on 0.1.0-beta.42 showed - Cyberpunk's upscale was FSR 2.1, built into the game (D121 answered); the 2.3 loader re-enters the leaf's export; the plane at infinity measured on two titles

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): three more starts of Cyberpunk on beta.42 - the game was still on FSR 2 (the settings file says so), and the first start was lost to a module list read while the game was loading: five such ends in the Agent's logs since 2026-11-05, and a decision for the owner

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* docs(handoff): the owner's screenshot of Cyberpunk's Graphics menu confirms it - Resolution Scaling is FSR 2.1

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 03:29:32 +00:00
poli0981andClaude Fable 5.2 e2de7cfbc0 docs(handoff): 0.1.0-beta.42's release facts - the tag's run, the digests, the delta, the wiki run; SonarCloud's one finding fixed (a strlen in a test helper); two transient failures recorded (#411)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 03:05:42 +00:00
poli0981andClaude Fable 5.2 a708ae8fed release: 0.1.0-beta.42 - no new patch site: the session's upscaler label, the patch registry, a native log that outlives the game, FSR's plane at infinity (D119-D122): its changelog file, HANDOFF's beta.42 section, owner-only item 40, the release lines (#410)
docs/changelog/0.1.0-beta.42.md takes [Unreleased]; HANDOFF's beta.41 section moves verbatim to docs/handoff/history/0.1.0-beta.41.md and a new START HERE says what the release decided, what to read in the Cyberpunk log and the design of beta.43's two hooks; owner-only item 40; the release lines in README, LIMITATIONS, ACCESSIBILITY and the accuracy block's copies. No legal version, schema, rules or layout change.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 02:09:22 +00:00
poli0981andClaude Fable 5.2 c74de2cc12 feat(advanced): FSR's depth plane at infinity is stored as +infinity and shown as an infinity sign - FLT_MAX or +infinity in cameraNear / cameraFar is a plane for FSR, where for Streamline it is 'not set' (beta.42 PR-4, D119) (#409)
FL_ADV_FLOAT_INFINITE (0x7F800000) in ffxNearBits / ffxFarBits only; fl::adv::FfxDepthPlaneBits on the ffx route only; InGameAdvancedSampler shows the row as the infinity sign. No field moved, the layout version is unchanged. The layout's comments carry the release numbers of D118 and D122.

Measured on fl_adv_fields.h with the Overlay's flags, old header beside new: no difference beyond the run-to-run spread (+25.0 to +25.4 ns per UPSCALE dispatch with Advanced on, +25.2 to +25.8 before; off unchanged).

Qodana's native report: six false 'never used' on the registry test's stand-in engines switched off for that file with the reason; one vector reserve fixed.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 01:51:25 +00:00
poli0981andClaude Fable 5.2 72ee9964b0 feat(overlay): the native log is whole after a game that simply exits, and says which module was patched and which descriptor types each ffx module was handed - the watchdog flushes when something new was logged (beta.42 PR-3, D121) (#408)
The log held the first half second of a session unless the Agent asked for it or a stop ran, and a game that exits does neither. FlushLogIfNew on the watchdog's tick; the init flush moved before the watchdog exists. Install lines name module!symbol from the inventory row (fl::inventory::NameOfRow; the event's text is 72 characters). fl_ffx_types.h keeps each descriptor type a hooked ffx-api module was handed, per entry point, and the watchdog logs it once (FFX_DISPATCH / FFX_QUERY), with UPSCALE_ROUTE and RUNTIME_CENSUS from words the detours already kept. A diagnostic: the log file and nothing else.

Measured on the header with the Overlay's flags: +3.8 ns per dispatch for a frame's three types; +0.9 ns first slot, +8.4 ns eighth.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 01:36:55 +00:00
poli0981andClaude Fable 5.2 ca7d9cb028 fix(overlay): the patch registry refuses a patch it cannot record and records a patch before it can run - a full table no longer leaves a live patch the stop does not know, and a stop that arrives while a hook goes in can no longer miss it (beta.42 PR-2) (#407)
fl_patch_registry.h: slots Free -> Claimed -> Live -> Retired and one sequence, fl::patch::Install - claim, create, enable, commit, re-read the flag - against a stop that clears the flag and then walks the committed slots, both sequentially consistent. PatchTarget, the loader detour and the three DXGI patches go through it; a patch that did not go in is logged once (PATCH_FAILED); the capacity is asserted against the inventory's row count; READY is written only over INIT and init's failure path stores nothing, so neither overwrites a stop's reason. No hook body changed.

ctest fl_unhook_inline drives the shipped sequence against real MinHook patches (+3 cases).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 01:17:10 +00:00
poli0981andClaude Fable 5.2 96cb45605d fix(metrics): a session's upscaler is the technology named on the most samples, with that technology's quality and render size - not the first one named beside the others' values (beta.42 PR-1, D120; decisions D119-D122) (#406)
A session that evaluated DLSS briefly and then ran FSR was stored as DLSS with FSR's quality byte and render size. Domain.Metrics.UpscalerChoice picks the modal named technology over the dominant stream; with more than one named, quality and size are read only from the samples that name it, a quality byte of the other family is withheld, and the session is marked settings-changed. One technology named keeps the whole stream, as before. The live card is unchanged.

HANDOFF and 18_GPU_VENDOR_APIS record D119-D122: train B's second step is beta.42 (no new patch site) and beta.43 (the two new hooks).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 01:00:26 +00:00
poli0981andClaude Fable 5.2 283e8b632e docs(handoff): owner-only items 38 and 39 passed on the owner's word - moved to owner-only-passed with what the sessions covered and what they did not; beta.42 begins with its plan (#405)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 00:31:56 +00:00
poli0981andClaude Fable 5.2 37c013e5e9 docs(handoff): what the owner's sessions on beta.40 and beta.41 show - the sensor-library fix over seven sessions, the first FSR 3.1 title read, Valor Mortis's crash was the game's own (#404)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-12 00:17:37 +00:00
poli0981andClaude Fable 5.2 b875be8f0d docs(handoff): 0.1.0-beta.41's release facts - the tag's run, the digests, the delta, the wiki run (#403)
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 23:51:28 +00:00
poli0981andClaude Fable 5.2 a6b95e8bc2 release: 0.1.0-beta.41 - the sensor-library fix, released at once; train B's numbers move by one (D118): its changelog file, HANDOFF's beta.41 section, owner-only item 39, the release lines (#402)
docs/changelog/0.1.0-beta.41.md holds the two [Unreleased] entries of #400 and #401. HANDOFF's START HERE for
0.1.0-beta.40 - with what the owner's first sessions showed - moved verbatim to
docs/handoff/history/0.1.0-beta.40.md; the new one records D118, NEXT and what is left open. Owner-only item 39:
the fix on the PC that showed the fault.

D118 (owner decision 2026-11-12): the fix takes the number D113 gave train B's second release; that one is
beta.42, the third beta.43, train C beta.44, train D beta.45.

The code loaded into a game is 0.1.0-beta.40's, line for line. Schema 23, rules 2026.11.2, Privacy Policy 2.16
and Disclaimer 2.15 unchanged.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 22:14:25 +00:00
poli0981andClaude Fable 5.2 d46a32c166 fix(telemetry): one sensor library a process - every LibreHardwareMonitor Computer is opened and closed through LhmLibrary, serialised, and parked rather than closed while another is in use (#401)
Found in the owner's first sessions as administrator with PawnIO (0.1.0-beta.40): after seven of ten sessions the
Advanced reader's processor device threw NullReferenceException twice and was left alone, and once the reader's
Open threw ObjectDisposedException and the session stored 17 series instead of 212. The Agent holds up to three
Computers at once; the library keeps the driver it reads the processor through and its bus mutexes per process,
and Computer.Close() lets go of them whoever else is reading. Read from the log's timing, not reproduced: it needs
that elevation.

LhmLibrary (one per process; LhmComputerAdapter is its user): opens and closes never run at once, and a Computer
closed while another is in use is parked and handed to the next user of the same groups; the last user's close
closes them all, even when one throws. The reads are not serialised - the readers keep their own threads.

20_OPEN_QUESTIONS M13 records the first elevated run (119 processor, 14 memory, 21 motherboard series read; no
value compared with another tool) and stays open: whether the faults are gone is the owner's next sessions to
show. HANDOFF records what the same ledger says of beta.40's in-game readings.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 22:00:47 +00:00
poli0981andClaude Fable 5.2 128321b129 docs(handoff): 0.1.0-beta.40's release facts; SonarCloud on #398 - two float comparisons told by their bits, and the accepted DangerousGetHandle said again where it moved to (owner-only item 38 (j)) (#400)
HANDOFF records the release: 2026-11-11T19:30:13Z, the signed tag on 5268332, run 43085044822, eight assets, the
seven digests, FL_BUILD_ID v0.1.0-beta.40, the delta of 2,413,166 bytes, the wiki run, the rehearsal.

InGameAdvancedSampler told one value from a range with == on two doubles (csharpsquid:S1244, twice); it now
compares the stored bits. ParentDump's DangerousGetHandle is the call accepted in beta.24, raised anew because
#398 moved it; it is unchanged, its reason is on the line and in 13_CI_CD, and accepting it again is the owner's.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 19:56:41 +00:00
poli0981andClaude Fable 5.2 52683321ae release: 0.1.0-beta.40 - Advanced metrics inside the game, the first of train B's three releases: its changelog file, HANDOFF's beta.40 section, owner-only item 38, the release lines (Advanced B PR-5) (#399)
docs/changelog/0.1.0-beta.40.md holds the [Unreleased] entries of #395-#398 under the release's heading. HANDOFF's
START HERE for 0.1.0-beta.39 moved verbatim to docs/handoff/history/0.1.0-beta.39.md; the new one records what the
train decided that the plan had not, that nothing of beta.41 begins before the owner has run item 38, what is left
open and the traps. Owner-only item 38: this release on real games.

The code loaded into a game changed in this release (the Overlay reads more members of calls it already hooked,
only while the Agent asks; no new patch site; shared-memory layout 5). Schema 23 and rules 2026.11.2 unchanged.
Privacy Policy 2.16, Disclaimer 2.15.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 19:06:31 +00:00
poli0981andClaude Fable 5.2 9c55dc1179 feat(advanced): a session that records Advanced metrics asks the Overlay for what the game tells DLSS and FSR and stores it - thirteen catalog entries read inside the game, two series, the session's in-game state; Privacy Policy 2.16; a crash dump tried three times (Advanced B PR-4, D111/D114) (#398)
ICaptureObserver.WantsInGameAdvanced: CaptureSession writes the control word either way as the ring is attached,
before its first guard result, and carries each publication of region 5 on CaptureProgress.InGameAdvanced as one
instance. A session that did not ask never reads the region.

Application.Capture.InGameAdvancedSampler turns a publication into session facts (reach Game: the DLSS output
size, presets, HDR input, auto-exposure, alpha upscaling, exposure; FSR's sharpening, largest jitter, exposure,
depth range, the RANGE of the field of view, flags) in the vendor's own words, and into two series
(adv.ffx.upscale.frame_time_delta, adv.ffx.memory_usage). Zero is "not read", never a value. No schema change:
facts_json, sensor_blobs and the .partial carry them. AdvancedSources.InGame says what became of them - read,
nothing seen, not hooked - and is absent in every row beta.39 wrote.

The App: the session's sources line and the Sources tab say it; fourteen strings and the Settings text in en, vi,
ja. Privacy Policy 2.16 (content approved with the plan), the accuracy statement, the guide, LIMITATIONS.

Also: SonarCloud's three findings on #397 (cpp:S6232 in the watchdog's QPC read, two cpp:S2259 in the new fl_guard
cases); ParentDump.WriteDump tries MiniDumpWriteDump up to three times - a running process changes under it
(ERROR_PARTIAL_COPY), which failed CrashDumpHelperTests under the gate's load for the third time in two days.

Verified end to end with the shipped Agent, the real Overlay and the harness: two sessions of one target, the
first (option off) storing no row and no adv. series, the second the exact values the target sends.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 18:53:18 +00:00
poli0981andClaude Fable 5.2 4051dba5c7 feat(overlay): the Overlay reads what a hooked game tells its upscaler, only while the Agent asks - DLSSOptions chained into the evaluation, the rest of the FSR upscale descriptor, FSR's own memory; no new patch site; each session's request its own; a gate that refuses a camera's position anywhere in code loaded into a game (Advanced B PR-3, D111/D114/D115/D116) (#397)
fl_adv_fields.h is the one place a vendor struct's extra members are read: sl::DLSSOptions at the struct version
that carries each member, ffxDispatchDescUpscale on the ffx-api route, the upscaler's memory query after the vendor
answered OK. The detours consult one process-local flag the watchdog copies from FlControlBlock.advancedRequested;
with it at 0 nothing more is read and region 5 is never written.

"On" is a request, not a 1: a value no session of the ring used before. The watchdog starts its summaries afresh at
a request it has not seen and names it in every publication (FlAdvancedState.request, the word that was reserved0);
ShmRingReader.AdvancedState returns only a publication that names its own. A second session of a game that kept
running no longer reads the first one's ranges and counts.

tools/overlay-fields-check.ps1 (build.ps1 gate): every member read is on tools/overlay-allowed-fields.txt and every
name there is read; no alias; no listed member read elsewhere in the Overlay or the Vulkan layer; a camera's
position, axes and matrices named nowhere, whatever the list says; sl::Constants and the frame-generation prepare
descriptor never copied or measured.

Measured (the header's functions, the Overlay's flags, 20 M calls x 3): +22.3-23.2 ns per ffx UPSCALE dispatch,
+9.5-11.7 ns per evaluation that chains DLSSOptions, +4.3-6.1 ns per memory query while asked; inside the bench's
noise when not asked.

ctest 33/33 (fl_adv_fields; fl_guard: every byte of region 5 zero until asked on a target that sends everything,
the exact values when asked, untouched after an off, a count that starts again at the next request).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 18:08:34 +00:00
poli0981andClaude Fable 5.2 546dd9bcbf feat(shm): layout 5 - region 5, FlAdvancedState, for what a hooked game tells its upscaler, and the control word that asks for it; sized once for train B's three releases, written by nobody yet (Advanced B PR-2, D111/D113) (#396)
fl_shm.h: FlAdvancedState (1472 bytes at 0x100; the ring moves to 0x6C0) and FlControlBlock.advancedRequested (took
reserved[0]). While the word is 0 the Overlay is to read and hook exactly what it does today, and the region stays
zero. One writer - the watchdog, once a second; a vendor Boolean or preset is stored plus one and a float as its
bits with a "none" pattern, so zero is "not read" and never a value. The members beta.41 and beta.42 fill are
declared now, so those releases bump no layout.

The Agent's side: ShmLayout (layout 5, the mirror), ShmRingReader.AdvancedState - null until published, then the
last consistent copy - and SetAdvancedRequested, which stores the word either way because the control block
outlives a session. ICaptureSink gains both with default bodies; nothing calls them yet. fl-layout-dump prints every
member and ShmLayoutMirrorTests checks each offset. No hook body changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 17:21:33 +00:00
poli0981andClaude Fable 5.2 3a4765ef55 docs(advanced): train B planned - three releases, the struct-version table from the upstream tags, the Agent's cost measured in a real session and NFR-2 amended to 2.5 % (Advanced B PR-1, D110-D117) (#395)
Train B (what only a hooked game can say) is three releases: beta.40 more fields of calls already hooked and no new
patch site; beta.41 slSetConstants + ffxConfigure; beta.42 slGetFeatureVersion + slGetFeatureRequirements. Inside the
game too, off means nothing more is read. docs/17_HOOK_ENGINE.md gains which struct version carries which field,
compared across sixteen Streamline and ten FidelityFX tags. tools/agent-cost.ps1 measures the shipped Agent in a
session: 1.93 % of one core off, 2.04 % on; the owner amended NFR-2 to 2.5 %. HANDOFF records beta.39's release facts.
ShmDrainIntegrationTests waits Init out and prints the guard's verdict.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 16:58:01 +00:00
poli0981andClaude Fable 5.2 e63cd496be release: 0.1.0-beta.39 - Advanced metrics, train A: its changelog file, HANDOFF's beta.39 section, owner-only item 37, the release lines (Advanced PR-8) (#394)
* release: 0.1.0-beta.39 - Advanced metrics, train A: its changelog file, HANDOFF's beta.39 section with what the train decided, the traps and the next work (ask M16, then train B's plan), owner-only item 37, the release lines (Advanced PR-8)

docs/changelog/0.1.0-beta.39.md holds [Unreleased] as it stood, under the release's summary: a setting, off by
default, and a page of its own; Privacy Policy 2.15 and Disclaimer 2.15 shown once; schema 23. docs/HANDOFF.md's START
HERE for beta.38 moves, verbatim, to docs/handoff/history/0.1.0-beta.38.md; the new one carries D109, what was
decided that the plan had not, what is left open - nothing was seen in the running App, and the first run as
administrator with PawnIO is still to come - and the traps this train paid for. Owner-only item 37 is what only the
owner's PC can show.

git diff v0.1.0-beta.38 -- src/native/FrameLedger.Overlay src/native/FrameLedger.VkLayer src/native/FrameLedger.Shm
src/native/FrameLedger.Injector is empty: the code loaded into a game is beta.38's, line for line.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(app): the page-load fixture loads Settings' System card once and awaits it - two pages built a moment apart could differ by a row, and the theme-switch comparison was red once on a hosted runner (Advanced PR-8)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 13:39:26 +00:00
poli0981andClaude Fable 5.2 15aaa65059 feat(app): the Advanced page - This PC, Session and Sources; its navigation item and Ctrl+6 there only while the option is on; every catalog entry in en, vi and ja; a guide page (Advanced PR-7, D101/D105/D107) (#393)
Pages/AdvancedPage over ViewModels/AdvancedViewModel. This PC is what the Agent reads now, asked only while the page
is on screen and never stored: the facts under the catalog's groups, every series with the time it was read at, and
one entry's last five minutes. Session is what a session recorded with the option on stored: the PC as it was, the
session's own facts, the statistics, and the entry's series over the session. Sources is every catalog entry with
whether it is read, or why not.

The navigation item is the menu's last and is bound to Services.AdvancedOption.IsOn - set at start and by the
Settings switch - and Ctrl+6 opens the page only while it is on; Ctrl+1..5 keep their pages. A session's summary has
"Advanced metrics..." where the session stored them. A viewer opens on Session and asks no Agent.

245 strings in en, vi and ja (ja marked for review): the English label of an entry is its summary in the catalog. A
family of more than eight series is drawn as its lowest, average and highest. ISessionRepository.ListAdvancedAsync;
SessionSeriesLoader.LoadAdvancedAsync. The guide gains a page; ACCESSIBILITY.md, LIMITATIONS.md, 08_UI §Advanced, the
accuracy block and Privacy Policy 2.15 follow. Nothing in src/native changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 13:26:09 +00:00
poli0981andClaude Fable 5.2 4984eea770 feat(ipc): GetAdvancedMachine - what the Agent's advanced reader last read, answered without touching a sensor, each request the page's hold on it; and the App's feed (Advanced PR-6) (#392)
* feat(ipc): GetAdvancedMachine - what the Agent's advanced reader last read, answered without touching a sensor, each request the page's hold on that reader; and the App's feed that asks while a page watches (Advanced PR-6, D101/D107)

Application.Ipc.AdvancedMachineResponder answers from IAdvancedHardware.Latest alone: the pipe's read loop awaits its
handler, so a driver call there would hold up Ping and StopSession. The first request acquires the shared reader and
each renews a five-second lease; a lease after the last one the hold is let go - a page that closes, an App that
exits and a pipe that drops all look the same. With telemetry.advanced off the answer is Error AdvancedOff and
nothing is held. The facts travel once, by their hash; a reading that is not a number is left out. Additive,
protocol 2.

App.Services.AdvancedMachineFeed asks once a second while a page holds Watch(), and keeps the facts, the last value
of every series with the tick it was read on, five minutes of each for the live charts and the sources' states.
It stores nothing, and no page watches it yet. Nothing in src/native changes.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* test(display): the monitor a game's window is on, asked by a named window as Observe is - and tested against the harness's window (Advanced PR-6; SonarCloud's coverage of PR-5's new code)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 12:33:55 +00:00
poli0981andClaude Fable 5.2 f780738bc2 feat(advanced): the switch - Settings > Recording > Advanced metrics (telemetry.advanced, off by default); on, a session holds the Agent's one reader and stores its series and facts; Privacy Policy 2.15 and Disclaimer 2.15 (Advanced PR-5, D101/D109) (#391)
telemetry.advanced becomes RecorderOptions.AdvancedMetrics at each session's start. The Agent registers one
AdvancedHardwareHub; SessionRecorder holds it for a session's length only when the option is on, and the session's
TelemetryPoller gets it on the same condition. A session's facts are read once, off the capture loop, fifteen seconds
after the game's process is known or at the session's end; which display is the game's comes from Windows
(WindowGeometryProbe.MonitorDeviceOf). Off: not held, not asked, no chunk 9 or 10, no session_advanced row.

Settings gains the row (off in a viewer), strings in en/vi/ja, the guide's settings page. Privacy Policy 2.15 names
what is read and stored only while the option is on, and what never is. Disclaimer 2.15 (the owner's answer of
2026-11-11, D109): PawnIO and the admin mode are no longer "for the CPU temperature" alone; the admin mode's
disclosure follows (agent-admin-dialog/2), with Settings, the first run, the guide and LIMITATIONS.md. The accuracy
block gains an item. Nothing in src/native changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 12:11:57 +00:00
poli0981andClaude Fable 5.2 e0bfbe6edf feat(recording): the ledger can hold Advanced metrics - schema 0023 (two tables, no column on sessions), adv.* series with NaN for a gap, two .partial chunks, and an unknown chunk is skipped (Advanced PR-4, D102/D106) (#390)
Produced by nothing yet: no Agent code composes the advanced reader, so no
session stores any of it and nothing a user sees or records changes.

- Schema 0023, the first CREATE TABLE since 0001: machine_profiles (the
  PC's facts, one row per distinct set, by the SHA-256 of their JSON) and
  session_advanced (one row per session recorded with the option on - its
  facts, each source's state, its series' statistics and directory). No
  column on sessions: every read of that row, the JSON export and the bug
  bundle carry all of it. A session recorded with the option off has no
  row, and that absence is how the App will say so.
- The advanced series are rows of sensor_blobs under names that start with
  adv., with NaN - not -1 - for a tick with no reading: a -12 V rail is
  negative, and the sensor library is read every fifth tick. A second
  encoder, ISeriesCodec.EncodeSparseFloat32, allows NaN and still refuses
  an infinity; EncodeFloat32 still throws on a NaN. FindSensorsAsync, which
  every summary, chart and export reads, leaves the adv. rows out;
  FindAdvancedSensorsAsync and FindAdvancedAsync read them.
- Application.Advanced.AdvancedJson: the one stored form. Writing is
  strict, reading is tolerant, and only a fact whose catalog entry is read
  may be written (D104 held at the door of the ledger).
- AdvancedSessionBuilder, both tiers: one series per advanced reading,
  aligned to t_ms; at most 256 a session, the rest counted. With no reading
  and no context it builds nothing, and a session is byte for byte what it
  was.
- TelemetrySample.Advanced, and the poller's optional IAdvancedHardware:
  held from Start to Dispose, each of its ticks attached to one sample and
  never repeated on the next.
- The .partial: chunk 9 (the readings of the sensor chunk before it, with
  their own names) and chunk 10 (the session's context), written only when
  there is something to write; a recovered session keeps both. D106: a
  chunk of a type the reader does not know, whose length and checksum are
  right, is skipped where it ended the read - so a build one release older
  no longer loses everything after the first new chunk.
- LedgerDatabaseTests' rewind helper undoes a script's CREATE TABLE and
  CREATE INDEX as well as its ADD COLUMN: 0023 would otherwise have failed
  every test that rewinds a ledger.
- Measured: one hour of 58 series shaped like the dev box's reader is 816
  KiB raw and 189 KiB stored (NFR-5 allows 3 MB an hour).
- docs: 06_DATA_MODEL (schema 0023, the NaN rule, chunks 9 and 10, D106),
  18_GPU_VENDOR_APIS, CLAUDE.md, CHANGELOG.

Hook overhead: not applicable - nothing in code loaded into a game
changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 09:37:53 +00:00
poli0981andClaude Fable 5.2 c4318b69d5 feat(telemetry): the reader behind Advanced metrics - one for the Agent, on its own thread, only while held; LibreHardwareMonitor's card, processor, memory and board groups; measured before anything composes it (Advanced PR-3, D103/D107) (#389)
Nothing calls it yet: no Agent code composes the reader, so nothing a user
sees or records changes.

- Application.Advanced: the port IAdvancedHardware (Acquire, Latest,
  ReadSessionFacts), AdvancedSnapshot, AdvancedSources and the status of
  each source - Reading, NotPresent, NotElevated, PawnIoMissing, Faulted.
  A source that is not there is a state, never an exception and never a
  value in its place.
- Infrastructure: AdvancedHardwareHub - one reader for the whole Agent on
  a thread of its own (fl-advanced), started by the first holder and
  closed a linger after the last lets go (D107), so a PC idling with
  FrameLedger in the tray reads nothing for this. AdvancedHardwareReader -
  one tick of NVAPI and of LibreHardwareMonitor. LhmAdvancedReader over a
  Computer of its own: every sensor of the graphics card, and as
  administrator with PawnIO the processor, memory and motherboard groups
  (D103). LhmGroups has a member for those four and no other.
- Storage is not opened (D103's outcome, written before any run):
  LibreHardwareMonitor 0.9.6's StorageGroup lists every drive through
  DiskInfoToolkit with no filter by bus and follows drives plugged in
  later. lhm.storage.* is Deferred in the catalog; a drive's temperature
  will read N/A with that reason.
- A device that throws twice is left alone and the others go on; NVAPI
  answering with no field twice running, after it had answered, is
  released and initialised again once (20_OPEN_QUESTIONS M12).
- FrameLedger.CaptureHost probe-advanced (unshipped): runs that reader on
  the caller's own rights and prints what it read and what it cost. On the
  dev box (RTX 5080, unelevated): 35 facts, 58 readings a tick; one NVAPI
  live read 0.4 ms of processor time; one refresh of the sensor library's
  graphics-card group 14 ms (62 ms elapsed) - 1.4% of one core at one a
  second, above NFR-2. So NVAPI is read every tick and the sensor library
  every fifth (0.28%), with no value repeated between.
- The same refresh is what the L2 telemetry layer does once a second in
  every session. That cost is not this train's: recorded as
  20_OPEN_QUESTIONS M16, with the owner's choices.
- PrivilegeFootprintTests: opening every group this project can name
  enables no powerful privilege (the real thing on the elevated CI runner).
- docs: 18_GPU_VENDOR_APIS (the reader, D103 as found, L2's line),
  20_OPEN_QUESTIONS (M13 half answered, M16 new), HANDOFF, CLAUDE.md,
  CHANGELOG.

Hook overhead: not applicable - nothing in code loaded into a game
changes.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 08:58:39 +00:00
poli0981andClaude Fable 5.2 42fb21bb18 feat(telemetry): the NVAPI bridge's ABI 3 - the reads "Advanced metrics" needs, measured on an RTX 5080, mapped to the catalog; and "never a setter" as a build gate (Advanced PR-2, D101/D104/D108) (#388)
FrameLedger.NvapiBridge.dll - loaded by the Agent into its own process,
never into a game - gains five reads. Nothing composes them yet: no Agent
code calls the reader, so no behaviour changes.

- Native (fl_nvapi_bridge.h/.cpp, ABI 3, additive): FlNvReadFacts (the
  card's type, bus, PCI ids, VBIOS, architecture, core counts, base and
  boost clocks, memory sizes, the driver's package, the chipset),
  FlNvReadLive (free video memory with evictions and promotions, the
  frame-buffer / video / bus utilisation, video clock, P-state, board and
  power-supply temperatures, the throttle word, encoder statistics),
  FlNvReadDisplays (HDR formats and luminances, variable refresh,
  adaptive sync per connected display), FlNvSessionDisplay (one display's
  variable-refresh state and one process's colour space) and
  FlNvProfileSettings (every setting of the global profile, or what a
  game's own profile sets). Each struct carries its size and a present
  mask; a clear bit is N/A, never 0. No ABI 2 struct changed size, and
  FL_NV_FIELD_VRAM stays clear, so which layer fills vram_adapter does
  not change.
- No identifier has an entry point (D104): no board serial, unique id,
  EDID, or list of encoding programs. A display is passed on by its place
  in the list, never by NVAPI's handle.
- tools/nvapi-readonly-check.ps1 (D108), wired into build.ps1: every
  NvAPI_* call of the bridge's source is on tools/nvapi-allowed-calls.txt
  and every name there is called - forty-four - and a name that sets,
  saves, restores, registers or controls, or returns an identifier, is red
  even when the list names it. Seventeen self-test cases, both directions;
  the self-test caught two holes in the rule as first written.
- Managed: mirrors of the five structs, INvapiAdvancedBridge (a port of
  its own; a DLL that predates ABI 3 answers Unavailable, never throws),
  NvapiAdvancedReader (the structs become the catalog's facts and
  readings - every nvapi. entry of the Agent's reach and no other, held
  in both directions) and Application.Advanced's value model: a fact is a
  number, the vendor's words, our own tokens the App will translate, or
  rows; one that was not read is absent.
- Measured the day it was written (ctest fl_nvapi_bridge, AVAILABLE
  branch, RTX 5080): all twenty facts answer; eight of ten live groups -
  the card has no board or power-supply temperature sensor; the display
  answers its four groups; the global profile lists 24 settings.
  20_OPEN_QUESTIONS M12 is half answered, with what is still open.
- The vendored nvapi.h misspells NV_ENCODER_STATISTICS_VER; the bridge
  names NV_ENCODER_STATISTICS_VER1.
- Fixed from #387: Qodana's one result, a redundant "!" in
  AdvancedMetricCatalogTests.
- docs: 18_GPU_VENDOR_APIS (ABI 3, the gate), 12_BUILD (gate 14e),
  CLAUDE.md, CHANGELOG.

Hook overhead: not applicable - nothing in code loaded into a game
changes (Overlay, VkLayer, Shm, Injector untouched).

Gate on the dev box: all passed in 301 s (native 32/32).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 08:30:54 +00:00
poli0981andClaude Fable 5.2 3e810d1aa6 docs: "Advanced metrics" planned - the owner's decisions D101-D108, the inventory of what NVAPI, LibreHardwareMonitor, Streamline and FidelityFX make available, and the catalog every later pull request reads (Advanced PR-1) (#387)
The owner asked for a setting, "Thông số nâng cao", with a navigation item
of its own that shows as much as the NVAPI and Streamline headers and PawnIO
make available, each value under what it belongs to: one game session, or
the PC. Planned 2026-11-11 with the owner's decisions asked while planning;
four trains, A (0.1.0-beta.39, the Agent's reads only) to D. This is the
first pull request of train A and changes no behaviour.

- FrameLedger.Application.Advanced: AdvancedMetricCatalog, 137 entries -
  86 the Agent will read (60 NVAPI, 10 LibreHardwareMonitor unelevated, 16
  through PawnIO as administrator) and 51 that are not read, each with its
  reason: a call already hooked, a hook not built, a route a spike decides,
  an identifier (D104), the player's own state (rule 4), a licence, a
  function the public headers do not declare. One list for the reader, the
  recorder, the page and the documents.
- docs/18_GPU_VENDOR_APIS.md section Advanced metrics: the decisions, what
  belongs to what, the four trains, what is left out by name, and the
  inventory as a table. AdvancedMetricCatalogTests fails when the table and
  the catalog differ in either direction (proven red on one changed row).
- docs/18: the Memory row said the vendored NVAPI headers do not declare
  NvAPI_GPU_GetMemoryInfoEx. They do (nvapi_lite_common.h:600, included by
  nvapi.h); the search had looked in nvapi.h alone. Corrected.
- docs/HANDOFF.md: the plan's order and the owner's four answers, D101-D108
  in the decisions index, and 0.1.0-beta.38's release facts, which its own
  release could not carry.
- docs/20_OPEN_QUESTIONS.md: M12-M15, each with its outcome written before
  the measurement - the NVAPI rows on real hardware, LibreHardwareMonitor's
  other groups elevated, a route to Reflex latency, the refused
  slGetFeatureFunction route re-examined.
- docs/02_SPEC.md: FR-5.6, and NFR-2's limits stated to hold with the
  option on (unmeasured). CLAUDE.md: the Application layer's new folder.

Gate on the dev box: all passed in 300 s.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 07:42:39 +00:00
poli0981andClaude Fable 5.2 b7d655078c release: 0.1.0-beta.38 - its changelog file (the App opens web and mail links only; Qodana Ultimate in two reports), HANDOFF's beta.38 section with the next work - the owner's "Thông số nâng cao" - owner-only item 36, the release lines (Qodana PR-3) (#386)
The documents of 0.1.0-beta.38, cut from [Unreleased] as it stood:

- docs/changelog/0.1.0-beta.38.md: what the release is for its user - one hardening, the link opener refusing
  anything but http, https and mailto, and nothing else seen, recorded or measured changing - then the entries of
  #384 and #385, moved verbatim. CHANGELOG.md's [Unreleased] is empty and the table has its row.
- The release lines: README, legal/ACCURACY.md and its two copies, LIMITATIONS.md, ACCESSIBILITY.md.
- docs/HANDOFF.md: START HERE is beta.38's. It opens with the NEXT work, in the owner's own words (2026-11-11, sent
  while this session was running): a setting "Thông số nâng cao" with a navigation item of its own, showing all the
  Streamline, NVAPI and PawnIO headers can give - recorded, not designed: it starts with a plan. Then what this
  session left open: LogNoteW (dead code in the Overlay, the owner's call), this release's own facts, the Qodana
  licence, C# 15's constructs, the resting alpha line. Owner-only item 36.
- docs/handoff/history/0.1.0-beta.37.md: the START HERE it replaces, moved whole - the plan's three parts, D94-D100.

No code changes. The legal texts keep their versions; the ledger stays at schema 22 and the rules at 2026.11.2. No
alpha follows: the alpha line rests (D98).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 02:41:59 +00:00
poli0981andClaude Fable 5.2 4fbc14e698 ci: Qodana reads the C++ as a second report - a Visual Studio solution CMake writes, the job native, fifty-four .editorconfig lines each with its reason, eleven small fixes, nothing changed in code loaded into a game (Qodana PR-2, D99/D100) (#385)
The second of two pull requests. Qodana for .NET - whose C++ is ReSharper C++ - is the only linter of JetBrains' that
can read this code: it is MSVC-only, and qodana-cpp and qodana-clang are Linux images. It reads C++ through a Visual
Studio solution, and src/native is a CMake tree, so the job has CMake write one.

The job
- src/native/CMakePresets.json: the preset x64-vs-analysis (generator "Visual Studio 18 2026", tests and tools on).
  Nothing ships from it, and the gate and the release never use it.
- qodana-native.yaml: the solution, RelWithDebInfo and x64; the build tree, third_party, the vendor stubs and a
  compiled shader left out by path; a gate of zero.
- qodana.yml: the job `native` - configure, build the one target whose output is a header (the blocklist floor), then
  the action with --config qodana-native.yaml. Its own results folder, cache keys, artifact (qodana-report-native)
  and Qodana Cloud project (frameledger-cpp, QODANA_TOKEN_NATIVE). No comment and no annotations: the pull request's
  are the managed job's. Advisory.

What its first run reported, and what was done (docs/13_CI_CD.md, section Qodana)
- 4,695 problems, 4,040 of them hints (off, as on the managed side). Of the 655 left, 494 are distinct: a file
  compiled into three targets is reported three times.
- Off in .editorconfig, each with its reason - every clang-tidy check and clang diagnostic the engine runs has a key
  of ReSharper's, so no .clang-tidy file is added:
  278 a choice this code made (struct padding, MSVC's keywords, a function pointer got by name cast to its type,
      "can be const / constexpr / static", an #include of what the file uses);
   53 a newer idiom proposed for working code;
   25 the engine misreading std::atomic_ref, zero-initialised API structures and tests of constexpr functions;
   25 fields nothing names in structures Windows defines (four files);
   19 %lu with NvU32 in one probe - unsigned int only under clang, the linter's compiler;
   78 what a probe or a test does on purpose (tools and tests);
    7 in code loaded into a game, kept as it is.
- Fixed, 11, none changing what a shipped binary does: std::size for four hand-written array lengths; two
  multiplications done in a narrower type than their result (a constant in the guard's sources, a test); two unused
  member functions, an unused type alias and two unused includes in the tests and the probes.

Code loaded into a game - the Overlay, the Vulkan layer, the shared-memory headers, the guard's rule matcher - is not
changed (D100: there only a defect is fixed, and the report found none). One thing there is not as it should be and
is RECORDED, not removed: dllmain.cpp's LogNoteW, which nothing calls. .editorconfig leaves that one file unread for
an unused function, in a section that says to remove it with the function; whether it goes is the owner's call
(HANDOFF).

"Zero" proved to mean "analysed": a branch that put back three of the things fixed here was reported with exactly
those three, by the inspections that had found them (run 42942390715).

Runs on this branch: 42942068024 (one problem left: LogNoteW), 42943094415 (none).

Docs: 13_CI_CD (the native report), 12_BUILD (the preset), CLAUDE.md (the layout, the C++ conventions), HANDOFF,
wiki/Releasing, CHANGELOG.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 02:27:27 +00:00
poli0981andClaude Fable 5.2 7cd933f62d ci: Qodana for .NET on the Ultimate licence reads the managed code - native on Windows, hints off, every process start read, ShellUrlOpener opens web and mail links only (Qodana PR-1, D99/D100) (#384)
The owner took an Ultimate licence for the Qodana Cloud organisation and asked for Qodana over every language. This is
the first of two pull requests: the existing job moves to Qodana for .NET (Rider's engine) and what it reports over the
managed code is settled. The C++ is the next one's, as a second report.

The job
- qodana.yaml: linter qodana-dotnet, withinDocker false (native mode on the job's Windows runner, the SDK from
  global.json; no container, no EnableWindowsTargeting, no bootstrap line). Everything a run needs is in the file.
- qodana.yml: the job is `managed` (Qodana (managed)), on windows-latest - checkout, setup-dotnet, the NuGet cache
  restored, dotnet build -c Release, then the action pinned by commit. Still advisory, still a gate of zero.

What its first run reported, and what was done (docs/13_CI_CD.md, section Qodana, has each)
- 3,953 problems, 3,607 of them hints - the level an editor offers without underlining anything; the Community linter
  never reported them. Hints are off as a severity (one group in qodana.yaml). 347 left.
- 271: a vi or ja value equal to the English one - off in .editorconfig for .resx (names, units, format strings).
- 23 + 10: two Markdown inspections (table padding; anchors computed otherwise than GitHub does) off in qodana.yaml;
  the application manifest's XML ignored for XmlHighlighting.
- 6 in PrimaryDisplay.cs are the linter's own fault (an internal assertion, in its log); two inspections are excluded
  for that file, and the file is as it was.
- 1 fixed: a test expression that carried a clause its first half implied.
- 36 from JetBrains' security rules: 35 process starts and 1 SQL text. Each was read - what it starts, where the path
  and the arguments come from (a table in the docs). None takes a value from outside the PC, except the game the user
  added. One place was hardened: ShellUrlOpener now refuses any address that is not http, https or mailto itself,
  where it trusted each caller to have checked (ShellUrlOpenerTests).
  The rules stay on. A rule's own id is honoured nowhere - a profile entry's `ignore`, an `exclude` path and
  `enabled: false` were each tried in a run of its own and every result stood: the OpenGrep plugin reports every rule
  through one inspection, OpenGrepInspection, and hands its server every rule file whatever the profile says. That
  inspection does take a path, so the eleven files of src the results were read in, and tests, are listed under its
  name in qodana.yaml: a new file that starts a process is reported.
- ProcessStartSitesTests, in the gate, is the same tripwire where a merge is blocked: it sweeps src for every way a
  process is started here - Process.Start, a ProcessStartInfo, CreateProcessWithTokenW, the logon task's
  registration - and fails on a file that has not been read and listed.

Also: ShmDrainIntegrationTests' drain under supervision supervises at least twice whatever the first drain holds. On
a busy machine the first scan was slow enough for fifty presents to be waiting, the loop ended after one evaluation,
and the test failed its own "supervised more than once" (once, in a local gate of this change).

Where an inspection is switched off (D100, amending D93's "never in qodana.yaml"): in the place the engine that raised
it reads - one of ReSharper's in .editorconfig, what has no key there in qodana.yaml.

Docs: 13_CI_CD (section Qodana rewritten above D93's account, which is kept as history), CLAUDE.md (the rule, the
layout), HANDOFF (D99, D100, this part's section, and the facts of the two releases of 2026-11-11: v0.1.0-beta.37 and
v0.1.0-alpha.4), wiki/Releasing, CHANGELOG.

Shipped code: ShellUrlOpener.cs alone. No link the App could open before stops opening.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 02:14:04 +00:00
poli0981andClaude Fable 5.2 428b723bb3 docs: 0.1.0-alpha.4's notes - 0.1.0-beta.37 on .NET 11.0.0, the last alpha for now; how its sync differs, owner-only item 35 (alpha.4 PR-1) (#382)
- docs/changelog/0.1.0-alpha.4.md: the experimental-build warning, why
  this alpha exists (to bring an installed alpha off the release
  candidate by an update), that the alphas pause after it, and what a
  paused alpha costs - no update, the anti-cheat list included; ten
  betas later an older alpha no longer finds it. Install the beta.
- CHANGELOG.md: its row.
- HANDOFF: the alpha.4 part of START HERE - every switch file conflicts
  this time and main's side is taken for each, leaving RELEASE_LINE -
  and owner-only item 35.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 00:29:52 +00:00
poli0981andClaude Fable 5.2 1a7fbf3ce7 release: 0.1.0-beta.37 - its changelog file (the main line on .NET 11, what an alpha is now, .NET 11's API and no C# 15 yet), HANDOFF's beta.37 section (D94-D98), owner-only item 34, the release lines (beta.37 PR-4) (#381)
- docs/changelog/0.1.0-beta.37.md: the [Unreleased] entries, under what
  a user should know before updating - the processor, Windows 10, the
  size of the download, the terms window (EULA 1.7, Disclaimer 2.14).
- HANDOFF: START HERE for beta.37 (the decisions asked while planning,
  the choices, what was measured, the traps); beta.36's section moved
  verbatim to docs/handoff/history/0.1.0-beta.36.md with its index row;
  owner-only item 34.
- The release lines: README, the accuracy statement and its copies,
  LIMITATIONS, ACCESSIBILITY.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 00:18:00 +00:00
poli0981andClaude Fable 5.2 94eec7831c refactor: .NET 11's API where it reads better, and why no C# 15 construct is used yet - the canary, the table, the dispatch trigger for CodeQL (D96) (beta.37 PR-3) (#380)
The owner asked for C# 15 in the release that moved the main line to
.NET 11. The rule approved with the plan: a construct is used only
where the build, CodeQL and Qodana all read it. Each of the four was
written once, in a file of its own, and shown to each tool (draft pull
request #377, closed unmerged; spike-notes section 29):

- The compiler and every analyzer take collection expression arguments,
  closed hierarchies and labeled jumps; a union draws CA1815 from the
  SDK's own rules.
- ReSharper's engine - InspectCode 2026.2.3.1, which Qodana 2026.2
  runs, and 2026.3.0 - reads none: `closed` and `union` are not parsed,
  the other two are called preview features. On a gate of zero every
  file using one would be an error.
- What CodeQL extracts of them is not established: a pull request's C#
  analysis is an overlay on main's database and prints no per-file
  result.

So beta.37 takes API, not syntax:
- sixteen char searches name their comparison through
  string.StartsWith/EndsWith/LastIndexOf(char, StringComparison), where
  they went through a span to satisfy .NET 10 and 11 alike;
- AgentLauncher finds the Agent it started with
  Process.TryGetProcessById instead of a caught ArgumentException.
Nothing changes in what the software does.

codeql.yml takes workflow_dispatch, so a branch can have a full
analysis. 13_CI_CD (C# 15 and the analyses), 20_OPEN_QUESTIONS (a new
row, with what answers it), CLAUDE.md, the decisions index.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-11 00:04:05 +00:00
poli0981andClaude Fable 5.2 1c60c87b92 legal: what an alpha is once .NET 11 has shipped - EULA 1.7, Disclaimer 2.14, Alpha notice 1.1; a resting alpha says so and names a prerelease only where it runs on one (D97, D98) (beta.37 PR-2) (#379)
Until 0.1.0-alpha.3 an alpha was "a beta's source built on a prerelease
.NET" and every text said so. 0.1.0-alpha.4 runs on .NET 11's release -
the betas' own from beta.37 - and the alpha line rests after it.

- EULA 1.7 (section 2B), Disclaimer 2.14 (section 0, which also gains
  the paragraph on Windows and the processor), Alpha notice 1.1 (what an
  alpha is, the hardware, the pause), the accuracy statement's line and
  its two copies. The texts were approved with the plan of 2026-11-11.
- Shared.AlphaLine: IsPrerelease(the runtime's description) and Rests,
  stamped from Directory.Build.props' FlAlphaLineRests (true on main
  since beta.37; set to false on the alpha branch when it tries a newer
  .NET). App/Services/AlphaTexts picks the banner and About's line from
  the two facts; a prerelease runtime is said whatever the stamp says.
- Strings in en, vi and ja: the banner and About's line no longer say
  "a prerelease .NET" unconditionally; three new keys.
- guide (the alphas are paused: install the beta; a paused alpha's
  anti-cheat list is not updated; ten betas after alpha.4 hide it from
  older alphas), LIMITATIONS, 11_UPDATER, 12_BUILD, 08_UI, the wiki.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-10 23:28:40 +00:00
poli0981andClaude Fable 5.2 49d7bd8e51 build: the main line on .NET 11 - SDK 11.0.100, net11.0, the three runtime-serviced packages at 11.0.0; Windows 10 stated as untested and unsupported (D94, D95) (beta.37 PR-1) (#376)
* build: the main line on .NET 11 - SDK 11.0.100, net11.0, the three runtime-serviced packages at 11.0.0; Windows 10 stated as untested and unsupported (D94, D95) (beta.37 PR-1)

The owner moved the main line to .NET 11 the day after it shipped (D94),
setting aside condition (a) of 20_OPEN_QUESTIONS section G's table - two
alphas on the released runtime. This is the alpha branch's switch applied
to main at released versions, and what follows from it.

- global.json 11.0.100 (allowPrerelease false), TFM
  net11.0-windows10.0.22621.0, Microsoft.Extensions.Hosting,
  Microsoft.Data.Sqlite and System.Diagnostics.EventLog 11.0.0;
  legal/licenses/nuget regenerated (78 packages -> 69).
- Two CA1307 sites the released SDK raised in tests written after the
  last alpha was cut take the (char, StringComparison) overload.
- IHost.RunAsync rethrows a failed BackgroundService on .NET 11: the
  Agent's --serve leaves through its crash report where it exited 0.
  Accepted, not wrapped; HostBackgroundFailureTests holds the runtime's
  behaviour, which nothing had observed.
- release.yml installs a .NET 10 runtime for vpk 1.2.161; Qodana's
  container installs the .NET 11 SDK first (interim, until the Ultimate
  linter); Dependabot's and the workflows' comments follow.
- Windows 10 22H2 is not blocked, and is stated as untested and
  unsupported (D95): NFR-8, LIMITATIONS, README, the guide, the wiki.
  A processor needs SSE4.2 and POPCNT.
- docs: 12_BUILD (the main line's .NET; the alpha line rests after
  alpha.4, D98), 13_CI_CD, 20_OPEN_QUESTIONS row closed, 10_LOGGING,
  spike-notes section 28, CLAUDE.md, the decisions index (D94-D96, D98).

Gate on the dev box: all passed in 338 s on SDK 11.0.100 (native 32/32,
2,561 tests, coverage floors held).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* build: what this pull request's runs measured on .NET 11 - CodeQL reads net11.0 whole, the delta from beta.36, Qodana's one finding fixed

- HostBackgroundFailureTests: the message is kept rather than the
  exception, so no `!` is needed - the one problem Qodana's interim run
  reported (run 42916491047, InspectCode 2026.2.3.1 on SDK 11.0.100).
- 13_CI_CD: what analyze (csharp, none) read of net11.0 (CodeQL 2.27.4,
  "scanned 1220 out of 1220 C# files", 363 references resolved, none
  unresolved), and the rule for C# 15 constructs (D96); the interim's
  first run.
- spike-notes section 28: the table of this pull request's runs - check,
  CodeQL, SonarCloud, Qodana, and the release rehearsal with the first
  delta across two runtime majors (72.6 MB, 372 of 443 files patched).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-10 23:13:00 +00:00
poli0981andClaude Fable 5.2 8767d30c7e docs: 0.1.0-beta.36's notes name the .NET runtime it carries (10.0.15), and HANDOFF records the release (#375)
The tag's run built with SDK 10.0.404, so the package carries runtime 10.0.15
where beta.35's carried 10.0.14 - 343 of 456 files patched in the delta. The
notes at the tag did not say so; the changelog file gains a marked paragraph.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-10 17:13:25 +00:00
poli0981andClaude Fable 5.2 deaa8f30df release: 0.1.0-beta.36 - its changelog file (Qodana in CI, what its first run reported settled), HANDOFF's beta.36 section (D93), owner-only item 33, the release lines (#374)
Nothing a user sees changes: the fixes of #373 are in the shipped assemblies,
and the version says which code is installed. No alpha (D92 stands).

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-10 16:34:10 +00:00
poli0981andClaude Fable 5.2 1342979353 ci: Qodana - JetBrains' code analysis from a workflow of its own, and what its first run reported settled (D93) (#373)
* ci: Qodana - a workflow of its own from JetBrains' template, Qodana Community for .NET over FrameLedger.slnx (D93)

The first run: the linter in its Linux container, EnableWindowsTargeting passed
to it, no quality gate yet - what it builds and what it reports decides the rest.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* ci: Qodana's gate is zero problems, and the inspections this project argues with are off in .editorconfig, each with its reason (D93)

Pushed before the fixes on purpose: the run this starts has the settings and
none of the fixes, so it shows the gate failing on what is left.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

* fix: what Qodana's first run reported, settled - about 400 fixes in 181 files, none changing behaviour; the documents (D93)

Redundant qualifiers, casts, using directives, nullable suppressions and unsafe
modifiers removed from the report's own positions; var for built-in types and
one-line block bodies brought to .editorconfig's style by dotnet format; every
XAML item template names its DataType and a binding that reached its page through
DataContext.ViewModel names the page's type; two unread fields, two overwritten
initialisers and an unused parameter removed; SpokenTextAutomationPeer no longer
assumes the base peer's name is a string.

docs/13_CI_CD.md §Qodana, CLAUDE.md, the wiki's workflow table, HANDOFF's
decisions index (D93) and the changelog entry.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
2026-11-10 16:13:04 +00:00
poli0981andClaude Opus 5.5 7e6ff64b3a release: 0.1.0-beta.35 - its changelog file (accessibility: the statement, a screen reader's names and announcements, headings and charts, contrast themes, focus), HANDOFF's beta.35 section (D88-D92), owner-only item 32, the release lines (beta.35 PR-7) (#372)
docs/changelog/0.1.0-beta.35.md takes [Unreleased] as it stood (#366-#371) under a
summary and the update paragraph: the legal texts keep their versions, schema 22
and rules 2026.11.2 unchanged. HANDOFF's beta.34 START HERE (with its alpha.3
lines) moves verbatim to docs/handoff/history/0.1.0-beta.34.md with its row; the
new START HERE carries D88-D92, the choices, what was measured and the traps (a
ContentDialog test needs the dispatcher's context, a static event hears every
test's, WPF UI's throwing brushes fall back to the Fluent fills inside the App, a
resource reference's callback runs on an equal value, a forward StaticResource
fails only when used, Python's "\b", RenderTargetBitmap's offset); owner-only item
32 is beta.35 on real hardware; D88-D92's rows, and D82's and D87's rows now point
at the history files they moved to. README, ACCURACY, the Disclaimer's accuracy
block and LIMITATIONS name 0.1.0-beta.35; ACCESSIBILITY.md's checks paragraph
rewrapped.

No alpha with this release (D92): alpha.4 moves the alpha line to .NET 11's
release as a task of its own.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 06:30:00 +00:00
poli0981andClaude Opus 5.5 731f6d3082 feat: focus and motion - Ctrl+1 to 5 put the keyboard in the page they open and say its title, a message box is owned by the window it opens over, and pages slide in only while Windows animates (beta.35 PR-6) (#371)
* feat: focus and motion - Ctrl+1 to 5 put the keyboard in the page they open and say its title, a message box is owned by the window it opens over, and pages slide in only while Windows animates (beta.35 PR-6)

PageFocus: after a page shortcut the page shown takes the focus once laid out (its first
control) and its first-level heading is announced; the shell tells it each page shown
from NavigationView.Navigated. PageMotion: the NavigationView's Transition is None while
SystemParameters.ClientAreaAnimation is off, followed as it changes (WPF UI 4.3.0 played
FadeInWithSlide whatever it was). MessageBoxes: WPF UI's MessageBox sets no owner
(measured: null before and after it is shown), so every box is owned by the App's
active window - six call sites, and a scan holds it.

Measured, and left alone: a ContentDialog already gives the focus back to its opener
(WPF UI's ContentDialogHostController stores and restores it); a test pins it. A first
measurement said otherwise because it ran with no dispatcher context, where the
dialog's own finally never resumed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* test(native): the tiny ring's case retries until one attempt both overlapped the writer and was lapped by it - CI's single attempt kept up with all 200000 records (beta.35 PR-6)

ci.yml's check on #371 failed in fl_ring, "a tiny ring, so the writer laps the
reader constantly": attempts 1, most overlapped drains 167058, dropped 0. The loop
repeated an attempt only when it had not overlapped the writer (#308); an attempt
that overlapped and was never lapped ended it, and CHECK(dropped > 0) failed. Both
now hold in the same attempt, the last, repeated up to kMaxAttempts (8): a reader
that never counts a drop still fails every attempt. 40 local runs of the case and
the native suite (32/32) pass; nothing outside the test changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 06:11:02 +00:00
poli0981andClaude Opus 5.5 cea7d87ff8 feat: Windows contrast themes followed - every colour the theme's own, the accent its highlight, the charts in its colours; and the title bar, the snackbars and a Danger button read (D91) (beta.35 PR-5) (#370)
* feat: Windows contrast themes followed - every colour the theme's own, by role from WPF UI's four contrast files, the accent its highlight and the highlight's text, the charts in its colours, the Remove button its own button, on and off at once (D91) (beta.35 PR-5)

While SystemParameters.HighContrast is on, the light or the dark Fluent theme (as the
contrast theme is) lies under Services/ContrastOverlay: every key of WPF UI 4.3.0's HC1,
HC2, HCBlack and HCWhite in the user's own SystemColors colour for its role
(Services/ContrastRoles: the colour a key has in all four, opaque colours only; ties and
placeholders by name), the transparent keys transparent, the 21 accent keys as the
highlight and the text on the accent as the highlight's text. No Mica, no
SystemThemeWatcher (it would apply WPF UI's files: 84 of 97 colours a red placeholder,
17 brushes that throw). A switch, an edit or turning it off is followed at once.

What the App draws itself reads ContrastTheme.Active: ChartPalette.ForContrast,
ChartTheme.TextOn, the artwork band, the Import list, the update footer. The Remove
button takes the DangerButton style (WPF UI's Danger is a fixed #F44336). Settings >
Theme says a contrast theme is in use.

Red first: the sweep under four default and two edited themes found the chips at
1.26-1.78:1, the Remove button at 3.06-4.42:1, the badges' black, and the pressed
menu item white on white.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix: what no sweep rendered reads - the agent's pill on the theme's own fills, every snackbar in the theme's colours, a Danger button's dark text on its red and the theme's own button under a contrast theme (beta.35 PR-5)

Measured 2026-11-09 against WPF UI 4.3.0's styles, outside any surface the sweeps render:
the title bar's agent pill was a ui:Badge with a bound palette appearance (light theme
2.78, 2.63, 2.16 and 3.68:1), the Success and Caution snackbars white on fixed green and
orange (2.78 and 2.16:1 in both themes), and a Danger button white on #F44336 in the dark
theme (3.68:1), out of any contrast theme's reach.

AgentPill (a Pill on the theme's success, caution and critical fills), the strip's
messages all Secondary, and an implicit ui:Button style: a Danger button's text is
TextFillColorDarkPrimaryBrush (5.3 and 5.7:1), and the theme's own button while a
contrast theme is on, message boxes included. The DangerButton style of the first
commit is gone; a bound badge appearance and the other five button appearances are
refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 05:46:12 +00:00
poli0981andClaude Opus 5.5 54a2c08601 feat: headings and charts a screen reader can use - page and section headings, every chart named with where its numbers are, its keys kept, a double-click that shows all of it, no series told apart by colour alone (D90) (beta.35 PR-4) (#369)
Every Title is AutomationProperties.HeadingLevel Level1 and every section Subtitle Level2; a Subtitle that shows
a value says None. AccessibilityTests holds every Title and Subtitle to a declared level, PagesLoadTests every
page to one first-level heading.

Every plot is Charts/ChartPlot, ScottPlot's WpfPlot with a peer: an Image of the localized type "chart", named
and with help text saying where its numbers are and which keys move it, focus drawn with WPF UI's focus style.
Charts/ChartKeys keeps the arrows and A with the chart after ScottPlot has acted (the page used to scroll or move
the focus on); a double-click shows all of the chart, as the guide said, instead of ScottPlot's benchmark overlay.
Charts/ChartPatterns gives every series a line pattern, marker or hatch by its index - Compare's five sessions,
the GPU temperature and hot spot, the two stutters, the two lows, the two limits - and
NoChartTellsTwoSeriesApartByColourAlone reads every chart's legend for two entries that differ by colour only.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 04:52:30 +00:00
poli0981andClaude Opus 5.5 926580c0f2 feat: a screen reader is told what appears - safety notices, the agent's and the update's banners, every short message, a game page's notice and each maintenance result, through UI Automation's notification (D89) (beta.35 PR-3) (#368)
Services/IAnnouncer over Services/UiaAnnouncer, the one place the App raises UI Automation's notification event:
only on its active window (a game that has the screen is never interrupted) and only when a screen reader listens.
An Alert (every one read): a safety notice as it arrives, a warning in the strip, the agent going offline, a game
page's refusal. A Status (a newer one replacing one not yet read): the strip's info and success, the agent back, an
update's step, a page's success, every Database maintenance result.

Every transient message goes through IMessageStrip now - MainWindowViewModel's six direct ISnackbarService calls
moved to it - and SnackbarStrip says what it shows. Services/ShellAnnouncements says only a change: the first state
is read with the window, a download's percent ticks are one step, a rebuilt window says nothing. The maintenance
line's LiveSetting, which announced nothing, is gone; no XAML may set one.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 04:36:48 +00:00
poli0981andClaude Opus 5.5 0b485d1293 feat: what a screen reader is given - every item, cell and row button named by what it shows, the display-only controls with a voice, the notices' text reachable, no silent Tab stop (D88) (beta.35 PR-2) (#367)
PagesLoadTests.Automation sweeps the UI Automation tree WPF builds for every page, window, dialog and control,
every tab shown: no Tab stop without a peer, no item named by its type, every focusable control named, no shown
text hidden in a template, no two buttons in a list sharing a name. All five went red on main first.

Fixes: every list item type speaks through ToString (Services/SpokenText) - Choice its label, a notice its title
and text, a document page its title, a session row its date, length and tier; FpsReadout, TriStateChip and
ChangeBadge get SpokenTextAutomationPeer (a Text element, no children) and leave the tab order, and the readout
says rule 6 aloud (Native, displayed and the factor together); a template column names its cells through
Controls/CellName over WPF UI's DefaultDataGridCellStyle, an empty cell read as "empty"; a row's button names its
row; the App's InfoBar template draws title and message through ContentPresenters; implicit styles take
ItemsControl, ui:Image and ui:CardControl out of the tab order; a CardExpander's expand button is named after its
header (Controls/ExpanderToggleName); the dialog hosts are no Tab stops. ScreenReaderNamesTests holds each piece.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 04:16:02 +00:00
poli0981andClaude Opus 5.5 fad7b84075 feat: the accessibility statement - ACCESSIBILITY.md under Help > Accessibility and on the wiki, the Accessibility problem form and its label, the documents corrected (D88) (beta.35 PR-1) (#366)
ACCESSIBILITY.md states WCAG 2.2 AA through WCAG2ICT as a goal and never a claim, what works with the keyboard
alone, a screen reader, a contrast theme or Magnifier, how it is checked (the owner's pass of 2026-11-09 on
beta.34 and alpha.3), the known limits with what to do instead, and how a barrier is reported and handled - by
severity, no promised times. Embedded and shown under Help > Accessibility and after Limitations in the guide's
window; a wiki page of its own.

The Accessibility problem issue form, a user-interface section in the PR template, and the labels accessibility,
safety and priority-high - the last two named by the Safety gap form since 2026-08-02 and never existing.

Documents corrected: charts take focus and ScottPlot's keys (ChartKeyboardTests measures it; 08_UI said
neither); the database-maintenance live region announces nothing; the Narrator gap is the owner's pass now;
20_OPEN_QUESTIONS G's accessibility row closed; NFR-9 widened (D88-D91); LIMITATIONS gains an Accessibility
section and loses its one struck sentence, which a screen reader reads as current.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-09 03:51:44 +00:00
dependabot[bot] 4cc4c3e9c2 Bump the minor-and-patch group with 1 update (#365)
Bumps Meziantou.Analyzer from 3.0.317 to 3.0.319

---
updated-dependencies:
- dependency-name: Meziantou.Analyzer
  dependency-version: 3.0.319
  dependency-type: direct:production
  update-type: version-update:semver-patch
  dependency-group: minor-and-patch
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
2026-11-09 07:03:16 +07:00
poli0981andClaude Opus 5.5 27f537a2d7 docs: 0.1.0-alpha.3's notes - its changelog file opening with the experimental-build warning, its row, HANDOFF's alpha.3 lines, owner-only item 31 (D87) (alpha.3 PR-1) (#363)
The third alpha is the source of 0.1.0-beta.34 on .NET 11 RC3 (D87: right after
beta.34, on RC3 because the train reached it before .NET 11's release on
2026-11-10). Its notes are on main as every release's are; the alpha branch takes
them with the sync that follows this pull request (13_CI_CD §The alpha line).

- docs/changelog/0.1.0-alpha.3.md: the experimental-build warning, the update
  from alpha.2 (the ledger copy, the Disclaimer 2.13 and the Privacy Policy 2.14
  once, the Alpha notice's box again, schema 22 - which an alpha.2 cannot open),
  installing over a beta and going back, and what differs from the beta - the
  same as alpha.1 from beta.32.
- CHANGELOG.md: its row in the Releases table.
- docs/HANDOFF.md: START HERE's "0.1.0-alpha.3" lines (the sync by a merge
  commit, the switch unchanged, the first alpha whose ledger moves) and
  owner-only item 31, the third alpha on real hardware.

No release line moves (README, ACCURACY, the Disclaimer's copy and LIMITATIONS
keep naming the newest beta).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 09:51:02 +00:00
poli0981andClaude Opus 5.5 44c031faca release: 0.1.0-beta.34 - its changelog file (DLSS 5 and DLSS 4.5 from evidence, filters, Games at once), HANDOFF's beta.34 section (D83-D87), owner-only item 30, the release lines (beta.34 PR-7) (#362)
docs/changelog/0.1.0-beta.34.md takes [Unreleased] as it stood (#356-#361)
under a summary and the update paragraph: the Disclaimer 2.13 and the
Privacy Policy 2.14 once, schema 22 (a beta.33 cannot open it), rules
2026.11.2. HANDOFF's beta.33 START HERE (with its alpha.2 lines) moves
verbatim to docs/handoff/history/0.1.0-beta.33.md with its row; the new START
HERE carries D83-D87, the choices, what was measured and the traps
(notice-check reads the index, the native token test's exact mask, WPF UI's
navigation cache and pane template, Microsoft.Data.Sqlite's synchronous
async); owner-only item 30 is beta.34 on real hardware; D87's row. README,
ACCURACY, the Disclaimer's accuracy block and LIMITATIONS name
0.1.0-beta.34.

One wording fix found while writing the notes, before the release: D83's
evidence is DLSS Frame Generation IDENTIFIED in the session (fg_mode =
dlssg), as DlssText reads it - its factor need not have been counted - and
five places written in #357 said "counted" (05_DETECTION twice, spike-notes
§26, the changelog entry, HANDOFF's D83 row), with DlssText's and the test's
comments. DlssTextTests now pins it: Dynamic applied with DLSS Frame
Generation identified and the count refused (non_uniform) reads DLSS 4.5.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 09:34:30 +00:00
poli0981andClaude Opus 5.5 f66d98eb78 docs: the tested-games list from a copy of the ledger of 2026-11-06 (D35) - fifteen more games measured, the first Xbox titles; the docs sweep across the train (beta.34 PR-6) (#361)
Owner request 2026-11-06, item 5: review, update and sync the documents and the
tested-games list, from the database.

- docs/LIST_GAME_TESTED.md: refreshed from a copy of the owner's ledger taken
  2026-11-06 (read-only; 131 sessions of 27 games, the same as the live ledger
  when the commit was made), as drafted in the plan the owner approved: fifteen
  more Tier-1 rows (Phantom Blade Zero under Dynamic, Hogwarts Legacy loading
  DLSS 5's runtime, Valor Mortis and inKONBINI from the Xbox app, ...), seven
  rows updated, the Tier-2 sessions and the guard's refusals since 2026-09-27,
  the driver range and the engine versions read since.
- docs/08_UI.md: the DLSS line's two new rows (D83, D84); Compare over every
  session.
- docs/HANDOFF.md: item 28's tested-list step done.
- docs/open-questions/closed-s.md: S30 notes the fifth Streamline id decoded,
  from the vendor's header rather than by guess (row O4's warning).
- guide/08-settings.md: a language change stays on Settings.
- CHANGELOG.md: the entry.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 09:17:49 +00:00
poli0981andClaude Opus 5.5 efaf585928 feat: filters on Games and Compare; the Compare page's layout; the selected pane item keeps its label (D85, D86) (beta.34 PR-5) (#360)
Owner request 2026-11-06, items 3 and 4, with a screenshot of Compare.

- Games (D85): Store, Engine, State, Installed and Played beside the search
  and sort, with Clear filters; the title counts what is shown of all while a
  filter or the search narrows it. Kept while the App runs, not saved.
- Compare (D85): every session (it was the newest 200, one game lookup each;
  now one read of the names, off the UI thread), filtered by game, tier and
  frame generation, with a count; a ticked session stays listed.
- Compare layout: the selection bar on the left above the list with Clear
  selection; columns sized to what they hold and one filler column; the date
  sorts by when; the settings-changed cell says so or shows a dash; the
  result table's labels sit over their columns (MetricHeader bound).
- The pane (D86): our copy of WPF UI 4.3.0's LeftFluent item template
  without the IsActive collapse, so the selected page keeps its label.
- Strings en/vi/ja (ja review); 08_UI, guide/07-screens, CHANGELOG, HANDOFF.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 09:05:13 +00:00
poli0981andClaude Opus 5.5 a78d7bd359 feat: Games opens at once from any page - the window keeps the library page, cards before the executable checks; a language change keeps the page on screen (beta.34 PR-4) (#359)
The owner, 2026-11-06: switching to Games stalls 0.5-1.5 s; and changing the language in Settings
brings back the last game page instead of Settings.

- Measured first (spike-notes 27, a scratch test, not committed; 65 games, the first executable
  check held 800 ms for a USB drive waking): first cards 822-830 ms -> 10-15 ms; a revisit
  181-215 ms -> 15-16 ms; the first build unchanged (the non-virtualizing grid, by decision).
- WindowPageCache: the window's own page provider keeps the GamesPage it built (set after
  SetNavigationControl, which hands the NavigationView the container's provider; WPF UI's
  NavigationCacheMode is never read when a page provider is set). One per window, so a language
  change's new window builds its own.
- GamesViewModel: the library read on the thread pool; the cards listed before the executable
  checks, each getting its Not installed pill when its check answers; a read keeps every card that
  would look the same (GameCardViewModel.LooksLike, Sync - never Clear); the page reads the library
  again each time it is shown (Attach). The list view is bounded (MaxHeight 640).
- ShellHost notes every page the live window shows (pane clicks included), so a rebuild opens on
  the page on screen - Settings after a change made there.
- Tests: GamesViewModelTests (cards before the checks, a page shown again, Sync), WindowPageCacheTests
  (kept per window; the shell remembers the page on screen).
- Docs: 16_WPFUI_SYNTAX, wiki App-UI-development, 08_UI, 09_I18N, spike-notes 27, CHANGELOG.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 08:49:55 +00:00
poli0981andClaude Opus 5.5 c9b0aab862 feat: DLSS 5 is named where the game ran its neural rendering (schema 0022, rules 2026.11.2, Privacy 2.14); Streamline headers re-vendored at v2.14.16 (D84) (beta.34 PR-3) (#358)
The owner, 2026-11-06: Hogwarts Legacy with Neural Rendering on loaded DLSS 5's runtime
(nvngx_dlssnr.dll 310.9.16.0, from the game's folder) and the session could say only "DLSS 5
runtime loaded". Then: check what changed in recent Streamline versions, fix it if needed.

- Streamline: the vendored closure was v2.12.0 byte for byte; v2.14.16 changes three of its
  eleven files (kFeatureDLSS_NR = 1004, the uplift buffer types 70-72, kStructVersion7, the
  version). Re-vendored byte for byte; README provenance, THIRD_PARTY_NOTICES row.
- Overlay (hook path): the slEvaluateFeature detour decodes kFeatureDLSS_NR into its own bit;
  RecordPresent sets FL_FEAT_SL_NEURAL_RENDERING (feature byte bit 7, an enumerator - no layout
  change) under the condition every Streamline fact shares; the evaluate row's family also claims
  FL_HOOK_SL_NEURAL_RENDERING (bit 12), published with the detour. Overhead: decode unchanged
  within noise (4.7 ns a call), +0.14..0.19 ns a present.
- Agent: Domain.Metrics.NeuralRendering.SharePct; sessions.dlss_nr_pct (schema 0022) only from a
  writer that published the family - an older Overlay's silence is never a measured 0 %.
  CensusNames records sl.dlss_nr.dll.
- Rules 2026.11.2: capability dlss_nr -> "Supports DLSS 5 Neural Rendering".
- App (D84): DLSS 5 heads the DLSS line where it ran ("DLSS 5 (neural rendering on 97.3% of
  rendered frames)"); its library listed with version and source; measured 0 -> "runtime loaded,
  not seen running"; nothing to tell -> "runtime loaded" as before. CSV basis/pct.
- CaptureHost dev census: a kFeatureDLSS_NR bucket.
- Privacy Policy 2.14 (neural-rendering libraries); the accuracy block's DLSS 5 sentence.
- Tests: native guard_test (the new harness mode through an injected Overlay; 0xF00D lights no NR
  fact; the token case's exact mask gains the family), NeuralRenderingTests, SessionAggregatorTests,
  DlssGenerationTests, DlssTextTests (a row shaped from session 132), LedgerDatabaseTests (0022),
  the repository round trip, the rules fixture, RuntimeModuleSourcesTests, SlCensusTests.
- Docs: 05_DETECTION, 06_DATA_MODEL, 17_HOOK_ENGINE, LIMITATIONS, guide/03, guide/10, the accuracy
  block, spike-notes 26 (+ owner-only item 30(a)'s decision table), HANDOFF (D84), CHANGELOG.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 08:39:42 +00:00
poli0981andClaude Opus 5.5 569bacac95 feat: a session under the NVIDIA App's Dynamic frame generation reads DLSS 4.5; Disclaimer 2.13 (D83) (beta.34 PR-2) (#357)
The owner, 2026-11-06: Phantom Blade Zero under the NVIDIA App's Dynamic multi frame generation,
up to x3, read "DLSS 4" - but Dynamic (and x6) exist only in DLSS 4.5. The DLSS line named a
generation from the measured multiplier alone, and "up to x3" never passes x3.0.

- Domain: DlssEvidence.DynamicFrameGeneration, DlssBasis.DynamicFrameGeneration; the rule's new
  row sits below a measured x5 and above every DLSS 4 row.
- App: DlssText reads Dynamic as DLSS Frame Generation counted AND the game's driver profile
  setting NGX_DLSSG_MODE to 4 (NvidiaDriverSettings.DlssgModeDynamic, NVIDIA's own name) AND the
  driver's FG word reporting the mode override applied (FG_MODE). The driver's own mode number
  (3 Dynamic / 1 Fixed on all sixteen stored sessions) is recorded, not keyed on. Label "DLSS 4.5
  (Dynamic multi frame generation, set in the NVIDIA App)", CSV basis=dynamic_frame_generation.
  A Fixed multiplier names nothing by itself (Witcher 3 under Fixed x6 still presented x2.0).
- Dlss_Basis_Dynamic in en/vi/ja (ja review).
- Docs: 05_DETECTION (row + amendment), 03_METRICS (the ceiling decode confirmed), LIMITATIONS,
  guide/03, guide/10, the accuracy block re-dated 2026-11-06 + Disclaimer 2.13, spike-notes 26
  (new), HANDOFF decisions index (D83), CHANGELOG.
- Tests: DlssGenerationTests, DlssTextTests (a row shaped from the owner's session 129 and four
  negatives).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-06 08:15:42 +00:00
poli0981andClaude Opus 5.5 f7bfe1e8f1 fix: an Xbox game's picture name that names a file's stream is passed over, not thrown on; the guard's note on GuardedInjectImpl sits above it again (beta.34 PR-1) (#356)
Two things a self-review of #351 and #352 found after their gates were green.

- XboxArtCache.Image: a ShellVisuals name in MicrosoftGame.config that points at a
  stream of a file, named like a drive ("Resources\a:b.png"), made the scale
  search's pattern "a:b.scale-*.png", and EnumerateFiles throws ArgumentException
  on a rooted pattern ("Second path fragment must not be a drive or UNC name") -
  which XboxArtCache.Find does not catch. GameArt.ForAsync's task then faulted:
  the game page showed no picture and logged an unobserved task, and a command
  that awaits the page's reload would reach the dispatcher's crash report. A name
  whose leaf holds ':' is now no picture, like a name outside the package. Red
  without the fix (the exact ArgumentException), green with it; no install of
  the owner's eleven titles has such a name (spike-notes §25).
- fl_guard.cpp: #352 put BlockingMitigation and TargetIsolationVerdict between
  GuardedInjectImpl and the comment that describes it ("A refusal returns here
  ... runs only for a target the full evaluation allowed"), so it read as the
  first line of BlockingMitigation's. It moves back above GuardedInjectImpl and
  names the target's isolation among what runs after the evaluation. No code
  change there.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 08:09:29 +00:00
poli0981andClaude Opus 5.5 405449283b docs: 0.1.0-alpha.2's notes - its changelog file opening with the experimental-build warning, its row, HANDOFF's alpha.2 lines, owner-only item 29 (D82) (alpha.2 PR-2) (#354)
The second alpha is the source of 0.1.0-beta.33 on .NET 11 RC3 (D82: right after
beta.33, on RC3 because the train reached it before .NET 11's release on
2026-11-10). Its notes are on main as every release's are; the alpha branch takes
them with the sync that follows this pull request (13_CI_CD §The alpha line).

- docs/changelog/0.1.0-alpha.2.md: the experimental-build warning, the update
  from alpha.1 (the ledger copy, the Privacy Policy 2.13 once, the Alpha notice's
  box again), installing over a beta and going back, and what differs from the
  beta - the same as alpha.1 from beta.32.
- CHANGELOG.md: its row in the Releases table.
- docs/HANDOFF.md: START HERE's "0.1.0-alpha.2" lines (the sync by a merge commit,
  the switch unchanged, the first delta an alpha's release has to find - #350's
  lookup) and owner-only item 29, the second alpha on real hardware.

No release line moves (README, ACCURACY, the Disclaimer's copy and LIMITATIONS
keep naming the newest beta).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 07:37:51 +00:00
755 changed files with 58930 additions and 2819 deletions

No files matched your search

+257
View File
@@ -59,6 +59,102 @@ dotnet_diagnostic.MA0011.severity = warning
# and GenerateDocumentationFile is on for the shipping ones.
dotnet_diagnostic.CS1591.severity = suggestion
# --- ReSharper inspections: what Qodana reads (docs/13_CI_CD.md §Qodana, D93 and D100) ---
# Qodana's linters run ReSharper's engine - Qodana for .NET since 2026-11-11, the Community linter before it - and the
# engine reads these keys; so does Rider or ReSharper in the editor, which is why they are here and not in
# qodana.yaml: the same list everywhere. An inspection is off only where it argues with a choice this project made;
# each line says which. Everything else it reports is fixed. (What is NOT ReSharper's - a whole severity, the
# IntelliJ platform's Markdown and XML inspections, JetBrains' security rules - has no key here and is in qodana.yaml.)
[*.{cs,xaml}]
# The compiler's missing-XML-comment warning, already a suggestion above (CS1591): InspectCode reports suggestions.
resharper_c_sharp_warnings_cs1591_highlighting = none
# Naming is this file's rule and the .NET analyzers'. ReSharper's own scheme would rename product names and units
# (OpenGL -> OpenGl, XeSS -> XeSs, Presents5s -> Presents5S) and every Chart_* resource key.
resharper_inconsistent_naming_highlighting = none
# `x is long v` / `x is > 0 and int v` is how this code unwraps a nullable AND names its type; `{ } v` or `var v`
# hides it, against csharp_style_var_for_built_in_types above.
resharper_convert_type_check_pattern_to_null_check_highlighting = none
resharper_pattern_always_matches_highlighting = none
# A code-behind class names its base so that the type reads whole without the half XAML generates.
resharper_redundant_extends_list_entry_highlighting = none
# An argument that repeats a default says which value the call relies on.
resharper_redundant_argument_default_value_highlighting = none
# `partial void OnXChanged(T value)`: the signature is CommunityToolkit.Mvvm's generator's, used or not.
resharper_unused_parameter_in_partial_method_highlighting = none
# Exception messages and log lines are English by design (docs/09_I18N.md); what a user reads comes from .resx,
# and tools/resx-audit.ps1 gates that.
resharper_localizable_element_highlighting = none
# WPF UI's styles (DefaultImageStyle ...) arrive with ui:ControlsDictionary at run time, from a package's compiled
# resources the inspection does not read. A key that does not exist fails PagesLoadTests.
resharper_xaml_static_resource_not_resolved_highlighting = none
# Proposals, not findings - ReSharper's suggestions for a style this project did not take. Its style is this file,
# `dotnet format` and the analyzers the build treats as errors.
resharper_convert_to_primary_constructor_highlighting = none
resharper_merge_into_pattern_highlighting = none
resharper_merge_conditional_expression_highlighting = none
resharper_use_object_or_collection_initializer_highlighting = none
resharper_use_utf8_string_literal_highlighting = none
resharper_redundant_explicit_params_array_creation_highlighting = none
resharper_arrange_object_creation_when_type_evident_highlighting = none
resharper_convert_to_using_declaration_highlighting = none
resharper_convert_closure_to_method_group_highlighting = none
resharper_redundant_type_declaration_body_highlighting = none
resharper_replace_with_field_keyword_highlighting = none
resharper_replace_conditional_expression_with_null_coalescing_highlighting = none
resharper_append_to_collection_expression_highlighting = none
resharper_use_collection_expression_highlighting = none
resharper_use_with_expression_to_copy_struct_highlighting = none
resharper_convert_to_auto_property_highlighting = none
resharper_convert_to_static_class_highlighting = none
resharper_use_index_from_end_expression_highlighting = none
# A notice for code moving to C# 14, which this code was written in.
resharper_c_sharp14_overload_resolution_with_span_breaking_change_highlighting = none
# `await using` needs ConfigureAwait(false) on the disposable under CA2007, which reads worse than the `using` it
# replaces; the streams these name are closed synchronously on purpose.
resharper_use_await_using_highlighting = none
# A member's accessibility is what its type offers, not a count of today's callers.
resharper_member_can_be_private_global_highlighting = none
resharper_member_can_be_private_local_highlighting = none
resharper_property_can_be_made_init_only_global_highlighting = none
resharper_property_can_be_made_init_only_local_highlighting = none
resharper_auto_property_can_be_made_get_only_global_highlighting = none
resharper_auto_property_can_be_made_get_only_local_highlighting = none
# "Never used" across the solution is not what decides for a member other code can reach: an enum's values mirror
# the native layout and the stored columns, an exception type carries the constructors CA1032 asks for, a row or a
# message is read by Dapper and System.Text.Json, an interface states a contract. The private kind (.Local) stays on.
resharper_unused_member_global_highlighting = none
resharper_unused_type_global_highlighting = none
resharper_unused_auto_property_accessor_global_highlighting = none
resharper_unused_member_in_super_global_highlighting = none
resharper_unused_method_return_value_global_highlighting = none
resharper_unused_parameter_global_highlighting = none
resharper_class_never_instantiated_global_highlighting = none
resharper_event_never_subscribed_to_global_highlighting = none
# The compiler's nullable analysis is the gate (warnings are errors). These three go further and call a guard
# redundant because an annotation promises non-null - where a dependency property, a JSON document or a WPF event
# hands over null all the same.
resharper_null_coalescing_condition_is_always_not_null_according_to_api_contract_highlighting = none
resharper_conditional_access_qualifier_is_non_nullable_according_to_api_contract_highlighting = none
resharper_condition_is_always_true_or_false_according_to_nullable_api_contract_highlighting = none
# A record's positional property is part of its equality and of what it prints, read by name or not.
resharper_not_accessed_positional_property_global_highlighting = none
resharper_not_accessed_positional_property_local_highlighting = none
# Every one compares a value with the one it was copied from (max == r) or with an exact sentinel (scale == 1.0).
resharper_compare_of_floats_by_equality_operator_highlighting = none
# A nested record may name a property as the type that holds it names a helper; the compiler tells them apart.
resharper_member_hides_static_from_outer_class_highlighting = none
# `default: break;` is how a switch over a kind says that the rest is left alone on purpose.
resharper_redundant_empty_switch_section_highlighting = none
# A binding path through a null yields no value, and the template shows nothing: that is the design.
resharper_xaml_possible_null_reference_exception_highlighting = none
[*.resx]
# A vi or ja value that equals the English one is deliberate: product and technology names, units and format strings
# are not translated, and a ja safety string is the English verbatim until a reviewer signs it (docs/09_I18N.md;
# tools/resx-audit.ps1 holds that rule). 271 results on Qodana for .NET's first run.
resharper_overridden_with_same_value_highlighting = none
[src/FrameLedger.App/**.cs]
# WPF codegen constraints, not style choices. The generated App.g.cs and
# *.g.cs partials declare these types public, so making them internal is a
@@ -72,3 +168,164 @@ dotnet_diagnostic.MA0049.severity = none # type name matches containing names
[tests/**/*.cs]
dotnet_diagnostic.CA1707.severity = none # underscores in test method names
dotnet_diagnostic.CA2007.severity = none
# A test's assertion lambda runs inside the `using` scope it captures from - and reads a variable the test changes
# on purpose - and a test's `using var x = new T { ... }` leaks nothing a test process keeps.
resharper_access_to_disposed_closure_highlighting = none
resharper_access_to_modified_closure_highlighting = none
resharper_using_statement_resource_initialization_highlighting = none
# A test that dereferences a null fails, which is what it is for.
resharper_assign_null_to_not_null_attribute_highlighting = none
resharper_possible_null_reference_exception_highlighting = none
# A test double keeps the shape of what it stands for: an accessor nothing reads, a collection only filled, a result
# only some tests look at, a value an arrange step sets before the lambda that reads it runs.
resharper_unused_auto_property_accessor_local_highlighting = none
resharper_collection_never_queried_local_highlighting = none
resharper_collection_never_updated_local_highlighting = none
resharper_unused_method_return_value_local_highlighting = none
resharper_redundant_assignment_highlighting = none
resharper_return_type_can_be_not_nullable_highlighting = none
# --- ReSharper C++ inspections: what Qodana's native report reads (docs/13_CI_CD.md §Qodana, D99 and D100) ---
# The native report runs ReSharper's C++ engine, and with it the clang-tidy checks and clang diagnostics that engine
# bundles (the keys that say clang_tidy). The native code is built by MSVC alone, with /W4 /WX; what clang says about
# code it never compiles is read here as advice. An inspection is off only where it argues with a choice this project
# made, and each line says which; everything else the report found was fixed.
[src/native/**.{cpp,h,inl}]
# Where the compiler pads a struct. Information, not a fault: the layouts that matter are asserted (static_assert on
# sizes and offsets in fl_shm.h, and the mirror test against fl-layout-dump).
resharper_cpp_clang_tidy_clang_diagnostic_padded_highlighting = none
# MSVC's own keywords - __try/__except (rule 5's guard around every hook body), __declspec, __stdcall. This code is
# built by MSVC and by nothing else.
resharper_cpp_clang_tidy_clang_diagnostic_language_extension_token_highlighting = none
# A function looked up by name - GetProcAddress, vkGetInstanceProcAddr, a vtable slot - comes back as a pointer to
# another function type and is cast to its real one: that is the API. Through void* is how the cast says it is meant.
resharper_cpp_clang_tidy_clang_diagnostic_cast_function_type_strict_highlighting = none
resharper_cpp_clang_tidy_bugprone_casting_through_void_highlighting = none
# Memory read as something else - a COM object as its vtable, a relay's bytes as an address - says reinterpret_cast,
# also where static_cast from void* would compile.
resharper_cpp_reinterpret_cast_from_void_ptr_highlighting = none
# It proposes std::cmp_less for a comparison whose sides differ in sign. The comparisons it found cast one side
# explicitly, and the compiler's own sign warnings (C4018, C4388, C4389) are errors in this build.
resharper_cpp_clang_tidy_modernize_use_integer_sign_comparison_highlighting = none
# A `default:` under a switch that names every enumerator is deliberate: a value outside the enumeration - a corrupt
# one, a future one - takes the refusing arm instead of falling out of the switch.
resharper_cpp_clang_tidy_clang_diagnostic_covered_switch_default_highlighting = none
# "Not eliding copy on return": a verdict returned from a const local is copied. It is a few words wide.
resharper_cpp_clang_tidy_clang_diagnostic_nrvo_highlighting = none
# A type that owns a handle deletes its copies or declares none; the moves are then not generated either, which is
# what is wanted. Four more `= delete` lines would say nothing the destructor did not.
resharper_cpp_clang_tidy_cppcoreguidelines_special_member_functions_highlighting = none
# An enumeration zero-initialised (`D3D_FEATURE_LEVEL got{}`) is an out parameter cleared before the call that fills it.
resharper_cpp_clang_tidy_bugprone_invalid_enum_default_initialization_highlighting = none
# An enumeration's width is part of a layout or of the atomic word it is stored in, not a size to be minimised.
resharper_cpp_clang_tidy_performance_enum_size_highlighting = none
# "Possibly unused #include": a file that calls std::memset includes <cstring>, also when <windows.h> already brought
# it in. What it reported was of that kind, or a header named again inside a conditional block, but for two that
# nothing used; those were removed.
resharper_cpp_unused_include_directive_highlighting = none
# A function another file calls by a declaration of its own (the harness's Hold* functions), an entry point the
# loader calls by name (DllMain), a placeholder a target links: "can be made static" reads one file.
resharper_cpp_use_internal_linkage_highlighting = none
# "Can be made const": a parameter or a method through which the code acts on a device, a ring or a handle is not
# const because the type system would allow it; and a local that is already const gains nothing from constexpr - for a
# std::string it would hold only while the text fits the small-string buffer.
resharper_cpp_parameter_may_be_const_ptr_or_ref_highlighting = none
resharper_cpp_member_function_may_be_const_highlighting = none
resharper_cpp_variable_can_be_made_constexpr_highlighting = none
# THE ENGINE MISREADS THESE. (1) `std::atomic_ref<T>{x}.fetch_or(bit)` is reported as an expression without side
# effects; the side effect is the call. (2) A Win32, D3D12 or ffx-api structure is zero-initialised and then filled
# field by field - the reserved fields must stay zero - and is handed to the API by its address or its header's: the
# engine reports the initialiser as always rewritten and the fields as assigned and never used. (3) A test of a
# constexpr function (`Check(F(a) == b && G(a) == c, ...)`) has the engine evaluate it and call one side redundant.
resharper_cpp_expression_without_side_effects_highlighting = none
resharper_cpp_initialized_value_is_always_rewritten_highlighting = none
resharper_cpp_assigned_value_is_never_used_highlighting = none
resharper_cpp_redundant_boolean_expression_argument_highlighting = none
# Proposals to restate working code in a newer idiom - a range-for, std::ranges, any_of, starts_with, std::min, a raw
# string, an if with an initialiser, emplace_back. The native code is written to be read beside the hook path: an
# index and a plain loop, the same in the guard, its tests and the probes.
resharper_cpp_clang_tidy_modernize_loop_convert_highlighting = none
resharper_cpp_clang_tidy_modernize_use_ranges_highlighting = none
resharper_cpp_use_range_algorithm_highlighting = none
resharper_cpp_clang_tidy_readability_use_anyofallof_highlighting = none
resharper_cpp_clang_tidy_modernize_use_starts_ends_with_highlighting = none
resharper_cpp_clang_tidy_readability_use_std_min_max_highlighting = none
resharper_cpp_clang_tidy_modernize_raw_string_literal_highlighting = none
resharper_cpp_clang_tidy_modernize_use_emplace_highlighting = none
resharper_cpp_clang_tidy_modernize_return_braced_init_list_highlighting = none
resharper_cpp_use_associative_contains_highlighting = none
resharper_cpp_too_wide_scope_highlighting = none
resharper_cpp_if_can_be_replaced_by_constexpr_if_highlighting = none
# A structure Windows defines and this SDK's headers do not give (RTL_PROCESS_MODULE_INFORMATION,
# PROCESS_MEMORY_COUNTERS_EX2), and a test's stand-in for a Vulkan dispatchable object, are declared field for field:
# the fields nothing names are what puts the others at their offsets.
[src/native/{FrameLedger.Injector/src/fl_guard_sources.cpp,FrameLedger.ProcessStats/src/fl_process_stats.cpp,tools/fl-probe-guard/main.cpp,tests/vklayer_test.cpp}]
resharper_cpp_declarator_never_used_highlighting = none
# fl::patch::Install takes its hook engine as a template parameter and calls Create / Enable / Disable on it. The
# test's two stand-in engines are used that way and no other, which ReSharper's "never used" does not follow through a
# dependent call: six reports, one per member (Qodana, 2026-11-12). The members are the test.
[src/native/tests/unhook_inline_test.cpp]
resharper_cpp_declarator_never_used_highlighting = none
# The linter's compiler, not the build's: NVAPI's header makes NvU32 `unsigned int` under clang and `unsigned long`
# under MSVC, which is the compiler that builds this probe - and `%lu` is right for it.
[src/native/tools/fl-probe-nvapi/main.cpp]
resharper_cpp_clang_tidy_clang_diagnostic_format_highlighting = none
# --- the probes, the harness and the tests ---
[src/native/{tools,tests}/**.{cpp,h}]
# printf, fprintf, fputs, fclose: a probe prints what it found and a test prints why it failed; neither has anything
# to do about a failed write to its own console.
resharper_cpp_clang_tidy_cert_err33_c_highlighting = none
# atoi on the harness's own arguments, which its tests write: a number that does not parse is 0, and the run that
# asked for it fails on its own assertion.
resharper_cpp_clang_tidy_bugprone_unchecked_string_to_number_conversion_highlighting = none
# A printf-shaped helper (`Note(fmt, ...)`): the attribute clang proposes is not MSVC's, and a parameter pack would
# be a second implementation of printf.
resharper_cpp_clang_tidy_modernize_avoid_variadic_functions_highlighting = none
resharper_cpp_clang_tidy_clang_diagnostic_missing_format_attribute_highlighting = none
# `&pointer` handed to an API that takes `void*` (ReadProcessMemory, a memcpy of a vtable slot, a Streamline array of
# constants): the conversion is the call's own.
resharper_cpp_clang_tidy_bugprone_multi_level_implicit_pointer_conversion_highlighting = none
# A test builds a handle or an address from a number, on purpose: a fake one, or one read from another process.
resharper_cpp_clang_tidy_performance_no_int_to_ptr_highlighting = none
# A macro that pastes its argument into a name (FL_DEV(vkGetDeviceQueue)) cannot put it in parentheses.
resharper_cpp_clang_tidy_bugprone_macro_parentheses_highlighting = none
# A test's fixture: a global with a std::string in it, a memcpy into a byte buffer that is not a string, two strings
# added in an assertion's message.
resharper_cpp_clang_tidy_bugprone_throwing_static_initialization_highlighting = none
resharper_cpp_clang_tidy_bugprone_not_null_terminated_result_highlighting = none
resharper_cpp_clang_tidy_performance_inefficient_string_concatenation_highlighting = none
# layout_test checks facts the compiler already knows - a constant's length, a bit-field's range - at run time, so
# that a failure prints which one: the branch that cannot be taken is the test passing.
resharper_cpp_unreachable_code_highlighting = none
resharper_cpp_clang_tidy_clang_diagnostic_unreachable_code_highlighting = none
resharper_cpp_clang_tidy_clang_diagnostic_tautological_value_range_compare_highlighting = none
# --- CODE LOADED INTO A GAME: the Overlay, the Vulkan layer, the shared-memory headers ---
# D100: here only a defect is fixed, in a pull request of its own with the hook's overhead measured (rule 5,
# 14_TESTING §Hook overhead). The report found none. What it did say about these files, each kept as it is:
[src/native/{FrameLedger.Overlay,FrameLedger.VkLayer,FrameLedger.Shm}/**.{cpp,h,inl}]
# Arms with the same body are different cases, each with its own comment saying why it reads as it does.
resharper_cpp_clang_tidy_bugprone_branch_clone_highlighting = none
# MinHook's MH_ALL_HOOKS is NULL, by its header.
resharper_cpp_zero_constant_can_be_replaced_with_nullptr_highlighting = none
# Kept on purpose, each with its reason beside it in dllmain.cpp: a retired enumerator whose bit must not be reused;
# slSetTag, deprecated by a header whose replacement no measured module exports.
resharper_cpp_enumerator_never_used_highlighting = none
resharper_cpp_clang_tidy_clang_diagnostic_deprecated_declarations_highlighting = none
# DEAD CODE, FOUND AND NOT REMOVED: nothing calls dllmain.cpp's LogNoteW. D100 changes code loaded into a game for
# a defect only, so taking it out is the owner's to order (HANDOFF, the native report's section) and not this
# report's to do. Until then this one file is not read for an unused function. Remove this section with it.
[src/native/FrameLedger.Overlay/src/dllmain.cpp]
resharper_cpp_declarator_never_used_highlighting = none
resharper_cpp_clang_tidy_clang_diagnostic_unused_function_highlighting = none
# The guard's rule matcher, which the Overlay compiles too. ResolveTolerance clears its out parameter with memset, as
# ResetRules does: a Tolerance is flags with default member initialisers, which is all "non-trivially default
# constructible" means here, and zero is what those initialisers say.
[src/native/FrameLedger.Injector/{src/fl_ac_rules.cpp,include/fl_ac_rules.h}]
resharper_cpp_clang_tidy_bugprone_raw_memory_call_on_non_trivial_type_highlighting = none
+100
View File
@@ -0,0 +1,100 @@
# ACCESSIBILITY.md (beta.35, D88) sends every accessibility barrier here. It is its own form rather than part of the bug
# report because a barrier needs different facts: what the reporter uses to read or operate the screen, and how far it
# stops them. Nothing here asks about the reporter themself.
name: Accessibility problem
description: Something in FrameLedger cannot be reached, read or seen — with the keyboard, a screen reader, a contrast theme, Magnifier or another aid
labels: [accessibility]
body:
- type: markdown
attributes:
value: |
Thank you for telling us. **You never need to tell us anything about yourself or a disability** — what you use
and what happened is enough. Screenshots and recordings help but are optional.
If FrameLedger refused to hook a game, stopped measuring or turned hooking off **without your knowing** — a
safety notice you could not perceive — say so below: it is a safety problem too, and it is handled first.
What FrameLedger aims for and what it cannot do yet: [ACCESSIBILITY.md](https://github.com/poli0981/frameledger/blob/main/ACCESSIBILITY.md).
- type: textarea
id: task
attributes:
label: What were you trying to do?
placeholder: 'e.g. switch hooking on for a game, read a session''s frame rate, answer the terms window'
validations:
required: true
- type: textarea
id: what-happened
attributes:
label: What happened instead?
description: What you heard, saw or could not do — and what you expected.
validations:
required: true
- type: textarea
id: steps
attributes:
label: How to get there
placeholder: |
1. Open Games (Ctrl+2)
2. Tab to …
3. …
- type: input
id: app-version
attributes:
label: FrameLedger version
description: Help ▸ About
validations:
required: true
- type: input
id: os
attributes:
label: Windows version
placeholder: 'Windows 11 26100.2314'
validations:
required: true
- type: dropdown
id: aid
attributes:
label: What do you use?
multiple: true
options:
- Keyboard only
- Narrator
- NVDA
- JAWS
- Windows Magnifier
- A Windows contrast theme
- Voice Access or other speech input
- Display scaling or a large text size
- Something else (say below)
validations:
required: true
- type: input
id: aid-version
attributes:
label: Its version, if you know it
placeholder: 'e.g. NVDA 2026.1'
- type: dropdown
id: severity
attributes:
label: How much does it stop you?
description: Optional — the maintainer sets it when reading the report (ACCESSIBILITY.md, How serious it is).
options:
- Not sure
- Critical — I cannot do the task at all
- Serious — only with great difficulty or a workaround
- Moderate — it is tiring or inconsistent
- Minor — it barely gets in the way
- type: textarea
id: more
attributes:
label: Anything else
description: A screenshot, a recording, or a diagnostic file from Help ▸ Report a bug (built on your PC; you choose what to attach).
+3 -3
View File
@@ -3,9 +3,9 @@
#
# The nuget entry also covers .config/dotnet-tools.json (the SonarScanner for
# .NET, since 2026-10-25). No dotnet-sdk entry, on purpose: global.json is a
# floor (10.0.100) with rollForward latestFeature, so every machine and the CI
# runner already build with the newest 10.0 SDK they have; raising the floor is
# a decision, not an update (docs/12_BUILD.md).
# floor (11.0.100 since beta.37, D94) with rollForward latestFeature, so every
# machine and the CI runner already build with the newest 11.0 SDK they have;
# raising the floor is a decision, not an update (docs/12_BUILD.md).
#
# COOLDOWN (2026-10-25): a version update waits seven days after a release is
# published, so a release that is pulled or found compromised in its first
+12
View File
@@ -30,6 +30,18 @@ review questions from `docs/19_SAFETY_AND_ANTICHEAT.md` §Review checklist.
**Measured overhead:** <!-- e.g. "0.31 µs/present, game Avg FPS delta -0.2%" -->
## User-interface changes only
Skip this section if the PR does not change what the App shows. Otherwise — `ACCESSIBILITY.md` §For contributors,
`docs/08_UI.md` §Accessibility:
- [ ] Every control that shows no text of its own has a name, from `.resx` in `en`, `vi` and `ja`
- [ ] Colours come from the theme only, and nothing is shown by colour alone
- [ ] Everything new can be reached and used with the keyboard, in reading order, and Esc leaves a new dialog
- [ ] `ACCESSIBILITY.md` (and `LIMITATIONS.md`) changed in this PR if what works or what does not moved
**Checked by hand:** <!-- e.g. "Tab through the new card with the mouse unplugged; Narrator reads its name and value" -->
## Safety-string changes only
- [ ] `Safety_*` resx keys have human-reviewed `ja` (never machine-drafted — `docs/09_I18N.md`)
+2 -2
View File
@@ -20,8 +20,8 @@ name: CI
on:
push:
# `alpha` since 2026-10-31 (D73, docs/13_CI_CD.md §The alpha line): the same source on a prerelease .NET, whose
# merges and syncs from main need the same gate under its own global.json.
# `alpha` since 2026-10-31 (D73, docs/13_CI_CD.md §The alpha line): the same source on the .NET that line is
# trying, whose merges and syncs from main need the same gate under its own global.json.
branches: [main, alpha]
pull_request:
workflow_dispatch:
+6
View File
@@ -16,6 +16,11 @@
name: CodeQL
# `workflow_dispatch` since 2026-11-11 (beta.37): on a pull request the C# analysis
# is an OVERLAY - it extracts the changed files on top of the database cached from
# `main` - so it cannot show what the extractor makes of the whole source, or of a
# construct it has not met (the move to .NET 11, a C# 15 feature). A dispatch on a
# branch is a full analysis of that branch (docs/13_CI_CD.md §C# 15 and the analyses).
on:
push:
branches: [main]
@@ -23,6 +28,7 @@ on:
branches: [main]
schedule:
- cron: '27 4 * * 1'
workflow_dispatch:
permissions:
contents: read
+178
View File
@@ -0,0 +1,178 @@
# Qodana (owner decisions D93, 2026-11-10, and D99, 2026-11-11; docs/13_CI_CD.md §Qodana).
#
# JetBrains' template for the Qodana Scan action, changed where FrameLedger differs
# and each change said here:
#
# - The linter is Qodana for .NET (Rider's engine) in NATIVE mode, on a Windows
# runner: the projects target Windows, and the linter then builds and loads them
# with the SDK this job installs from global.json - no container, and no
# EnableWindowsTargeting. qodana.yaml names the linter and the mode; nothing is
# passed as an action argument. (Until beta.37 this was the Community linter in a
# Linux container, the one Linux job this repository had.)
# - The action is pinned to a commit, as every action that is not GitHub's own is
# here.
# - Permissions are the ones the job uses: it reads the repository, writes a check
# run with the annotations and one comment on the pull request. The template's
# `contents: write` is for pushing quick-fixes, which this job never does.
# - A workflow of its own, as JetBrains advises, and ADVISORY: `main`'s required
# contexts stay `check` and the two `analyze` jobs. A problem Qodana finds, a
# Qodana Cloud outage or a lapsed licence is a red job beside a green `check`,
# never a blocked merge.
# - Two jobs, two reports (D99): `managed` reads the solution - the C#, the XAML and
# what Rider's plugins read around it - and reports to the Qodana Cloud project
# `frameledger`; `native` reads the C++ of src/native through a Visual Studio
# solution CMake generates for it, with settings of its own (qodana-native.yaml),
# and reports to `frameledger-cpp`. Neither reads what the other does. The pull
# request's comment and annotations are the managed job's alone.
# - `main` only: the `alpha` branch is main's source under another release line
# (docs/13_CI_CD.md §The alpha line), and main's is what is analysed.
name: Qodana
on:
workflow_dispatch:
pull_request:
branches: [main]
push:
branches: [main]
permissions: {}
concurrency:
group: qodana-${{ github.ref }}
cancel-in-progress: true
jobs:
managed:
name: Qodana (managed)
# A pull request from this repository, or a push or a dispatch. QODANA_TOKEN
# reaches neither a fork's pull request nor Dependabot's, and their token cannot
# write the check run or the comment.
if: >-
github.event_name != 'pull_request' ||
(github.event.pull_request.head.repo.full_name == github.repository &&
github.event.pull_request.user.login != 'dependabot[bot]')
runs-on: windows-latest
timeout-minutes: 60
permissions:
contents: read
checks: write
pull-requests: write
steps:
- uses: actions/checkout@v7
with:
# The pull request's own commit, not its merge commit (empty on a push or
# a dispatch, which then check out the commit that started them).
ref: ${{ github.event.pull_request.head.sha }}
# A full history is required for pull request analysis.
fetch-depth: 0
# Native mode runs the linter in this job's environment, so the SDK it loads
# the projects with is this one. No MSVC step here on purpose: it exports
# Platform=x64, which MSBuild reads as a solution platform (build.ps1's
# Clear-VcxprojPlatform).
- name: Set up .NET
uses: actions/setup-dotnet@v6
with:
global-json-file: global.json
# Restored only: `check` saves this cache, under the same key.
- name: Restore the NuGet cache
uses: actions/cache/restore@v6
with:
path: ~/.nuget/packages
key: nuget-${{ runner.os }}-${{ hashFiles('**/Directory.Packages.props', 'global.json') }}
restore-keys: nuget-${{ runner.os }}-
# Built before it is inspected: a build that fails is then one red step with
# the compiler's own message, not a report of thousands of unresolved symbols.
# The native DLLs are absent here, so the .targets that copy them warn and the
# build goes on, as on a fresh clone.
- name: Build what the linter loads
shell: pwsh
run: |
dotnet build FrameLedger.slnx -c Release
if ($LASTEXITCODE -ne 0) { throw 'the build failed' }
# pr-mode false: the whole solution every time, so a pull request and `main`
# are judged by the same list. upload-result: the SARIF and the linter's log
# as an artifact of the run, for a reader with no Qodana Cloud account.
# Caches are saved from `main` only (a pull request reads them), under keys of
# this job's own: one per branch would crowd the repository's cache quota,
# which NuGet's share.
- name: Qodana Scan
uses: JetBrains/qodana-action@3e8d76ff0e0241618abd808682f426afe84bab3f # v2026.2.2
with:
pr-mode: false
upload-result: true
artifact-name: qodana-report
cache-default-branch-only: true
primary-cache-key: qodana-managed-${{ github.ref_name }}-${{ github.sha }}
additional-cache-key: qodana-managed-main
env:
QODANA_TOKEN: ${{ secrets.QODANA_TOKEN }}
QODANA_ENDPOINT: 'https://qodana.cloud'
native:
name: Qodana (native)
# The same condition as `managed`, for the same reason: QODANA_TOKEN_NATIVE
# reaches neither a fork's pull request nor Dependabot's.
if: >-
github.event_name != 'pull_request' ||
(github.event.pull_request.head.repo.full_name == github.repository &&
github.event.pull_request.user.login != 'dependabot[bot]')
runs-on: windows-latest
timeout-minutes: 60
# No annotations and no comment from this job (below), so it writes nothing.
permissions:
contents: read
steps:
- uses: actions/checkout@v7
with:
ref: ${{ github.event.pull_request.head.sha }}
fetch-depth: 0
# global.json sits above the solution, and the linter resolves it.
- name: Set up .NET
uses: actions/setup-dotnet@v6
with:
global-json-file: global.json
# The Visual Studio generator finds the toolset itself: no MSVC step. Configure
# fetches MinHook, Catch2 and jsmn into the build tree, as the gate's does.
- name: Generate the solution and the one generated header
shell: pwsh
working-directory: src/native
run: |
cmake --version
cmake --preset x64-vs-analysis
if ($LASTEXITCODE -ne 0) { throw 'cmake configure failed' }
cmake --build --preset x64-vs-analysis --target fl_ac_floor_gen
if ($LASTEXITCODE -ne 0) { throw 'generating the blocklist floor failed' }
if (-not (Test-Path ../../build/native/x64-vs-analysis/FrameLedgerNative.slnx)) {
Get-ChildItem ../../build/native/x64-vs-analysis -Filter 'FrameLedgerNative.*' | ForEach-Object { Write-Host $_.Name }
throw 'CMake did not write FrameLedgerNative.slnx (qodana-native.yaml names it)'
}
# A report of its own: its own settings file, results folder, artifact, cache
# keys and Qodana Cloud project. The pull request's comment and annotations are
# the managed job's; a second job of the same linter would overwrite them.
- name: Qodana Scan
uses: JetBrains/qodana-action@3e8d76ff0e0241618abd808682f426afe84bab3f # v2026.2.2
with:
args: --config qodana-native.yaml
results-dir: ${{ runner.temp }}/qodana-native/results
cache-dir: ${{ runner.temp }}/qodana-native/caches
pr-mode: false
upload-result: true
artifact-name: qodana-report-native
cache-default-branch-only: true
primary-cache-key: qodana-native-${{ github.ref_name }}-${{ github.sha }}
additional-cache-key: qodana-native-main
post-pr-comment: false
use-annotations: false
env:
QODANA_TOKEN: ${{ secrets.QODANA_TOKEN_NATIVE }}
QODANA_ENDPOINT: 'https://qodana.cloud'
+6 -1
View File
@@ -24,7 +24,7 @@
#
# THE ALPHA LINE (2026-10-31, owner decisions D73-D76; 12_BUILD §The alpha line): the
# RELEASE_LINE file says `alpha` on the long-lived `alpha` branch, where the same source
# builds on a prerelease .NET. Its tags are `vX.Y.Z-alpha.N` and nothing else, no other
# builds on the .NET the line is trying. Its tags are `vX.Y.Z-alpha.N` and nothing else, no other
# line may carry one, and its packages go to Velopack's `alpha` channel
# (releases.alpha.json) - which no beta reads, as no alpha reads releases.win.json
# (11_UPDATER §Channels). The version gate below enforces all three.
@@ -67,10 +67,15 @@ jobs:
steps:
- uses: actions/checkout@v7
# The SDK global.json names builds everything. The .NET 10 line beside it is for `vpk` (the Pack step): 1.2.161
# ships net8.0, net9.0 and net10.0 builds and none that rolls forward to another major, so on a machine with
# .NET 11 alone the tool would not start. The alphas were packed under an 11 SDK because the hosted image
# happened to carry a .NET 10 runtime; this states the dependency (2026-11-11, beta.37, D94).
- name: Set up .NET
uses: actions/setup-dotnet@v6
with:
global-json-file: global.json
dotnet-version: 10.0.x
- name: Set up MSVC (x64)
uses: ilammy/msvc-dev-cmd@0b201ec74fa43914dc39ae48a89fd1d8cb592756 # v1.13.0
+1 -1
View File
@@ -1,5 +1,5 @@
# The GitHub Wiki is a copy of the repository's own pages (D53/D54, the owner, 2026-10-07): wiki/ (Home, the sidebar,
# the developer pages) and guide/ (the user guide, which is the End-user part) with LIMITATIONS.md, assembled and checked
# the developer pages) and guide/ (the user guide, which is the End-user part) with LIMITATIONS.md and ACCESSIBILITY.md, assembled and checked
# by tools/wiki-build.ps1 - the check build.ps1 runs as `wiki-check`. A page edited on the Wiki tab is overwritten here;
# edit wiki/ or guide/ instead (docs/13_CI_CD.md §wiki.yml).
#
+171
View File
@@ -0,0 +1,171 @@
# Accessibility
FrameLedger should be usable by everyone who plays games on a Windows PC — with a keyboard alone, with a screen reader,
with a Windows contrast theme or Magnifier, and without telling colours apart. This page says what FrameLedger aims
for, what works today, how that was checked, what does not work yet and what to do instead, and how to tell us about a
barrier. It is written for the person who has just met one; the details for developers are in
[`docs/08_UI.md`](docs/08_UI.md#accessibility).
FrameLedger is a desktop program: its own window, its dialogs, its document windows, the icon by the clock and the
notices Windows shows for it. It draws nothing inside a game, so it cannot make a game itself any easier to use.
*Beta software: this page describes version 0.1.0-beta.49. In the app: Help ▸ Accessibility.*
## What FrameLedger aims for
- **WCAG 2.2 at level AA**, applied to desktop software as W3C's
[WCAG2ICT](https://www.w3.org/TR/wcag2ict-22/) guidance explains. This is a goal that guides the work, not a claim:
no formal evaluation has been made, and FrameLedger does not say it conforms to any standard.
- **Every task can be done with the keyboard alone.**
- **A screen reader can read every control and every value**, in the language FrameLedger is shown in, and hears what
changes — a notice, a result.
- **Text and controls keep their contrast** in the light and the dark theme and under Windows contrast themes, and
**colour is never the only way something is shown.**
- **Everything stays sharp at any display scale**, and when the window moves between screens.
## What works today
### The keyboard
| Keys | What they do |
|---|---|
| **Ctrl+1** … **Ctrl+5** (also with the numeric keypad) | Open Dashboard, Games, Compare, Logs and Settings, and move the keyboard into that page; a screen reader hears its title. **Ctrl+6** opens Advanced the same way while Advanced metrics is switched on, and does nothing while it is off. |
| **Alt** or **F10**, then the arrow keys and **Enter** | Move to the menus. **Alt** with a menu's underlined letter opens it (*File*, *Tools*, *Help* in English). |
| **Tab** and **Shift+Tab** | Move through a page or a dialog in reading order. |
| The **arrow keys**, **Home**, **End**, **Page Up**, **Page Down** | Move inside a list or a table — a game's sessions, the games list, the Compare list, the import checklist. |
| **Enter** or **Space** | Press the button, switch or box that has the focus. **Enter** on a game in the list opens its page. |
| **Esc** | Closes a dialog the way its **Close** or **Cancel** button does: nothing is changed. A dialog that has no Close button waits for an answer, and the terms window has its own **Decline** and **Accept**. |
| **Enter** in the hooking warning | Switches hooking on once the phrase is typed exactly; before that it does nothing. |
| On a chart: the **arrow keys**, **Ctrl** with an arrow, **A** | Tab reaches every chart. The arrows move its view, Ctrl with an arrow zooms, A shows all of it again (so does a double-click); the page around it stays put. |
The same keys are in the user guide, [The screens](guide/07-screens.md#keyboard-and-mouse).
### Screen readers
- **Every page, menu, dialog and setting can be reached and used with Narrator.** The owner checked this on
2026-11-09, keyboard only with the mouse disconnected, on 0.1.0-beta.34 and 0.1.0-alpha.3, and found no fault.
- **Every page has headings**: its title is a heading of the first level, each section one of the second, so a screen
reader's heading keys move from section to section.
- **Every chart says what it is and where its numbers are written as text**, and which keys move it.
- **Every button, switch and box that shows no text of its own has a name**, from the same translations as the screen.
- **Lists, drop-down options and table cells are read by the words they show**, and a button repeated on each row names
its row — *Open: 11/9/2026 14:03*, *Play: Cyberpunk 2077*, *Turn off: Cyberpunk 2077*.
- **The frame-rate readout says the game's own frames, the displayed frames and the frame-generation factor together**
(*Native 62 FPS, displayed 118 FPS, frame generation ×1.9*) — never the larger number alone; a ray-tracing chip says its
value and where it came from (*Ray Tracing: Yes, measured*).
- **What appears is said as it appears**: a safety notice (hooking refused, measuring stopped, hooking turned off), the
capture agent going away and coming back, an update's steps, a short message at the bottom of the window, a notice on a
game's page and a result of **Tools ▸ Database maintenance** — in the words shown, while FrameLedger's window is the
active one. While a game has the screen nothing interrupts it; the notice stays in the window, and Windows' own
notifications from the icon by the clock are read by the screen reader itself.
- **The notices above the page can be read where they are shown**, and the button that answers one says which notice it
answers.
- **Tab goes from control to control**: pictures, lists and the rows of Settings are not stops of their own.
- **Dialogs say they are dialogs**, keep the focus inside them until they close, and then give it back to the control
that opened them. A message box opens over FrameLedger's window, never behind it.
### Seeing the screen
- **Light, dark, or the same as Windows** (Settings ▸ Theme). FrameLedger takes every colour from the theme, and every
text is measured against what is painted under it, in both themes.
- **Windows contrast themes** (Settings ▸ Accessibility ▸ Contrast themes) — Aquatic, Desert, Dusk, Night sky, or one
whose colours you changed: every colour in the window becomes one of the theme's, whatever Settings ▸ Theme says, as
soon as you turn one on, edit it or turn it off. The charts draw in its colours too, their lines and marks still told
apart by pattern, and Settings ▸ Theme says a contrast theme is in use.
- **The colour a game's picture lends its page** is used only as far as every text on it still reads, and never under a
contrast theme.
- **Windows display scaling** at any size, on every screen; **Magnifier** (Windows logo key + Plus) works as with any
program.
- Labels such as *Hooking on* and *Not installed*, and the ray-tracing chips, always carry their words; a chip that reads
*N/A* is also drawn with a dashed edge.
- **No chart tells two series apart by colour alone**: its lines are solid, dashed or dotted, its points circles,
triangles or squares, its shaded stretches striped or dotted.
### Languages
FrameLedger is shown in English, Vietnamese or Japanese (Settings ▸ Language), and the names a screen reader reads come
from the same translations. The documents — the user guide, this page, the legal texts — are in English.
## How it is checked
- **On every build**, by the project's own tests: what a screen reader is given for every page, window and dialog —
every place Tab stops has a name, no list or table item is read by its program name, no text is hidden from a screen
reader inside a control, no two buttons in a list share a name; each thing that appears is said once, in its words;
every page has one first-level heading and every section title a level; no chart tells two series apart by colour
alone; a chart's keys move it and stay its own; every control that shows no text has a name; no colour is written
into a screen; every text reads against its background in both themes and after a theme switch, and under Windows'
four contrast themes and two edited ones is drawn in the theme's own colours at 4.5:1 or better; **Esc** closes the
open dialog without an answer, and a dialog gives the focus back; **Ctrl+1** … **Ctrl+5** open the pages in their
order and move the keyboard into them, and **Ctrl+6** the Advanced page only while its option is on; a page slides in only while Windows' animation effects are on; the program
declares it follows each screen's scaling.
- **By hand**: the owner's pass of 2026-11-09 above, repeated for each release that changes a screen.
- **Not tested yet**: NVDA, JAWS, Voice Access, touch, Windows 10, and other people's PCs. They should work through the
same Windows interface Narrator uses, but nobody has checked. Contrast themes are checked by the tests above and have
not yet been looked at on a screen.
## Known limitations
Each one with what to do instead. A barrier that is not on this list is worth reporting.
- **The Advanced page's "This PC" does not announce its readings as they change.** They change every second; a
screen reader reads the table when you move through it, and each row says its reading, its value and the time it
was read at.
- **A chart's points cannot be read one by one.** A screen reader hears the chart's name and where its numbers are:
a session's figures and *Statistics per series* in its summary, the table under the Compare chart, the Sessions table
and the list of changes under a trend chart, and **File ▸ Export ▸ Session CSV** or **Session JSON**.
- **In the user guide and the other documents the app shows, headings are read as ordinary text**: Windows' document
view gives them no heading role. Their links and their order work as on any page.
- **The short messages at the bottom of the window close after four seconds**, and cannot be read again once closed. A
screen reader is told each one as it appears.
- **Windows' text size setting** (Settings ▸ Accessibility ▸ Text size) **is not followed.** Use display scaling
(Settings ▸ System ▸ Display ▸ Scale) or Magnifier.
- **Pausing capture is only in the menu of the icon by the clock.**
- **The windows Windows itself shows** — choosing a file, the administrator prompt — are Windows', not FrameLedger's.
## Report an accessibility problem
1. Open an issue with the
[Accessibility problem form](https://github.com/poli0981/frameledger/issues/new?template=accessibility.yml).
2. Say what you were trying to do and what happened instead; your FrameLedger version (Help ▸ About) and your Windows
version; and what you use — the keyboard alone, Narrator, NVDA, JAWS, Magnifier, a contrast theme — with its version.
The steps to get there help, and so does a screenshot or a recording, but neither is required. **You never need to
tell us anything about yourself or a disability.**
3. If you want to attach details from your PC, **Help ▸ Report a bug** builds them into a file on your PC and shows you
what is inside first; nothing is uploaded by itself.
**A safety notice you could not perceive** — hooking refused, measuring stopped or hooking turned off, without your
knowing — is a safety problem as well: say so in the report, and it is handled first.
### How serious it is
You do not have to choose; the maintainer sets it when reading the report.
| Severity | What it means | For example |
|---|---|---|
| **Critical** | A task cannot be done at all. | Hooking cannot be switched off with the keyboard; the terms window or the hooking warning cannot be answered; a safety notice cannot be read. |
| **Serious** | It can be done only with great difficulty or a workaround. | A value can be read only by exporting the session. |
| **Moderate** | It is tiring or inconsistent. | A name that differs from what the screen shows; an extra stop for Tab. |
| **Minor** | It barely gets in the way. | A tooltip the keyboard cannot reach, when its words are shown elsewhere too. |
### What happens next
FrameLedger has one maintainer, so there are no promised times. Every report is read and answered, without doubting what
you describe. Critical problems come before other work; a workaround, when there is one, is written on the issue; the
fix ships in a release, and the issue is closed once you say it works for you — or when a fix has been out for a while
and nobody has answered.
## Who looks after this
The project's owner, poli0981, keeps this page and the checks behind it, and reviews both for each release that changes
a screen. A change to FrameLedger that moves anything on this page changes the page in the same pull request.
## For contributors
A change to the App keeps what the tests above hold: a name, from the translations in all three languages, for every
control that shows no text; colours only from the theme. Say in the pull request what you checked with the keyboard
and with Narrator. The rules and the checks are in [`docs/08_UI.md`](docs/08_UI.md#accessibility) and
[`CONTRIBUTING.md`](CONTRIBUTING.md).
## Feedback
To improve this page, open an issue or a pull request. A barrier goes through the form above, so that it is not lost.
+96
View File
@@ -35,12 +35,108 @@ numeric core and does not move; the heading is the full version — corrected 20
## [Unreleased]
### Added
- **A session records what it was played with — a controller, the keyboard and mouse, both, or neither** (D155–D163,
the owner's decisions of 2026-11-14; *Settings ▸ Controllers and input*, on unless you switch it off). While a game
is recorded, FrameLedger's agent notes — from Windows, never from the game — which game controllers are connected
and in use, their battery level, and whether the keyboard or mouse is in use, and the session stores how the time
the game was in front divides between them, how often each controller connected and dropped, and its battery. It
works for a game FrameLedger does not measure too. **No button, stick, key or text is kept; nothing is sent to a
controller; and nothing is read until you have accepted the Privacy Policy that describes it** — after this
update, sessions recorded before FrameLedger has been opened once have no such figures. The screens that show the
figures come with the same release. Xbox-compatible controllers and the DualSense are read for use; another
controller is listed and never counted as in use. **None of this had been read from a real controller when it was
written** — it was checked against data written by hand (`LIMITATIONS.md` §Controllers and input).
- **The terms are shown once: Privacy Policy 2.22 and Disclaimer 2.18.** Both said FrameLedger never reads input.
They now say what is read — that a controller, or the keyboard or mouse, is in use — what is kept, and what never
is (`docs/legal-drift-history.md`).
- **For developers: the arithmetic of controller statistics, with nothing reading a device yet** (D155–D163, the
owner's decisions of 2026-11-14; `docs/21_INPUT_DEVICES.md`, `docs/03_METRICS.md` §Input). `FrameLedger.Domain.Input`
says whether an XInput state or a DualSense report is a pad in use and what its battery is, adds a battery's readings
up, counts switches, and decides when two entries are one pad seen twice; `Application.Input` holds one session's
slices and turns them into a reading a second; `InputSessionBuilder` turns a session's readings into its totals, its
controllers as JSON and its timeline. A telemetry sample can carry a reading; none does. **Every byte and reading
under test was written by hand — no report from a real controller has passed through this code.**
- **For developers: a session can store what it was played with — schema 0024** (D160; `docs/06_DATA_MODEL.md`). Nine
columns on `sessions` (the time in front and its four classes, the switches of method, the sources that answered,
the controllers as JSON) and three on `games` (what a store lists of controller support), ADD COLUMN only; the
`.partial` file gains chunk 11, so a session recovered after the Agent died keeps its readings; the finalizer
stores them for hooked and unhooked sessions alike; the input series are `sensor_blobs` rows named `input.*`, which
the sensors' read leaves out. Nothing produces a reading yet. **A ledger this build has opened cannot be opened by
0.1.0-beta.49 or earlier.**
- **For developers: the input reader, which the Agent does not start yet** (D155, D158, D159; `docs/21_INPUT_DEVICES.md`
§The reader). `Application.Input.InputReader` takes a slice about ten times a second from
`Infrastructure.Input.Win32InputBackend` — raw input for joysticks and gamepads delivered to a window that is never
shown, three of XInput's getters, the time of the last keyboard or mouse input, the window in front — on the thread
of `InputDeviceHub`, which exists only while a session has a view of it. It opens no device, sends nothing to one,
and registers for no keyboard or mouse. A new gate, `tools/input-readonly-check.ps1`, holds the reader's Windows
calls to a committed list and refuses, on the list or off it, a call that sends to a device, injects input or reads
a key; no other managed source and no native source may name an input function. **With no controller at hand, what
the reader does with one connected is not established**: its tests use a Windows whose answers they write, and one
suite runs it on the real Windows of the machine and asserts only that it opens, answers and closes.
### Changed
- **Advanced metrics is closed: the ten readings that were "being studied" are now "not read", with the reason**
(D154, the owner's decision). What a game sets and asks through Streamline's own functions for a feature — DLSS
frame generation's options and state, the video memory DLSS estimates for itself, the options of ray
reconstruction, Image Scaling and Dynamic Vibrance, the PC-latency markers — is reached only by handing the game
a function of FrameLedger's in place of Streamline's, which is refused. The *Sources* list says so, in English,
Vietnamese and Japanese, and the guide no longer says a later release reads more. Nothing that was read stops
being read.
- **The two lines of the graphics card's time say what each is, not which holds the other** (D152, the owner's
decision). *GPU active render (idle time excluded)* and *GPU render, start to end (idle time included)* are now
*GPU active render (the driver's own figure)* and *GPU render (between the driver's start and end marks)*, in
English, Vietnamese and Japanese, and the guide says why: in Cyberpunk 2077 without frame generation the
driver's active time is longer than its two marks are apart. The numbers are unchanged.
- **The guide's page on Advanced metrics says what two of the driver's switches mean** (`guide/12`): the latency
report is there in a game that has Reflex built in, whether Reflex is on or off in its menu; `bUseGameSleep`
says the game calls Reflex's sleep — NVIDIA asks for that even with Reflex off — and `bLowLatencyMode` is the
one that says the mode is on; frame generation needs Reflex, not the reverse. From NVIDIA's Streamline guide
for Reflex, and from the owner's ten sessions on 0.1.0-beta.49.
### Documentation
- **Owner-only item 47 was run and read** (`docs/HANDOFF.md`): in Cyberpunk 2077 without frame generation the
driver's active GPU time is longer than its own two marks are apart on 4 917 of 5 394 frames (1 849 µs beside
1 571 µs) and on none of 129 900 frames in five other games — the difference is the driver's; samples without a
report are now confined to a game's start; a session with the option off has no latency line; and the stop of
D141 fired in a real game for the first time (Kingdom Come: Deliverance II, one call of 5.3 ms while it
started).
- **Streamline's programming guides for Reflex and for DLSS Ray Reconstruction (2.14.16) read against the code**
(`docs/spike-notes/30-reflex-latency.md`, *Added 2026-11-14*). Ray reconstruction was already measured the way
the guide describes; two comments in the Overlay's source and one line of `17_HOOK_ENGINE` said super
resolution and ray reconstruction run together, and they do not. `03_METRICS` and `05_DETECTION` no longer
promise "PC latency as reported by Reflex" (D147). No code changed: comments only.
## Releases
Newest first. Each file holds that release's section exactly as it stood here.
| Version | Date | Notes |
|---|---|---|
| `0.1.0-beta.49` | 2026-11-14 | [`docs/changelog/0.1.0-beta.49.md`](docs/changelog/0.1.0-beta.49.md) |
| `0.1.0-beta.48` | 2026-11-14 | [`docs/changelog/0.1.0-beta.48.md`](docs/changelog/0.1.0-beta.48.md) |
| `0.1.0-beta.47` | 2026-11-14 | [`docs/changelog/0.1.0-beta.47.md`](docs/changelog/0.1.0-beta.47.md) |
| `0.1.0-beta.46` | 2026-11-13 | [`docs/changelog/0.1.0-beta.46.md`](docs/changelog/0.1.0-beta.46.md) |
| `0.1.0-beta.45` | 2026-11-12 | [`docs/changelog/0.1.0-beta.45.md`](docs/changelog/0.1.0-beta.45.md) |
| `0.1.0-beta.44` | 2026-11-12 | [`docs/changelog/0.1.0-beta.44.md`](docs/changelog/0.1.0-beta.44.md) |
| `0.1.0-beta.43` | 2026-11-12 | [`docs/changelog/0.1.0-beta.43.md`](docs/changelog/0.1.0-beta.43.md) |
| `0.1.0-beta.42` | 2026-11-12 | [`docs/changelog/0.1.0-beta.42.md`](docs/changelog/0.1.0-beta.42.md) |
| `0.1.0-beta.41` | 2026-11-12 | [`docs/changelog/0.1.0-beta.41.md`](docs/changelog/0.1.0-beta.41.md) |
| `0.1.0-beta.40` | 2026-11-12 | [`docs/changelog/0.1.0-beta.40.md`](docs/changelog/0.1.0-beta.40.md) |
| `0.1.0-beta.39` | 2026-11-11 | [`docs/changelog/0.1.0-beta.39.md`](docs/changelog/0.1.0-beta.39.md) |
| `0.1.0-beta.38` | 2026-11-11 | [`docs/changelog/0.1.0-beta.38.md`](docs/changelog/0.1.0-beta.38.md) |
| `0.1.0-alpha.4` | 2026-11-11 | [`docs/changelog/0.1.0-alpha.4.md`](docs/changelog/0.1.0-alpha.4.md) |
| `0.1.0-beta.37` | 2026-11-11 | [`docs/changelog/0.1.0-beta.37.md`](docs/changelog/0.1.0-beta.37.md) |
| `0.1.0-beta.36` | 2026-11-10 | [`docs/changelog/0.1.0-beta.36.md`](docs/changelog/0.1.0-beta.36.md) |
| `0.1.0-beta.35` | 2026-11-09 | [`docs/changelog/0.1.0-beta.35.md`](docs/changelog/0.1.0-beta.35.md) |
| `0.1.0-alpha.3` | 2026-11-06 | [`docs/changelog/0.1.0-alpha.3.md`](docs/changelog/0.1.0-alpha.3.md) |
| `0.1.0-beta.34` | 2026-11-06 | [`docs/changelog/0.1.0-beta.34.md`](docs/changelog/0.1.0-beta.34.md) |
| `0.1.0-alpha.2` | 2026-11-03 | [`docs/changelog/0.1.0-alpha.2.md`](docs/changelog/0.1.0-alpha.2.md) |
| `0.1.0-beta.33` | 2026-11-03 | [`docs/changelog/0.1.0-beta.33.md`](docs/changelog/0.1.0-beta.33.md) |
| `0.1.0-alpha.1` | 2026-10-31 | [`docs/changelog/0.1.0-alpha.1.md`](docs/changelog/0.1.0-alpha.1.md) |
| `0.1.0-beta.32` | 2026-10-31 | [`docs/changelog/0.1.0-beta.32.md`](docs/changelog/0.1.0-beta.32.md) |
+106 -11
View File
@@ -6,7 +6,7 @@ Read this file first, then the reading order at the bottom. **`docs/19_SAFETY_AN
## Architecture in one paragraph
A C++ DLL (`FrameLedger.Overlay.dll`) is injected into games the user explicitly enabled. It hooks the presentation path (DXGI for D3D11/D3D12, plus OpenGL) and the upscaler/RT APIs, writes fixed-size frame records into a lock-free shared-memory ring, and never blocks. A C# Agent drains that ring at ~10 Hz, enriches with GPU telemetry from vendor APIs, and writes sessions to SQLite. A WPF UI reads SQLite. Vulkan uses an implicit **layer** instead of hooking. If injection is refused or fails there is **no measurement**: the session records duration, hardware telemetry and the reason, and every measured field reads `N/A` (Tier 2, owner decision 2026-08-28).
A C++ DLL (`FrameLedger.Overlay.dll`) is injected into games the user explicitly enabled. It hooks the presentation path (DXGI for D3D11/D3D12, plus OpenGL) and the upscaler/RT APIs, writes fixed-size frame records into a lock-free shared-memory ring, and never blocks. A C# Agent drains that ring at ~10 Hz, enriches with GPU telemetry from vendor APIs, and writes sessions to SQLite. A WPF UI reads SQLite. Vulkan uses an implicit **layer** instead of hooking. If injection is refused or fails there is **no measurement**: the session records duration, hardware telemetry and the reason, and every measured field reads `N/A` (Tier 2, owner decision 2026-08-28) — **and, since owner decision D161 (2026-11-14), what the Agent reads from outside the game**: the game process's memory (beta.12, D43) and which controllers, or the keyboard and mouse, the session was played with (`docs/21_INPUT_DEVICES.md`). Nothing measured inside the game is among them.
## Non-negotiable rules
@@ -20,6 +20,8 @@ A C++ DLL (`FrameLedger.Overlay.dll`) is injected into games the user explicitly
> **Amended 2026-10-03 (owner decision D38): the evidence moved from before the grant to its trial.** "≥ 2 successful Tier-1 sessions" above could never be met by a game blocked at add time — since beta.8 that is every blocked game — nor after a ledger reset. Asked with that, the owner chose: a grant needs no session; the first two hooked sessions under it are a trial, and any end during the trial (a crash, a safety unhook, a floor stop, a new finding, or a stored hooked session with no frames or an abnormal exit) ends the exception **for good** (`games.ac_exception_trial_failed_at`, schema 0017, never cleared). Every other condition above is unchanged, and so is the last sentence.
3. **Never implement evasion.** No manual mapping, no PE header erasure, no thread hiding, no signature obfuscation, no unlinking from the PEB module list. Injection uses documented `LoadLibraryW`. The DLL keeps its real name, real exports, and version info. A performance tool must be *visible* to anti-cheat, not hidden from it. Any PR that makes FrameLedger harder to detect is rejected on principle.
4. **Never read or write game memory outside our own hooks.** We read arguments passed to APIs we hooked and COM/handle objects we legitimately own. No pattern scanning for game internals, no reading player/entity state, no writing to game memory. Ever.
> **Amended 2026-11-13 (owner decision D132): a driver's read-only getter, with the device of the swap chain we hook.** Asked with this rule quoted, the owner ruled that calling NVIDIA's two latency getters (`NvAPI_D3D_GetLatency`, `NvAPI_D3D_GetSleepStatus`) with the device behind a swap chain the Overlay hooks is inside this rule: it reads the driver's report, not game memory, and writes nothing. ~~Nothing shipped does it yet~~ **Shipped since 0.1.0-beta.47 (owner decisions D136, D138–D141)**, after the spike on `hook-harness`, a stub and the installed driver (`spike-notes` §31): only while a session records Advanced metrics (no switch of its own), about once a second on the presenting thread, through an `nvapi64.dll` the game had loaded — one in the game's own folder is refused, and a fault or a call of 2 ms or more ends it until the game exits (`17_HOOK_ENGINE` §Calls the Overlay makes into a driver). **It is the only call of its kind; another needs a ruling of the same kind.** Getters only, never a setter, a marker or a sleep, and never a call that loads `nvapi64.dll` into a game.
5. **The hook must never crash the game.** Every hook body: SEH-guarded, allocation-free, lock-free, no logging, no exceptions crossing the boundary. On repeated faults, self-disable (`17_HOOK_ENGINE` §Fault policy).
6. **FPS display rule (product requirement):** wherever FPS is shown and Frame Generation is active, show *Native FPS*, *Displayed FPS*, and the FG factor together (`62 → 118 FPS (×1.9 FG)`). Never a single inflated number. See `docs/03_METRICS.md`.
@@ -42,14 +44,14 @@ A C++ DLL (`FrameLedger.Overlay.dll`) is injected into games the user explicitly
| Concern | Choice |
|---|---|
| Managed runtime | .NET 10 (LTS, pinned via `global.json`), C# 14, TFM `net10.0-windows10.0.22621.0` with `SupportedOSPlatformVersion=10.0.19045.0` (Win10 22H2 floor, NFR-8), x64 only — `docs/12_BUILD.md`. **The `alpha` branch** (D73–D76, 2026-10-31) builds the same source on a prerelease .NET — .NET 11 RC3, C# 15 — and differs from main only where `docs/12_BUILD.md` §The alpha line lists; code on main must keep compiling as C# 14 on .NET 10, so no C# 15 syntax |
| Managed runtime | .NET 11 (STS, pinned via `global.json`: the 11.0 SDK band) and C# 15 since 0.1.0-beta.37 (owner decisions D94 and D96, 2026-11-11; .NET 10 and C# 14 until beta.36), TFM `net11.0-windows10.0.22621.0` with `SupportedOSPlatformVersion=10.0.19045.0` (NFR-8 as D95 amended it: Windows 11 is the supported platform; Windows 10 22H2 is not blocked, and is neither tested nor supported), x64 only — `docs/12_BUILD.md` §The main line's .NET. No preview feature (`runtime-async`, the updated memory-safety rules). **No C# 15 construct is used yet** — collection expression arguments, `closed`, `union`, labeled jumps: a construct is used only where the build, CodeQL and Qodana all read it, and Qodana's engine reads none of the four (`docs/13_CI_CD.md` §C# 15 and the analyses). .NET 11's API is used freely. **The `alpha` branch** (D73–D76, 2026-10-31) builds a released beta's source to try a newer .NET; it rests since 0.1.0-alpha.4 (D98) and then differs from main in `RELEASE_LINE` alone — `docs/12_BUILD.md` §The alpha line |
| Native | **C++20, MSVC v143+, `/MT` static CRT, `/GS`, `/guard:cf`, no RTTI, no C++ exceptions in hook paths** |
| Hooking | **MinHook** (BSD-2-Clause) for inline hooks; direct vtable-entry swap for COM interfaces |
| Vulkan | **Implicit Vulkan layer** (`VK_LAYER_FRAMELEDGER_overlay`), not hooking — `17_HOOK_ENGINE` §Vulkan |
| UI | WPF + WPF UI (`WPF-UI` pinned = 4.3.0) + CommunityToolkit.Mvvm — `docs/16_WPFUI_SYNTAX.md` |
| App composition | .NET Generic Host + DI in `FrameLedger.App` |
| Charts | ScottPlot 5 |
| Markdown in the App | **Markdig** (BSD-2-Clause, pinned ~~1.4.0~~ 1.5.0 since beta.31 — 1.4.2 in beta.30) → WPF `FlowDocument`, HTML off, no image fetched — the Legal Gate, the user guide, Help ▸ Limitations, the update notes, About (beta.12; row added 2026-10-04) |
| Markdown in the App | **Markdig** (BSD-2-Clause, pinned ~~1.4.0~~ 1.5.0 since beta.31 — 1.4.2 in beta.30) → WPF `FlowDocument`, HTML off, no image fetched — the Legal Gate, the user guide, Help ▸ Limitations, Help ▸ Accessibility (beta.35, D88), the update notes, About (beta.12; row added 2026-10-04) |
| GPU telemetry | **The game process's own memory since beta.12 (D43):** `FrameLedger.ProcessStats.dll` (C++, loaded by the Agent, never into a game) reads its dedicated/shared GPU memory from the `GPU Process Memory` counters and its working sets from `GetProcessMemoryInfo` — `18_GPU_VENDOR_APIS` §The game process's memory. Layered: DXGI + PDH counters (all vendors) → LibreHardwareMonitorLib (MPL-2.0, all vendors) → **NVAPI (MIT, headers + `amd64/nvapi64.lib` vendored 2026-08-05** at `src/native/third_party/nvapi/`, consumed via the `fl_nvapi` target; NVIDIA extras + Reflex). **No AMD/Intel vendor SDK** — `docs/18_GPU_VENDOR_APIS.md`. ~~**No telemetry source exists in code yet**~~ **`LhmTelemetrySource` (L2) exists since 2026-09-03 (`20_OPEN_QUESTIONS` §M5 row R1, eight GPU fields unelevated); L1 (`BaselineTelemetrySource`: DXGI identity + the PDH adapter-memory counter, the first CsWin32 consumer), `CompositeTelemetrySource` and the 1 Hz `TelemetryPoller` since 2026-09-09; L3 (`NvapiTelemetrySource` over the native `FrameLedger.NvapiBridge.dll` — the only shipped binary linking `nvapi64.lib`, loaded by the Agent into its own process and never into a game) since 2026-09-10 (PR-E2)** — `ctest fl_nvapi_probe` was the only NVAPI consumer until then, and it exists so the vendoring is verified rather than asserted. This row said no source existed for six days after one did, recorded 2026-09-09 rather than repaired quietly. This row previously claimed the vendoring in the present tense while `third_party/` held only `vulkan-headers`; `legal/` had already caught the identical claim about itself and gained a bidirectional gate, which is what made the real vendoring a red-then-green |
| CPU/board sensors | LibreHardwareMonitorLib ≥ 0.9.6 (PawnIO) — **optional**, elevated only |
| Tier-2 fallback | **There is none, and Tier 2 is not a measurement.** Owner decision 2026-08-28: the ladder is two rungs — hooked, or duration + sensors + the reason. ~~ETW-based, no injection~~; ~~Intel PresentMon console binary~~ dropped 2026-08-27 after §S31 retired it. Whether a shipped build ever regains a no-injection measurement is `docs/20_OPEN_QUESTIONS.md` §G |
@@ -67,11 +69,18 @@ A C++ DLL (`FrameLedger.Overlay.dll`) is injected into games the user explicitly
FrameLedger.slnx # XML solution format (SDK default since .NET 10)
NOTICE # copyright + the GPLv3 section 7 additional terms (beta.12, D45)
CONTRIBUTING.md FORKING.md CODE_OF_CONDUCT.md # for people who build, contribute or redistribute
LIMITATIONS.md ACCESSIBILITY.md # what the software cannot do; the accessibility statement (D36, D88)
global.json # pins the SDK band — see 12_BUILD
RELEASE_LINE # `main`, or `alpha` on the alpha branch (D73, 2026-10-31) — 12_BUILD §The alpha line
build.ps1 # the quality gate; CI runs this identical script
SonarQube.Analysis.xml # SonarCloud's settings for ci.yml's sonarcloud job (D68, 2026-10-25; it replaced
.config/dotnet-tools.json # .sonarcloud.properties) and the scanner it runs — 13_CI_CD §SonarCloud
qodana-native.yaml # the same for the native job: the C++ of src/native, through the Visual Studio solution
# CMake writes for it (preset x64-vs-analysis), a report of its own (D99)
qodana.yaml # Qodana's settings for the managed job of .github/workflows/qodana.yml (D93; Qodana for
# .NET on the Ultimate licence since 2026-11-11, D99): the linter, the solution, a gate
# of zero problems, and what is switched off that .editorconfig has no key for —
# ReSharper's inspections are .editorconfig's resharper_* keys (D100) — 13_CI_CD §Qodana
src/
native/
FrameLedger.Overlay/ # C++20 DLL injected into the game (hooks + ring writer + optional overlay)
@@ -96,7 +105,70 @@ src/
# composes it from, SessionEndReason, the census/NGX/marker facts;
# and Recording/ since 2026-09-10 (PR-D): SessionRecorder,
# SessionAggregator + FgLadder, SessionFinalizer, the .partial
# writer and PartialRecovery, ExitStatusMapper, the crash policy
# writer and PartialRecovery, ExitStatusMapper, the crash policy;
# and Advanced/ since 2026-11-11 (D101): AdvancedMetricCatalog, the one
# list "Advanced metrics" reads - what is read, and what is not with the
# reason (18_GPU_VENDOR_APIS §Advanced metrics) - and the value model
# (AdvancedValue / AdvancedFact / AdvancedReading / AdvancedSnapshot) and
# the port IAdvancedHardware. Infrastructure implements it: AdvancedHardwareHub
# (one reader for the Agent, on its own thread, only while held - D107) over
# NvapiAdvancedReader (the bridge's ABI 3) and LhmAdvancedReader (LhmGroups:
# the card; as administrator with PawnIO the processor, memory and board -
# never storage, D103). The Agent composes it since the switch, `telemetry.advanced`
# (PR-5, 2026-11-11: off by default; on, a session holds the hub), and
# Ipc/AdvancedMachineResponder answers GetAdvancedMachine from the hub's last
# snapshot - each request is the page's hold on it (PR-6, D107). Recording/
# AdvancedSessionBuilder stores what a session's samples carried (schema 0023:
# machine_profiles + session_advanced, adv.* series with NaN for a gap, D102).
# Train B (0.1.0-beta.40, D111): Capture/InGameAdvancedSampler turns the shared
# memory's region 5 - what a hooked game told DLSS and FSR, read by the Overlay
# only while the session asks (ICaptureObserver.WantsInGameAdvanced) - into session
# facts and two adv.* series; reach `Game` in the catalog; no schema change.
# Since 0.1.0-beta.43 also what the game tells Streamline each frame and how it
# configures FSR frame generation, from two calls hooked only once Advanced has been
# asked for; AdvancedInGameCall says of each: read, never called, or not hooked.
# Since 0.1.0-beta.44 also what Streamline answers when the game asks about a
# feature (its version, what it requires), from two more calls hooked the same
# way; ~~a game asks while it starts, so "never called" is the usual answer there~~
# (the games run since ask again later, and are read).
# Since 0.1.0-beta.45 also whether Streamline says a feature is supported and
# loaded, the game switching it, its resources allocated and freed and Streamline
# shut down, from six more calls hooked the same way (D130); slInit is not hooked
# and its entry says why (AdvancedReach.BeforeAttach, D129).
# Since 0.1.0-beta.47 (train C, D136-D141) also what NVIDIA's driver reports about
# a frame's stages - NOT an argument of a hooked call: the Overlay calls two
# read-only getters of the driver itself (reach GameDriver); seven adv.* series
# and one table, and AdvancedDriverRead says what became of the read.
# Since 0.1.0-beta.48 (D142-D144) also two figures the driver works out itself, of
# the same report - no further call: an entry of their own, nvapi.reflex.gpu_times,
# two adv.* series; and the table lists every frame-generation multiplier seen.
# Since 0.1.0-beta.49 (D148-D151) the Overlay also counts each frame of that report
# by its GPU span and the driver's active time - for its native log ONLY
# (LATENCY_FRAMES, a batch per state of the sleep status; nothing published or
# stored); its watchdog's periodic tick is on a grid (fl_watchdog_tick.h), and
# Recording/PendingInGameReadings keeps the reading that waits for a sample.
# Controllers and input (0.1.0-beta.50, owner decisions D155-D163, docs/21_INPUT_DEVICES):
# Input/ - InputReader (a slice ten times a second over the port IInputBackend),
# InputTickAccumulator (one session's slices into a reading a tick), InputTerms (the
# reading waits for Privacy Policy 2.22, D156), the ports IInputDevices /
# IInputActivitySource; Recording/InputSessionBuilder (totals, the controllers as JSON,
# the input.* series; schema 0024). Domain/Input holds the arithmetic and the two
# parsers (an XInput state, a DualSense report). Infrastructure/Input is the Windows
# side: Win32InputBackend (raw input for joysticks and gamepads, three XInput getters,
# GetLastInputInfo - it opens no device and sends nothing to one) and InputDeviceHub
# (one thread for the Agent, only while a session has a view). Read from OUTSIDE the
# game, both tiers; `capture.input_devices`, on by default.
# Controllers and input (0.1.0-beta.50, owner decisions D155-D163, docs/21_INPUT_DEVICES):
# Input/ - InputReader (a slice ten times a second over the port IInputBackend),
# InputTickAccumulator (one session's slices into a reading a tick), InputTerms (the
# reading waits for Privacy Policy 2.22, D156), the ports IInputDevices /
# IInputActivitySource; Recording/InputSessionBuilder (totals, the controllers as JSON,
# the input.* series; schema 0024). Domain/Input holds the arithmetic and the two
# parsers (an XInput state, a DualSense report). Infrastructure/Input is the Windows
# side: Win32InputBackend (raw input for joysticks and gamepads, three XInput getters,
# GetLastInputInfo - it opens no device and sends nothing to one) and InputDeviceHub
# (one thread for the Agent, only while a session has a view). Read from OUTSIDE the
# game, both tiers; `capture.input_devices`, on by default.
FrameLedger.Infrastructure/ # SQLite, shm reader, vendor APIs, injector interop, parsers,
# and Capture/ (PR-C): ShmRingAttacher/ShmCaptureSink over the
# reader, HeldProcessLivenessSource, TargetResolver (path only,
@@ -119,7 +191,9 @@ src/
# is built as of 2026-09-14 (PR-5 Games/Dashboard, PR-6 summary +
# charts, PR-7 trend/compare, PR-8a Settings/safety notices/Logs/
# --diag, PR-8b tray + the Agent's maintenance flags, PR-9 first run
# + Legal Gate FR-11); Strings.resx en/vi/ja. `--data-dir <copy>`
# + Legal Gate FR-11); Strings.resx en/vi/ja. The Advanced page
# (2026-11-11, D101/D105: Pages/AdvancedPage, there and on Ctrl+6 only
# while `telemetry.advanced` is on - 08_UI §Advanced). `--data-dir <copy>`
# (beta.15, D52) = a VIEWER: no Agent started or contacted,
# nothing on the PC changed, the profile's own folder refused
FrameLedger.CaptureHost/ # UNSHIPPED, and since P2 PR-C (2026-09-09) a THIN SHELL: verbs,
@@ -147,7 +221,20 @@ tools/ # accuracy-check, changelog-check, chokepoint-che
# versioninfo-check, vklayer-blastradius, fps-impact-runbook,
# wiki-build (beta.17: builds and checks the wiki; the wiki-check gate),
# buildid-check (beta.18: the release's build id in the shipped natives),
# docs-split-check (beta.24, D66: CHANGELOG/HANDOFF and their parts agree)
# docs-split-check (beta.24, D66: CHANGELOG/HANDOFF and their parts agree),
# agent-cost (D110, 2026-11-11: a developer tool, not a gate - what the shipped
# Agent costs per second of session, Advanced metrics off and on),
# nvapi-readonly-check (D108, 2026-11-11: every NVAPI call of the bridge is on
# nvapi-allowed-calls.txt, and none there writes or identifies; since beta.47 also
# the Overlay - no NvAPI call by name, and the only interface ids it may name are
# the two on overlay-nvapi-allowed.txt),
# overlay-fields-check (D116, 2026-11-12: every member of a vendor's struct the
# Overlay reads for Advanced metrics is in fl_adv_fields.h and on
# overlay-allowed-fields.txt; a camera's position, axes and matrices nowhere),
# input-readonly-check (D155-D163, 2026-11-14: every Windows function the input reader
# calls is declared in Infrastructure/Input/Win32Input.cs and on input-allowed-calls.txt;
# none sends to a device, injects input or reads a key; raw input is registered for
# joysticks and gamepads only; no native source names an input function)
# (six names added 2026-10-04)
# (PowerShell). This line used to name three, one of which
# — resx-audit — ~~does not exist~~ exists since 2026-09-13 with
@@ -158,6 +245,9 @@ tools/ # accuracy-check, changelog-check, chokepoint-che
# (--vulkan since #140, --opengl since #141)
# fl-probe-* -> guard, hook profile, interposer, NVAPI, process
# stats, signer, Vulkan layer; fl-baseline-probe;
# and fl-probe-reflex (2026-11-13, D132: the spike of Reflex route D,
# spike-notes §31 - unshipped, with a stand-in nvapi64.dll among the stubs;
# built from the Overlay's own fl_reflex.h since beta.47);
# vendor-stubs/ -> fixture DLLs with the vendors' export names
rules/detection-rules.json # engine/platform/capability + anticheat blocklist
guide/ # the user guide (beta.12, D44): ~~eight~~ short English pages for players, embedded in the
@@ -165,7 +255,7 @@ guide/ # the user guide (beta.12, D44): ~~eight~~ short
# (twelve since beta.17, D54: the screens, every setting, troubleshooting, a glossary;
# and the guide is the wiki's End-user part)
wiki/ # the wiki's own pages (beta.17, D53): Home, _Sidebar, the developer pages. wiki.yml
# publishes them with guide/ and LIMITATIONS.md to the Wiki tab at a release;
# publishes them with guide/, LIMITATIONS.md and ACCESSIBILITY.md to the Wiki tab at a release;
# tools/wiki-build.ps1 assembles and checks them (build.ps1 wiki-check)
docs/ legal/ legal/licenses/
```
@@ -174,13 +264,13 @@ Dependency direction: `App/Agent → Application → Domain`; **`Application →
## Coding conventions
**C# —** `Nullable enable`, `TreatWarningsAsErrors`, `AnalysisLevel latest-all`, file-scoped namespaces, Roslynator + Meziantou + VS Threading analyzers, `dotnet format` clean. Async suffixed `Async`, `ConfigureAwait(false)` off the UI. All user-visible strings from `.resx`. Timestamps UTC (unix-ms in SQLite); QPC ticks only inside the capture pipeline.
**C# —** `Nullable enable`, `TreatWarningsAsErrors`, `AnalysisLevel latest-all`, file-scoped namespaces, Roslynator + Meziantou + VS Threading analyzers, `dotnet format` clean. Qodana reads the C# and the XAML in an advisory workflow of its own, with a gate of zero problems *(since 2026-11-10, D93; Qodana for .NET on the Ultimate licence since 2026-11-11, D99 — `13_CI_CD` §Qodana)*: fix what it reports, or switch the inspection off with the reason on the line above, **in the place the engine that raised it reads** *(D100)* — one of ReSharper's (C#, XAML, `.resx`, the C++) in `.editorconfig`; what has no key there (the hints as a severity, the Markdown and XML inspections, the files JetBrains' security rules were read in) in `qodana.yaml` — never anywhere else, never without a reason. A new file that starts a process fails `ProcessStartSitesTests`, and Qodana's security rule reports it: read what it starts and where the path and the arguments come from, then list it in the test, in `qodana.yaml` and in `13_CI_CD` §Qodana. Async suffixed `Async`, `ConfigureAwait(false)` off the UI. All user-visible strings from `.resx`. Timestamps UTC (unix-ms in SQLite); QPC ticks only inside the capture pipeline.
**C++ —** `clang-format` (LLVM base, 4-space, 120 col) enforced in CI. No STL containers that allocate in hook paths. No `std::mutex` in hook paths. `-D_HAS_EXCEPTIONS=0` in the Overlay target. Every hook entry point wrapped per the `FL_HOOK_GUARD` macro (`17_HOOK_ENGINE`). Static analysis: ~~`/analyze` + clang-tidy (`bugprone-*`, `cert-*`, `concurrency-*`)~~ CodeQL (`codeql.yml`, a required check); `/analyze` and clang-tidy are not configured *(corrected 2026-10-04)*. SonarCloud analyses the C++ too, from CMake's compilation database, in `ci.yml`'s advisory `sonarcloud` job *(since 2026-10-25, D68 — `13_CI_CD` §SonarCloud)*.
**C++ —** `clang-format` (LLVM base, 4-space, 120 col) enforced in CI. No STL containers that allocate in hook paths. No `std::mutex` in hook paths. `-D_HAS_EXCEPTIONS=0` in the Overlay target. Every hook entry point wrapped per the `FL_HOOK_GUARD` macro (`17_HOOK_ENGINE`). Static analysis: ~~`/analyze` + clang-tidy (`bugprone-*`, `cert-*`, `concurrency-*`)~~ CodeQL (`codeql.yml`, a required check); `/analyze` and clang-tidy are not configured *(corrected 2026-10-04)*. SonarCloud analyses the C++ too, from CMake's compilation database, in `ci.yml`'s advisory `sonarcloud` job *(since 2026-10-25, D68 — `13_CI_CD` §SonarCloud)*. Qodana reads it as well, in a report of its own *(since 2026-11-11, D99: ReSharper's C++ inspections and the clang-tidy checks they run, advisory, a gate of zero — `13_CI_CD` §Qodana)*: fix what it reports, or switch the inspection off in `.editorconfig` with the reason above it; **in code loaded into a game only a defect is fixed**, in a pull request of its own with the overhead measured *(D100)*.
**Licence header —** every first-party source file (`.cs .cpp .h .inl .ps1 .xaml`) opens with two comment lines, `SPDX-License-Identifier: GPL-3.0-only` and `Copyright (C) 2026 poli0981 - additional terms under GPLv3 section 7: see NOTICE` (beta.12, owner decision D45). GPLv3 §7 requires additional terms to be named in the source files, and `NOTICE` holds them — (b) preserve the notices, (c) mark modified versions, (e) no trademark rights in the name. `tools/notice-check.ps1` is a `build.ps1` gate; `-Fix` writes a missing header. `src/native/third_party/` is excluded.
**Struct mirroring —** ~~`FlFrameRecord`, `FlShmHandshake`, `FlWriterState` and `FlControlBlock`~~ `FlShmHandshake`, `FlWriterState`, `FlControlBlock`, `FlDisplayState` (layout 4, beta.10), `FlFrameRecord` and `FlTolerance` (D33, beta.9) each exist twice *(corrected 2026-10-04)*: `src/native/FrameLedger.Shm/include/fl_shm.h` / `fl_tolerance.h` (**normative**) and `src/FrameLedger.Shared/ShmLayout.cs` as `[StructLayout(LayoutKind.Sequential)]`. `ShmLayoutMirrorTests` asserts size and every field offset on both sides **against JSON emitted by ~~`tools/fl-layout-dump`~~ `src/native/tools/fl-layout-dump`**, never a transcribed table, and walks the field list in both directions so a field added on either side alone fails.
**Struct mirroring —** ~~`FlFrameRecord`, `FlShmHandshake`, `FlWriterState` and `FlControlBlock`~~ `FlShmHandshake`, `FlWriterState`, `FlControlBlock`, `FlDisplayState` (layout 4, beta.10), `FlAdvancedState` (layout 5, 2026-11-11, Advanced metrics train B), `FlFrameRecord` and `FlTolerance` (D33, beta.9) each exist twice *(corrected 2026-10-04)*: `src/native/FrameLedger.Shm/include/fl_shm.h` / `fl_tolerance.h` (**normative**) and `src/FrameLedger.Shared/ShmLayout.cs` as `[StructLayout(LayoutKind.Sequential)]`. `ShmLayoutMirrorTests` asserts size and every field offset on both sides **against JSON emitted by ~~`tools/fl-layout-dump`~~ `src/native/tools/fl-layout-dump`**, never a transcribed table, and walks the field list in both directions so a field added on either side alone fails.
It also asserts **blittability** (`RuntimeHelpers.IsReferenceOrContainsReferences`), because offsets cannot see it: measured 2026-08-05, swapping the `fixed byte BuildId[32]` for this repo's own `[MarshalAs(ByValTStr)] string` idiom keeps `Marshal.SizeOf` at 64 and every offset correct while `Unsafe.SizeOf` collapses to **40** — a mirror that passes the obvious checks and cannot be read out of a memory-mapped view.
@@ -196,6 +286,8 @@ The version handshake is wired end to end as of 2026-08-05: `FlGuardBuildId` giv
- Hook-path changes state measured overhead in the PR body (`14_TESTING` §Hook overhead).
- Any new hook is listed in `17_HOOK_ENGINE` §Hook inventory and justified against rule 4.
- New user-visible strings exist in `en`, `vi`, `ja`.
- A change to what the App shows keeps `ACCESSIBILITY.md` §For contributors (names, theme colours, nothing by colour alone,
keyboard reach) and updates that page when it moves what works or what does not (D88).
- Any deviation from a doc updates that doc in the same PR.
## Reading order
@@ -219,7 +311,10 @@ The version handshake is wired end to end as of 2026-08-05: `FlGuardBuildId` giv
10. Remaining (`09`–`15`).
11. **`LIMITATIONS.md`** (repository root; shown in the App under Help ▸ Limitations) — what the WHOLE software cannot
do, in the user's words, each item linked to its details. When a change moves a limit, the page changes in the same
PR (D36, 2026-09-27). **`docs/LIST_GAME_TESTED.md`** — the real games FrameLedger has been run with and what each
PR (D36, 2026-09-27). **`ACCESSIBILITY.md`** (repository root; Help ▸ Accessibility) — the accessibility statement:
WCAG 2.2 AA through WCAG2ICT as a goal, never a claim; what works with the keyboard, a screen reader and contrast
themes, what does not, how a barrier is reported (D88, 2026-11-09); `docs/08_UI.md` §Accessibility holds its checks.
**`docs/LIST_GAME_TESTED.md`** — the real games FrameLedger has been run with and what each
session measured, from a copy of the owner's ledger and the repository's own evidence (D35).
**Before writing native or capture code, read `docs/20_OPEN_QUESTIONS.md`** (its open entries; since beta.24 the
+10 -3
View File
@@ -11,12 +11,15 @@ merged, and under what licence your contribution is accepted.
- **Open an issue before anything larger than a fix**, so the approach can be agreed before the work is done.
- **A game with anti-cheat that FrameLedger fails to detect** is a safety bug: use the *Safety gap* issue form.
- **A vulnerability in FrameLedger itself** goes the private route in `SECURITY.md`, never a public issue.
- **Something that cannot be reached, read or seen** — with the keyboard, a screen reader, a contrast theme — uses the
*Accessibility problem* form; `ACCESSIBILITY.md` says what the App aims for.
- Everyone who takes part follows the `CODE_OF_CONDUCT.md`.
## Building
You need Windows 10 22H2 or later (x64), Visual Studio 2022 with the C++ workload (MSVC v143), CMake 3.28 or later,
PowerShell 7, and the .NET SDK that `global.json` pins. Then:
You need Windows 11 (x64), Visual Studio 2022 or later with the C++ workload (MSVC v143 or later), CMake 3.28 or later,
PowerShell 7, and the .NET SDK that `global.json` pins — .NET 11 since 0.1.0-beta.37, which Microsoft does not support
on Windows 10 Home or Pro. Then:
```powershell
./build.ps1 check
@@ -45,13 +48,17 @@ The definition of done is the one in `CLAUDE.md`:
- a change to the hook path states its measured overhead in the PR (`docs/14_TESTING.md` §Hook overhead);
- a new hook is listed in `docs/17_HOOK_ENGINE.md` §Hook inventory and justified against rule 4;
- every new user-visible string exists in `en`, `vi` and `ja` (`docs/09_I18N.md`);
- a change to what the App shows keeps `ACCESSIBILITY.md` §For contributors — a name for every control that shows no
text, theme colours only, nothing shown by colour alone, everything reachable by keyboard — and says in the PR what was
checked with the keyboard and Narrator; if it moves what works or what does not, `ACCESSIBILITY.md` changes with it;
- a change that makes a document wrong updates that document in the same PR;
- a change under `src/` adds an entry under `## [Unreleased]` in `CHANGELOG.md`;
- every new source file carries the licence header — `./tools/notice-check.ps1 -Fix` writes it. A file you wrote
yourself may name you in its second line instead, in the same form:
`Copyright (C) <year> <your name> - additional terms under GPLv3 section 7: see NOTICE`.
The pull-request template lists the safety checks for changes to the capture layer.
The pull-request template lists the safety checks for changes to the capture layer, and the accessibility checks for
changes to the user interface.
## What will not be merged
+23 -9
View File
@@ -8,12 +8,14 @@
- TargetFramework's platform version selects which Windows API
projections are available to compile against.
- SupportedOSPlatformVersion is the floor CA1416 enforces.
NFR-8 requires Windows 10 22H2 = build 19045. Letting the supported version
default to the TFM's 19041 would silently accept installs three releases
below the stated floor.
NFR-8 names Windows 10 22H2 = build 19045 as the oldest Windows the code is
held to (since 2026-11-11, D95: not blocked there, and neither tested nor
supported - Microsoft supports .NET 11 on Windows 11). Letting the supported
version default to the TFM's 19041 would silently accept installs three
releases below that floor.
A bare `net10.0-windows` is wrong for a different reason: it resolves to
net10.0-windows7.0, and CA1416 would then error on Mica, MiniDumpWriteDump
A bare `net11.0-windows` is wrong for a different reason: it resolves to
net11.0-windows7.0, and CA1416 would then error on Mica, MiniDumpWriteDump
and IDXGIAdapter3::QueryVideoMemoryInfo — every Windows-10-era API this
application is built on.
@@ -25,7 +27,9 @@
nothing and CA1416 still enforces the 19045 floor at the API level.
-->
<PropertyGroup>
<TargetFramework>net10.0-windows10.0.22621.0</TargetFramework>
<!-- .NET 11 since 0.1.0-beta.37 (owner decision D94, 2026-11-11; docs/12_BUILD.md). net10.0 until then, and
net11.0 on the alpha branch from 2026-10-31. -->
<TargetFramework>net11.0-windows10.0.22621.0</TargetFramework>
<SupportedOSPlatformVersion>10.0.19045.0</SupportedOSPlatformVersion>
<!-- x64 only (CLAUDE.md pinned stack, NFR-8). Previously asserted in prose
@@ -69,17 +73,27 @@
<IncludeSourceRevisionInInformationalVersion>false</IncludeSourceRevisionInInformationalVersion>
<!-- D73 (2026-10-31): the RELEASE LINE, from the RELEASE_LINE file beside VERSION - `main`, the betas and releases on
the .NET that global.json pins, or `alpha`, a beta's source built on a prerelease .NET from the long-lived `alpha`
branch, where this file and global.json are two of the few lines that differ (docs/12_BUILD.md §The alpha line).
the .NET that global.json pins, or `alpha`, a beta's source built to try a newer .NET from the long-lived `alpha`
branch, where this file and global.json are two of the few lines that differ while it does (docs/12_BUILD.md
§The alpha line).
Not a property a command line sets: the file is the one source, release.yml refuses a tag that disagrees with it,
and the target below refuses a value that is neither line. Every assembly carries it (the ItemGroup below), and
Shared.BuildLine reads it at run time. -->
<FrameLedgerReleaseLineFile>$(MSBuildThisFileDirectory)RELEASE_LINE</FrameLedgerReleaseLineFile>
<FlReleaseLine Condition="Exists('$(FrameLedgerReleaseLineFile)')">$([System.IO.File]::ReadAllText('$(FrameLedgerReleaseLineFile)').Trim())</FlReleaseLine>
<!-- D98 (2026-11-11): THE ALPHA LINE RESTS - the betas build on the same .NET as the newest alpha (0.1.0-alpha.4,
.NET 11.0.0), so no alpha is cut until a newer .NET is tried. An alpha built while this says true tells its user
that the line is paused and to install the beta: a resting alpha gets no update, its anti-cheat list included
(Shared.AlphaLine, App/Services/AlphaTexts). Stamped on both lines, read by an alpha alone. When the `alpha`
branch starts trying a newer .NET it sets this to false there - one more line of its switch (docs/12_BUILD.md
§The alpha line). -->
<FlAlphaLineRests>true</FlAlphaLineRests>
</PropertyGroup>
<ItemGroup>
<AssemblyMetadata Include="FrameLedger.ReleaseLine" Value="$(FlReleaseLine)" />
<AssemblyMetadata Include="FrameLedger.AlphaLineRests" Value="$(FlAlphaLineRests)" />
</ItemGroup>
<Target Name="FrameLedgerReleaseLineCheck" BeforeTargets="BeforeBuild" Condition="'$(FlReleaseLine)' != 'main' And '$(FlReleaseLine)' != 'alpha'">
@@ -87,7 +101,7 @@
</Target>
<!-- The NuGet audit (2026-10-25, docs/13_CI_CD.md §Dependabot and the NuGet audit): every restore checks the WHOLE
package graph, transitive packages included, against the advisory database, at every severity. These are the
package graph, transitive packages included, against the advisory database, at every severity. These were the
SDK's defaults for net10.0, written down so that a later SDK's defaults cannot move them. With
TreatWarningsAsErrors, a vulnerable package is NU1901-NU1904 and the restore - and so the gate - is red, which
is what happened with SQLitePCLRaw 2.1.11 (Directory.Packages.props). -->
+11 -5
View File
@@ -6,7 +6,8 @@
All versions verified against nuget.org on 2026-08-02; re-verified and bumped 2026-09-21; bumped again
2026-10-25 (beta.29 PR-2) with the updates Dependabot had never proposed (docs/13_CI_CD.md §Dependabot and
the NuGet audit), and 2026-10-26 (beta.30 PR-3): Velopack and Markdig, the SQLitePCLRaw pin dropped; and again
2026-10-26 (beta.31, the owner's request): the .NET 10.0.14 servicing packages and the updates beta.30 had left.
2026-10-26 (beta.31, the owner's request): the .NET 10.0.14 servicing packages and the updates beta.30 had left;
and 2026-11-11 (beta.37, D94): the three packages serviced with the runtime follow it to .NET 11.0.0.
-->
<PropertyGroup>
<ManagePackageVersionsCentrally>true</ManagePackageVersionsCentrally>
@@ -38,7 +39,9 @@
</ItemGroup>
<ItemGroup Label="Hosting and logging">
<PackageVersion Include="Microsoft.Extensions.Hosting" Version="10.0.14" />
<!-- The three packages serviced with the runtime follow it: this one, Microsoft.Data.Sqlite and
System.Diagnostics.EventLog (docs/12_BUILD.md). -->
<PackageVersion Include="Microsoft.Extensions.Hosting" Version="11.0.0" />
<PackageVersion Include="Serilog" Version="4.4.0" />
<PackageVersion Include="Serilog.Extensions.Hosting" Version="10.0.0" />
<PackageVersion Include="Serilog.Sinks.Console" Version="6.1.1" />
@@ -47,7 +50,7 @@
<ItemGroup Label="Storage">
<!-- No EF: Dapper over Microsoft.Data.Sqlite (CLAUDE.md pinned stack). -->
<PackageVersion Include="Microsoft.Data.Sqlite" Version="10.0.14" />
<PackageVersion Include="Microsoft.Data.Sqlite" Version="11.0.0" />
<PackageVersion Include="Dapper" Version="2.1.89" />
<!--
No transitive pin on SQLitePCLRaw since beta.30 (2026-10-26). From the scaffold (2026-08-02) to beta.29 this held
@@ -57,6 +60,9 @@
its own; 10.0.13 declares SQLitePCLRaw 2.1.13, and the pin had become a conflict NuGet does not report - it put a
package built against the 2.x line on SQLitePCLRaw's 3.x line. The NuGet audit (Directory.Build.props) still
turns a vulnerable SQLite red (docs/13_CI_CD.md §Dependabot and the NuGet audit).
Since beta.37 (2026-11-11): Microsoft.Data.Sqlite 11 declares SQLitePCLRaw 3.0.5 itself - the 3.x line, built for
it this time, whose engine is SourceGear's SQLite package; still no pin here. The alphas shipped that graph from
2026-10-31.
-->
</ItemGroup>
@@ -69,12 +75,12 @@
<PackageVersion Include="LibreHardwareMonitorLib" Version="0.9.6" />
<!-- The crash witness (04_CAPTURE §Crash & exit classification): Application log events 1000/1001. In-box
for WPF apps, a package for a class library on the Windows TFM. Microsoft, MIT. -->
<PackageVersion Include="System.Diagnostics.EventLog" Version="10.0.14" />
<PackageVersion Include="System.Diagnostics.EventLog" Version="11.0.0" />
</ItemGroup>
<ItemGroup Label="Analyzers (build-time only)">
<PackageVersion Include="Roslynator.Analyzers" Version="5.0.1" />
<PackageVersion Include="Meziantou.Analyzer" Version="3.0.317" />
<PackageVersion Include="Meziantou.Analyzer" Version="3.0.319" />
<PackageVersion Include="Microsoft.VisualStudio.Threading.Analyzers" Version="18.12.1" />
</ItemGroup>
+163 -16
View File
@@ -5,7 +5,7 @@ hooking on for — and reading the graphics calls the game makes. That is why it
resolution, the upscaler, frame generation and ray tracing apart. It is also where its limits come from. This
page lists them in plain words; each item links to the document that has the details.
*Beta software: this page describes version 0.1.0-beta.33.*
*Beta software: this page describes version 0.1.0-beta.49.*
## Games it will not measure
@@ -36,7 +36,8 @@ page lists them in plain words; each item links to the document that has the det
When a game is not measured — any of the reasons above — its session still records **how long it ran, the
machine's sensors (GPU temperature, load, power; CPU load; memory), how much memory the game itself used (read from
Windows, outside the game) and why nothing was measured.** Every
Windows, outside the game), what it was played with ([Controllers and input](#controllers-and-input)) and why nothing
was measured.** Every
frame-rate, upscaler, frame-generation and ray-tracing value reads **N/A**, never 0. How the game's window was
shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
@@ -51,6 +52,11 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
game). Unity's crash handler and Chromium's `crashpad_handler` start with every game that has them, so their presence
says nothing. Sessions recorded before 0.1.0-beta.14 keep how they were stored — a few of them may read *crashed*
because Windows logged a memory report about the game. ([details](docs/04_CAPTURE.md))
- **A game that faults while it is closing can read *ended normally* in a session measured from inside the game.**
FrameLedger looks for Windows' crash report when it saves the session, and for such a session that is before
Windows has written it; the same fault in a session that was not measured from inside reads *crashed*. Seen on one
game whose graphics-driver library faults at every exit, measured or not (2026-11-12), and on a second whose
Steam crash handler faults at every measured exit. Not changed yet.
## Frame rate and frame generation
@@ -90,7 +96,9 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
same games as *Render → output resolution* in the accuracy statement; elsewhere it says it was not measured.
- The upscaler is read from the game's calls: DLSS and NIS through NVIDIA Streamline, and FSR through AMD's
FidelityFX libraries. **Intel XeSS is not read** (its licence forbids it), FSR 2 and upscalers built into the
game are not identified, and Vulkan and OpenGL games read N/A. Where no hook saw DLSS but the NVIDIA driver
game are not identified — Cyberpunk 2077's FSR 2.1 is one: its sessions say "unknown" beside a measured render
size — and Vulkan and OpenGL games read N/A. A game that sends FSR upscaling through two unrelated libraries in
one run is counted from the first alone (none is known). Where no hook saw DLSS but the NVIDIA driver
reports it, the page says it is the driver's report.
- **The quality mode** (Quality, Balanced, DLAA, …) is shown only when it is known: the NVIDIA App's override
where the NVIDIA driver reports it applied (marked *· NVIDIA App*, since 0.1.0-beta.24), or the mode the game
@@ -106,14 +114,28 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
show neither the driver's report nor the NVIDIA App's mode. ([Metrics](docs/03_METRICS.md))
- **Which DLSS ran** (since 0.1.0-beta.24) is named only from what the session showed: the model preset the NVIDIA
driver reports (L or M: DLSS 4.5; J or K: DLSS 4), a measured multi-frame factor (×5 or ×6: DLSS 4.5; ×3 or ×4:
DLSS 4), or a library whose major version names one (2.x, 3.x). A 310 library with no such evidence shows its
version only. DLSS 5 is shown only as *runtime loaded*, by a library name not yet seen on a real game. Where each
DLSS 4), the NVIDIA App's *Dynamic* frame generation that the driver reports applied (DLSS 4.5, whatever multiplier
it chose — since 0.1.0-beta.34), or a library whose major version names one (2.x, 3.x). A *Fixed* ×5 or ×6 set in
the App names DLSS 4.5 only when the frames show it. A 310 library with no such evidence shows its
version only. **DLSS 5**
(since 0.1.0-beta.34) is named where the game was seen running its neural rendering through Streamline, with the share
of rendered frames; its library loaded and nothing seen reads *runtime loaded* — *not seen running* only where the
Overlay looked. A game that runs neural rendering outside Streamline is not seen. Where each
library was loaded from is read from its folder, for sessions recorded since 0.1.0-beta.24.
- **Ray tracing** is *Yes*, *No* or *N/A* — measured from the game's ray-tracing calls on Direct3D 12, and
N/A where FrameLedger could not look. **Path tracing is never stated as a fact**: there is no API signal
for it. You can correct any of these by hand, and your correction is labelled as yours.
- HDR output, NVIDIA Reflex latency, the video-memory budget Windows gives a game, and shader-compilation
stutter are **not measured yet** — they read N/A. ([Accuracy](legal/ACCURACY.md))
- HDR output, end-to-end latency (click to screen), the video-memory budget Windows gives a game, and
shader-compilation stutter are **not measured yet** — they read N/A. ([Accuracy](legal/ACCURACY.md))
- **NVIDIA's own latency report** is read only with *Advanced metrics* on, only on an NVIDIA card, only in a
Direct3D 12 game that uses NVIDIA Reflex: how long the driver says each stage of a frame took, about once a
second. They are the driver's numbers, not FrameLedger's measurement. **The stages overlap and do not add up to
the total**, and the graphics card's stage includes the time the card waited — in one game it is longer than the
whole frame; the driver's own figure for the card's active time is shown beside it where the driver gives one.
One point is one reading, drawn at the sample that carried it — up to a second after it was taken; with the
sensor interval set above one second, some readings are not stored.
FrameLedger asks the driver for them from inside the game; a read that is slow or faults is switched off until the game is closed, and the session says so.
([Advanced metrics](guide/12-advanced-metrics.md))
## The game's memory
@@ -128,6 +150,35 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
N/A for video memory. AMD and Intel graphics cards have not been tested.
([Metrics](docs/03_METRICS.md))
## Controllers and input
Since 0.1.0-beta.50 a session records what it was played with. All of it is read from Windows, from outside the
game — so it is there for a game FrameLedger does not measure as well.
- **It says a controller was in use while the game was in front — not what the game did with it.** A controller
held while a game ignores it reads as a controller in use.
- **Keyboard and mouse are one figure.** FrameLedger reads only the time of the last keyboard or mouse input; it
cannot tell the two apart, and never reads a key.
- **Input a program makes counts as keyboard and mouse**: Steam Input mapping a controller to keys, a macro tool, a
remapper. A controller mapped that way reads as *both*.
- **Only the time the game is in front counts.** A game whose window belongs to another process than the one
FrameLedger follows reads as never in front, and has no shares.
- **Xbox-compatible controllers and the DualSense are read for use.** Any other controller — a DualShock 4, a
Switch Pro controller — is listed with its connections and disconnections and is never counted as in use.
- **A controller that Windows presents as an Xbox pad is "Xbox"**, whoever made it. Its model is named only when it
cannot be mistaken for another one that is connected.
- **Battery.** An Xbox controller on a cable reads "wired"; on Bluetooth it may report nothing. A DualSense on
Bluetooth reports its battery only where Steam or the game has switched it to its long report — FrameLedger never
asks a controller for anything. An Xbox controller's level is four words, a DualSense's is in steps of ten per
cent.
- **Steam Input shows a controller twice** — the one you hold and an Xbox pad Steam presents for it. FrameLedger
folds the two into one when they move together; with another remapper it may list two.
- **Nothing is read until you have accepted the Privacy Policy that describes it.** After an update, sessions
recorded before FrameLedger has been opened once have no such figures, and say so in their notes.
- **None of this had been read from a real controller when 0.1.0-beta.50 was written.** It was checked against data
written by hand; what real controllers do is in the developer's notes once the owner has tried them.
([Input devices](docs/21_INPUT_DEVICES.md))
## Display mode (full-screen, borderless, windowed)
- **Exclusive full-screen is known only for a measured Direct3D (DXGI) game** — the game's own swap chain
@@ -168,6 +219,84 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
- **CPU temperature** needs *Run the agent as administrator* and the separately installed PawnIO driver, and
has not been verified on real hardware yet. CPU load is the time the processor was busy — not Task
Manager's "utility" figure, which can exceed 100%. ([Telemetry](docs/18_GPU_VENDOR_APIS.md))
- **Advanced metrics** (since 0.1.0-beta.39; Settings ▸ Recording, off by default) stores more of what the graphics
driver and the sensors report with each session. Its limits:
- **A session recorded while it was off is not filled in later**, and switching it on applies from the next session.
- **What a game tells DLSS and FSR** (since 0.1.0-beta.40) is kept only for a game FrameLedger measures from
inside, and only where the game hands it over in a call FrameLedger intercepts. **Most games set their DLSS
options through a call it cannot see and show nothing**; the FSR rows need FSR 3.1 or later through AMD's own
library. The session says which it was — nothing seen, or not measured inside the game — and neither means the
game uses no upscaler. They are the game's words, not a check of what the upscaler did; the field of view is a
lowest and highest value, never a line over time; FSR's memory is a point where the game asked, so most sessions
have none. FSR's upscaling values have been read from real games; **the DLSS options have not been confirmed on
one yet.**
- **What a game tells Streamline each frame, and how it configures FSR frame generation** (since 0.1.0-beta.43)
are read from two calls FrameLedger hooks **only once a session with Advanced metrics on has started in that
game** — a session recorded with the option off puts neither hook in. Streamline's values are those of **the
first viewport that called**: a game with several (split screen, a picture-in-picture view) shows one, and the
count of calls covers all. The frame-generation rows are **what the game asked for** — a game that asks for
generation and gets none still reads "on", and the session's summary is where what was generated is measured. A
setting the game changed during the session reads "on and off". The session's line says, for each of the two,
whether it was read, hooked and never called by the game, or not hooked (the game did not load the library).
Streamline 1.x is not read. Streamline's values have been read from three real games and the frame-generation
configuration from one (2026-11-12); **a configuration that reaches FSR through AMD's 2.3 loader has not been
read from a real game yet.** **A game may give a value in a unit of its own**: Cyberpunk 2077 gives Streamline
its field of view in degrees where NVIDIA's header says radians. A field of view too large to be radians is
shown as the game gave it, with a line saying so, and is never converted (since 0.1.0-beta.44; a session
recorded before keeps the line it was stored with, which labels the degrees as radians). The count of resets is
the number of calls in which the game said so — menus and loading screens raise it.
- **Which version each Streamline feature is, and what it requires** (since 0.1.0-beta.44) are Streamline's
answers to two questions a game asks while it starts — and, in the games run so far, again later. FrameLedger
hooks the two calls only once a session with Advanced metrics on has started in that game. **A game that asks
only at its start shows nothing here**, and the session's line then says the call was hooked and the game did
not ask again; the four Unreal Engine games run on 2026-11-12 asked again and were read, and Cyberpunk 2077
asked again for the requirements and not for the versions. A second session of a game that kept running keeps
what the first one read. The rows are Streamline's answer as the game received it — the versions
of Windows and of the driver are the ones Streamline found — not FrameLedger's own check of the PC. Twelve
features are kept; a thirteenth is not. The lists the answer points to (the buffers, the Vulkan extensions) are
never read, so the rows give how many buffers a feature needs and not which. **The version Streamline gives for
a feature can differ from the file in the game's folder** — three games answered 2.14.16 for DLSS frame
generation while shipping an older plugin; which file that newer one is was not looked into. The games run so
far do not ask Streamline about DLSS upscaling unless they upscale through it, so a game can list frame
generation here and no upscaler.
- **Whether each Streamline feature is supported and loaded, and what the game did with it** (since
0.1.0-beta.45) come from six more calls, hooked the same way — only once a session with Advanced metrics on has
started in that game. **What the game did before that is not seen**: a feature it switched on while it started
has no "switched" row, and resources it allocated then are not counted. The counts are **calls the game made**,
not resources, and each stops at fifteen (shown as `≥ 15`); they and the switching start again with each
session, while the two answers are kept from an earlier session of a game that kept running. "Supported" is
Streamline's answer for the graphics card the game asked about; which card that was is never read, so on a PC
with two cards the row does not say. A refusal is shown by NVIDIA's own name for it
(`eErrorFeatureNotSupported`), and it is the game that was refused, not FrameLedger. None of this says a
feature ran or drew anything. **Read on five games on 2026-11-13**: each asked whether its features are
supported, three asked whether they are loaded, two switched features off and on, one freed a feature's
resources once and one shut Streamline down once. **None allocated a feature's resources after FrameLedger was
in it**, so that row has not been seen in a real game.
- **What a game started Streamline with is not read, and will not be**: the game makes that call once, before
FrameLedger can be inside it — also when FrameLedger starts the game itself. The Advanced page lists the entry
with that reason.
- **The driver's own figures** — the card's firmware and architecture, its memory in detail, what a display can do
with HDR and variable refresh, the driver's settings for the game — exist on **NVIDIA** cards only. On other cards
they read N/A; the card's sensors (fans, voltages, loads) are read on every brand, and none has been tested on an
AMD or Intel card.
- **The sensors are read every five seconds**, NVIDIA's figures every second; between two readings a sensor has no
value, and its last one is not repeated.
- **The processor's, the memory's and the motherboard's sensors** need *Run the agent as administrator* and the
separately installed PawnIO driver, and **have never been read on real hardware**. Without them they read N/A
with the reason.
- **A drive's temperature is not read at all**: the sensor library cannot list drives without touching removable
ones. Power supplies, fan controllers and batteries are not read either.
- **No serial number and no display identification data (EDID) is read**, so two PCs with the same parts look the
same.
- **None of these readings has been checked against another tool**, and they are **not in an export or a bug
report**.
- **The Advanced page's "This PC" is read only while that page is open** (or a game is being recorded), is not
stored, and keeps five minutes of each reading in memory. A reading with more than eight parts — every
processor core — is drawn as its lowest, average and highest; the table lists each part.
- **A session's chart needs its per-second series**, which are kept for the newest sessions of each game only
(Settings ▸ Recording); an older session keeps its statistics.
([Telemetry](docs/18_GPU_VENDOR_APIS.md#advanced-metrics-owner-decisions-d101d108-2026-11-11))
- **Whether the graphics card was held back by its power limit or by heat** (since 0.1.0-beta.14) is known only on
NVIDIA cards, and only as it was at each once-a-second reading. On other cards it reads N/A.
- **Frames per watt** (since 0.1.0-beta.14) counts the game's own frames — never generated ones — against the graphics
@@ -178,7 +307,8 @@ shown is recorded too, from outside the game. ([Capture](docs/04_CAPTURE.md))
Off by default. When on, **Windows asks every time the agent starts** and names an unknown publisher, because
FrameLedger is not signed. It lets FrameLedger open games that run as administrator and read the CPU
temperature; it does not bypass anti-cheat, does not open protected processes, and does not make games you
temperature — with Advanced metrics on, also the processor's, the memory's and the motherboard's sensors — where
PawnIO is installed; it does not bypass anti-cheat, does not open protected processes, and does not make games you
start from FrameLedger run as administrator. **The risk:** FrameLedger is installed in your user folder, so a
program running under your account could replace its files and would then run as administrator the next time
you accept the prompt. Turn it on only if you trust everything that runs under your account.
@@ -186,7 +316,13 @@ you accept the prompt. Turn it on only if you trust everything that runs under y
## Platform and what has been tested
- **Windows 10 22H2 or Windows 11, 64-bit only.**
- **Windows 11, 64-bit only.** FrameLedger is built on .NET 11 since 0.1.0-beta.37, which Microsoft supports on
Windows 11 and, on Windows 10, only in its LTSC and IoT editions. **Windows 10 22H2 Home and Pro are not blocked,
not tested and not supported**: FrameLedger may run there, and nobody has checked that it does
([Build](docs/12_BUILD.md)).
- **It does not start at all on a processor without SSE4.2 and POPCNT** — .NET 11's requirement, which every
processor Windows 11 accepts meets. On such a PC an update from 0.1.0-beta.36 or older installs and then does not
start; installing 0.1.0-beta.36 again brings FrameLedger back, with your data as it was.
- The games FrameLedger has actually been run with, and what was measured on each, are listed in
[LIST_GAME_TESTED.md](docs/LIST_GAME_TESTED.md). Everything was measured on one machine (an NVIDIA
GeForce RTX 5080 with an Intel Core i7-14700KF on Windows 11); other hardware is untested.
@@ -197,6 +333,15 @@ you accept the prompt. Turn it on only if you trust everything that runs under y
anti-cheat, never to judge it, and recording or measuring a game vouches for nothing. Four itch.io downloads this
project once used as examples turned out to carry malware ([details](docs/spike-notes.md)).
## Accessibility
What works with the keyboard alone, a screen reader, a contrast theme or Magnifier, what does not yet, and how to report
a barrier is on its own page: [Accessibility](ACCESSIBILITY.md) (in the app: Help ▸ Accessibility). The largest limits:
- **A chart's points cannot be read one by one.** A screen reader hears its name and where its numbers are written —
the session's figures, the tables, **File ▸ Export**; its keys move it.
- **Windows' text size setting is not followed** — use display scaling or Magnifier.
## Play
- **Through GOG GALAXY or itch, you finish in their app.** GOG GALAXY runs the game as its own shortcuts do, and the
@@ -208,7 +353,9 @@ you accept the prompt. Turn it on only if you trust everything that runs under y
- **Some games will not start directly** (seen on two Steam games, 2026-10-19 and 20). Lies of P, started by *Start
directly*, hands itself to Steam within seconds — since 0.1.0-beta.27 the game Steam then starts is recorded —, and
Clair Obscur: Expedition 33 closes within seconds, with or without FrameLedger's hook. Each such start ends a
measured session early, which the crash safety counts: two turn hooking off for the game. Start such a game through
measured session early, which the crash safety counts: two turn hooking off for the game — and since
0.1.0-beta.44 the count starts again each time you switch hooking back on (before, one more such end was enough
once a game had been switched off). Start such a game through
its store (Settings ▸ Games ▸ *Play button*); the notice after Play says when a Steam game may hand itself to Steam.
([details](docs/spike-notes/23-owner-runs-2026-10-20.md))
@@ -269,13 +416,13 @@ releases unless you choose *Beta* in Settings ▸ Updates, and apply a downloade
*Only for an alpha — a version ending in `-alpha.N`, an experimental build on a new .NET
([Alpha builds](guide/11-alpha-builds.md)).*
- **It runs on a .NET that Microsoft has not finished** — .NET 11's release candidate first — and has not been tested
the way a beta is: everything on this page applies to it, and so can problems no beta has shown
- **It is built to try a newer .NET — at times one Microsoft has not finished** — and has not been tested the way a
beta is: everything on this page applies to it, and so can problems no beta has shown
([Alpha notice](legal/ALPHA_NOTICE.md)).
- **Windows 10 22H2 Home and Pro are not supported** by Microsoft for .NET 11; an alpha may run there, but nobody
supports it. It does not start at all on a processor without SSE4.2 and POPCNT.
- **A failure in the capture agent's own background work ends the agent with a crash report** on an alpha, where a
beta's agent stops and exits quietly: .NET 11 reports it that way.
- **The alphas are paused since 0.1.0-alpha.4, and a paused alpha gets no updates** — the list of anti-cheat software
it refuses included, which arrives only inside an update. The betas run on the same .NET 11 since 0.1.0-beta.37:
install the latest beta over the alpha. Once ten betas have followed 0.1.0-alpha.4, an older alpha no longer finds
it as an update ([Updater](docs/11_UPDATER.md)).
- **It shares the beta's data folder.** It copies the database into `backups` before it first opens it, and does not
open it if the copy cannot be made; the newest three copies are kept and older ones it made are deleted.
- **It updates only to newer alphas**, and its update channel is locked; going back to the beta means installing the
+65 -13
View File
@@ -6,14 +6,15 @@
**New here? Start with the [user guide](guide/README.md)** — install, record a game, read your results, stay safe with
anti-cheat, and look after your data, a page each, then every screen, every setting and what to do when something goes
wrong. What FrameLedger cannot do is on one page: [Limitations](LIMITATIONS.md). Both are in the app too (Help ▸ User
guide, Help ▸ Limitations), and on the project's [wiki](https://github.com/poli0981/frameledger/wiki) with the developer
pages.
wrong. What FrameLedger cannot do is on one page: [Limitations](LIMITATIONS.md). How it works with the keyboard alone, a
screen reader or a contrast theme, and how to report a barrier: [Accessibility](ACCESSIBILITY.md). All three are in the
app too (Help ▸ User guide, Help ▸ Limitations, Help ▸ Accessibility), and on the project's
[wiki](https://github.com/poli0981/frameledger/wiki) with the developer pages.
<!-- accuracy-block:begin -->
> ⚠ **What FrameLedger actually measures today — 2026-10-19.** The software is a beta; its latest
> **beta is `0.1.0-beta.33`** (2026-11-03), an unsigned installer built from that tag with its
> checksums published beside it (alphas — experimental builds on a prerelease .NET — are published apart). The source holds the desktop app
> ⚠ **What FrameLedger actually measures today — 2026-11-11.** The software is a beta; its latest
> **beta is `0.1.0-beta.49`** (2026-11-14), an unsigned installer built from that tag with its
> checksums published beside it (alphas — experimental builds that try a newer .NET — are published apart). The source holds the desktop app
> (library, store import, charts, settings) and the background Agent, which records a session when a
> game in the library runs (unless you switched its recording off), injects only into games you enabled
> and only past the safety guard, and
@@ -44,7 +45,9 @@ pages.
> every other case — FSR at a dynamic resolution among them — no mode is shown.
> - **DLSS version:** the version of each DLSS library the game loaded and whether it came from the game or from
> the NVIDIA driver; "DLSS 4" or "DLSS 4.5" only when the session shows it (the model preset the driver reports,
> or a measured multi-frame factor); DLSS 5 only as "runtime loaded".
> a measured multi-frame factor, or the NVIDIA App's Dynamic frame generation that the driver reports applied);
> "DLSS 5" only when the game was seen running its neural rendering, otherwise "runtime loaded" when its library was
> loaded.
> - **Render → output resolution:** measured where the vendor's own per-frame call carries the render
> size (AMD FidelityFX dispatches, and Streamline titles whose resource tags state it); `N/A` where it
> does not: DLSS titles that bypass Streamline, Streamline titles whose tags carry no size, and Vulkan
@@ -84,8 +87,18 @@ pages.
> mean, median and peak. The game's other processes (launchers, helpers) are not counted.
> - **Video memory:** in use on the whole graphics card, recorded from Windows and GPU-driver telemetry
> and charted. **Not measured at all:** the video-memory budget Windows gives a game, which frame
> spikes were shader compilation, PC latency (Reflex), and HDR. Stutter count and stutter time are
> spikes were shader compilation, end-to-end PC latency, and HDR. Stutter count and stutter time are
> measured from frame times.
> - **NVIDIA's latency report (only with Advanced metrics on; Direct3D 12; a game that uses NVIDIA Reflex):** how
> long the driver says each stage of a frame took — simulation, render submit, present, driver, the operating
> system's queue, the graphics card — read from the graphics driver about once a second and shown as the driver
> reported it: **the driver's numbers, not a measurement by FrameLedger**, an average of the frames of one read,
> never checked against another tool, and not the delay from a click to the screen. **The stages overlap and do
> not add up to the total.** The graphics card's stage runs from its start to its end and includes the time the
> card waited — in one game it was longer than the whole frame — so since beta.48 two figures the driver works
> out itself are shown beside it: the card's active time, and the time from one frame's end to the next. A
> driver may give neither, and then there is no line. Read in two real games, on one driver and one card
> (2026-11-14). A game that does not use Reflex has none. The session summary's latency fields stay N/A.
> - **Pacing, VSync and the graphics card (since beta.14):** the share of the time below 30 FPS, below 60 FPS and below
> the monitor's refresh rate, measured from frame times with small margins so that a game held at its limit is not
> counted as below it (a rate less than 2 % under the line reads partly as at it), and how much one frame's time
@@ -94,11 +107,48 @@ pages.
> temperature in every session, and, on NVIDIA cards only, how much of the time its power limit or its temperature held
> it back, read once a second; and frames per watt — the game's own frames per second for each watt the graphics card
> drew, never counting generated frames, `N/A` where the game's own frame rate is not known.
> - **Controllers and input (since beta.50; on unless you switch it off in Settings, and only once you have accepted
> the Privacy Policy that describes it):** how much of the time a game was in front it was played with a
> controller, with the keyboard or mouse, with both or with neither; each controller's connections and drops; its
> battery. All of it is read from Windows, from outside the game: it says a controller was in use while the game
> was in front, not what the game did with it. Keyboard and mouse are one figure — only the time of the last input
> is read — and input a program injects (Steam Input mapping a controller to keys, a macro tool) counts as keyboard
> and mouse. An Xbox-compatible controller and a DualSense are read for use; another maker's controller is listed
> and never counted as in use. A DualSense on Bluetooth reports its battery only where Steam or the game has
> switched it to its long report, and an Xbox controller on Bluetooth may report none. **None of this had been
> read from a real controller when it was written (2026-11-14)**: it was checked against data written by hand.
> - **Processor and memory:** how busy the processor was (time busy, all cores together — not the
> frequency-scaled figure Task Manager draws) and how much system memory was in use, read from
> Windows once a second and charted. Processor temperature is read only when the Agent runs as
> administrator — an option, off by default, for which Windows asks at every start of the Agent — with
> PawnIO installed, and that reading has not yet been checked on real hardware; everywhere else it is `N/A`.
> - **Advanced metrics (since beta.39; off unless you switch it on in Settings):** with it on, a session also stores
> what the graphics driver and the sensor library report about the hardware — the graphics card's memory, load,
> clocks, fans, voltages and power in detail, what a display can do, the driver's settings for the game — as they
> report it: NVIDIA's figures once a second, the sensor library's every five seconds. None of these readings has
> been checked against another tool. The processor's, the memory's and the motherboard's need the Agent as
> administrator with PawnIO and have never been read on real hardware; a drive's temperature is not read. The
> Advanced page shows them as they were stored, and what the PC reads at the moment only while that page is open.
> - **What a game tells its upscaler (since beta.40; only with Advanced metrics on, only for a game measured from
> inside):** the output size, presets and exposure a game hands DLSS, and the sharpening, jitter, exposure, depth
> range, field-of-view range, flags and frame time it hands FSR, are read from the arguments of the calls
> FrameLedger intercepts and shown as the game stated them — not checked against what the upscaler did. Most games
> set their DLSS options through a call that cannot be intercepted and show none; the FSR values need FSR 3.1 or
> later through AMD's own library. Since beta.43 also what a game tells NVIDIA Streamline about each frame (jitter,
> motion-vector scale, depth range, field-of-view range, how often it reset the upscaler's history — for the first
> viewport that called, where a game has several) and how it configures FSR frame generation (the settings it asked
> for, not whether frames were generated). Since beta.44 also what Streamline answers when a game asks about a
> feature — its version, and what it requires (graphics APIs, V-Sync, GPU scheduling, Windows and driver versions):
> Streamline's answer as the game received it, not a check of the PC. A game that asks only while it starts has
> none and the session says so; the Unreal Engine games run so far ask again later, and were read. Since beta.45
> also whether Streamline says each feature is supported and whether it is loaded, and what the game did with it
> after FrameLedger attached — switched it on or off, allocated and freed its resources (counted to fifteen), shut
> Streamline down: calls the game made, not a statement that a feature ran or produced anything. FSR's upscaling
> values, Streamline's constants (seven games), its answers about a feature (seven games), the
> frame-generation configuration (one game) and five of beta.45's six calls (five games) have been read from
> real games; no game run so far allocated a feature's resources after FrameLedger attached, and the DLSS
> options have not been confirmed on one yet. A game may give a value in a unit of its own: one gives
> Streamline its field of view in degrees, and it is shown as given.
> - **Safety:** every pre-injection check runs before injection, including the signed-by-a-known-vendor
> half of the suspicious-module rule. During every capture the Agent re-runs the checks every 30 s and
> stops capturing on a refusal: a Direct3D or OpenGL game's hooks are removed, and the Vulkan layer
@@ -148,7 +198,7 @@ FrameLedger is built to keep that risk small and to be honest about it:
| **Off by default** | Injection is disabled for every game until you enable it individually, after a consent prompt |
| **Hard refusal** | Known anti-cheat/anti-tamper components are detected before injection and every 30 s during a session. Detected ⇒ FrameLedger refuses, or stops capturing at the next scan — so up to 30 s can pass before it reacts to anti-cheat that loads mid-session. For Direct3D/OpenGL that means hooks are removed; for Vulkan it means the layer goes passthrough, because a layer cannot leave the loader chain of a running game. **There is no override — not in settings, not in a config file, not on the command line**<br><br>**One narrow exception since beta.9 (owner decisions D33, D38):** per game, after its own disclosure and only while a Settings option (off by default) is on, a game whose only finding is one anti-cheat that runs entirely in user mode — no driver file in its folder, on no title list — may be hooked under a user-mode exception. ~~and that FrameLedger has measured successfully at least twice~~ *(corrected 2026-10-03: since beta.11, D38, a grant needs no earlier session; its first two hooked sessions are a trial, and one that does not go well ends it for good.)* The guard still runs every check under it, refuses anything else it finds, and ends it on anything new; a ban remains possible (`legal/DISCLAIMER.md` §2A).<br><br>✅ **Both halves run since 2026-09-10.** The *pre-injection* refusal runs every documented check, and during every capture the Agent re-runs them every 30 s: the component inside the game removes its hooks within one frame of being told to, and within 65 s if it stops hearing from the scanner. *(Until 2026-09-15 this cell said the in-session re-scan did not run at all, which stopped being true when the Agent's capture loop landed.)* For a Vulkan title the stop is the layer's own: it checks the scanner's counter on every present and goes passthrough within one present of the stop flag, or after 65 s without a scan — a title that has stopped presenting is not stopped by that deadline, and presents nothing to be observed either. |
| **No evasion, ever** | FrameLedger does not hide, rename, obfuscate, or disguise itself. It keeps its real name, real exports, and version info. It is meant to be plainly visible to any security software that looks. This is an architectural rule, not a setting |
| **Read-only** | It never reads or writes game memory, never modifies game behavior, never touches saves or input, and never changes GPU clocks, fans, or power limits |
| **Read-only** | It never reads or writes game memory, never modifies game behavior, never touches saves, never reads what you type or which buttons you press — from outside the game it notes only whether a controller, or the keyboard or mouse, is in use — and never changes GPU clocks, fans, or power limits |
| **Always a way out** | FrameLedger refuses rather than pushes through. When the guard finds anti-cheat, or the hook fails, or you have not enabled a game, it **stops** — and records the session anyway: start, end, duration, whatever hardware sensors your machine provides, and the reason there is nothing else. Every measured value reads `N/A`; it never estimates one.<br><br>**The way out is that it does not inject, not that it measures another way.** An earlier version of this row promised a no-injection measurement mode. It was designed around Intel PresentMon, that was dropped on 2026-08-27 after measurement, and no replacement was chosen — so **frame times require hooking**. Said in this row rather than a footnote because the row is in the SAFETY table, and a reader deciding whether to enable a game needs the real alternative, not a comfortable one |
**FrameLedger is for offline and single-player games.** If you enable it for anything with an online or competitive component, that is your call and your responsibility. The developer cannot reverse a ban. Please read [`legal/DISCLAIMER.md`](legal/DISCLAIMER.md) and [`docs/19_SAFETY_AND_ANTICHEAT.md`](docs/19_SAFETY_AND_ANTICHEAT.md) before enabling it for anything.
@@ -166,7 +216,8 @@ The tier is recorded on every session and shown in the UI. Metrics unavailable a
## Requirements
- Windows 10 (22H2) or Windows 11, 64-bit
- Windows 11, 64-bit. Windows 10 (22H2) is not blocked, but FrameLedger is built on .NET 11 since 0.1.0-beta.37, which Microsoft does not support on Windows 10 Home or Pro, and FrameLedger has never been tested there
- A processor with SSE4.2 and POPCNT — any that Windows 11 accepts. Without them FrameLedger does not start
- A 64-bit DirectX 11/12 or OpenGL game, or a Vulkan game FrameLedger starts, for hooked (Tier-1) capture. **32-bit games — including most DirectX 9 titles — cannot be measured at all**: the component FrameLedger loads is x64 and an x64 DLL cannot enter a 32-bit process. Such a title records duration and hardware telemetry and nothing else. *(This line previously said "supported at Tier 2 only", which meant frame times without injection. That tier no longer exists.)*
- Optional: [PawnIO](https://pawnio.eu/) for CPU temperature, read only when the Agent runs as administrator (since 0.1.0-beta.3; not yet checked on real hardware). CPU load, memory in use and all GPU telemetry work without it
@@ -174,7 +225,7 @@ Elevation is **optional — for everything.** Hooked capture is the normal path
## Install
> **The latest beta is `v0.1.0-beta.33` (2026-11-03); before it `v0.1.0-beta.32` (2026-10-31), `v0.1.0-beta.31` (2026-10-26), `v0.1.0-beta.30` (2026-10-26), `v0.1.0-beta.29` (2026-10-25), `v0.1.0-beta.28` (2026-10-20), `v0.1.0-beta.27` (2026-10-20), `v0.1.0-beta.26` (2026-10-19), `v0.1.0-beta.25` (2026-10-19), `v0.1.0-beta.24` (2026-10-19), `v0.1.0-beta.23` (2026-10-14), `v0.1.0-beta.22` (2026-10-13), `v0.1.0-beta.21` (2026-10-13), `v0.1.0-beta.20` (2026-10-12), `v0.1.0-beta.19` (2026-10-12), `v0.1.0-beta.18` (2026-10-07), `v0.1.0-beta.17` (2026-10-07), `v0.1.0-beta.16` (2026-10-06), `v0.1.0-beta.15` (2026-10-06), `v0.1.0-beta.14` (2026-10-06), `v0.1.0-beta.13` (2026-10-04), `v0.1.0-beta.12` (2026-10-04), `v0.1.0-beta.11` (2026-10-03), `v0.1.0-beta.10` (2026-09-28), `v0.1.0-beta.9` (2026-09-26), `v0.1.0-beta.8` (2026-09-25), `v0.1.0-beta.7` (2026-09-23), `v0.1.0-beta.6` (2026-09-23), `v0.1.0-beta.5` (2026-09-22), `v0.1.0-beta.4` (2026-09-21), `v0.1.0-beta.3` (2026-09-21), `v0.1.0-beta.2` (2026-09-17) and `v0.1.0-beta.1` (2026-09-16).** They are pre-releases: read the notes of the one you install — `CHANGELOG.md` links each release's own file under `docs/changelog/` — for what it measures, what it does not yet, and how to update, before installing. A source build is `docs/12_BUILD.md`.
> **The latest beta is `v0.1.0-beta.49` (2026-11-14); before it `v0.1.0-beta.48` (2026-11-14), `v0.1.0-beta.47` (2026-11-14), `v0.1.0-beta.46` (2026-11-13), `v0.1.0-beta.45` (2026-11-12), `v0.1.0-beta.44` (2026-11-12), `v0.1.0-beta.43` (2026-11-12), `v0.1.0-beta.42` (2026-11-12), `v0.1.0-beta.41` (2026-11-12), `v0.1.0-beta.40` (2026-11-12), `v0.1.0-beta.39` (2026-11-11), `v0.1.0-beta.38` (2026-11-11), `v0.1.0-beta.37` (2026-11-11), `v0.1.0-beta.36` (2026-11-10), `v0.1.0-beta.35` (2026-11-09), `v0.1.0-beta.34` (2026-11-06), `v0.1.0-beta.33` (2026-11-03), `v0.1.0-beta.32` (2026-10-31), `v0.1.0-beta.31` (2026-10-26), `v0.1.0-beta.30` (2026-10-26), `v0.1.0-beta.29` (2026-10-25), `v0.1.0-beta.28` (2026-10-20), `v0.1.0-beta.27` (2026-10-20), `v0.1.0-beta.26` (2026-10-19), `v0.1.0-beta.25` (2026-10-19), `v0.1.0-beta.24` (2026-10-19), `v0.1.0-beta.23` (2026-10-14), `v0.1.0-beta.22` (2026-10-13), `v0.1.0-beta.21` (2026-10-13), `v0.1.0-beta.20` (2026-10-12), `v0.1.0-beta.19` (2026-10-12), `v0.1.0-beta.18` (2026-10-07), `v0.1.0-beta.17` (2026-10-07), `v0.1.0-beta.16` (2026-10-06), `v0.1.0-beta.15` (2026-10-06), `v0.1.0-beta.14` (2026-10-06), `v0.1.0-beta.13` (2026-10-04), `v0.1.0-beta.12` (2026-10-04), `v0.1.0-beta.11` (2026-10-03), `v0.1.0-beta.10` (2026-09-28), `v0.1.0-beta.9` (2026-09-26), `v0.1.0-beta.8` (2026-09-25), `v0.1.0-beta.7` (2026-09-23), `v0.1.0-beta.6` (2026-09-23), `v0.1.0-beta.5` (2026-09-22), `v0.1.0-beta.4` (2026-09-21), `v0.1.0-beta.3` (2026-09-21), `v0.1.0-beta.2` (2026-09-17) and `v0.1.0-beta.1` (2026-09-16).** They are pre-releases: read the notes of the one you install — `CHANGELOG.md` links each release's own file under `docs/changelog/` — for what it measures, what it does not yet, and how to update, before installing. A source build is `docs/12_BUILD.md`.
1. Download the latest `FrameLedger.App-win-Setup.exe` from [Releases](https://github.com/poli0981/frameledger/releases). It installs into `%LOCALAPPDATA%\FrameLedger.App`; your data stays in `%LOCALAPPDATA%\FrameLedger`, and uninstalling asks before touching it.
2. SmartScreen may warn — releases are not code-signed (free, open-source project). Verify the SHA-256 checksum published with each release, then **More info → Run anyway**.
@@ -201,7 +252,8 @@ FrameLedger is GPL-3.0-only free software, and you are welcome to build it, chan
- **Build it:** [`docs/12_BUILD.md`](docs/12_BUILD.md) — `./build.ps1 check` is the whole quality gate, the same one CI runs.
- **Contribute:** [`CONTRIBUTING.md`](CONTRIBUTING.md) — the definition of done, what will not be merged, and that a
contribution is licensed GPL-3.0-only with the [`NOTICE`](NOTICE) terms (no CLA). Conduct:
[`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md). Security reports: [`SECURITY.md`](SECURITY.md).
[`CODE_OF_CONDUCT.md`](CODE_OF_CONDUCT.md). Security reports: [`SECURITY.md`](SECURITY.md). Accessibility — the goal,
the checks and how a barrier is reported: [`ACCESSIBILITY.md`](ACCESSIBILITY.md).
- **Fork or redistribute:** [`FORKING.md`](FORKING.md) — what the GPL asks of anyone who distributes builds, and the names
a fork must change so it does not share FrameLedger's update feed, install folder, data, pipe or Vulkan layer.
- **Find your way around:** the [wiki](https://github.com/poli0981/frameledger/wiki)'s developer pages are short, and each
@@ -219,7 +271,7 @@ FrameLedger is GPL-3.0-only free software, and you are welcome to build it, chan
| `FrameLedger.Overlay.dll` | Inside the game | C++20 hooks + lock-free shared-memory writer. Records only; never analyzes, allocates, or blocks |
| `FrameLedger.VkLayer.dll` | Inside the game | Vulkan implicit layer (Vulkan titles use this instead of injection) |
GPU telemetry is layered so the project never depends on a proprietary vendor licence: a vendor-neutral DXGI/performance-counter baseline, LibreHardwareMonitor (MPL-2.0) for sensors on all vendors, and NVIDIA's NVAPI SDK (MIT) for NVIDIA-only extras (telemetry and the driver's DLSS record today; Reflex latency is not read yet). No Intel or AMD GPU *telemetry* SDK is bundled — see `docs/18_GPU_VENDOR_APIS.md` for why. The one AMD component in the tree is five MIT headers from the FidelityFX SDK, used for types only so the upscaler hook can read an FSR title's own dispatch descriptor; nothing AMD-built is linked or redistributed.
GPU telemetry is layered so the project never depends on a proprietary vendor licence: a vendor-neutral DXGI/performance-counter baseline, LibreHardwareMonitor (MPL-2.0) for sensors on all vendors, and NVIDIA's NVAPI SDK (MIT) for NVIDIA-only extras (telemetry and the driver's DLSS record; and, with Advanced metrics on, the driver's own latency report, read inside a Direct3D 12 game that uses NVIDIA Reflex). No Intel or AMD GPU *telemetry* SDK is bundled — see `docs/18_GPU_VENDOR_APIS.md` for why. The one AMD component in the tree is five MIT headers from the FidelityFX SDK, used for types only so the upscaler hook can read an FSR title's own dispatch descriptor; nothing AMD-built is linked or redistributed.
---
+8 -2
View File
@@ -25,8 +25,14 @@
-->
<Property Name="sonar.exclusions">src/native/third_party/**,**/Strings.Designer.cs,src/native/tools/hook-harness/dxr_raygen_dxil.h,src/native/tools/vendor-stubs/**,tests/fixtures/**,legal/licenses/**,docs/vendor-exports.json,src/FrameLedger.Infrastructure/Persistence/Migrations/**,build/**,out/**,**/bin/**,**/obj/**,**/TestResults/**</Property>
<!-- Tests repeat their arrange steps on purpose; duplication there is not a defect. -->
<Property Name="sonar.cpd.exclusions">tests/**,src/native/tests/**</Property>
<!--
Tests repeat their arrange steps on purpose; duplication there is not a defect.
HookFamilies.cs is the Domain's mirror of FlHookFamily (ShmLayout.cs): Domain references nothing, so the enum is
written twice by design and MetricEnumMirrorTests fails when the two differ. Six members added at once
(0.1.0-beta.45) were 39 duplicated lines of 1,063 new ones, 3.7 % against the gate's 3 %, and failed it on the
release's commit with no issue open. The mirror is left out of the detection; FlHookFamily itself is not.
-->
<Property Name="sonar.cpd.exclusions">tests/**,src/native/tests/**,src/FrameLedger.Domain/Metrics/HookFamilies.cs</Property>
<Property Name="sonar.sourceEncoding">UTF-8</Property>
+30 -1
View File
@@ -467,7 +467,7 @@ function Invoke-ProjectGates {
Invoke-Checked 'notice-check (self-test)' { & $noticeTool -SelfTest }
Invoke-Checked 'notice-check' { & $noticeTool -RepoRoot $repo }
# beta.17 (D53/D54): the GitHub Wiki is built from wiki/, guide/ and LIMITATIONS.md, and wiki.yml publishes it when a
# beta.17 (D53/D54): the GitHub Wiki is built from wiki/, guide/, LIMITATIONS.md and (beta.35) ACCESSIBILITY.md, and wiki.yml publishes it when a
# release is made. A link or a section that leads nowhere, a page nothing leads to, a sidebar out of the guide's
# order, a path that names nothing or a version written into a page is red here, not on the published wiki.
# Self-test first (nine cases, both directions), then the live assembly into a temporary folder.
@@ -479,6 +479,35 @@ function Invoke-ProjectGates {
# beta.24 (D66): CHANGELOG.md and docs/HANDOFF.md keep their paths as the index of their parts under docs/changelog/
# and docs/handoff/. A released section left in CHANGELOG.md, a release file whose heading is not its name or that
# links relatively, or an index and its files that do not name each other both ways is red here.
# 2026-11-11 (D108): "the bridge never calls a setter" was a sentence in docs/18 a reviewer checked by reading one
# file. With Advanced metrics the bridge calls forty-four NVAPI entry points, and the header declares a setter beside
# almost every getter: each call is held to tools/nvapi-allowed-calls.txt, in both directions, and a writer or a
# reader of an identifier (D104) is red even when the list names it. Text only - it runs with or without the native
# build.
Write-Step 'nvapi-readonly-check'
$nvapiTool = Join-Path $repo 'tools/nvapi-readonly-check.ps1'
Invoke-Checked 'nvapi-readonly-check (self-test)' { & $nvapiTool -SelfTest }
Invoke-Checked 'nvapi-readonly-check' { & $nvapiTool -RepoRoot $repo }
# 2026-11-12 (D116): the same for the code loaded into a game. Advanced metrics, train B, reads more members of
# the vendors' structs than the frame record ever needed, and the same structs hold where the player's camera is.
# Every such read is in one header, each member on tools/overlay-allowed-fields.txt in both directions; a camera's
# position, axes and matrices are named nowhere in the Overlay or the Vulkan layer, whatever the list says. Text
# only.
Write-Step 'overlay-fields-check'
$fieldsTool = Join-Path $repo 'tools/overlay-fields-check.ps1'
Invoke-Checked 'overlay-fields-check (self-test)' { & $fieldsTool -SelfTest }
Invoke-Checked 'overlay-fields-check' { & $fieldsTool -RepoRoot $repo }
# 2026-11-14 (D155-D163): what is asked of Windows about controllers, the keyboard and the mouse. The reader's
# calls are declared in one file and held to tools/input-allowed-calls.txt in both directions; a call that sends to
# a device, injects input or reads a key is red even when the list names it; no managed source outside the reader
# and no native source names an input function; raw input is registered for joysticks and gamepads only. Text only.
Write-Step 'input-readonly-check'
$inputTool = Join-Path $repo 'tools/input-readonly-check.ps1'
Invoke-Checked 'input-readonly-check (self-test)' { & $inputTool -SelfTest }
Invoke-Checked 'input-readonly-check' { & $inputTool -RepoRoot $repo }
Write-Step 'docs-split-check'
$splitTool = Join-Path $repo 'tools/docs-split-check.ps1'
Invoke-Checked 'docs-split-check (self-test)' { & $splitTool -SelfTest }
+1 -1
View File
@@ -162,7 +162,7 @@ The profile's own folder is refused there.)*
- **ADR-8:** **No evasion, ever** (`19_SAFETY` §What we will never build). This is an architectural constraint, not a preference — it defines what "correct" means for this codebase.
- **ADR-9:** Elevation demoted from required to optional; the Agent runs as a standard user by default.
**Chosen, not inherited, since beta.10 (owner decision D34, 2026-09-27):** Settings ▸ Capture agent ▸ *Run the agent as
administrator* (`capture.run_elevated`, off by default, a versioned disclosure — `agent-admin-dialog/1` — before it turns
administrator* (`capture.run_elevated`, off by default, a versioned disclosure — `agent-admin-dialog/2` *(`/1` until 2026-11-11, when Advanced metrics widened what the option adds — D101)* — before it turns
on). With it on, **Windows is asked every time the Agent starts** — a UAC prompt owned by the App's window when the App
starts it, flashing on the taskbar when the logon task does — and there is no silent path: no "highest privileges" task,
no service. The elevated Agent is started for one user (`--serve --for-user <SID>`) and refuses to run as another
+29 -3
View File
@@ -53,6 +53,7 @@ Requirement IDs (`FR-x`, `NFR-x`) are referenced by other docs, commits, and tes
- FR-5.3 Charts handle ≥ 500k points via min/max decimation; zoom/pan; PNG export.
- FR-5.4 Native/Displayed toggle wherever FG data exists.
- FR-5.5 Stutter markers annotated with cause when known (`PSO compile`, `VRAM budget exceeded`).
- FR-5.6 *(owner decision D101, 2026-11-11)* **Advanced metrics**: an option, off by default, and a page of its own that shows what the vendor interfaces make available — NVAPI, LibreHardwareMonitor through PawnIO, and in later releases Streamline and FidelityFX inside the game — each value under what it belongs to, the PC or one session, and each value that is not read listed with the reason. Off means nothing more is read or stored. `18_GPU_VENDOR_APIS` §Advanced metrics is the design and the inventory. *(Not built: the catalog exists, nothing reads it yet.)*
### FR-6 History & comparison
- FR-6.1 Sessions table per game: date, duration, tier badge, Native, Displayed, FG×, lows, resolution + upscaler, RT/PT/RR chips, crash badge, tags.
@@ -101,10 +102,30 @@ Tray icon states (idle / capturing T1 / capturing T2 / paused); menu; toasts for
### FR-15 In-game overlay *(v1.1, opt-in, off by default)*
Draw live FPS/frametime/upscaler info inside the game. Only for games already hooked; a separate toggle; never on by default. Deferred until the data path is stable (`15_ROADMAP` P5).
### FR-16 Controllers and input *(owner decisions D155–D163, 2026-11-14)*
What a session was played with, read by the Agent from outside the game — so for hooked and unhooked sessions alike (`21_INPUT_DEVICES`, `03_METRICS` §Input).
- FR-16.1 The share of the time the game was in front that it was played with a controller, with the keyboard or mouse (one class), with both, with neither; the number of switches between a controller and the keyboard and mouse.
- FR-16.2 Each controller: its maker and model by its USB ids, how it is connected, its connections and disconnections during the session, the time it was in use.
- FR-16.3 Each controller's battery: the time at each level, the level that lasted longest, the first and the last, a rate where one can be worked out.
- FR-16.4 Each maker's share of the time controllers were in use, and the number of switches between makers. A pad that is one controller seen twice (Steam Input) counts once.
- FR-16.5 Whether a game supports controllers, as evidence and never a verdict: what the store's own file on the disk says, which input libraries the game loaded, how it was played (D163).
- FR-16.6 An option, **on unless switched off**, that reads nothing until the Privacy Policy describing it has been accepted on the PC (D156). Off means nothing is read or stored.
- FR-16.7 Never read or stored: which button, stick, trigger or key; text; the mouse's position; a serial number, a Bluetooth address, a device path. Nothing is ever sent to a controller (D159). Held by `tools/input-readonly-check.ps1`.
### FR-16 Controllers and input *(owner decisions D155–D163, 2026-11-14)*
What a session was played with, read by the Agent from outside the game — so for hooked and unhooked sessions alike (`21_INPUT_DEVICES`, `03_METRICS` §Input).
- FR-16.1 The share of the time the game was in front that it was played with a controller, with the keyboard or mouse (one class), with both, with neither; the number of switches between a controller and the keyboard and mouse.
- FR-16.2 Each controller: its maker and model by its USB ids, how it is connected, its connections and disconnections during the session, the time it was in use.
- FR-16.3 Each controller's battery: the time at each level, the level that lasted longest, the first and the last, a rate where one can be worked out.
- FR-16.4 Each maker's share of the time controllers were in use, and the number of switches between makers. A pad that is one controller seen twice (Steam Input) counts once.
- FR-16.5 Whether a game supports controllers, as evidence and never a verdict: what the store's own file on the disk says, which input libraries the game loaded, how it was played (D163).
- FR-16.6 An option, **on unless switched off**, that reads nothing until the Privacy Policy describing it has been accepted on the PC (D156). Off means nothing is read or stored.
- FR-16.7 Never read or stored: which button, stick, trigger or key; text; the mouse's position; a serial number, a Bluetooth address, a device path. Nothing is ever sent to a controller (D159). Held by `tools/input-readonly-check.ps1`.
## Non-functional requirements
- **NFR-1 Game-side overhead:** ≤ 1 µs per present, ≤ 8 MB resident, measured game FPS impact ≤ 0.5% vs uninstrumented.
- **NFR-2 Agent overhead:** ≤ 1% of one core during capture, ≤ 150 MB working set, ~0% idle.
- **NFR-2 Agent overhead:** ~~≤ 1% of one core during capture~~ **≤ 2.5 % of one core during capture** *(owner decision D117, 2026-11-11: the 1 % had never been measured; measured in a real Agent it is 1.93 % with Advanced metrics off and 2.04 % with it on — `20_OPEN_QUESTIONS` §M16, `tools/agent-cost.ps1` — nearly all of it the once-a-second read of the graphics card's sensors, which stays)*, ≤ 150 MB working set *(measured: 113 MiB off, 119 MiB on)*, ~0% idle. *(D101, 2026-11-11: the limits hold with Advanced metrics on — FR-5.6. Its reads run on a thread of their own, the slow ones seldom, and with no game running only while its page is open (D107), so "~0% idle" is not spent on it. Measured on the development PC in a session that is not hooked; a hooked session, an elevated Agent with PawnIO's groups and a real game are the owner's run, `HANDOFF` §Owner-only.)*
- **NFR-3 Stability:** faults **originating in our own hook bodies** are contained by `FL_HOOK_GUARD`, counted, and self-disable the Overlay after 3 occurrences; the safety-unhook path completes within one frame. The Overlay adds no allocation, lock, syscall or logging to the present path.
> This deliberately does **not** say "must never crash a game" (`20_OPEN_QUESTIONS` §H8). SEH containment cannot reliably catch stack overflow, cannot intercept `__fastfail` — and `-D_HAS_EXCEPTIONS=0` converts a would-be STL throw into exactly that (`spike-notes.md` §H3) — and does not help when the game installs a vectored exception handler that runs first. An absolute promise the mechanism cannot support does not belong in a requirement, and must stay out of user-facing text: the Disclaimer already says injection carries risk, and that wording is the honest one.
@@ -112,8 +133,13 @@ Draw live FPS/frametime/upscaler info inside the game. Only for games already ho
- **NFR-5 Storage:** raw series ≤ 3 MB per hour compressed; 50 games × 50 sessions ≤ 500 MB at default retention.
- **NFR-6 Accuracy:** per `03_METRICS` §Accuracy budget, stated per tier.
- **NFR-7 Resilience:** power loss loses at most the last 60 s of raw data; DB never corrupts (WAL + transactions); Agent restart recovers `.partial` sessions as `interrupted`.
- **NFR-8 OS support:** Windows 10 22H2+ and Windows 11, x64. No ARM64 in v1.
- **NFR-9 Accessibility:** full keyboard navigation, adequate contrast in both Fluent themes, Per-Monitor V2 DPI awareness. *(What each clause means for this app, and the check that holds it: `08_UI` §Accessibility, P4 PR-8.)*
- **NFR-8 OS support:** ~~Windows 10 22H2+ and Windows 11, x64.~~ **Windows 11, x64** — the Windows that Microsoft
supports .NET 11 on, which FrameLedger is built on since 0.1.0-beta.37. **Windows 10 22H2 is not blocked** (the code
is still held to it: `SupportedOSPlatformVersion` 10.0.19045.0, CA1416) **and is neither tested nor supported**
there; a processor needs SSE4.2 and POPCNT (x86-64-v2), without which .NET 11 does not start. No ARM64 in v1.
*(Amended 2026-11-11, owner decisions D94 and D95 — `12_BUILD` §The main line's .NET. Windows 10 was never tested
on any version: `ACCESSIBILITY.md`, `08_UI` §Accessibility.)*
- **NFR-9 Accessibility:** full keyboard navigation, adequate contrast in both Fluent themes, Per-Monitor V2 DPI awareness. *(What each clause means for this app, and the check that holds it: `08_UI` §Accessibility, P4 PR-8.)* *(Widened 2026-11-09, beta.35, owner decisions D88–D91: and a screen reader reads every control and value and is told what changes; Windows contrast themes are followed; nothing is shown by colour alone. The goal behind all of it is WCAG 2.2 AA as W3C's WCAG2ICT applies it to desktop software — a goal, not a conformance claim — stated to users in `ACCESSIBILITY.md`.)*
- **NFR-10 Offline-first:** everything except updates and opt-in metadata works with zero connectivity.
- **NFR-11 Auditability:** the entire set of hooks is enumerable in `17_HOOK_ENGINE` §Hook inventory and matches the code; a reviewer must be able to verify in minutes that nothing reads game memory.
+72 -3
View File
@@ -36,7 +36,7 @@ Let `D` = session duration (first → last present), `F_app` = frames the *game*
| **Frametime σ** | population stddev of `ft_app` (ms) | 1 |
| **Stutter count** | `ft_app[i] > 2 × rollingMedian(ft_app, 19)` (centered, window in **frames**) | 1 |
| **Stutter time %** | `Σ ft_app[stutter] / Σ ft_app × 100` | 1 |
| **PC latency** | mean/p95 of `reflexLatencyUs` when Reflex reports it | **1 only** |
| **PC latency** | mean/p95 of `reflexLatencyUs` when Reflex reports it *(nothing fills `reflexLatencyUs`: a report arrives after its frame's record is written. Since train C, 2026-11-13, the driver's stage report is read as seven Advanced series beside the frames — `18_GPU_VENDOR_APIS` §The inventory — and these session fields stay N/A; asked again on 2026-11-14 with two real games read, the owner kept them N/A (D147): the driver's total runs from simulation start to the graphics card's end, not from an input to the display)* | **1 only** |
| **PSO stutter %** | share of stutter frames with `psoCreatedThisFrame > 0` | **1 only** |
| **Time below 30 / 60 FPS** (beta.14) | share of `Σ ft_app` in frames that count as below X — `ft > 1.10 × 1000/X`, or the nine-frame rolling median `> 1.02 × 1000/X` (§Pacing) | 1 |
| **Time below refresh** (beta.14) | the same over **presents**, X = `display_monitor_hz` — the display's rate, so over what the display received | 1 |
@@ -535,7 +535,10 @@ F_app = Σ fgEvaluations (APPLICATION frames, counted at the
> in a game's menu is not read — it reaches the runtime through `sl::DLSSGOptions`, the route `HANDOFF` §2b refused —
> so the measured row is its evidence. The dynamic ceiling (`NGX_DLSSG_DYNAMIC_MULTI_FRAME_COUNT_MAX`) is decoded as
> `NGX_DLSSG_MULTI_FRAME_COUNT` is (frames generated per rendered frame) and the automatic target as "max refresh
> rate"; NVIDIA's header names neither, and owner-only item 20(a) checks both against the App.
> rate"; NVIDIA's header names neither, and owner-only item 20(a) checks both against the App. *(2026-11-06, beta.34:
> the ceiling's decode held on the owner's Phantom Blade Zero — a ceiling of 5 held states up to ×6.0, a ceiling of 2
> nothing above ×3.0 — `spike-notes` §26. And the context line is no longer the only place the Dynamic mode shows: the
> DLSS line names DLSS 4.5 where the driver reports Dynamic applied, D83, `05_DETECTION` §DLSS generation.)*
`fgEvaluations` is recorded per present by ~~the NGX/Streamline/FFX/XeSS FG hooks~~ the `slGetNewFrameToken`
detour; where Streamline never issued a token, by the FidelityFX dispatches — new PREPARE `frameID`s, else UPSCALE
@@ -684,6 +687,18 @@ From the upscaler hooks we get, per frame:
> Measured on the hook harness with the vendor stub answering as the SDK documents (`fl_guard` "FSR's quality
> mode", by size through the SDK 2.x loader and by ratio through the 1.1.x monolith); a real title is owner-only
> item 21 (Lies of P at Quality, then Performance).
- **One technology's name, quality and size** *(2026-11-12, owner decision D120)*. A session's — and a segment's —
upscaler is **the technology named on the most samples of the dominant stream** (a tie goes to the lower value, as
the quality byte's does), not the first one any record names. While a session names one technology, its quality
and render size are read from the whole stream as before: under frame generation the name sits on one present in
N. When it names **more than one**, the quality and the size are read **only from the samples that name the
chosen one**, a quality byte of the other family is withheld (`UpscalerChoice.QualityFor`), and the session is
marked "settings changed during session". The session this was found on (Dying Light: The Beast, 81 seconds:
DLSS evaluated for most of it, an FSR dispatch for a few seconds) had been stored as "DLSS" with FSR's Balanced
byte and FSR's render size — three true readings of two technologies under one name. What the chosen technology
did not say stays N/A (rule 7). `Domain.Metrics.UpscalerChoice`; the live card still names the first technology
the session saw (`FgLadder.UpscalerIdentity`), which for a session in progress is a different question and was
not changed.
- Resolution changes mid-session (user changed settings, or dynamic resolution scaling) produce **segments**: the session stores a segment list `(startFrame, renderW/H, outputW/H, upscaler, quality)`. Aggregates report the dominant segment plus a "settings changed during session" flag — averaging across a settings change is the classic way benchmark numbers become meaningless.
> **Two segmentation axes exist, they compose in ONE order, and neither document said so until
@@ -961,6 +976,60 @@ min / max in `sensor_stats_json` (§Sensor aggregates). MiB stored; shown in GB
~~`vramUsedMb` from `IDXGIAdapter3::QueryVideoMemoryInfo(LOCAL)` inside the game = **this game's** usage and budget. **MiB, truncating, and it must use the same divisor as `vramBudgetMb`** — the two are compared, and mismatched rounding would put a systematic bias into `budget_exceeded_pct`. Residual: a flip within 1 MiB of the budget, 0.004% of a 24 GiB card. Stored as its own series and clearly labelled apart from the adapter-wide figure from `18_GPU_VENDOR_APIS`. Aggregates: avg, max, and `budget_exceeded_pct` (share of samples where `CurrentUsage > Budget`, i.e. the driver was likely evicting — a genuinely useful stutter explanation).~~
## Input — what a session was played with, both tiers (owner decisions D155–D163)
**Owner request, 2026-11-14:** the share of a session played with a controller and with the keyboard, how often a
controller connected and dropped, the battery state that lasted longest, the switches between makers' controllers.
What is read and what never is: [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md). The arithmetic is
`Domain.Input` and `Application.Recording.InputSessionBuilder`. **Not built end to end yet** — nothing reads a device
until the pull requests after the one that wrote this section; until then these are definitions under test.
**What it can say, and what it cannot.** Everything is read from outside the game. So a number here says that a
controller, or the keyboard or mouse, was *in use while the game's window was in front* — never what the game did with
it, and never which button or key.
**The unit is a slice** of about a tenth of a second, and a slice counts for a session only when a window of the
session's game was in front and the session was not paused (a pause is a gap, as for frames). A slice with another
program in front is not classified: typing in a browser beside the game is not the game played with the keyboard.
| Class | A slice in front in which |
|---|---|
| Controller | a controller was in use and the time of the last keyboard-or-mouse input did not move |
| Keyboard & mouse | that time moved and no controller was in use. One class: only a time is read (D158) |
| Both | both |
| No input | neither |
- **A controller is in use** when a button is down, a trigger is past its threshold, or a stick is outside its dead
zone. A stick held counts. One that drifts inside the dead zone does not — a pad lying on a desk is not played with.
- **A share** is a class's time over the time the game was in front (`input_foreground_ms`). The four add up to it.
- **A switch of method** is a change between *controller* and *keyboard & mouse* from one tick (a second) to a later
one. A tick played with both, or with neither, is no one thing and does not end the run before it: putting the
controller down and picking it up again is not a switch.
- **Connections** are counted from the session's second slice: a controller that is there when the session starts has
not connected. A Bluetooth pad that goes to sleep has disconnected.
- **A battery** is timed by the clock, in front or not: the time at each level, the level that lasted longest, the
first and the last. Each source keeps its own steps — XInput's four words, a DualSense's per cent in tens — and none
is converted into another's. **A rate**, in per cent an hour, is the fall from the first to the last reading on
battery over the time read on battery; it is given only for a level in per cent that fell, was not put on power in
between, and was read on battery for ten minutes or more.
- **A maker's share** is the time its controllers were in use over that time for every maker. **A switch of maker**
is counted like a switch of method, between ticks in which one maker's controllers alone were in use.
- **One pad seen twice is one controller.** With Steam Input in front of a controller Windows shows the controller the
player holds and an Xbox pad Steam presents for it. One of XInput's pads that was in use at the same moments as
another controller — nine tenths of the busier one's time, over ten seconds or more — is folded into it: into a
controller that is not XInput's at any time, into another of XInput's pads only in a session in which Steam's
virtual gamepad was seen. Two players pressing in step are two players.
**The timeline** is one series, `input.class`: what each tick was mostly played with, as the class's number (0 no
input, 1 controller, 2 keyboard & mouse, 3 both), and no value for a tick with the game not in front. Each of the
first eight controllers has `input.dev.N.active` (the part of the tick's time in front it was in use; no value where
it was not connected) and, with a battery that has a level, `input.dev.N.battery`.
**Known limits.** Keyboard and mouse cannot be told apart. Input a program injects counts as keyboard and mouse — Steam
Input mapping a pad to keys, a macro tool — so a pad mapped that way reads as *both*. A game whose window belongs to
another process reads as never in front (the focus sample's limit). Whether a controller's own input moves Windows'
last-input time has not been checked on a real PC.
## Sensor aggregates
Per session over 1 Hz samples: `avg` (mean of non-null), `max` — and since beta.12 the `median` (the linear-interpolation
@@ -1004,7 +1073,7 @@ processor's, not the whole PC's — averaged over the session's sensor ticks. `D
| RT active | measured per frame | not available |
| Path tracing | heuristic, confidence-scored, never asserted | not available |
| ~~Per-process VRAM~~ Game process memory (beta.12) | ~~exact~~ dedicated / shared GPU memory as Task Manager counts it — a known 512 MiB allocation read +512.0 MiB (`spike-notes` §16); private working set equal to Windows' own counter; read from outside the game | ~~not available~~ the same, from outside; an unpinned hold sums the executable's processes |
| PC latency | as reported by Reflex | not available |
| PC latency | ~~as reported by Reflex~~ **not available** (D147, *corrected 2026-11-14*): the driver's report ends at the end of the GPU's work, not at the display. The stages it does give are Advanced metrics — the driver's numbers, unchecked | not available |
| Display mode (beta.10) | per 100 ms sample: exclusive fullscreen as the swap chain answers it (DXGI; OpenGL and Vulkan cannot say), borderless, windowed, minimised; the window's size from user32 | per 1 s sample, from the window alone: windowed, minimised, or *fullscreen or borderless* — never split |
| GPU temp / load / power | vendor API accuracy, ±1 s sampling | same |
| GPU clocks / fan / memory temperature (beta.14) | vendor API accuracy, ±1 s sampling | same |
+8 -3
View File
@@ -26,6 +26,9 @@ public interface IFrameSource : IAsyncDisposable
> DrainResult Drain(Span<FlFrameRecord> into, IList<ulong> gapIndices);
> void PublishGuardResult(uint completedEvaluations, bool unhookRequested);
> void SetPaused(bool paused); void RequestLogFlush();
> FlAdvancedState? AdvancedState { get; } void SetAdvancedRequested(bool requested); // region 5, train B (2026-11-12): the
> // word is written either way as the ring is attached, before the first guard result; the region is read only by
> // a session whose observer WantsInGameAdvanced, and carried on CaptureProgress.InGameAdvanced
> }
> public sealed class CaptureSession(...) // the loop: gate → inject → attach → drain, until something ends it
> {
@@ -347,6 +350,7 @@ What the Agent checks **before** asking the guard is the thing the native side s
|---|---|---|---|
| **Session loop** (`CaptureSession.RunAsync`, one async loop per session) | `Task.Delay(100 ms)` drain ticks; the 30 s guard scan is awaited inline | The ring reader, for the life of the session — drain, `PublishGuardResult`, `SetPaused`, `RequestLogFlush` — and the record buffer it fills | Samples foreground and the display mode per tick (beta.10), drains the telemetry queue per tick (PR-D), writes `.partial` itself (PR-D) |
| **Telemetry** (`fl-telemetry`, `TelemetryPoller`, one per session) | 1 Hz, never faster than 500 ms | Its own `ConcurrentQueue` of `TelemetrySample` | Reads the composite source only; L2 has a thread of its own inside the library, and the composite is read from this thread and no other |
| **Input** (`fl-input`, `InputDeviceHub`, **one for the Agent**, since 0.1.0-beta.50) | A slice every 100 ms, only while a session has a view of it — with none there is no thread, no window, no registration | The message-only window raw input is delivered to, and the registration for joysticks and gamepads; the reader's device table | Raw input's buffer, three XInput getters, `GetLastInputInfo`, the window in front. Hands each slice to every session's view under one lock; the telemetry thread takes a view's sum once a tick. Opens no device, sends nothing to one (`21_INPUT_DEVICES`) |
| **Watcher** (built 2026-09-10, PR-F: `CaptureOrchestrator.RunAsync` on a `PeriodicTimer(1 s)`) | 1 Hz process snapshot | The watcher's events and the running-session table — ~~`PollOnceAsync` is their only writer; there is no `Channel`, because there is no second consumer~~ **since 2026-09-13 (P3 PR-1b) the table has one lock**: `PollOnceAsync` on the timer, `LaunchAsync` and `StopSession` from the pipe's tasks (HANDOFF §P3 D12 as built — two commands do not justify a channel); the sessions themselves never take it | The `games` rows (one read per poll) and never the ring; each session it starts runs on its own task |
| **Finalize** (PR-D) | Once, on the session loop's task | The one SQLite connection, behind a `SemaphoreSlim(1)` | `ApplicationStopping` cancels the loop; finalize gets a grace window, then the `.partial` stays for recovery |
| **Pipe** (built 2026-09-13, P3 PR-1: `PipeServer.RunAsync` accepting, one reader and one writer task per client) | Per frame | Each client's bounded outbound queue (drop-oldest, so a publisher never waits) and the answer to each request | `Hello` / `GetStatus` / `Ping` read `AgentStatus`, an immutable snapshot the publisher swaps whole; **never the ring, never the recorder**. The events themselves are handed over ON the session loop's task by `SessionEventPublisher` (`ISessionObserver`, called after the recorder's own work each tick) — the loop enqueues a frame and returns |
@@ -369,7 +373,7 @@ the pipe's tasks touches the ring or the recorder, and the session loop never wa
## Telemetry poller
1 Hz (the session's `telemetry.interval_ms`, 500–2000 ms) on its own thread: `CompositeTelemetrySource.TryRead` (`18_GPU_VENDOR_APIS` — DXGI/PDH baseline, LibreHardwareMonitor sensors, NVAPI extras on NVIDIA) ~~+ optional CPU sample when the Agent is elevated and PawnIO is present~~ + *(corrected 2026-10-04)* the machine's reading on the same tick — CPU load and memory in use without elevation, CPU temperature only when elevated with PawnIO — and the game process's memory (beta.12, D43). Never called from the game process, never faster than 500 ms. Samples carry the session QPC epoch for timeline alignment.
1 Hz (the session's `telemetry.interval_ms`, 500–2000 ms) on its own thread: `CompositeTelemetrySource.TryRead` (`18_GPU_VENDOR_APIS` — DXGI/PDH baseline, LibreHardwareMonitor sensors, NVAPI extras on NVIDIA) ~~+ optional CPU sample when the Agent is elevated and PawnIO is present~~ + *(corrected 2026-10-04)* the machine's reading on the same tick — CPU load and memory in use without elevation, CPU temperature only when elevated with PawnIO — and the game process's memory (beta.12, D43). **Since 0.1.0-beta.50 (D155, D156) also what the session was played with**: when the option `capture.input_devices` is on and its terms are accepted, the poller opens a view of the Agent's input reader at its start, tells it which processes are the session's game (`Follow`), and attaches to each sample what that view saw since the tick before — `TelemetrySample.Input` (`21_INPUT_DEVICES` §The reader). A tick only the input view answered is still a sample. Never called from the game process, never faster than 500 ms. Samples carry the session QPC epoch for timeline alignment.
## Session recorder — state machine
@@ -428,7 +432,8 @@ because a layer cannot leave a running game's loader chain.
> - **Recovery never throws.** A file it cannot finalize is set aside as `<guid>.partial.failed` (never retried, kept for
> a bug report) and reported `Failed`; the next file is recovered, and the watcher starts either way.
> - The Agent's host logs through Serilog since the same day: a background service that stops the host is now in
> `agent-*.log`, where before it was in no file at all.
> `agent-*.log`, where before it was in no file at all. *(Since beta.37, on .NET 11, that service's exception also
> ends the Agent through its crash report instead of an exit 0 — `10_LOGGING` §Crash handling.)*
> **Built 2026-09-10 (P2 PR-D): `Application.Recording`.** The state machine above is the intent; what
> runs is one class per box. `SessionRecorder` owns a session end to end — the row's identity and time
@@ -536,7 +541,7 @@ Crash-within-60s-of-injection happening twice for the same game ⇒ **hooking au
|---|---|
| Game process, per present | ≤ 1 µs; no syscall, no allocation, no lock, no logging (`17_HOOK_ENGINE`) |
| Game process, resident | ≤ 8 MB (DLL + ring) |
| Agent, during capture | ≤ 1% of one core average, ≤ 150 MB working set |
| Agent, during capture | ~~≤ 1%~~ ≤ 2.5 % of one core average *(NFR-2 as D117 amended it, 2026-11-11: measured 1.93 %, `20_OPEN_QUESTIONS` §M16)*, ≤ 150 MB working set |
| Agent, idle | ~0% (1 Hz process poll only) |
| Measured game FPS impact | ≤ 0.5% vs uninstrumented, verified per release (`14_TESTING` §Hook overhead) |
+34 -5
View File
@@ -22,8 +22,8 @@ Derived in the Agent from the record stream; the mapping is in `03_METRICS`. Sum
| Render resolution vs output resolution (incl. mid-session changes) | upscaler parameter reads + `ResizeBuffers` |
| Frame Generation mode + factor | NGX/SL/FFX FG feature evaluation, then cadence — **Tier 1 only**. ~~then (Tier 2 only) PresentMon `FrameType`~~ retired (§S31 row P2), and Tier 2 no longer produces frames to run cadence over either |
| Ray Tracing active | AS builds + `DispatchRays` (both, to catch inline RayQuery) |
| Ray Reconstruction | NGX `RayReconstruction` feature evaluated |
| Reflex on + PC latency | NVAPI Reflex hooks |
| Ray Reconstruction | ~~NGX `RayReconstruction` feature evaluated~~ Streamline's `slEvaluateFeature(kFeatureDLSS_RR)` — it runs **instead of** super resolution, in DLSS's own quality mode (Streamline's guide; measured 2026-08-15), so the upscaler is DLSS and this is a yes/no beside it. A game that calls NGX directly reads N/A (`03_METRICS` §RT/PT/RR) *(corrected 2026-11-14)* |
| ~~Reflex on + PC latency~~ The driver's latency report and sleep status (0.1.0-beta.47, Advanced metrics only) | ~~NVAPI Reflex hooks~~ two read-only getters of the driver, called by the Overlay (`17_HOOK_ENGINE` §Calls the Overlay makes into a driver). **PC latency — input to display — is not read** (D147): the report ends where the GPU's work ends, and NVIDIA's own figure is taken from outside the game (`spike-notes` §30, route E) *(corrected 2026-11-14)* |
| Per-process VRAM | `QueryVideoMemoryInfo` |
| PSO compilation events | pipeline-creation hooks |
| Display mode — exclusive fullscreen / borderless / windowed / minimised, with the window, back buffer and monitor (beta.10) | `GetFullscreenState` + `GetDesc` polled on the presenting chain (≤ 2 Hz, shared-memory region 4), and the window's geometry read by the Agent **out of process** — so this row is **both tiers**: a Tier-2 session has the window alone, and never splits exclusive from borderless (`03_METRICS` §Display mode) |
@@ -482,12 +482,27 @@ never the path), and the super-resolution preset the driver reports applied:
|---|---|---|
| Super-resolution render preset **L or M** that the NVIDIA driver reports applied, in a session where super resolution ran | DLSS 4.5 (2nd-gen transformer) | `ngx_driver_words` — the SR word's `PRESET` bit and `renderPreset`; presets L/M arrived with DLL 310.5.0 (NVIDIA's DLSS 4.5 release, driver 591.74) |
| A DLSS Frame Generation state measured at **×5 or more** | DLSS 4.5 (Dynamic / 6× multi-frame generation) | `fg_mix` / `fg_factor` with `fg_mode = dlssg` |
| The NVIDIA App's **Dynamic** frame generation — set in the game's profile and reported applied by the driver — in a session where DLSS Frame Generation was identified *(beta.34, owner decision D83)* | DLSS 4.5 (Dynamic multi frame generation, set in the NVIDIA App) | `driver_profile`'s `NGX_DLSSG_MODE` = 4 (dynamic), `ngx_driver_words`' FG word with `FG_MODE` (bit 13), `fg_mode = dlssg` |
| Preset **J or K** reported applied | DLSS 4 (transformer) | as above |
| A DLSS Frame Generation state at **×3 or ×4** | DLSS 4 (multi-frame generation) | as above |
| `nvngx_dlss.dll` **2.x** | DLSS 2 | `runtime_modules` |
| `nvngx_dlss.dll` **3.x**, or `nvngx_dlssg.dll` **1.x–3.x** | DLSS 3 | `runtime_modules` |
| A **310.x** library alone, a CNN preset (A–F), ×2 | *no label* — the versions are shown | DLSS 4 and 4.5 ship the same major number; the preset is what names the model |
| `nvngx_dlssnr.dll` loaded | "DLSS 5 runtime loaded" — never a generation of these features | the module name is community-sourced until owner-only item 20(d) sees it load |
| ~~`nvngx_dlssnr.dll` loaded~~ | ~~"DLSS 5 runtime loaded" — never a generation of these features~~ | ~~the module name is community-sourced until owner-only item 20(d) sees it load~~ *(replaced 2026-11-06 by the two rows below, D84)* |
| DLSS 5's neural rendering **evaluated** — `kFeatureDLSS_NR` (Streamline 2.14) on some of the frames the game rendered through Streamline, seen by an Overlay that decodes it *(beta.34, owner decision D84)* | DLSS 5 (neural rendering on N % of rendered frames) — **heads the table** | `dlss_nr_pct` (schema 0022), from `FL_FEAT_SL_NEURAL_RENDERING` under `FL_HOOK_SL_NEURAL_RENDERING` |
| `nvngx_dlssnr.dll` loaded, neural rendering not seen evaluated | "DLSS 5 runtime loaded" — and "…, not seen running" where an Overlay that decodes it drained Streamline batches and none carried it (0 %); never a generation | `runtime_modules` (the name seen on Hogwarts Legacy, 2026-11-06) and `dlss_nr_pct` |
**Amended 2026-11-06 (beta.34, owner decision D83): Dynamic is DLSS 4.5's.** The NVIDIA App's *Dynamic* frame
generation and its ×6 mode exist only in DLSS 4.5, so a session the driver reports running Dynamic is DLSS 4.5 whatever
multiplier the mode chose: the owner's Phantom Blade Zero under *Dynamic, up to ×3* held ×2.0 and ×3.0 — nine hooked sessions
from 2026-11-03 to 11-05 — and read *DLSS 4 (a ×3.0 frame-generation state)* until then. The row needs all three: the
profile's Dynamic (`NGX_DLSSG_MODE` = 4, NVIDIA's own name for the value), the driver's report that its frame-generation
override applied in the game's process (the FG word's `FG_MODE` bit), and DLSS Frame Generation identified in the session
(`fg_mode = dlssg`; its factor need not have been counted) —
a profile set to Dynamic that the driver never applied is a setting, not evidence. The driver's own mode number is not
read: it answers 3 under Dynamic and 1 under *Fixed* on every stored session (`spike-notes` §26), and NVIDIA's header
names neither. A *Fixed* ×5 or ×6 is not this row — Witcher 3 under the App's Fixed ×6 still presented ×2.0 — and
names DLSS 4.5 only where a state is measured at ×5 or more, which ranks above it.
Never a default preset inferred from a version: what a DLL "would" run by default is the game's and the driver's to
decide. A preset the driver applied to a super resolution it never saw created and evaluated names the model of
@@ -495,7 +510,21 @@ nothing and is not used (the owner's Silent Hill 2 session: preset L applied, no
the ×6.0 state's). **Sources:** NVAPI's `NV_NGX_DLSS_OVERRIDE_GET_STATE_PARAMS_V1` (the vendored
`src/native/third_party/nvapi/nvapi.h`); presets L/M with DLL 310.5.0 — PCGH's DLSS 4.5 test and NVIDIA's driver
591.74 release notes; DLSS 4.5's Dynamic MFG and 6× — NVIDIA's announcement (driver 595, RTX 50); DLSS 5 — NVIDIA's
GTC 2026 announcement and its 2026-09-03 launch; `nvngx_dlssnr.dll` — community reports only.
GTC 2026 announcement and its 2026-09-03 launch; ~~`nvngx_dlssnr.dll` — community reports only~~ `nvngx_dlssnr.dll` —
loaded by the owner's Hogwarts Legacy with Neural Rendering on, 310.9.16.0 from the game's folder (2026-11-06,
`spike-notes` §26), beside its Streamline plugin `sl.dlss_nr.dll`; `kFeatureDLSS_NR = 1004` — Streamline `v2.14.16`'s
public `sl_core_types.h` (re-vendored for it, `src/native/third_party/streamline/README.md`).
**Amended 2026-11-06 (beta.34, owner decision D84): DLSS 5 is named where it ran.** Until then a session could say only
that DLSS 5's runtime was LOADED — and Hogwarts Legacy, Neural Rendering on, loaded it. The Overlay's
`slEvaluateFeature` detour now decodes `kFeatureDLSS_NR` into its own per-frame fact (`FL_FEAT_SL_NEURAL_RENDERING`,
the feature byte's last bit, under the condition every Streamline fact shares) and publishes
`FL_HOOK_SL_NEURAL_RENDERING` with the detour; the Agent stores the share of Streamline batches that carried it
(`dlss_nr_pct`) **only from a writer that published that family** — an older Overlay drains the same batches and never
sets the bit, and its silence must not read as 0 %. Above every other row because DLSS 5 is named for that feature as
DLSS 4.5 is for its own: the session's super resolution, ray reconstruction and frame generation are the libraries the
line lists after it. Whether the game routes neural rendering through `slEvaluateFeature` at all is owner-only item
30(a)'s run; a title that does not reads *runtime loaded*, never *not running*.
## Anti-cheat pre-scan (static)
@@ -529,7 +558,7 @@ Three things this section previously implied that are not true, and are worth st
> the user's and stays; a `detected` entry is refreshed; **an empty field with no entry has nothing to protect
> and is filled and badged `detected`** — without that clause a freshly added game could never receive an
> engine at all, since `EnsureAsync` writes no provenance. A null detected value erases nothing.
> `capability_flags` is written whole every run (a JSON array of the rule ids — `dlss`, `dlss_g`, `dlss_rr`,
> `capability_flags` is written whole every run (a JSON array of the rule ids — `dlss`, `dlss_g`, `dlss_rr`, `dlss_nr` (beta.34),
> `streamline`, `fsr`, `xess`, `xefg`) because the Supports row has no user-edit surface and a stale "Supports
> DLSS-G" after the DLL left is the confusion §Capability hints exists to prevent.
- Strings scans bounded to 8 MB, local only.
+157 -2
View File
@@ -2,7 +2,7 @@
`%LOCALAPPDATA%\FrameLedger\ledger.db`. Pragmas: `journal_mode=WAL`, `synchronous=NORMAL`, `foreign_keys=ON`, `busy_timeout=5000`. `Microsoft.Data.Sqlite` + Dapper; all writes in explicit transactions.
**Writer ownership:** Agent writes `sessions`, `session_segments`, `frame_blobs`, `sensor_blobs`, `hardware_snapshots`, the hook-state columns on `games`, **and since P4 PR-1 (2026-09-14) the *detected* columns on `games`** — `engine`, `engine_version`, `platform` under the per-field provenance rule (`05_DETECTION` §Caching: a `user` field is never touched), `capability_flags` whole, and the cache key `detection_rules_version` / `detection_exe_size_bytes` / `detection_exe_mtime_ms`, and since schema 0011 (2026-09-25) the executable's own facts `exe_machine` / `exe_file_version` / `exe_product_version` / `library_versions`, written whole like the flags — through `IGameRepository.ApplyDetectionAsync` only. The UI and the Agent were both already writers of `games`; this adds columns to the Agent's half, not a third writer (HANDOFF §P4). **And since P4 PR-4 the UI's library import writes `platform`, `store_id` and `game_version` through `ApplyStoreMetadataAsync`** — the same per-field provenance rule, badged `detected` (a store's record is not the user's typing), on columns the UI already owned. **And since 2026-09-23 the Agent's relocator merges two entries for one executable under two drive letters** (`IGameMerge`, HANDOFF D26, `19_SAFETY` §A moved drive is the same executable): in one transaction it re-parents the dropped entry's `sessions` to the survivor (before the delete — `sessions.game_id` cascades), carries a block or a crash auto-disable still in force to the survivor (`hook_enabled` only ever to 0; `hook_blocked_reason` is copied, never originated or cleared), adds the crash counts, deletes the dropped `games` row, and then moves the survivor to the file's path (after the delete — `exe_path` is UNIQUE). It was the one Agent write that deletes a `games` row until beta.19; it never touches a consent column. **And since beta.19 (2026-10-12, D55) the Agent removes an entry whose game is no longer installed** (`IGamePrune`, `SqliteGamePrune`, asked by the detection sweep through `UninstalledEntryPrune`): one `DELETE` that holds every stored condition in the same statement — the same row at the same path, `removed_at` NULL, no session stored (`NOT EXISTS` on `sessions` — nothing cascades), no `hook_blocked_reason`, `hook_prescan_state` not `blocked`, no `ac_exception_at` and no `ac_exception_trial_failed_at` — behind `retention.remove_uninstalled_games`; a block, a grant and D38's mark are never removed with their entry (`19_SAFETY` §What a finding does to the game, *What clears it*). **And since 2026-09-23 the UI writes `record_sessions`** (schema 0009, FR-1.6): the entry's recording switch, which the Agent only reads — off, the watcher does not watch for the program. The script's one data change switches it off for entries an old import made for Steam's own tools (`SteamLibrarySource.KnownTools`, the ids as of the script); a twin merge keeps it off when either twin had it off. **And since 2026-09-25 the Agent's pre-scan of the library (`19_SAFETY` §What a finding does to the game) writes the hook-state columns of every entry, not only of those the user asked to hook** — `hook_prescan_state`, and on a finding `hook_enabled = 0` and `hook_blocked_reason` — and its own key, `hook_prescan_rules_version` / `hook_prescan_exe_size_bytes` / `hook_prescan_exe_mtime_ms` (schema 0010), through `IGameConsentStore.RecordPreScanAsync` only: columns that were already the Agent's, never over a block, never a new row. **And since 2026-09-26 (schema 0014, D33) the Agent writes the user-mode exception's columns** — `ac_exception_*` on `games` through `IGameConsentStore` only (eligibility from the sweep, a grant from the `SetAntiCheatException` command, every end), and `sessions.ac_exception_family` with the session — the grant only after the user accepted its disclosure in the UI, which asks and never writes. **And since 2026-10-03 (beta.11, D40) the Agent deletes sessions on request** — every session of one game or of every game, `DeleteSessions` over the pipe, through `ISessionRepository.DeleteSessionsAsync` in one transaction (segments, blobs and annotations cascade), never while one is being recorded; the UI asks and never deletes a session row itself *(for Delete all sessions; FR-1.4's removal is the App's own cascade, above — corrected 2026-10-04)*. **Maintenance (P4 PR-7) adds no writer:** the retention sweep on demand is the Agent's, asked over the pipe; the UI's `VACUUM` rewrites pages with the same contents and changes no row (§Retention built note). UI writes `games` (user-editable fields — since P3 PR-3 that is the FR-1.3 metadata, the `*_default` tri-states, and `removed_at`; ~~never a `hook_*` column~~ **and two downgrades** *(corrected 2026-10-04)*: *Change executable* (2026-09-21, `GameLibrary.ChangeExecutableAsync`) turns hooking off, clears the consent and pre-scan columns and ends an exception grant — it never writes a block or a grant — and *Remove game ▸ DeLine truncated
**Writer ownership:** Agent writes `sessions`, `session_segments`, `frame_blobs`, `sensor_blobs`, `hardware_snapshots`, *(since schema 0023, 2026-11-11)* `machine_profiles` and `session_advanced`, the hook-state columns on `games`, **and since P4 PR-1 (2026-09-14) the *detected* columns on `games`** — `engine`, `engine_version`, `platform` under the per-field provenance rule (`05_DETECTION` §Caching: a `user` field is never touched), `capability_flags` whole, and the cache key `detection_rules_version` / `detection_exe_size_bytes` / `detection_exe_mtime_ms`, and since schema 0011 (2026-09-25) the executable's own facts `exe_machine` / `exe_file_version` / `exe_product_version` / `library_versions`, written whole like the flags — through `IGameRepository.ApplyDetectionAsync` only. The UI and the Agent were both already writers of `games`; this adds columns to the Agent's half, not a third writer (HANDOFF §P4). **And since P4 PR-4 the UI's library import writes `platform`, `store_id` and `game_version` through `ApplyStoreMetadataAsync`** — the same per-field provenance rule, badged `detected` (a store's record is not the user's typing), on columns the UI already owned. **And since 2026-09-23 the Agent's relocator merges two entries for one executable under two drive letters** (`IGameMerge`, HANDOFF D26, `19_SAFETY` §A moved drive is the same executable): in one transaction it re-parents the dropped entry's `sessions` to the survivor (before the delete — `sessions.game_id` cascades), carries a block or a crash auto-disable still in force to the survivor (`hook_enabled` only ever to 0; `hook_blocked_reason` is copied, never originated or cleared), adds the crash counts, deletes the dropped `games` row, and then moves the survivor to the file's path (after the delete — `exe_path` is UNIQUE). It was the one Agent write that deletes a `games` row until beta.19; it never touches a consent column. **And since beta.19 (2026-10-12, D55) the Agent removes an entry whose game is no longer installed** (`IGamePrune`, `SqliteGamePrune`, asked by the detection sweep through `UninstalledEntryPrune`): one `DELETE` that holds every stored condition in the same statement — the same row at the same path, `removed_at` NULL, no session stored (`NOT EXISTS` on `sessions` — nothing cascades), no `hook_blocked_reason`, `hook_prescan_state` not `blocked`, no `ac_exception_at` and no `ac_exception_trial_failed_at` — behind `retention.remove_uninstalled_games`; a block, a grant and D38's mark are never removed with their entry (`19_SAFETY` §What a finding does to the game, *What clears it*). **And since 2026-09-23 the UI writes `record_sessions`** (schema 0009, FR-1.6): the entry's recording switch, which the Agent only reads — off, the watcher does not watch for the program. The script's one data change switches it off for entries an old import made for Steam's own tools (`SteamLibrarySource.KnownTools`, the ids as of the script); a twin merge keeps it off when either twin had it off. **And since 2026-09-25 the Agent's pre-scan of the library (`19_SAFETY` §What a finding does to the game) writes the hook-state columns of every entry, not only of those the user asked to hook** — `hook_prescan_state`, and on a finding `hook_enabled = 0` and `hook_blocked_reason` — and its own key, `hook_prescan_rules_version` / `hook_prescan_exe_size_bytes` / `hook_prescan_exe_mtime_ms` (schema 0010), through `IGameConsentStore.RecordPreScanAsync` only: columns that were already the Agent's, never over a block, never a new row. **And since 2026-09-26 (schema 0014, D33) the Agent writes the user-mode exception's columns** — `ac_exception_*` on `games` through `IGameConsentStore` only (eligibility from the sweep, a grant from the `SetAntiCheatException` command, every end), and `sessions.ac_exception_family` with the session — the grant only after the user accepted its disclosure in the UI, which asks and never writes. **And since 2026-10-03 (beta.11, D40) the Agent deletes sessions on request** — every session of one game or of every game, `DeleteSessions` over the pipe, through `ISessionRepository.DeleteSessionsAsync` in one transaction (segments, blobs and annotations cascade), never while one is being recorded; the UI asks and never deletes a session row itself *(for Delete all sessions; FR-1.4's removal is the App's own cascade, above — corrected 2026-10-04)*. **Maintenance (P4 PR-7) adds no writer:** the retention sweep on demand is the Agent's, asked over the pipe; the UI's `VACUUM` rewrites pages with the same contents and changes no row (§Retention built note). UI writes `games` (user-editable fields — since P3 PR-3 that is the FR-1.3 metadata, the `*_default` tri-states, and `removed_at`; ~~never a `hook_*` column~~ **and two downgrades** *(corrected 2026-10-04)*: *Change executable* (2026-09-21, `GameLibrary.ChangeExecutableAsync`) turns hooking off, clears the consent and pre-scan columns and ends an excepLine truncated
## Schema (v2 — hook architecture)
@@ -320,6 +320,9 @@ CREATE TABLE sessions (
-- schema 0021 (2026-10-19, beta.24, D64): each loaded runtime module's folder class — 05_DETECTION §DLSS generation
runtime_module_sources TEXT, -- {"nvngx_dlss.dll":"graphics_driver", …}: game | graphics_driver | windows; never a path
-- schema 0022 (2026-11-06, beta.34, D84): DLSS 5's neural rendering — 05_DETECTION §DLSS generation
dlss_nr_pct REAL, -- 0–100 of the Streamline batches that evaluated kFeatureDLSS_NR; NULL = nothing could tell
-- schema 0020 (2026-10-19, beta.24, D62): every frame-generation state the session held — 03_METRICS §The mix
fg_mix TEXT, -- {"Rows":[{"Kind":"generated","Factor":4.0,"NativeFps":…,"DisplayedFps":…,
-- "Seconds":…,"Share":…,"Windows":…}, …]}: states by rising factor, then
@@ -339,7 +342,25 @@ CREATE TABLE sessions (
display_source TEXT, -- swapchain | opengl | window: the strongest witness any sample had
display_window_w INTEGER, display_window_h INTEGER, -- client area, in the mode that lasted longest (the last seen, beta.27)
display_buffer_w INTEGER, display_buffer_h INTEGER, -- the swap chain's back buffer, same mode
display_monitor_w INTEGER, display_monitor_h INTEGER, display_monitor_hz INTEGER
display_monitor_w INTEGER, display_monitor_h INTEGER, display_monitor_hz INTEGER,
-- schema 0024 (2026-11-14, beta.50, D155-D163): what the session was played with, both tiers (03_METRICS §Input) -
-- read by the Agent from outside the game. Milliseconds. NULL before, with the option off, and before the terms
-- that describe the reading were accepted - N/A, never 0. From the samples, so a recovered session has them.
input_observed_ms INTEGER, -- the reader ran and the session was not paused
input_foreground_ms INTEGER, -- of it, a window of the game was in front: the four below add up to this
input_controller_ms INTEGER, -- a controller in use, the keyboard and mouse not
input_km_ms INTEGER, -- the keyboard or mouse in use (one class - only a time is read), no controller
input_both_ms INTEGER,
input_idle_ms INTEGER, -- no input at all
input_method_switches INTEGER, -- changes between a controller and the keyboard and mouse
input_source TEXT, -- the sources that answered: rawinput, xinput, lastinput, joined by +
input_devices TEXT -- InputDevicesRecord's JSON: {"V":1,"Devices":[{"Index","Family","VendorId",
-- "ProductId","Model","Connection","XInputSlot","Virtual","Remapped",
-- "ConnectedMs","ActiveMs","Connects","Disconnects","Battery":{"Scale",
-- "States":[{"State","Level","Ms"}],"First","Last","DrainPercentPerHour"}}],
-- "Families":[{"Family","ActiveMs","Share"}],"FamilySwitches","VirtualPad",
-- "Health":{...counts...}} - no serial number, no address, no path
);
CREATE INDEX ix_sessions_game_started ON sessions(game_id, started_at DESC);
@@ -396,6 +417,15 @@ CREATE TABLE sensor_blobs (
-- beta.14 (D49) also gpu_core_clock|gpu_mem_clock (MHz), gpu_fan (RPM), gpu_mem_temp (°C)
-- and gpu_power_limit|gpu_thermal_limit (100 or 0 per tick, NVIDIA only) — a series name
-- needs no migration
-- since schema 0023 (2026-11-11, D102) also the ADVANCED series, every one under a name
-- that starts with `adv.` (Application.Advanced.AdvancedMetricCatalog): NaN, not -1,
-- for a tick with no reading, and read by FindAdvancedSensorsAsync only
-- since schema 0024 (2026-11-14, D155-D163) also the INPUT series, every one under a
-- name that starts with `input.`: input.class (what the tick was mostly played with:
-- 0 no input, 1 controller, 2 keyboard and mouse, 3 both), input.dev.N.active (the
-- part of the tick's time in front controller N was in use, 0-1) and
-- input.dev.N.battery (its level, in its own scale) - NaN for a tick with nothing to
-- say, and read by FindInputSensorsAsync only
-- cpu_load / ram_mb / cpu_temp and sessions.avg_cpu_load, avg_cpu_temp, max_cpu_temp,
-- avg_ram_mb were in schema 0001 with NO producer until 2026-09-21 (Telemetry.SystemTelemetrySource
-- on the poller's tick). Rows before that date hold NULL, which is what they measured; no migration.
@@ -461,6 +491,9 @@ switch's rule); integers are invariant-culture within an inclusive range; choice
| `capture.run_elevated` | bool | `0` | | The App when it starts the Agent, and an Agent the logon task started — both BEFORE the Agent exists, through `Infrastructure.Startup.RunElevatedSetting` (one row, `PRAGMA query_only`, no version gate, anything but `1` is off) (D34, beta.10): on, Windows is asked for administrator rights at every Agent start (`01_ARCHITECTURE` ADR-9) |
| `capture.min_session_s` | int | `30` | 5–600 | Agent, per session (FR-3.6) |
| `telemetry.interval_ms` | int | `1000` | 500–2000 | Agent, per session (FR-3.5) |
| `capture.input_devices` | bool | `1` | | Agent, per session (FR-16, owner decisions D155 and D156, 2026-11-14): **Controllers and input**. On — and only once the Privacy Policy that describes the reading (2.22 or later) is in `legal_acceptance` (`Application.Input.InputTerms`) — a session has a view of the Agent's input reader and stores what it was played with (schema 0024); off, or before that acceptance, nothing of it is read or stored, and a session held back by the terms says so in its notes. Any row that is not exactly `0` is on. Settings ▸ Recording (`21_INPUT_DEVICES`) |
| `capture.input_devices` | bool | `1` | | Agent, per session (FR-16, owner decisions D155 and D156, 2026-11-14): **Controllers and input**. On — and only once the Privacy Policy that describes the reading (2.22 or later) is in `legal_acceptance` (`Application.Input.InputTerms`) — a session has a view of the Agent's input reader and stores what it was played with (schema 0024); off, or before that acceptance, nothing of it is read or stored, and a session held back by the terms says so in its notes. Any row that is not exactly `0` is on. Settings ▸ Recording (`21_INPUT_DEVICES`) |
| `telemetry.advanced` | bool | `0` | | Agent, per session (FR-5.6, owner decision D101, 2026-11-11): **Advanced metrics**. On, a session holds the Agent's shared reader and stores its `adv.*` series and its `session_advanced` row (schema 0023); off — and any row that is not exactly `1` — nothing more is read or stored, and a session recorded while off has no row. Settings ▸ Recording (`18_GPU_VENDOR_APIS` §The switch) |
| `retention.raw_sessions_per_game` | int | `20` | 0–10000; **0 = unlimited** | Agent, per session (§Retention) |
| `retention.remove_uninstalled_games` | bool | `1` | | Agent, at every detection sweep (beta.19, D55, FR-1.8): on, an entry whose executable has been gone for ten minutes — its drive connected, the folder it was installed in still there — and that has no session and no anti-cheat block or exception is removed; off, it stays and reads *Not installed* (D41's behaviour) |
| `update.channel` | choice | ~~`stable`~~ `auto` (beta.14, D47) | `auto` · `stable` · `beta` | UI (P4 PR-5 reads it: `beta` = GitHub pre-releases too; `auto` follows the running copy, a pre-release asking for pre-releases — `11_UPDATER` §The update audit) |
@@ -488,6 +521,48 @@ Because Tier-1 and Tier-2 sessions carry different fields, every query that comp
> axis (a gap's interval is stored as 0), so a series still drifts ahead of its sensors past each gap.
> Sensor series are aligned to `t_ms` (milliseconds from `qpc_epoch`) and a tick with no reading is stored as
> **−1** — never 0, which would be a reading — because temperatures, loads, power and memory are never negative.
>
> **A reading taken inside a game stands at the sample that carried it** (0.1.0-beta.49, owner decision D150): the
> Overlay publishes about once a second and the Agent samples exactly once a second, so a publication waits for
> the next sample — two may wait, the older is carried first (`PendingInGameReadings`). An `adv.*` point that
> came from a game can therefore stand up to a second after its read; a tick with nothing waiting is NaN.
>
> **`t_ms` is the time axis; `hz` is the fallback** (written here 2026-11-14, owner decision D146 — the code cited
> this rule before the page had it). Every series of a session is stored beside `t_ms`, which carries each
> sample's own time, and the App draws against it; `hz` is read only for a row that has no `t_ms`
> (`SessionSeriesLoader`). **Until 0.1.0-beta.48 `hz` was the constant 1**, whatever the samples did — and they
> were not one second apart: the poller waited its whole interval *after* a poll's work. In the owner's ledger of
> 2026-11-14, under a setting of 1000 ms, 180 of the 186 sessions long enough to tell have a median of 1.32 s
> between samples; six have 1.0 s. **What takes the third of a second is not established.** Unelevated — the layers
> composed as the Agent composes them, each read under a stopwatch — a poll takes about a millisecond. The two
> sessions with no `cpu_temp` series (an Agent that was not elevated) are among the six, which points at the
> sensor library's processor update, run on the poll's thread as administrator with PawnIO; but four sessions
> that have the series kept one second, and the call itself was not timed (it needs elevation). Since 0.1.0-beta.48 the poller keeps a fixed period
> (`TelemetryPoller.UntilNextTick`) and `hz` is the rate the session's samples had, measured from the samples
> (`SessionFinalizer.SampleRate`: the gaps between the first and the last over the time between them, to a
> thousandth; 1 for fewer than two) — so it is right for a session recovered from a `.partial` file as well.
> Rows written before keep their 1.
>
> **The advanced series are the exception, and say so by their name (schema 0023, owner decision D102).** A series
> whose name starts with `adv.` stores **NaN** for a tick with no reading: a motherboard's −12 V rail is negative, so
> −1 cannot mark a gap, and most of its ticks are gaps — the sensor library is read every fifth. They go through a
> second encoder, `ISeriesCodec.EncodeSparseFloat32`, which allows NaN and still refuses an infinity; **the rule
> above is not relaxed**: `EncodeFloat32` throws on a NaN as before, because a frame time has no gaps. Both decode
> with `DecodeFloat32`. `ISessionRepository.FindSensorsAsync` — what every summary, chart and export reads — leaves
> the `adv.` rows out (`series NOT LIKE 'adv.%'`), since those readers drop every negative value as "no reading";
> `FindAdvancedSensorsAsync` returns them with `t_ms` first. One session stores at most 256 of them
> (`AdvancedSessionBuilder.MaxSeries`); what is beyond is counted in its directory, not stored.
> **Size, measured 2026-11-11** on values shaped like the dev box's reader (17 series with a reading every tick, 41
> with one every fifth): one hour is 816 KiB raw and **189 KiB stored** — well inside NFR-5's 3 MB an hour.
>
> **The input series follow the same rule under their own name (schema 0024, owner decisions D155–D163).** A series
> whose name starts with `input.` stores NaN for a tick with nothing to say — the game was not in front, the
> controller was not connected, the tick carried no reading — through the same sparse encoder. `FindSensorsAsync`
> leaves them out as well (`series NOT LIKE 'input.%'`): a class's number is not a sensor's value, and the sensors'
> statistics and chart would take it for one. `FindInputSensorsAsync` returns them with `t_ms` first. A session
> stores `input.class` and, for each of its first eight controllers (`InputSessionBuilder.MaxDevices`),
> `input.dev.N.active` and — where the battery has a level — `input.dev.N.battery`: at most seventeen, a few KiB an
> hour stored.
## The `.partial` file
@@ -510,6 +585,17 @@ Because Tier-1 and Tier-2 sessions carry different fields, every query that comp
> | 6 `Note` | wall clock + UTF-8 text: a state transition | each transition |
> | 7 `Touches` | N × `i64` QPC at which the host touched the target | each flush |
> | 8 `SensorsWide` | the same samples under a 32-bit presence mask that carries the game process's memory (D43) | each flush, since beta.12 *(row added 2026-10-04)* |
> | 9 `SensorsAdvanced` | the advanced readings of the samples of the type-8 chunk before it (D101): `u16` names, then per name `u16` bytes and UTF-8; `u32` samples, then per sample `i64` QPC, `u16` readings and per reading `u16` name, `f64` value. Each chunk carries its own names, so it reads without any earlier chunk | each flush **in which a sample carries any** — with Advanced metrics off, never; the file is then byte for byte what it was (2026-11-11) |
> | 10 `AdvancedContext` | UTF-8 JSON (`AdvancedJson.OfContext`): the PC's facts, the session's, its series' directory, each source's state; the last one wins | the first flush after the recorder set it, and again only when it changed (2026-11-11) |
> | 11 `SensorsInput` | the input readings of the samples of the type-8 chunk before it (D155–D163, `InputSampleCodec`): `u32` samples, then per sample `i64` QPC, seven `i32` (span, observed, in front, controller, keyboard and mouse, both, no input), `u8` flags, the reader's health (`u8` answers, six `i64` counts, `i32` faults), `u8` controllers and 23 bytes a controller — what it is (maker, the two USB ids, connection, XInput pad) and its part of the tick (connected, time in use, battery, connections, drops). Each sample describes every controller it counts, so a chunk reads without any earlier one | each flush **in which a sample carries one** — with the option off, never; the file is then byte for byte what it was (2026-11-14) |
>
> **A chunk of a type the reader does not know is skipped, not the end of the file (owner decision D106,
> 2026-11-11).** Its length and checksum have been verified by then, so it is a whole chunk a newer build wrote.
> Until that day `_ => false` ended the read there, and a build one release older lost everything after the first
> new chunk — the price types 9 and 10 would have charged a beta.38 reading a beta.39 file. A chunk of a KNOWN type
> that does not parse still ends the read: that is a torn or a wrong file, not a newer one.
> (`PartialSessionFileAdvancedTests`: an unknown type between two known ones; the same with its checksum broken;
> a kill at every byte offset of a file that holds both new types.)
>
> Recovery (`PartialRecovery`): a file whose guid is already a row is deleted (the finalize had landed);
> one under the minimum session length is discarded; one with no readable header is dropped; the rest
@@ -715,6 +801,75 @@ Sequential embedded SQL (`Migrations/0001_init.sql`, `0002_*.sql`, …), applied
> from its crash file (which carries no module list). `LatestVersion` is 21;
> `LedgerDatabaseTests.ScriptTwentyOneAddsTheModuleSourcesColumnAndAnEarlierSessionHasNone`, the repository round trip.
> **`0023_advanced_metrics.sql` (2026-11-11, owner decisions D101 and D102) — what Advanced metrics stores.** Two
> tables and **no column on `sessions`** — the first `CREATE TABLE` since 0001. Every read of a `sessions` row, the
> JSON export and the bug bundle carry the whole row, and a driver profile does not belong in a bug bundle.
> `machine_profiles(id, hash UNIQUE, facts_json, captured_at)` holds the facts of the PC, one row per distinct set
> (`hash` = SHA-256 of `facts_json`); it stands beside `hardware_snapshots`, not in it, because that table's hash
> feeds the trend's hardware-change markers and a set of facts that exists only while an option is on would raise a
> marker each time the option is switched. `session_advanced(session_id PRIMARY KEY → sessions ON DELETE CASCADE,
> machine_profile_id → machine_profiles, facts_json, stats_json, series_json)` holds one row per session **recorded
> with the option on**: its facts and each source's state, the statistics of its advanced series (the form of
> `sensor_stats_json`), and the directory of the series that are members of a family. **A session recorded with the
> option off has no row** — that absence is how the App says "the option was off when this was recorded" rather than
> "nothing answered" — and is never given one. The JSON is `Application.Advanced.AdvancedJson`'s (camel case, nulls
> left out, a status by its name); writing is strict, reading is tolerant, and **only a fact whose catalog entry is
> read may be written** (`AdvancedJson.Storable`), so no identifier reaches either table (D104). **Since
> 0.1.0-beta.40 (train B) the same row holds what the game told its upscaler** — the session facts of the entries
> whose reach is `Game`, after the driver's, and `sources.inGame` (`Reading`, `NothingSeen`, `NotHooked`; absent in
> a row beta.39 wrote, which reads as "not said") — and its two series are `adv.ffx.upscale.frame_time_delta` and
> `adv.ffx.memory_usage` in `sensor_blobs`. **Since 0.1.0-beta.43** `sources` may also hold `streamlineConstants`
> and `fsrFrameGeneration` (`Read`, `NeverCalled`, `NotHooked`; `Application.Advanced.AdvancedInGameCall`) — what
> became, in that session, of the two calls hooked only for Advanced metrics; both absent in a row written before,
> and in a session not measured inside the game. **Since 0.1.0-beta.44** also `streamlineFeatureVersion` and
> `streamlineFeatureRequirements`, the same three values for the two questions a game asks Streamline about a
> feature — where `NeverCalled` is the usual one, a game asking while it starts — and the facts
> `sl.feature.versions` and `sl.feature.requirements`, rows named `<feature> · <member>` in the vendor's words.
> **Since 0.1.0-beta.45** six more keys, the same three values each — `streamlineFeatureSupported`,
> `streamlineFeatureLoaded`, `streamlineSetFeatureLoaded`, `streamlineAllocateResources`, `streamlineFreeResources`,
> `streamlineShutdown`, absent in every row written before — and the facts `sl.feature.support` and
> `sl.feature.lifecycle`, rows named `<feature> · <call>` (`slShutdown` alone for the shutdowns); a count that
> reached its ceiling is the text `≥ 15`. No column and no script: schema 23 as it was. A session deleted
> takes its row with it; a profile no session points to is left, as a hardware snapshot is; the retention sweep
> removes a session's `adv.` series with its other raw series and keeps its row. Written by the Agent only, in the
> session's own transaction (`SqliteSessionRepository.InsertAdvancedAsync`). `LatestVersion` is 23;
> `LedgerDatabaseTests.ScriptTwentyThreeCreatesTheTwoAdvancedTablesAndAnEarlierSessionHasNoRow` (whose helper now
> undoes a script's `CREATE TABLE` and `CREATE INDEX` as well as its `ADD COLUMN`), the repository's four cases.
> **Nothing produces a row yet**: the option that switches the reader on is the next pull request's.
> **A beta.38 cannot open a ledger a later build migrated.**
> **`0024_input_devices.sql` (0.1.0-beta.50, 2026-11-14, owner decisions D155–D163) — what a session was played
> with, and what a store lists.** ADD COLUMN only, twelve columns. **Nine on `sessions`**: `input_observed_ms`,
> `input_foreground_ms` and the four classes that add up to it (`input_controller_ms`, `input_km_ms`,
> `input_both_ms`, `input_idle_ms`), `input_method_switches`, `input_source`, and `input_devices` — the session's
> controllers as `Application.Recording.InputDevicesRecord`'s JSON (`03_METRICS` §Input). **On `sessions` and not in
> a table of their own, on purpose** (D160) — the opposite of 0023's choice, for the opposite reason: the trend and
> the comparison read a session's row, and these numbers are meant to be in the JSON export and the bug bundle,
> which carry the whole row; the Privacy Policy's §3A says so. What may never be a column or a key of that JSON: a
> controller's serial number, its Bluetooth address, its device path
> (`LedgerDatabaseTests.ScriptTwentyFourOnlyAddsColumnsAndNamesNoIdentifier`,
> `InputSessionBuilderTests.TheStoredJsonHasNoKeyThatCouldCarryASerialNumberAnAddressOrAPath`). **Three on
> `games`**: `controller_support` (`full`, `partial`, or `none` for a game the store knows and lists no support
> for), `controller_support_source` (`steam-appinfo`) and `controller_support_at` — the store's own word, from its
> file on the disk, never a verdict of FrameLedger's (D163). Written by `SessionFinalizer.Build` from the session's
> samples, before the tier branch — both tiers, and a session recovered from its crash file, which carries the
> readings in chunk 11. NULL before; with the option off; before the terms that describe the reading were accepted
> (D156); on a game no store file describes. The series are `sensor_blobs` rows named `input.*` (§Blob encoding).
> `LatestVersion` is 24; `LedgerDatabaseTests.ScriptTwentyFourAddsTheInputColumnsAndAnEarlierSessionAndGameHaveNone`,
> the repository round trip. **Nothing produces a reading yet, and nothing writes the three `games` columns yet**:
> the reader, the switch and the store file are the pull requests after this one.
> **A beta.49 cannot open a ledger a later build migrated.**
> **`0022_dlss_nr.sql` (beta.34, 2026-11-06, owner decision D84) — DLSS 5's neural rendering, measured.** One REAL column,
> ADD COLUMN only: `dlss_nr_pct`, the share (0–100) of the presents that drained a Streamline evaluation on which the game
> evaluated `kFeatureDLSS_NR` (`Domain.Metrics.NeuralRendering`). Written by the recorder's aggregation
> (`SessionAggregator.ApplyRt`) **only when the writer published `FL_HOOK_SL_NEURAL_RENDERING`** — an Overlay older than
> the script drains the same batches and never sets the bit, so its NULL is "could not tell", never 0. NULL before, and
> never rebuilt (`frame_blobs` keep no feature byte); a session recovered from its crash file is re-aggregated from its
> raw records and keeps it. `LatestVersion` is 22;
> `LedgerDatabaseTests.ScriptTwentyTwoAddsTheNeuralRenderingColumnAndAnEarlierSessionHasNone`, the repository round trip.
> **A beta.33 or alpha.2 cannot open a ledger beta.34 migrated** (the alpha keeps its copy first, D73).
> **Every open used to migrate, and one that should not have did — 2026-09-16.** The Agent's `--console sessions`,
> a verb that prints, was run against the owner's ledger while the file on disk was at schema 2 and applied 0003 and
> 0004 to it. `LedgerDatabase.OpenReadOnlyAsync` exists since that day: an existing file, `PRAGMA query_only`, no
+122 -5
View File
@@ -14,15 +14,17 @@ in beta.9 without a row here)*:
Created by the Overlay on init, opened by the Agent. Name is deliberately plain and identifiable (`19_SAFETY` — no obfuscated object names).
Layout — **~~four~~ five regions, each exactly one 64-byte cache line before the ring** (layout 4 since
2026-09-27, beta.10: region 4 is new and the ring moved from `0xC0` to `0x100`):
Layout — **~~four~~ ~~five~~ six regions** (layout 4 since 2026-09-27, beta.10: region 4 is new and the ring moved
from `0xC0` to `0x100`; **layout 5 since 2026-11-11**, Advanced metrics train B: region 5 is new — twenty-three cache
lines, where the four before it are one each — and the ring moved from `0x100` to `0x6C0`):
```
[0x0000] FlShmHandshake 64 B write-once by the Overlay at init
[0x0040] FlWriterState 64 B Overlay-written (writeIndex touched every present)
[0x0080] FlControlBlock 64 B Agent-written
[0x00C0] FlDisplayState 64 B Overlay-written, at most every 500 ms per chain (layout 4)
[0x0100] FlFrameRecord[capacity] (64 B each, power-of-two capacity)
[0x0100] FlAdvancedState 1472 B Overlay-written by its watchdog, once a second, only while the Agent asks (layout 5)
[0x06C0] FlFrameRecord[capacity] (64 B each, power-of-two capacity)
```
The header is split in two on purpose. The immutable handshake and the
@@ -72,7 +74,8 @@ struct alignas(64) FlControlBlock { // Agent-written
uint32_t overlayEnabled; // @8 in-game overlay draw toggle (v1.1)
uint32_t guardTicks; // @12 completed guard evaluations, NOT a timer
uint32_t logFlushRequested; // @16 a COUNTER: bumped by the Agent at session end; the watchdog flushes the native log on a change (17_HOOK_ENGINE §Native logging, 2026-09-06)
uint32_t reserved[11]; // @20..63 must be zero
uint32_t advancedRequested; // @20 0, or this session's request for Advanced metrics (D111): non-zero, new to this ring; written at attach, either way
uint32_t reserved[10]; // @24..63 must be zero
};
struct alignas(64) FlDisplayState { // Overlay-written (layout 4, 2026-09-27, beta.10)
@@ -93,9 +96,122 @@ static_assert(sizeof(FlShmHandshake) == 64);
static_assert(sizeof(FlWriterState) == 64);
static_assert(sizeof(FlControlBlock) == 64);
static_assert(sizeof(FlDisplayState) == 64);
static_assert(FL_SHM_DISPLAY_OFFSET == 0xC0 && FL_SHM_RING_OFFSET == 0x100 && FL_SHM_RING_OFFSET % 64 == 0);
static_assert(sizeof(FlAdvancedState) == 1472); // region 5: fl_shm.h is its declaration, member by member
static_assert(FL_SHM_DISPLAY_OFFSET == 0xC0 && FL_SHM_ADVANCED_OFFSET == 0x100 && FL_SHM_RING_OFFSET == 0x6C0 && FL_SHM_RING_OFFSET % 64 == 0);
```
**Region 5 — what a hooked game told its upscaler (layout 5, 2026-11-11; Advanced metrics train B, owner decisions
D111 and D113; written by the Overlay's watchdog since 0.1.0-beta.40).** `18_GPU_VENDOR_APIS` §Train B is the design.
What the region is, as a contract between the two sides:
- **Off means nothing is read and nothing is written.** `FlControlBlock.advancedRequested` is the Agent's word that
this session records Advanced metrics — written at attach, either way, because the control block outlives a session
(`ShmRingReader.SetAdvancedRequested`). While it is 0 the Overlay reads no field and installs no hook beyond what
it did before, and writes nothing here: a ring nobody ever asked is zero from end to end (`fl_guard` checks every
byte, with a target that sends everything), and `validMask == 0` means "nothing was read", never "the game used
none of it".
- **"On" is a request, not a 1** *(2026-11-12)*. A game outlives a session, and so do the summaries the Overlay keeps
inside it — a range, a count, a mask of what was ever seen. So the word is a value no session of this ring used
before (above the word as it stands and above the request the region last answered; never 0), the watchdog starts
its summaries afresh at a value it has not seen, and every publication names the request it answers
(`FlAdvancedState.request`, the word that was `reserved0`). The reader returns only a publication that names the
value it wrote: what a second session of the same game finds in the region before the first tick is the first
session's, and is not read. After an "off" the region keeps its last publication, untouched.
- **One writer, once a second — on a grid since 0.1.0-beta.49 (owner decision D150).** The detours keep
process-local accumulators; the watchdog copies them here at its **periodic** tick, one a second by the
performance counter (`fl_watchdog_tick.h`). Until then it published at *every* tick, and a tick was a second
and its own work after the one before — or at once, whenever the LoadLibrary detour woke the thread. So a
session's publications fell behind the Agent's samples a little with every tick, and a woken tick's publication
took the driver read with it and armed none: the publication after it was empty (0.4 to 16 % of a session's
samples carried no report, HANDOFF item 46; under a loader that never stops, 35 720 publications in eight
seconds and nine reads). A tick woken early still follows the request, installs and takes the census; it writes
nothing here. **A request first seen on a woken tick is published by the periodic tick that follows, however
soon that is**: a session's first publication may hold a moment's worth — one dispatch of a range that has two
ends — and every one after it a second's. A reader takes ranges and "seen" words as *since the request*, which
they are, and never reads a single publication as complete. No game thread writes the region, so the seqlock has a single writer: `seq` odd, the fields,
`seq` even.
- **A publication waits for a sample, and is not replaced while it waits** (0.1.0-beta.49, owner decision D150;
the Agent's half). `SessionRecorder` looks at the region and at the telemetry poller about ten times a second
and attaches a publication's readings to the newest sample it drains. It used to keep one set and overwrite it:
a publication that landed just after a sample left that sample without a report, and at the next change of
step the set that had waited was overwritten — one sample in five for minutes in the owner's session 223. Two
sets now wait, the older is carried first (`PendingInGameReadings`): a session lacks a report at one sample at
most when the two fall out of step, and loses no reading after it. With the sensor interval above a second
more publications arrive than samples and the oldest is pushed out, as before.
- **The reader keeps its last consistent copy** (`ShmRingReader.AdvancedState`): null until something was published
for its request;
the same copy while `seq` has not moved, without copying 1472 bytes again; and that copy still when the writer is
inside on every attempt — a watchdog killed with its process leaves `seq` odd for good, and what it published
before is still what the game said.
- **Zero is "not read", never a value.** A Streamline `Boolean` and a DLSS preset are stored as the vendor's value
plus one; a float as its bits, with `FL_ADV_FLOAT_NONE` for "none seen" — and, in `ffxNearBits` / `ffxFarBits` alone, `FL_ADV_FLOAT_INFINITE` (`0x7F800000`, +infinity's own bits) for a plane at infinity *(0.1.0-beta.42, owner decision D119: an engine with an infinite depth hands FSR `FLT_MAX` or +infinity for one plane; for Streamline `FLT_MAX` is "not set" and stays refused)* — the vendors' own sentinels are refused in
the detour and never reach the region; a size of 0 is "not given".
- **A per-frame value is summarised where it is seen**: a flag is an "ever seen" mask, a jitter the largest magnitude,
a field of view its lowest and highest (D114), the reported frame time a count and a sum the watchdog drains.
- **Sized once for the train's three releases.** The members ~~0.1.0-beta.41 and beta.42~~ ~~0.1.0-beta.42 and
beta.43~~ 0.1.0-beta.43 and beta.44 fill *(renumbered twice on 2026-11-12 — D118: a fix release took beta.41;
D122: the second step became two releases, and beta.42 fills none of them. The comments in `fl_shm.h` and
`ShmLayout.cs` carry the new numbers since beta.42)* (`ffxFg*`,
`sl*`, `features`) are declared now and stay zero until then, so those releases bump no layout. **The `sl*` group
is filled since 0.1.0-beta.43** by the `slSetConstants` hook (`17_HOOK_ENGINE` §Hook inventory), under
`FL_ADV_VALID_SL_CONSTANTS`: the call count of every viewport, and for the first viewport seen the resets, the
jitter's largest magnitude, the motion-vector scale, the two depth planes, the field of view's range, the aspect
ratio, and one bit per `Boolean` in each of two words — seen true, seen false. `FlHookFamily` gains
`FL_HOOK_SL_CONSTANTS` (bit 13): the hook is in, which it is only in a process that was asked at least once.
**The `ffxFg*` group is filled since the same release** by the `ffxConfigure` hook, under
`FL_ADV_VALID_FFX_FG_CONFIG`: the frame-generation configures of the one module counted, three switches as "seen
off / seen on" bit pairs (enabled, asynchronous, generated frames only), every flag any configure carried, and
the last generation rectangle as left, top, width, height — **four `-1` for a rectangle with a negative side or
one past 32,767**, never a clipped one. `FL_HOOK_FFX_CONFIGURE` (bit 14) is published when every ffx-api module
the process loaded has its `ffxConfigure` patched.
- **The feature slots are written since 0.1.0-beta.44**: `features` is twelve slots of twenty-four words — word 0
the `sl::Feature` id plus one (0 is an empty slot), word 1 which of the two questions has an answer (bit 0 the
version, bit 1 the requirements), then the two versions, the flags, three counts and four more versions, as
`fl_shm.h` lays them out. Slots fill from 0 in the order the game first asked about each feature; **a new request
does not empty them** — the one group that is a fact about the process and not about a session — and
`FL_ADV_VALID_SL_FEATURES` stays set with them. `FL_HOOK_SL_FEATURE_VERSION` (bit 15) and
`FL_HOOK_SL_FEATURE_REQUIREMENTS` (bit 16) are published with their patches, in a process that was asked.
- **Word 1 of a slot carries more since 0.1.0-beta.45** (no layout change): bit 2 `slIsFeatureSupported` answered,
with the `sl::Result` it gave in bits 8..15 (0 is `eOk`; a refusal is kept, it is the answer; `0xFF` a code a byte
cannot carry); bit 3 `slIsFeatureLoaded` answered and bit 4 what it said; bit 5 the game switched the feature off
through `slSetFeatureLoaded`, bit 6 on. **Bits 5 and 6 are a session's and a new request clears them**; every
other bit of the word is a fact about the process and stays, as the slot does. A feature Streamline refused has
a slot with only that word in it. `FL_HOOK_SL_FEATURE_SUPPORTED` (bit 17), `FL_HOOK_SL_FEATURE_LOADED` (18) and
`FL_HOOK_SL_SET_FEATURE_LOADED` (19) are published with their patches.
- **And two counts** (the same release): bits 16..19 of word 1 are how many times the game allocated the feature's
resources (`slAllocateResources`), bits 20..23 how many times it freed them (`slFreeResources`) — each stops at
15 — and `slShutdowns` (@308, taken from the three reserved words that ended the `sl::Constants` line; two remain)
counts `slShutdown` calls, with `FL_ADV_VALID_SL_SHUTDOWN` (bit 6 of the mask) set once one was seen. All three
are a session's and a new request clears them. `FL_HOOK_SL_ALLOCATE_RESOURCES` (bit 20),
`FL_HOOK_SL_FREE_RESOURCES` (21) and `FL_HOOK_SL_SHUTDOWN` (22) are published with their patches.
- **Reflex latency words (train C; written by the Overlay since 0.1.0-beta.47; owner decisions D136, D138).** The
ten reserved words that ended the first line (@24..63) are `latState`, `latReads`, `latFrames` and seven sums:
simulation, render submit, present, driver, OS render queue, GPU render, and the total from simulation start to
GPU render end. **The layout is still 5**: no offset and no size moved, and the handshake's build id refuses a
mismatched pair before either side reads the other. Unlike every other group here they are **not what the game
passed to a call the Overlay hooks**: the Overlay makes a call of its own, about once a second, on the presenting
thread, to the driver's two read-only getters, through an `nvapi64.dll` the game had already loaded
(`17_HOOK_ENGINE` §Memory / latency). `latState`'s low byte is a `FlLatencyState` — no device, no module, refused
(a module in the executable's own folder, D140), no getter, reading, error, slow (D141), faulted, stalled — and
above it the driver's sleep status as switches *seen on* and *seen off* and the last frame-generation multiplier.
`latFrames` and the sums are **of the newest read not yet published** — zero when no read was made since the
publication before — so a reader divides one publication's sum by its frames and never adds publications up. A
sum of all ones is a span some summed frame did not report. `FL_ADV_VALID_LATENCY` (bit 7 of the mask) is set once
a read returned OK for the request; it is not one of "the game's calls were read", and `latState` is published
whether or not it is set.
- **Four more latency words (0.1.0-beta.48; owner decisions D142, D144).** The first line was full, so they are
four of the seven reserved words that ended the `sl::DLSSOptions` line (@100..115; three remain, @116..127) —
nothing of DLSS's, and **the layout is still 5** for the same reasons. `latGpuActiveSumUs` and `latGpuFrameSumUs`
are the two times the driver works out itself — the graphics card's render time *excluding the idles in
between* (`gpuActiveRenderTimeUs`) and one frame's GPU end to the next's (`gpuFrameTimeUs`) — summed over the
frames of the newest read that gave one; `latGpuCounts` holds how many did, the first in its low half and the
second in its high half, because a frame can give one and not the other. A frame that gives zero (an older
driver leaves both zero) or more than a second is in neither the sum nor the count, and is still summed for the
seven spans. `latMultipliersSeen` is the session's: bit n is set once the sleep status gave frame-generation
multiplier n since the request, bit 31 standing for 31 and more; `latState` still carries the last one.
- The Vulkan layer never writes it. `ShmLayoutMirrorTests.TheAdvancedStateMirrorMatches` checks every member's
offset against `fl-layout-dump`; `ShmRingReaderTests` the three reads and that the request is stored either way.
**Region 4 — the presenting swap chain's display facts (layout 4, 2026-09-27, beta.10).** The owner asked for the game's
window size and display mode (exclusive fullscreen / borderless / windowed) and the share of a session each took. The two
facts only the process knows — whether the chain is in exclusive fullscreen, and which window it presents to — are
@@ -402,6 +518,7 @@ Unchanged in spirit from v1, bumped to `v2` for the new message set.
| `Shutdown` | UI→A | graceful stop → `ShutdownAck`, then the host stops (sessions finalize inside its grace window). *(Built 2026-09-13, PR-1b: `IAgentLifetime`, the host's under `--serve`, a logged no-op under `--console`.)* |
| `DeleteSessions` | UI→A | **Added 2026-10-03 (beta.11, D40).** `{ gameId? }` — every session of that game, or of every game when it is null, with their segments, frame and sensor series and annotations (they cascade) → `DeleteSessionsAck { gameId?, games, sessions }`. `Error SessionRunning` and nothing deleted while a session of that game — of any game, for all — is being recorded (`CaptureOrchestrator.IsRecording` / `IsRecordingAny`). The client names no rows beyond the game: they are the Agent's (§Writer ownership). Additive: an older Agent answers `Error UnknownType`, which the App says as "restart the agent". Carries no safety fact. |
| `SweepRetention` | UI→A | **no payload** — a trigger, like `UpdateRules`: the Agent applies `06_DATA_MODEL` §Retention to every game with the `retention.raw_sessions_per_game` it reads itself → `SweepRetentionAck { keep, games, sessions }` (`keep` 0 = unlimited, and then nothing was swept). *(Built 2026-09-15, P4 PR-7, for Tools ▸ Database maintenance: `Application.Recording.RetentionSweep` over `ISessionRepository.SweepRetentionAllAsync`, composed into `AgentCommandHandler` under `--serve`. The client names no number because the rows are the Agent's (§Writer ownership). Additive: an older Agent answers `Error UnknownType`, which the dialog reads as "restart the Agent". Carries no safety fact.)* |
| `GetAdvancedMachine` | UI→A | **Added 2026-11-11 (Advanced PR-6, owner decisions D101/D107).** `{ factsHash? }` → `AdvancedMachineAck { ready, takenAtUnixMs, factsHash?, factsJson?, readings[{ series, value }], series[{ series, metricId, label }], nvapi, nvapiCode?, sensors, elevatedSensors }` — what the Agent's shared advanced reader last read about this PC (`18_GPU_VENDOR_APIS` §The page's feed). **The handler reads no sensor**: the answer is the reader's last snapshot, one reference, because the pipe's read loop awaits its handler and a driver call there would hold up `Ping` and `StopSession`; the one thing it awaits is the option, a `settings` row asked for at most every two seconds. **A request is the page's hold on the reader**: the first acquires it and each renews a five-second lease, so the App asks once a second while the page is open and nothing has to say that it closed (`Application.Ipc.AdvancedMachineResponder`). `ready: false` — held, not yet read — carries nothing else. `readings` are that tick's only: a slow source is read every few ticks and the client keeps the last value it saw; a value that is not a number is left out. `factsJson` is the PC's facts as the ledger stores them (`AdvancedJson.OfFacts`), left out when the request's `factsHash` is the Agent's own. The three states are `AdvancedSourceStatus` names; a name a client does not know reads as unknown. `Error AdvancedOff` while `telemetry.advanced` is off — nothing is read, and a hold a previous request took is let go at once. Additive, protocol 2: an older Agent answers `Error UnknownType`, which the page shows as such. Carries no safety fact and changes nothing. |
| `SessionStarted` | A→UI | `{ sessionGuid, gameId, gameName, pid, tier, startedAt, hold? }` — **published once per session**: at the attach (tier 1, `startedAt` the attach), or — **since 2026-09-23** — at the first tick of a session held unhooked (tier 2, `startedAt` the session's start, `pid` 0 when the loop never opened the process, as a hooking-off hold never does). ~~a session that never attaches has no `SessionStarted`, only its `SessionCompleted` (tier 2)~~ — that was the beta.5 wire, and every hooking-off, refused or unreadable game read "Nothing is being captured." for as long as it ran. **`hold` (optional, tier 2 only)** is why nothing is measured: `{ reason, guardReason?, family?, signal?, hookingTurnedOff }` — the loop's `SessionEndReason` and, when the guard said anything but allow, its reason, family and signal: the same words the row's `capture_notes` will carry, so the live card and the summary read one sentence (`Formats.Tier2HoldReason` / `Tier2Reason`). The refusal's own event (`CaptureRefused` and the rest, below) goes out with this start rather than when the game exits, and is not repeated at the end |
| `SessionHeld` | A→UI | **Added 2026-09-23.** 1 Hz while a client listens, for a session held unhooked only: `{ sessionGuid, elapsedS, gpuTempC?, cpuTempC?, gpuLoadPct?, cpuLoadPct?, gameVramDedicatedMb?, gameRamPrivateMb? }` — how long it has run and the machine's sensors (and since beta.12 the game process's own memory, read from outside it — optional, an older client ignores them). **Never `SessionProgress`**: its `presents5s`, `presentedQualifier` and `fgMode` are required, and zeros there would put measured-looking numbers on the wire for a session that measures nothing (rule 6 has nothing to show and shows nothing) |
| `SessionProgress` | A→UI | 1 Hz: `{ sessionGuid, elapsedS, presents5s, presentedFps5s?, presentedQualifier, nativeFps5s?, displayedFps5s?, fgFactor?, fgMode, upscaler?, upscalerQuality?, upscalerDriverReported?, renderW/H?, outputW/H?, rtActive?, gpuTempC?, cpuTempC?, cpuLoadPct?, vramProcMb?, latencyUs?, fgRefusal?, fgRuntimeCensus?, gameVramDedicatedMb?, gameRamPrivateMb? }` (the last two added 2026-10-03, beta.12, optional: the game process's dedicated GPU memory and private working set, MiB, from the newest telemetry tick; `vramProcMb` stays reserved for the in-process figure and has no producer) — a 5 s rolling window over the ring through the SAME Domain calculators the row uses (`SessionProgressCalculator`, D13). **`upscalerQualityOverride` added 2026-10-19 (beta.24, D63), optional:** the mode an NVIDIA App override set when the driver reports it applied in the target and DLSS super resolution ran there — `performance`, `balanced`, `quality`, `ultra_performance`, `ultra_quality`, `dlaa`, or `override` when the driver's report alone does not settle the mode (the profile is the stored session's); absent otherwise. **`upscalerDriverReported` added 2026-09-21, optional:** `dlss` when the NVIDIA driver reports an NGX super-resolution feature created and evaluated in the target (`03_METRICS` §The driver-reported rung), so the live card says `DLSS (driver-reported)` where no hook can name the upscaler; identity only. **`cpuLoadPct` added 2026-09-21, optional** (CPU time busy over the last telemetry tick, 0–100; `cpuTempC` has a producer from the same date, present only when the Agent is elevated with PawnIO). **Two fields added 2026-09-14, both optional (an older client ignores them, §C):** `fgRefusal` is the row's `fg_refusal` (schema 0003) — present when `fgMode` names a technology and `fgFactor` is absent, so the live card can say "DLSS-G active — factor not counted" rather than N/A; `fgRuntimeCensus` is the raw census so the `fg_runtime_loaded` chip can name the module. **Rule 6 at the wire (2026-09-13):** `nativeFps5s` / `displayedFps5s` / `fgFactor` are present only when frame generation was measured in the window; otherwise `presentedFps5s` stands alone with `presentedQualifier` (`census_not_run` · `no_fg_runtime` · `fg_runtime_loaded` · `none_withheld`, the row's `presented_qualifier` from the same function). Every unmeasured field is absent, never 0. Suppressed when no client is connected |
+251 -17
View File
@@ -1,6 +1,9 @@
# 08 — UI
WPF + **WPF UI (Wpf.Ui, Fluent)**. Shell = `ui:FluentWindow` with custom `ui:TitleBar`, a thin classic **Menu bar** (file-level operations), and a left **`ui:NavigationView`** (`PaneDisplayMode="LeftFluent"`) as primary navigation. Light/Dark/System themes via `ApplicationThemeManager` + `SystemThemeWatcher`; Mica backdrop on Windows 11 with automatic solid fallback on Windows 10. All XAML follows `16_WPFUI_SYNTAX.md` — that doc is mandatory reading before writing views.
WPF + **WPF UI (Wpf.Ui, Fluent)**. Shell = `ui:FluentWindow` with custom `ui:TitleBar`, a thin classic **Menu bar** (file-level operations), and a left **`ui:NavigationView`** (`PaneDisplayMode="LeftFluent"`) as primary navigation *(since beta.34, D86, its items use
our copy of WPF UI 4.3.0's LeftFluent item template — `FlLeftFluentNavigationViewItemTemplate` in `Styles/FrameLedger.xaml` —
whose only change is that the selected page keeps its label; WPF UI's animates it away, and the owner's screenshot of
Compare showed the icon alone)*. Light/Dark/System themes via `ApplicationThemeManager` + `SystemThemeWatcher`; Mica backdrop on Windows 11 with automatic solid fallback on Windows 10. *(beta.35, D91: a Windows contrast theme wins over all three while it is on — §Contrast.)* All XAML follows `16_WPFUI_SYNTAX.md` — that doc is mandatory reading before writing views.
## Shell
@@ -35,13 +38,18 @@ WPF + **WPF UI (Wpf.Ui, Fluent)**. Shell = `ui:FluentWindow` with custom `ui:Tit
> a refusal is a message box and exit 2. The viewer's folder is its own single-instance key, so it runs beside the
> profile's App. D6 is unchanged: the Agent's `--data-dir` stays under `--console`.
**NavigationView items:** Dashboard · Games · Compare · Logs — `Settings` lives in `FooterMenuItems`. Icons from `SymbolIcon` (Fluent System Icons; pick from the `SymbolRegular` enum — see 16 §Icons).
**NavigationView items:** Dashboard · Games · Compare · Logs · *Advanced* (since 2026-11-11, D105: last in `MenuItems`, `Visibility` bound to `Services.AdvancedOption.IsOn` — there only while `telemetry.advanced` is on, §Advanced) — `Settings` lives in `FooterMenuItems`. Icons from `SymbolIcon` (Fluent System Icons; pick from the `SymbolRegular` enum — see 16 §Icons).
**Menu bar**
- **File:** Add game… · Import library (Steam / GOG / Epic / itch.io / Xbox *— since beta.33*) · Export ▸ (Session CSV / Session JSON / Chart PNG) · Exit
- **Tools:** Agent status… · Update detection rules · Vulkan layer registration… · Open data folder · Database maintenance…
- **Help:** User guide (beta.12) · Check for updates · Report a bug… · ~~Documentation~~ Online documentation · Limitations · About FrameLedger (~~App / Developer / Third-party licenses tabs~~ — built beta.12 as *This app* / *Third-party licences* / *Legal documents*, below)
- **Help:** User guide (beta.12) · Check for updates · Report a bug… · ~~Documentation~~ Online documentation · Limitations · Accessibility (beta.35) · About FrameLedger (~~App / Developer / Third-party licenses tabs~~ — built beta.12 as *This app* / *Third-party licences* / *Legal documents*, below)
> **Help ▸ Accessibility, beta.35 (owner decision D88, 2026-11-09):** `ACCESSIBILITY.md` — the accessibility statement —
> as the build embeds it (`Services.AccessibilityDocument`), in a `Windows/DocumentWindow` of its own like Limitations,
> and after Limitations in the user guide's window so the guide's links to it open there. Information, not one of
> FR-11's documents. What it says and the checks behind it are §Accessibility below.
>
> **Help ▸ Limitations, beta.10 (D36):** `LIMITATIONS.md` as the build embeds it (`Services.LimitationsDocument`), in a
> ~~read-only, scrolling WPF UI `MessageBox` with *Open on GitHub*~~ window of its own since beta.12 (`Windows/DocumentWindow`,
> rendered — below; corrected 2026-10-04). It is information, not one of FR-11's documents:
@@ -141,7 +149,11 @@ The Menu sits in its own row **below** the TitleBar (never inside the drag regio
## Notifications policy
- **In-app, transient** (rules updated, export done, non-fatal capture warning): `ISnackbarService` snackbar, 4 s.
- **In-app, transient** (rules updated, export done, non-fatal capture warning): `ISnackbarService` snackbar, 4 s — shown
only through `IMessageStrip` (`Services/SnackbarStrip`) since beta.35, which says it to a screen reader as it shows it
(D89; `AnnouncementsTests.OnlyTheStripShowsASnackbarAndNoXamlSetsALiveSetting`).
- **Every one of these is said to a screen reader as it appears (beta.35, owner decision D89)** — §Accessibility, Screen
readers.
- **In-app, persistent** (Agent offline, ~~PawnIO missing,~~ update downloaded, a viewer's folder, the safety notices, *an alpha build* — since 2026-10-31, D74): `ui:InfoBar` banner pinned above content with action button. *(Corrected 2026-10-07: no PawnIO banner was ever built — CPU temperature reads N/A, and Settings ▸ System says it needs the admin mode and PawnIO; the viewer's banner (D52) and the safety notices were missing from the list.)*
- **System-level, app possibly minimized** (session saved with "View" action, capture error while in tray, update available): tray balloon/toast via H.NotifyIcon.
- One event → exactly one channel; never duplicate the same event across snackbar *and* tray.
@@ -285,6 +297,13 @@ All screens are `Page`s registered with the navigation page provider and navigat
### Games (grid → detail)
- `GridView`-style panel of cover cards (`ui:CardAction`): name, platform badge, engine badge, total playtime, last played, mini trend sparkline (avg Native FPS last 10 sessions). Sort/filter/`ui:AutoSuggestBox` search.
> **The filters, beta.34 (owner request 2026-11-06, D85)** — specified here since P3 and never built: a row under the
> search and sort with **Store**, **Engine** (the stores and engines the library holds, each list naming its
> "everything" first; *Added by hand* for an entry with no store, *Engine not identified* for one with none),
> **State** (the card's one state, as its pill says it), **Installed** and **Played**, and *Clear filters*. While a filter
> or the search narrows the list the title counts what is shown of all (*Games · 12 of 65*). Kept while the App runs —
> the page is the window's (`WindowPageCache`) — and not saved. Tests:
> `GamesViewModelTests.TheFiltersNarrowTheLibraryAndTheHeaderCountsWhatIsShown`.
> **The library, beta.11 (owner requests 2026-10-03).** *The empty band on the right:* the cards were 300 px in a plain
> `WrapPanel`, so at ~1350 px three fitted and a card's width was left unused; the grid is now
> `Controls.UniformWrapPanel` — as many columns as fit at 300 px or wider, every card an equal share of the width, each row
@@ -305,7 +324,14 @@ All screens are `Page`s registered with the navigation page provider and navigat
> message at once in its one snackbar, so only the last name stayed on screen. And only the newest read of the library is
> shown: reads overlap — a session that ends, an import, a removal — and the one that started first could finish last,
> its executable checks waiting on a slow drive, and put back a removed game's card (`GamesViewModel.LoadAsync`,
> `GamesViewModelTests.AnOlderReadThatFinishesLastIsNotShownOverANewerOne`)*). **The page reloads by itself** while it is on screen when a session ends (`SessionCompleted`) or an
> `GamesViewModelTests.AnOlderReadThatFinishesLastIsNotShownOverANewerOne`)*). *(beta.34, owner request 2026-11-06 —
> "switching to Games stalls 0.5–1.5 s", `spike-notes` §27: the window keeps the page it built and shows it again on
> every later visit, reading the library again behind it; the library is read on the thread pool; the cards appear
> before any executable is checked — 39 of the owner's 41 Steam games are on a USB drive that sleeps — and each card
> gets its *Not installed* pill when its check answers; a read keeps every card that would look the same; the list
> view is bounded (`MaxHeight` 640) so it builds the rows on screen. `GamesViewModelTests`
> `.TheCardsComeBeforeTheExecutableChecksAndThePillAfter`, `.APageShownAgainReadsTheLibraryAndKeepsTheCardsThatLookTheSame`,
> `WindowPageCacheTests`.)* **The page reloads by itself** while it is on screen when a session ends (`SessionCompleted`) or an
> import adds games (`LibraryChanges` — navigating to the page already on screen does nothing in WPF UI 4.3.0, so an
> import from the menu never showed until the page was left). **Import** (FR-1.2): the games already in the library — at
> this path or under another drive letter — are hidden by default (*Show the N games already in the library* brings them
@@ -625,7 +651,10 @@ Stat cards (Avg uses `FpsReadout`, Median, 1% Low, 0.1% Low, Min/Max, σ, stutte
> frame-generation state)*, *DLSS 3 (library 3.7.10)* — else *DLSS*; then each DLSS library with its version and where
> it was loaded from (*Super Resolution 310.5.0, from the graphics driver, preset L (2nd-gen transformer model)*), then
> *DLSS 5 runtime loaded* when it was. The game page's *Last session measured* repeats the label; the CSV export gains
> `# dlss:`. `DlssTextTests`.
> `# dlss:`. `DlssTextTests`. *(beta.34: two more rows — the NVIDIA App's Dynamic frame generation the driver reports
> applied is *DLSS 4.5 (Dynamic multi frame generation, set in the NVIDIA App)* (D83), and *DLSS 5 (neural rendering on
> N % of rendered frames)* heads the line where the game was seen running it, its library listed like the others, *runtime
> loaded, not seen running* where the Overlay looked and saw none (D84) — `05_DETECTION` §DLSS generation.)*
> **beta.27: one window per session, and no DLSS line that says only *DLSS*.** Opening a session whose summary window is
> open — from a Sessions tab, a notification, the game page — brings that window forward, restored if minimised, rather
@@ -773,10 +802,20 @@ Stat cards (Avg uses `FpsReadout`, Median, 1% Low, 0.1% Low, Min/Max, σ, stutte
### Compare
Session picker (2–5, cross-game allowed) → overlaid percentile curves + stat table with best-value highlighting; FG columns always show Native and Displayed separately. Export PNG.
> **beta.34 (owner request 2026-11-06, with a screenshot; D85).** The list holds **every session** — it was the newest
> 200, read on the UI thread with one game lookup each; now one read of the names, on the thread pool — and filters by
> **Game**, **Tier** and **Frame generation** (with / without; a session that measured none is neither), with a count
> (*6 of 131 sessions*); a ticked session stays listed whatever the filters say. The selection bar — *Pick 2 to 5
> sessions*, the count, *Clear selection*, *Compare* — sits above the list on the left: it was a window's width away. The
> columns are sized to what they hold with one filler column after them (three star columns spread a wide window into
> gaps, and the Tier column was exactly its header's width, so *Tier* and *FPS* touched); the date sorts by when; the
> settings-changed cell says *Changed mid-session* beside its icon, or a dash (it was blank). The result table's session
> labels sit over their columns: the header row has the metric column's 140 px first, as every row does. Tests:
> `CompareViewModelTests.TheFiltersNarrowTheListAndATickedSessionStaysOnIt`, the page in `PagesLoadTests`' surfaces.
**Mixed-tier guard (FR-6.2):** selecting sessions from different capture tiers shows a blocking `ContentDialog` explaining that they measured different things; proceeding stamps the chart with a tier legend and greys out metrics unavailable in the lower tier. Sessions with `settings_changed_midsession` carry a warning icon and are excluded from trend lines by default.
> **Built 2026-09-14 (P3 PR-7).** `CompareViewModel` over the last 200 sessions across games: a `DataGrid`
> **Built 2026-09-14 (P3 PR-7).** `CompareViewModel` over ~~the last 200 sessions~~ every session *(since beta.34)* across games: a `DataGrid`
> picker with a checkbox column (Compare enabled at 2–5), the FR-6.4 warning icon per row; **Compare** asks
> `IMixedTierPrompt` (a `ContentDialog`, default button Cancel) when the selection mixes tiers and compares
> nothing on a decline; a comparison that proceeds carries `Compare_Mixed_Legend` as the plot's title and an
@@ -811,7 +850,77 @@ Tail view of `ui-*.log` / `agent-*.log` (toggle), level filter, search (`ui:Auto
> The status line says shown / total / file. Tests: `LogTailTests`, `LogsViewModelTests`, and the page renders in
> `PagesLoadTests`.
### Advanced
> **Built 2026-11-11 (Advanced PR-7; owner decisions D101, D105, D107).** `Pages/AdvancedPage` over
> `ViewModels/AdvancedViewModel`, a `TabControl` of three tabs by what a value belongs to:
>
> - **This PC** — `Services.AdvancedMachineFeed`, watched between `Loaded` and `Unloaded` and while the page is visible
> (a window hidden to the tray lets go too), which is what makes the Agent read with no game running and stop when the
> page closes (D107). A status line says what the last request came to (`Advanced_State_*`: read at a time, waiting,
> off, no agent, an older agent, a viewer); *About this PC* is the facts under the catalog's groups; *Right now* is a
> `DataGrid` of every series with **the time it was read at** — a sensor read every fifth tick is never shown as new;
> *Over time* is one catalog entry's last five minutes (`Charts/AdvancedSeriesChart`, a date axis).
> - **Session** — a selector of the newest fifty sessions that have a `session_advanced` row
> (`ISessionRepository.ListAdvancedAsync`), then the PC's facts as they were and the session's own, the stored
> statistics, and the entry's series over the session (`SessionSeriesLoader.LoadAdvancedAsync`: `t_ms` is the clock,
> NaN is left out, a negative reading is kept). A session whose raw series retention removed keeps its statistics
> and says the series are gone. A session's summary has **Advanced metrics…** where the session has a row
> (`SessionSummaryViewModel.HasAdvanced`); it asks for that session through `Services.AdvancedSelection` and
> navigates, and a page already on screen takes the request at once.
> - **Sources** — every entry of `AdvancedMetricCatalog`: its label, what it belongs to, its source, and *Read*, or
> the reason it is not (`AdvReach_*`), or — live — why its source is silent (`AdvSource_*`). An entry read inside
> the game (reach `Game`, 0.1.0-beta.40) says *Read inside a game FrameLedger measures, when the game makes the
> call* (`Advanced_Status_InGame`): it is not a state of the PC. What became of those entries in one session is in
> that session's sources line on the Session tab (`Advanced_InGameLine` with `AdvInGame_*`): read, the game made none
> of the calls, or the session was not measured inside the game; a session stored before beta.40 has no such
> sentence. **Since 0.1.0-beta.43** a session measured inside the game has one more
> (`Advanced_InGameCallsLine` with `AdvInGameCall_*`): for what the game tells Streamline each frame and for how it
> configures FSR frame generation, each *read*, *hooked - the game did not make the call*, or *not hooked - the
> game did not load the library that has the call*. "Nothing here" then says which of the two it is.
> **Since 0.1.0-beta.44** one more (`Advanced_InGameFeatureCallsLine`): for which version each Streamline feature
> is and for what each requires, the same three — except that hooked and not seen reads *hooked - the game asks
> this while it starts, and did not ask again after FrameLedger attached* (`AdvInGameCall_NotAskedAgain`,
> `AdvancedLabels.InGameStartupCall`): the usual case, and not a finding about the game.
> **Since 0.1.0-beta.45** one more, and one only for six calls (`Advanced_InGameMoreCallsLine`,
> `AdvancedPresenter.MoreCalls`): *Streamline's other calls - read: …; hooked, the game did not make the call: …;
> not hooked: …*, each call under its own name (`slIsFeatureSupported`, a function's name and so not translated)
> in the group that is true of it, and the word for "none" where a group is empty. A session stored before has
> none of the six and no such sentence. The catalog's `sl.init.preferences` carries a reason of its own
> (`AdvReach_BeforeAttach`): *Not read: the game makes this call once, before FrameLedger is inside it*.
>
> **Words.** Every catalog entry, group, reason, source state and driver token is a `Strings.resx` key made from its own
> name (`Adv_<id>`, `AdvGroup_*`, `AdvReach_*`, `AdvSource_*`, `AdvToken_*`) in `en`, `vi`, `ja`; the English label of
> an entry **is** its summary in the catalog, and a test holds the two equal. A key that is not there — a token a
> newer build stored — reads as the stored text, never as nothing (`ViewModels/AdvancedLabels`).
> `ViewModels/AdvancedPresenter` is the pure mapping, each rule with a test.
>
> **Charts.** A family with more than eight members is drawn as its lowest, average and highest
> (`AdvancedPresenter.MaxLines`), and the page says so; the table beside it lists every member. Lines differ by
> pattern as well as colour (`ChartPatterns`).
>
> **A viewer** (`--data-dir`) has no Agent: *This PC* says so, the page opens on *Session*, and nothing is asked. The
> navigation item is there when the copy's option is on **or the copy holds an advanced session** (`App.RunAsync`).
>
> **Accessibility** (`ACCESSIBILITY.md` §For contributors): one first-level heading, a level on every section; the two
> templates are implicit by type, so no page-local resource key is named; every row type overrides `ToString` so an
> item is read by its text; no `LiveSetting` — the readings change every second and are not announced; each chart has
> its name and says its numbers are in the table beside it. Tests: `AdvancedPresenterTests`, `AdvancedViewModelTests`,
> `PagesLoadTests` (the page among every surface, both themes, the contrast themes), `AccessibilityTests`
> (`CtrlOneToFive…`), `GuideDocumentsTests`.
### Settings
> **Advanced metrics (2026-11-11, owner decision D101; Advanced PR-5).** The Recording section has **Advanced metrics**
> (`telemetry.advanced`, off by default, off in a viewer like every setting the Agent reads): one `ui:CardControl` with a
> `ToggleSwitch`, under *Raw series kept per game*. Its text says what is read, that off reads and stores nothing more,
> that it applies from the next session and that an earlier session is not filled in. The Agent reads it at each
> session's start, so the row gets the usual "Applied. The capture agent reads it at the next session" strip. The
> switch also sets `AdvancedOption.IsOn` at once, so the navigation item and Ctrl+6 follow it without a restart
> (§Advanced, PR-7). With it, three texts that said what
> an Agent running as administrator adds with PawnIO now name the processor's, the memory's and the motherboard's
> sensors beside the CPU temperature: the Agent section's elevation line, the first run's Agent step, and the admin
> mode's disclosure (`agent-admin-dialog/2`). Tests: `SettingsViewModelTests.AdvancedMetricsIsOffByDefaultAndItsToggleIsPersisted`,
> `ViewerModeTests.EverySettingTheAgentReadsIsOffInAViewer`, `GuideDocumentsTests.TheSettingsPageOfTheGuideNamesEverySetting`.
>
> **beta.23 (owner decisions D60/D61).** The Games section has, under the Play button, **Show game artwork**
> (`ui.game_art`, on) and **Colour the game page from its artwork** (`ui.game_art_tint`, on; disabled while the first is
> off) — the game page's artwork and its band's colour (§Games › Game detail, *The header's cover*). Neither downloads
@@ -844,7 +953,7 @@ Tail view of `ui-*.log` / `agent-*.log` (toggle), level filter, search (`ui:Auto
> (`Settings_RemoveUninstalled_Applied`), not at the next session — the detection sweep reads it at every pass. Tests:
> `ViewerModeTests.EverySettingTheAgentReadsIsOffInAViewer`, `GuideDocumentsTests.TheSettingsPageOfTheGuideNamesEverySetting`.
Sections per FR-10 using the WPF UI settings idiom: `ui:CardControl`/`ui:CardExpander` rows with `ToggleSwitch`, `ComboBox`, `ui:NumberBox` (numeric inputs — min session length, sensor interval, retention). Language switch applies immediately (09_I18N). Theme selector (Light/Dark/System) drives `ApplicationThemeManager` (16 §Theming). **Capture section (top, most important):** the global **hooking kill switch** (`ToggleSwitch`, prominent) with a one-line summary of what it does; a list of games currently hook-enabled with a per-game revoke; Vulkan layer registration state + Register/Unregister; a link to the safety documentation. Turning the global switch off must take effect immediately, including for the Vulkan layer.
Sections per FR-10 using the WPF UI settings idiom: `ui:CardControl`/`ui:CardExpander` rows with `ToggleSwitch`, `ComboBox`, `ui:NumberBox` (numeric inputs — min session length, sensor interval, retention). Language switch applies immediately (09_I18N). Theme selector (Light/Dark/System) drives `ApplicationThemeManager` (16 §Theming) *(beta.35, D91: a Windows contrast theme wins while it is on, §Contrast, and a line under the selector says so, bound to `ContrastTheme.IsOn` so it comes and goes with the page shown)*. **Capture section (top, most important):** the global **hooking kill switch** (`ToggleSwitch`, prominent) with a one-line summary of what it does; a list of games currently hook-enabled with a per-game revoke; Vulkan layer registration state + Register/Unregister; a link to the safety documentation. Turning the global switch off must take effect immediately, including for the Vulkan layer.
**System (2026-09-21, FR "show the user's hardware"):** a card above Legal stating this PC as a session's hardware snapshot would record it right now — processor, graphics card, graphics driver, memory, Windows build, primary display — through the same `IHardwareSnapshotSource` the Agent stamps sessions with, read on the thread pool when the page opens, a field that could not be read as N/A, with **Copy** (the same lines as text) and a note that nothing is sent anywhere and what CPU temperature needs. Its own `SystemInfoViewModel`, not `SettingsViewModel`'s. The same snapshot is one tertiary line under the session summary's header (it was loaded for the export and shown nowhere), and `--diag` prints it; the bug bundle's `sysinfo.json` is deliberately NOT given it — its keys are the issue link's prefill.
@@ -852,7 +961,7 @@ Agent section: elevation state with a plain explanation of what elevation *optio
> **The admin mode, beta.10 (owner decision D34; `01_ARCHITECTURE` ADR-9).** The Agent section has a switch, *Run the agent
> as administrator* (`capture.run_elevated`, off by default). Turning it on opens a `ContentDialog`
> (`Services.AgentAdminPrompt`, text `FrameLedger.Shared`'s `Safety_AdminMode_*`, version `agent-admin-dialog/1` in its
> (`Services.AgentAdminPrompt`, text `FrameLedger.Shared`'s `Safety_AdminMode_*`, version `agent-admin-dialog/2` *(`/1` until 2026-11-11, when Advanced metrics widened what the option adds — D101)* in its
> corner) whose primary button is enabled only by ticking "I understand…" and whose default button keeps standard rights;
> declined, nothing is written. Either way the row is written and the Agent restarts — asked to stop over the pipe, its
> instance lock waited out, the connection's next round starting it the new way — **except under a running session**,
@@ -885,7 +994,7 @@ Agent section: elevation state with a plain explanation of what elevation *optio
> `Infrastructure.Startup.RunAtLogon` — this user's `HKCU\…\Run` value, quoted executable path, never HKLM —
> and the Run entry is the truth the toggle loads from; "Minimize to tray on close" is stored now and honoured
> when the tray exists. **Privacy:** ~~`privacy.online_metadata`, off by default, with the network sentence~~ — the card is not shown since 2026-09-16 (no code makes the request; the setting stays reserved).
> **Updates (alpha, D73, 2026-10-31):** in an alpha build the channel picker is locked and its line reads *An alpha build updates only to newer alphas, whatever is chosen here…* — an alpha's packages are on Velopack's own channel (`11_UPDATER` §Channels). **The alpha's shell (D74):** an **ALPHA** badge left of the Agent pill in the title bar for as long as it runs (its tooltip names `RuntimeInformation.FrameworkDescription`), the window and taskbar titled *FrameLedger Alpha*, and a Warning InfoBar above every page at every start, closable for that run, whose button opens the Alpha notice; About's version line names the runtime on every line, and `--diag` prints `release_line:` and the runtime's description.
> **Updates (alpha, D73, 2026-10-31):** in an alpha build the channel picker is locked and its line reads *An alpha build updates only to newer alphas, whatever is chosen here…* — an alpha's packages are on Velopack's own channel (`11_UPDATER` §Channels). **The alpha's shell (D74):** an **ALPHA** badge left of the Agent pill in the title bar for as long as it runs (its tooltip names `RuntimeInformation.FrameworkDescription`), the window and taskbar titled *FrameLedger Alpha*, and a Warning InfoBar above every page at every start, closable for that run, whose button opens the Alpha notice *(since beta.37, D97 and D98: its text is one of three — the experimental-build warning; the same naming "a prerelease .NET" where the runtime's description is a prerelease's; or, in an alpha cut while the alpha line rests, that the line is paused and to install the beta — `Services/AlphaTexts`, `Shared.AlphaLine`)*; About's version line names the runtime on every line, and `--diag` prints `release_line:` and the runtime's description.
>
> **Updates:** `update.channel` is read by PR-5's updater (beta = GitHub pre-releases; since beta.14 *Automatic (Beta)* or *(Stable)* first, the default, following the running copy — D47), and `update.auto_check` — the startup check, on by default — sits beside it. **Logging:** `log.debug` moves this app's `LoggingLevelSwitch` at
> once; the Agent has no reader for the key yet, and the row says so. **Legal:** reopen is a "not yet" line
@@ -968,8 +1077,9 @@ Agent section: elevation state with a plain explanation of what elevation *optio
(`ScrollViewer.CanContentScroll="False"` on the `Page` — `16_WPFUI_SYNTAX` §Gotchas: until beta.14 the wheel scrolled
nothing on five pages, and the Logs tail did not follow the newest line); a scroller nested in a dialog hands the wheel
on at its ends (`ui:PassiveScrollViewer`); over a chart a plain wheel scrolls the page and **Ctrl+wheel zooms the chart**
(`Charts/ChartWheel`, every plot through `ChartTheme.Attach`) — drag still pans and double-click still resets, as
ScottPlot does. The import checklist scrolls inside its list and nowhere else, so its intro and its count stay put.
(`Charts/ChartWheel`, every plot through `ChartTheme.Attach`) — drag still pans and ~~double-click still resets, as
ScottPlot does~~ *(corrected 2026-11-09: ScottPlot's left double-click toggled its render-time benchmark overlay; since
beta.35 a double-click shows all of the chart, `ChartTheme.ResetOnDoubleClick`, `ChartKeyboardTests`)*. The import checklist scrolls inside its list and nowhere else, so its intro and its count stay put.
## Accessibility
@@ -977,17 +1087,39 @@ NFR-9 names three things — full keyboard navigation, adequate contrast in both
and this section is what they mean for this app, each with the test that holds it (P4 PR-8, 2026-09-15). Until then
the requirement was one bullet under §UX rules, and nothing checked it.
> **The users' half is `ACCESSIBILITY.md` (beta.35, owner decision D88, 2026-11-09):** the goal — WCAG 2.2 AA as W3C's
> WCAG2ICT applies it to desktop software, a goal and never a conformance claim — what works, what does not and what to do
> instead, and how a barrier is reported (the *Accessibility problem* form, `labels: [accessibility]`) and handled (by
> severity, no promised times). It is embedded and shown under Help ▸ Accessibility. **This section is the design and the
> checks; the statement says nothing this section does not hold**, and a change that moves either changes both
> (`AccessibilityDocumentTests` holds the statement's own rules: embedded as written, a goal and no claim, a form that
> exists, no struck text in a page the App shows). NFR-9 was widened the same day to screen readers and contrast themes
> (D88–D91).
### Keyboard
| Keys | Does | Where |
|---|---|---|
| **Ctrl+1 … Ctrl+5** (and the numeric keypad's 1–5) | Dashboard · Games · Compare · Logs · Settings | `MainWindow` `InputBindings` → `MainWindowViewModel.NavigateToCommand` → `ShellShortcuts` |
| **Ctrl+1 … Ctrl+5** (and the numeric keypad's 1–5); **Ctrl+6** since 2026-11-11 (D105) — the Advanced page, only while `AdvancedOption.IsOn`: `ShellShortcuts` refuses it otherwise, so the number stays free | Dashboard · Games · Compare · Logs · Settings — and since beta.35 the page's first control takes the focus once it is shown, its title (the page's first-level heading) said to a screen reader; before, the focus stayed where it was, often on the page that had just gone | `MainWindow` `InputBindings` → `MainWindowViewModel.NavigateToCommand` → `ShellShortcuts`, then `Services/PageFocus` (the shell tells it each page shown, from `NavigationView.Navigated`) |
| **Alt**, **F10** | Focuses the menu bar; arrows and Enter walk File / Tools / Help | WPF `Menu`, nothing custom |
| **Alt** + a menu's access key | Opens that menu — `_File` `_Tools` `_Help`; `_Tệp` `_Công cụ` `Trợ _giúp`; `ファイル(_F)` `ツール(_T)` `ヘルプ(_H)` *(row added 2026-11-09: the keys existed since P3 and this table did not name them)* | the `Menu_*` strings' `_` |
| **Tab / Shift+Tab** | Moves through a page in reading order | layout order; no `TabIndex` overrides |
| **Arrows, Home/End, Page Up/Down** | Move within a `DataGrid` (Sessions, Compare picker, Import checklist) | `DataGrid` defaults |
| On a focused chart: **arrows**, **Ctrl**+arrow, **A** *(beta.35, D90)* | Pan, zoom, show all — ScottPlot's keys; the key is then the chart's, so the page neither scrolls nor moves the focus on (Tab still leaves) | `Charts/ChartPlot`, `Charts/ChartKeys` |
| **Esc** | Closes the open `ContentDialog` or `MessageBox` with no result — the same as its Close/Cancel button | `DialogKeyboard`, one class handler per dialog type registered at start-up |
| **Enter** | In the consent dialog's phrase box: enables hooking once the phrase is typed exactly, does nothing before (beta.30, D69); elsewhere in a `ContentDialog`, the default button | WPF's default button (`ConsentPrompt`: `DefaultButton = Primary`) and `ConsentDialogContent.SwallowsEnter` |
**Focus after a dialog, and motion (beta.35).** A `ContentDialog` gives the focus back to the control that opened it: WPF
UI 4.3.0's `ContentDialogHostController` stores `Keyboard.FocusedElement` when the dialog opens and focuses it again, at
`DispatcherPriority.Input`, when it closes — the App adds nothing, and
`FocusAndMotionTests.AContentDialogGivesTheFocusBackToTheControlThatOpenedIt` holds it (a first measurement said
otherwise; it ran on a thread with no dispatcher context, where the dialog's own `finally` never resumed). A WPF UI
`MessageBox` sets no owner — null before and after it is shown — so `Services/MessageBoxes` gives it the App's active
window: it opens over it and hands the focus back to it, not to another program; every box goes through it
(`EveryMessageBoxGoesThroughMessageBoxes`). A page slides in only while Windows' animation effects are on
(`SystemParameters.ClientAreaAnimation`, followed as it changes): `Services/PageMotion` sets the `NavigationView`'s
`Transition` to `None` while they are off, where WPF UI 4.3.0 played `FadeInWithSlide` whatever they were.
**Esc is the Close button, never the primary one.** On the consent dialog (FR-2.1) that is ~~"Keep it off"~~ "Keep hooking
off" *(the button's text, corrected 2026-10-07)*: nothing is
stamped. A dialog that offers no Close button is not dismissed by Esc, and the Legal Gate (FR-11) is a window of its own
@@ -1009,7 +1141,54 @@ handled key — so whether Esc alone would close a dialog through `IsCancel` is
— corrected 2026-10-04): every `Button`, `ToggleButton`, `ToggleSwitch`, `CheckBox`, `RadioButton`, `ComboBox`,
`TextBox`, `NumberBox`, `PasswordBox`, `Slider` and `HyperlinkButton` in the App's XAML either has content or a name. A new
unnamed one is red.
- Status lines that change after an action (the Database maintenance result) are `AutomationProperties.LiveSetting="Polite"`.
- **What a screen reader is given, swept on every surface (beta.35, D88).** `PagesLoadTests.Automation` hosts every page,
window, dialog and control as the App hosts it, every tab shown in turn, and reads the UI Automation tree WPF builds — the
tree Narrator reads: **no Tab stop without an automation peer** (a plain `Control`, WPF UI's `Image`, `Card` and
`CardControl` take the focus with no peer or an empty name), **no item named by its type** (WPF names a list's, a
drop-down's and a grid row's item by `ToString()`, and a template column's cell, which has no text of its own, by
`Item: <row>, Column Display Index: n`), **every focusable control named**, **no shown text hidden in a template** (a
`TextBlock` whose templated parent is not a `ContentPresenter` is no control element), and **no two buttons in a list
sharing a name**. All five went red on `main` first — the type-named items filled 190 lines over the surfaces and their
tabs, the silent stops 37, and the safety InfoBar's title and message were hidden — and the fixes are: every item type speaks (`ToString()` through `Services/SpokenText`: `Choice` its
label, a notice its title and text, a document page its title, a session row its date, length and tier…); the display-only
controls (`FpsReadout`, `TriStateChip`, `ChangeBadge`) have a `Controls/SpokenTextAutomationPeer` — a Text element named
by their words, no children — and leave the tab order; a template column names its cells through
`Controls/CellName` (a cell style based on WPF UI's `DefaultDataGridCellStyle`, an empty cell read as *empty*); a row's
button names its row (*Open: …*, *Play: …*, *Turn off: …*); the App's InfoBar template draws its title and message through
`ContentPresenter`s; implicit styles take a plain `ItemsControl`, WPF UI's `Image` and `CardControl` out of the tab order;
a `CardExpander`'s expand button is named after its header (`Controls/ExpanderToggleName` — WPF UI 4.3.0 leaves it
unnamed). **The FPS readout says rule 6 aloud**: *Native 62 FPS, displayed 118 FPS, frame generation ×1.9*, a steady factor
with its share, a stand-alone number with its qualifier (`FpsPresentation.Spoken`). `ScreenReaderNamesTests` holds each
piece.
- **Headings (beta.35, owner decision D90).** A page's title (`FontTypography="Title"`) is
`AutomationProperties.HeadingLevel="Level1"`, a section title (`Subtitle`) `Level2`; a `Subtitle` that shows a value — the
Dashboard's three figures, a stat card's value, the FPS readout's line — says `None` on purpose.
`AccessibilityTests.EveryTitleAndSubtitleDeclaresAHeadingLevel` (every Title and Subtitle declares one),
`PagesLoadTests.EveryPageHasOneFirstLevelHeadingItsTitle`. A rendered document's headings (the guide, LIMITATIONS, this
statement, the legal texts) cannot be: a `FlowDocument` `Paragraph` has no automation peer, so it has no heading level to
give — stated in `ACCESSIBILITY.md` §Known limitations.
- **Charts (beta.35, D90).** Every plot is a `Charts/ChartPlot` — ScottPlot's `WpfPlot` with a peer: an Image whose
localized type is *chart*, named by its `AutomationProperties.Name` (`A11y_Chart_*`), its help text where its numbers are
as text (`A11y_ChartData_*`) and its keys (`A11y_Chart_Keys`); no children. Its focus is drawn with WPF UI's
`DefaultControlFocusVisualStyle`. **No two series of a chart differ by colour alone** (`Charts/ChartPatterns`: lines
cycle solid / dashed / dotted / densely dashed, then take markers; point series cycle circle / triangle / square /
diamond; the power and thermal limit spans are striped and dotted) — `PagesLoadTests.NoChartTellsTwoSeriesApartByColourAlone`
reads every chart's legend on every surface and the Compare chart with five sessions.
- **What appears is said as it appears (beta.35, owner decision D89).** `Services/IAnnouncer`, over UI Automation's
notification event in `Services/UiaAnnouncer` — the one place the App raises it, on its **active** window only (a game
that has the screen is never interrupted; the tray's Windows notifications cover a hidden window) and only when a
screen reader listens (`AutomationPeer.ListenerExists`). An **Alert** (`ImportantAll`, every one read): a safety notice as
it arrives (`SafetyNotices`), a warning in the strip, the capture agent going offline, a game page's refusal or warning.
A **Status** (`MostRecent` per activity, a newer one replacing one not yet read): the strip's info and success, the agent
back (`A11y_Agent_Connected`), an update's step (`Services/ShellAnnouncements` says a *change* only — the first state is
read with the window, a download's percent ticks are one step, a rebuilt window re-reads the same state silently), a
game page's success, every Database maintenance result line. The words are the window's, in its language.
`AnnouncementsTests`, `DatabaseMaintenanceViewModelTests.AResultIsSaidToAScreenReaderAsItAppears`.
- ~~Status lines that change after an action (the Database maintenance result) are `AutomationProperties.LiveSetting="Polite"`.~~
*(Corrected 2026-11-09: this announces nothing. WPF raises `LiveRegionChanged` only when the app raises it — "you
explicitly raise an event", Microsoft's* What's new in accessibility *for 4.7.1 — and `TextBlockAutomationPeer` raises
none (dotnet/wpf); the App raised no automation event anywhere. The setting marks the line as a live region and no
more. It came out the same day, for the announcement above; no XAML may set one.)*
### Contrast
@@ -1092,6 +1271,45 @@ theme's background, as it was before; the badge and the list carry its informati
> under it, in both themes. `EveryTextFollowsAThemeSwitch` hosts its pages in the presenter too since, and went red with
> them: a page made after a switch stayed black. All three went red on the unfixed XAML before they went green.
> **Windows contrast themes, followed (beta.35, owner decision D91).** WPF UI 4.3.0 cannot be asked to do it:
> `ApplicationTheme.HighContrast` names no dictionary; its four high-contrast files (HC1 Dusk, HC2 Night sky, HCBlack
> Aquatic, HCWhite Desert) each hold 84 red `#FFFF0000` placeholders among 97 colours and 17 brushes that throw when the
> file is read alone; their colours are Windows 11's defaults, not the user's; and they put the window text on the accent.
> So while `SystemParameters.HighContrast` is on, `Services/WpfThemeApplier` applies the light or the dark Fluent theme as
> the contrast theme is (by its window colour), with no Mica and no `SystemThemeWatcher`, and `Services/ContrastOverlay`
> over it: every key of those four files in the user's colour for its role (`Services/ContrastRoles` — a key's role is the
> colour it has in all four, opaque colours only, because inside the App a brush that throws alone falls back to the
> Fluent theme's translucent fill: `MenuBarItemBackgroundPressed` read as the white of the window text, and a pressed menu
> item was white on white; ties and placeholders by the key's name), the keys transparent there transparent, three
> Fluent-only keys by role, the 21 accent keys WPF UI writes into the application's resources as the highlight, and the
> text on the accent as the highlight's text. The colours are `SystemColors` (`ContrastPalette.FromSystem`) — the user's
> own, edited or not, Windows 10's as well — and turning a theme on, editing it or turning it off is followed at once
> (`SystemContrastSource`). What the App draws itself reads `ContrastTheme.Active`: the charts (`ChartPalette.ForContrast`,
> the theme's colours only, the series told apart by `ChartPatterns`), a badge's text (`ChartTheme.TextOn`, the theme's
> colour that reads best on the fill), the mix bar, the artwork band (no tint), the Import list (no dimmed rows), the
> unsigned-update footer (the secondary text style, not an opacity). WPF UI's Danger button paints a fixed `#F44336` in its
> style, out of any resource's reach, so every `ui:Button` takes the App's implicit style (`Styles/FrameLedger.xaml`): a
> Danger button is the theme's own button while one is on, the message boxes' Delete included. Held by `ContrastThemeTests` — the four files still placeholders; every key a role or
> transparent, none red; the roles the same over either Fluent theme; with each default theme's own colours the overlay
> is WPF UI's colours key for key, the text on the accent apart; an edited theme's colours reach every key; every Fluent
> colour key covered; every resource the App reads resolves; `ThemePlan`; the applier on and off at once — and
> `PagesLoadTests.EveryTextIsInTheContrastThemesColoursAndReadable`: every surface rendered under the four default themes
> and two edited ones, every text the theme's colour on the theme's colour at 4.5:1 or better. It went red first on the
> chips (window text on the highlight, 1.26–1.78:1), the Remove button (3.06–4.42:1 on `#F44336`) and the badges' black.
> **A sixth hole, closed 2026-11-09 (beta.35), found by measuring what no sweep renders.** The title bar and the snackbars
> are no surface, and the 3:1 sweep let 3.68:1 pass. The capture agent's pill was a `ui:Badge` whose appearance was
> BOUND — Success, Info, Caution, Danger, WPF UI's `Palette*Brush` under `BadgeForeground` — so `ContrastTests`, which reads
> literal attributes, never saw it: in the light theme 2.78, 2.63, 2.16 and 3.68:1. It is `AgentPill` now (a `Pill` on
> `SystemFillColor{Success,Caution,Critical}BackgroundBrush`, the theme's text), and a bound badge appearance is refused.
> WPF UI's Success and Caution snackbars are fixed colours under white in both themes, 2.78:1 and 2.16:1: the strip shows
> every message as Secondary, the theme's, 13.97:1 and 16.53:1, its icon saying the kind. And a Danger button's text was
> the theme's button text on `#F44336` — white in the dark theme, 3.68:1: the App's implicit button style puts
> `TextFillColorDarkPrimaryBrush` there, 5.3:1 and 5.7:1. `ContrastThemeTests.TheTitleBarsMarksReadInEveryState`,
> `ADangerButtonReadsInBothThemesAndIsTheThemesOwnButtonUnderAContrastTheme`,
> `NoButtonOrBadgeTakesAnAppearanceThatPaintsAFixedColour` and
> `AnnouncementsTests.EveryMessageOfTheStripIsDrawnInTheThemesOwnColours` hold the three.
### DPI
`app.manifest` declares Per-Monitor V2 (`AccessibilityTests.TheManifestIsPerMonitorV2`). A chart reads the pointer
@@ -1100,8 +1318,24 @@ ScottPlot reads once when the control is made, before it has a window.
### Known gaps, stated
- ScottPlot plots are not keyboard-operable (no focus, no pan/zoom by keys). The numbers behind them are on the
- ~~ScottPlot plots are not keyboard-operable (no focus, no pan/zoom by keys).~~ *(Corrected 2026-11-09, measured by
`ChartKeyboardTests` in a window off the screen: a plot takes focus — ScottPlot's control is focusable and a Tab stop —
and the keys ScottPlot reads reach it: the arrows pan, Ctrl with an arrow zooms, A shows all of it. ~~What it is not is
usable: it has no automation peer, so Narrator says nothing on reaching it, and it marks no key handled, so the page
around it acts on the same arrow too.~~ Since beta.35 it has a peer and keeps its keys, §Screen readers.)* A chart's
points are still not read one by one. The numbers behind them are on the
session summary's stat cards and in File ▸ Export (CSV/JSON); the hardware-change markers' text is WPF text in the
list under each trend chart (beta.30).
- No screen-reader pass with Narrator has been done; the names above are necessary, not proven sufficient. That pass
is the owner's.
- ~~No screen-reader pass with Narrator has been done; the names above are necessary, not proven sufficient. That pass
is the owner's.~~ **The owner's pass, 2026-11-09: keyboard only with the mouse disconnected, and Narrator, on
0.1.0-beta.34 and 0.1.0-alpha.3 — every page, menu, dialog and setting reached and used, no fault.**
- **What that pass could not show**, found the same day by measuring the UI Automation tree WPF builds (in memory, on
.NET 10.0.14 — not by listening): ~~a drop-down's items are named by a record's printed form (`Choice { Value = dark,
Label = Dark }`) and plain lists' items by their type; a control with no automation peer that still takes focus — a
chart, the FPS readout, the RT/PT/RR chips, the game's artwork — is a Tab stop Narrator says nothing about; text inside
a control template (the safety notices' title and message, the readout, the chips) is hidden from the control view;
nothing is announced; no text is a heading; several charts tell series apart by colour alone~~ *(closed the same day,
§Screen readers above)*; ~~page transitions play with Windows' animation effects off~~ *(not since beta.35,
§Keyboard)*; ~~Windows contrast themes are not
followed~~ *(followed since beta.35, D91 — §Contrast)*. `ACCESSIBILITY.md` §Known limitations
states each to users with what to do instead.
+4
View File
@@ -17,6 +17,10 @@ Languages: **en** (neutral/default) · **vi** · **ja**. Same `.resx` pattern as
- One `Strings.resx` (+ `Strings.vi.resx`, `Strings.ja.resx`) per assembly that has user-visible text (`App`, `Shared` ~~for error codes~~ for the safety family, the `Safety_*` keys — corrected 2026-10-04). Access through a generated strongly-typed class; **no string literals in XAML/C# for user-visible text** (analyzer TODO: Meziantou MA0011/manual review in PR checklist). **As built (P3 PR-2, 2026-09-13):** the class is `Strings.Designer.cs`, written by `tools/resx-gen.ps1 -Resx <file>` and **committed** — not the SDK's `ResXFileCodeGenerator`, which the SDK-style WPF project does not run — with `Culture`, `ResourceManager`, `Keys`, and one property per key; the audit below fails when it is stale. The first family is `src/FrameLedger.App/Strings.resx` (65 keys); `Shared`'s arrives with PR-4's `Safety_*` text, because an empty family is what the audit refuses.
- XAML binding via `{x:Static res:Strings.Key}` ~~plus a `LocalizeExtension` for the few dynamic cases~~ *(corrected 2026-10-04: no `LocalizeExtension` exists; text that varies is composed in the view model from `Strings.*`)*.
- **Runtime switching:** setting change → set `CultureInfo.CurrentUICulture` + `Thread` defaults → re-create the main window shell (cheap; state preserved via VM). Persist in `settings`.
**The new window opens on the page that was on screen** *(beta.34, owner request 2026-11-06)* — Settings, after a
change made there. `ShellHost` remembered only the pages opened from code (`PageNavigator`), never a click in the
pane, so a language changed on Settings brought back the last game page opened; it now notes every page the live
window's NavigationView shows (`ShellHost.NoteShown`).
**Two kinds of text did not follow, until beta.15:** a list of choices built once in a static (Settings' themes and update
channels, the Logs page's two lists, the Games page's sort order — `x:Static` read the list built in the language the
App started in), and the singletons that outlive the rebuilt window (the tray's menu, the title bar's Agent pill and the
+11 -1
View File
@@ -64,6 +64,16 @@
> and never the Agent's usage code 1. The catch exists because an exception escaping `Main` runs `Main`'s `finally`
> first, which closes the logger; the AppDomain handler alone would dump with no Fatal line.
>
> **A background service that fails, since 0.1.0-beta.37 (.NET 11, D94).** `--serve` awaits the Generic Host's
> `RunAsync` and catches nothing. On .NET 10 a `BackgroundService` whose `ExecuteAsync` threw stopped the host and
> `RunAsync` completed, so the Agent exited 0 with the fault in the log alone (the host's own line, in `agent-*.log`
> since 2026-09-23). On .NET 11 `RunAsync` rethrows that exception — an `AggregateException` when several services
> failed — so it reaches `Main`'s catch above: Fatal, the dump, and the runtime's exit code. Nothing was changed to get
> this and nothing wraps it: an Agent that stopped because a piece of it broke should not report success. The App
> hosts with `StartAsync` and `StopAsync` inside its own try/catch and is not affected. `HostBackgroundFailureTests`
> holds the runtime's behaviour — a host with one throwing service fails `RunAsync` with that exception; the three
> alphas shipped with it unobserved (`spike-notes` §24).
>
> Tests: `CrashDumpWriterTests` (a real dump of the test process starts `MDMP` and carries the UTC time in its name; the
> prune keeps five and touches no other file; an unusable directory is null and one line), `CrashReporterTests` (the
> order, no report on No, three further exceptions show nothing and are counted, a failed dump still asks, a failed
@@ -83,7 +93,7 @@ Tails the active files (shared read), level filter, text search, pause autoscrol
2. **Bundle builder** creates `FrameLedger-bugreport-YYYYMMDD-HHmm.zip` in a user-chosen location:
- `logs/` (last 7 days, redacted copies)
- `sysinfo.json` (app + agent + **overlay build id**, OS build, CPU/GPU name, driver version, telemetry source, Vulkan layer state, PawnIO present?, elevation state, locale)
- `overlay-<pid>-*.log` for the last hooked sessions, plus hook fault details and the ring's dropped/fault counters — these are what make injection bugs diagnosable at all. *(The file exists since 2026-09-06 — `17_HOOK_ENGINE` §Native logging: written at init, on the Agent's request at session end, and on the stop; `FAULT` lines carry the exception code and the hook's name, `UNHOOK_*` lines say per patch whether it was restored or left to a later hooker.)*
- `overlay-<pid>-*.log` for the last hooked sessions, plus hook fault details and the ring's dropped/fault counters — these are what make injection bugs diagnosable at all. *(The file exists since 2026-09-06 — `17_HOOK_ENGINE` §Native logging: written at init, on the Agent's request at session end, and on the stop — and since 0.1.0-beta.42 by the Overlay's watchdog whenever something new was logged, so it is whole after a game that simply exits, with each install named by module and symbol and each descriptor type an ffx module was handed; `FAULT` lines carry the exception code and the hook's name, `UNHOOK_*` lines say per patch whether it was restored or left to a later hooker.)*
- `settings.json` (sanitized — no paths)
- optional checkbox: last session metadata + aggregates JSON (never raw blobs by default)
- crash dumps included only when the user ticks the checkbox (size warning shown)
+1 -1
View File
@@ -98,7 +98,7 @@ flow, and the tests run the flow over a fake client because nothing here may tou
never sees a beta, under either setting of `update.channel` — and in an alpha build that picker is locked, its line
saying so (`08_UI` §Settings). **One limit follows from the source, read from Velopack's code 2026-10-31:** it asks
GitHub for one page of ten releases, newest first, so ten releases of one line in a row would hide the other line's
newest from its update check. The cadence is one alpha per beta (`13_CI_CD` §The alpha line). Moving between the lines
newest from its update check. The cadence is one alpha per beta (`13_CI_CD` §The alpha line) — *while the line tries a .NET; it rests since 0.1.0-alpha.4 (D98, 2026-11-11), betas then follow one another with no alpha between, and once ten of them have been published after alpha.4 a copy of alpha.3 or older no longer finds it: the alpha's notes, the guide and its banner say to install the beta.* Moving between the lines
is a deliberate install, never an update: the other line's `Setup.exe` over the installed copy, which Velopack answers
with its overwrite prompt — the packId, the install folder and the data folder are the same.)*
- **Hooks** (`VelopackHooks`, in the hand-written `Program.Main` so `VelopackApp.Run()` precedes any window):
+87 -32
View File
@@ -1,10 +1,11 @@
# 12 — Build & development environment
Mixed toolchain: .NET 10 for managed projects, MSVC for the native layer.
Mixed toolchain: .NET 11 for managed projects *(since 0.1.0-beta.37, owner decision D94, 2026-11-11; .NET 10 until
0.1.0-beta.36 — §The main line's .NET)*, MSVC for the native layer.
## Prerequisites
- Windows 11 (dev), **.NET 10 SDK**, **Visual Studio 2026** with: .NET desktop development, **Desktop development with C++** (MSVC v143+, Windows 11 SDK ≥ 10.0.22621), C++ ATL not required.
- Windows 11 (dev), **.NET 11 SDK** (11.0.100 or a later 11.0 feature band), **Visual Studio 2026** with: .NET desktop development, **Desktop development with C++** (MSVC v143+, Windows 11 SDK ≥ 10.0.22621), C++ ATL not required.
- **Vulkan SDK** (~~for the layer +~~ validation layers during development, and to run `tools/vklayer-blastradius.ps1`) *(corrected 2026-10-04: the layer builds against the vendored `third_party/vulkan-headers`; CI installs no SDK)*.
- CMake ≥ 3.28 (native projects use CMake; consumed by the solution via a build target — see below).
- Optional: PawnIO for CPU-temperature testing; an NVIDIA GPU for the NVAPI/Reflex paths (dev machine: RTX 5080). AMD/Intel telemetry paths need no SDK — if you have such hardware, testing L1/L2 coverage there is valuable.
@@ -19,6 +20,11 @@ cmake --preset x64-release
cmake --build --preset x64-release
```
A third preset, **`x64-vs-analysis`** *(2026-11-11)*, builds nothing that ships: it has CMake write a Visual Studio
solution (`build/native/x64-vs-analysis/FrameLedgerNative.slnx`, generator "Visual Studio 18 2026", tests and tools
on) for Qodana's native report, which reads C++ through a solution (`13_CI_CD` §Qodana). It needs Visual Studio 18;
the gate and the release never use it.
Targets:
- `FrameLedger.Overlay` → `FrameLedger.Overlay.dll`
- `FrameLedger.VkLayer` → `FrameLedger.VkLayer.dll` + `VkLayer_FRAMELEDGER_overlay.json`
@@ -29,6 +35,7 @@ Targets:
- `hook-harness` → the dummy **D3D11 + D3D12** app, WARP + composition swapchain so it runs headless on CI (`17_HOOK_ENGINE` §Test harness). D3D12 is device → command queue → swapchain, which is the acquisition asymmetry a D3D11-only fixture cannot exercise. ~~Vulkan and OpenGL modes to follow as those hooks land.~~ **`--vulkan --hold-presenting N` since 2026-09-06 (P1 item 3):** a real swapchain on a hidden window through the real loader, reached by `LoadLibraryW("vulkan-1.dll")` so only the vendored headers are needed and CI (no loader) still builds it; exit 77 = cannot run here, which the ctest reads as SKIP. **`--opengl --hold-presenting N` since the same day (P1 item 4):** a real context on a hidden window, `SwapBuffers`'d through gdi32 as a title does, on Microsoft's generic renderer — so it runs on CI.
- `fl-probe-hookprofile` → build-profile probes for `/guard:cf` and `-D_HAS_EXCEPTIONS=0` (`20_OPEN_QUESTIONS` §H1/§H3)
- `fl-probe-guard` → measures the Windows APIs the guard is built on, unelevated (`spike-notes.md` §1). Not the guard, and takes no injection rights.
- `fl-probe-reflex` → the instrument of `spike-notes` §31 (2026-11-13, owner decision D132) — **since the owner said go (D136) it is built from the Overlay's own header, `FrameLedger.Overlay/include/fl_reflex.h`**, so its ctest runs the shipped call sequence with Control Flow Guard enforced; its `--real` also prints which image the two getter addresses lie in (`nvapi64_impl.dll`, not the `nvapi64.dll` that is asked). What it was written as: the call sequence Reflex route D would put into the Overlay — find `nvapi64.dll` only if the process already holds it, resolve the driver's two getters by interface id, call them — in an unshipped probe built with the Overlay's flags. `--stub` runs it against `fl_stub_nvapi64` → `nvapi64.dll` (a stand-in with knobs for a missing getter, an empty report, an error and a call that does not return; its outputs in a directory of their own, so that its `nvapi64.lib` is never beside the vendored one's name) and is ctest `fl_reflex_probe_stub`; `--real` runs it against the installed driver on an NVIDIA adapter, with the probe setting the latency markers a game would, in its own process — a measurement, never a test. It takes the vendored NVAPI headers' include directory and **does not link `nvapi64.lib`**. Nothing shipped includes its header.
- `fl_stub_sl_interposer` → `sl.interposer.dll`, and `fl_stub_sl_common` → `sl.common.dll` (`src/native/tools/vendor-stubs/`). **Fixtures, never shipped**, built under `FL_BUILD_TOOLS`; nothing `12_BUILD` publishes references them. They export the **measured** vendor names from `docs/vendor-exports.json` so the Overlay's symbol resolution is observable, and the second is a **decoy** exporting the same name from a different module — without it, "we resolve module-scoped" is a property no test can falsify. Their **output filenames are the vendor's and their version blocks are ours**: `fl_add_version_resource` takes the real filename so a file called `sl.interposer.dll` in our build tree still says plainly that it is FrameLedger's, which `19_SAFETY` requires of everything we build.
- `fl_stub_ffx_dx12` → `amd_fidelityfx_dx12.dll`, `fl_stub_ffx_upscaler` → `amd_fidelityfx_upscaler_dx12.dll`, `fl_stub_ffx_fg` → `amd_fidelityfx_framegeneration_dx12.dll` — the three AMD **leaves** — and `fl_stub_ffx_loader` → `amd_fidelityfx_loader_dx12.dll`, a **forwarding stand-in** bound to the two 2.x leaves through their FrameLedger-named *direct* entry — the way the real loader was measured to bypass the leaves' exports — so `--hold-presenting-ffx` can push every dispatch through a loader that is hooked along with the leaves and prove the count reads 1× (2026-09-04). Same rules as the Streamline stubs: fixtures, never shipped, vendor filenames with our version blocks. `fl_stub_ffx_fsr3_host` → `ffx_fsr3_x64.dll` (2026-09-05): the FSR 3.0 **host** facade, exporting the four names `SpeaksFsr3Host` probes for through the vendored `fsr3-v3.0.4` header's own `FFX_API`; `--ffx-topology fsr3host|fsr3host+mono` drives it, and `hookinventory-check` Pass C reads it as the export parser's second positive control.
- **`fl_fidelityfx_headers`** → an INTERFACE target over the five vendored FidelityFX ffx-api headers (`src/native/third_party/fidelityfx/`, tag v2.3.0, MIT by exception). Three include directories because upstream's headers include each other by relative path; include directories only and nothing to link — no `.lib` is vendored — and `hookinventory-check` Pass C reads the Overlay's export table because `ffx_api.h` declares its entry points `__declspec(dllexport)` unconditionally.
@@ -48,9 +55,9 @@ Compiler/linker flags (enforced in CMake, not per-target ad hoc): `/std:c++20 /M
## Managed build & native integration
- **`global.json` pins the SDK band** (`10.0.x`, `rollForward: latestFeature`). Without it `dotnet build` picks the newest installed SDK — on a machine with a .NET 11 preview installed that silently changes analyzer behaviour under `TreatWarningsAsErrors`, and makes local and CI disagree for reasons nobody can see in a diff. *(On the `alpha` branch it pins the prerelease SDK instead, on purpose — §The alpha line, 2026-10-31.)*
- **`global.json` pins the SDK band** (`11.0.x` since beta.37 — `10.0.x` before — `rollForward: latestFeature`, `allowPrerelease: false`). Without it `dotnet build` picks the newest installed SDK — on a machine with a preview of the next .NET installed that silently changes analyzer behaviour under `TreatWarningsAsErrors`, and makes local and CI disagree for reasons nobody can see in a diff. *(The `alpha` branch carries the same file while its line rests, and pins the SDK it is trying when it does not — §The alpha line.)*
- `Directory.Build.props`:
- **`TargetFramework` = `net10.0-windows10.0.22621.0`** and **`SupportedOSPlatformVersion` = `10.0.19045.0`**. These are two different knobs and must not be conflated: the TFM platform version selects which Windows API projections are available to compile against, while the supported version is the floor CA1416 enforces. NFR-8 requires Windows 10 **22H2** = build **19045**; letting the supported version default to the TFM's would silently accept installs below the stated floor. Two constraints pin the target to 22621 rather than something closer to the floor: the SDK rejects `SupportedOSPlatformVersion` above `TargetPlatformVersion` (NETSDK1135), and 19045 is not a targeting-pack version in any case — the packs are cut at 19041, 20348, 22000, 22621, 26100. The Windows 11 SDK ≥ 10.0.22621 is already a prerequisite above, so this costs nothing. A bare `net10.0-windows` is wrong for a different reason: it resolves to `net10.0-windows7.0`, and CA1416 would then error on Mica, `MiniDumpWriteDump` and `QueryVideoMemoryInfo` — every Windows-10-era API this app is built on.
- **`TargetFramework` = `net11.0-windows10.0.22621.0`** *(`net10.0-…` until beta.36)* and **`SupportedOSPlatformVersion` = `10.0.19045.0`**. These are two different knobs and must not be conflated: the TFM platform version selects which Windows API projections are available to compile against, while the supported version is the floor CA1416 enforces. NFR-8 names Windows 10 **22H2** = build **19045** as the oldest Windows the code is held to *(since D95, 2026-11-11: not blocked there, and neither tested nor supported — Microsoft supports .NET 11 on Windows 11)*; letting the supported version default to the TFM's would silently accept installs below that floor. Two constraints pin the target to 22621 rather than something closer to the floor: the SDK rejects `SupportedOSPlatformVersion` above `TargetPlatformVersion` (NETSDK1135), and 19045 is not a targeting-pack version in any case — the packs are cut at 19041, 20348, 22000, 22621, 26100. The Windows 11 SDK ≥ 10.0.22621 is already a prerequisite above, so this costs nothing. A bare `net11.0-windows` is wrong for a different reason: it resolves to `net11.0-windows7.0`, and CA1416 would then error on Mica, `MiniDumpWriteDump` and `QueryVideoMemoryInfo` — every Windows-10-era API this app is built on.
- `Platforms`/`PlatformTarget` = `x64`. "x64 only" is asserted in CLAUDE.md and NFR-8 but was previously enforced nowhere except an ad-hoc `-r win-x64` on the publish command.
> **`RuntimeIdentifier` is NOT set here, and this bullet claimed it was until 2026-08-06.** The props file's own comment explains why: forcing a RID onto class libraries pulls RID-specific assets into their restore for no benefit, so each executable sets it — `FrameLedger.App`, `FrameLedger.Agent` and now `FrameLedger.CaptureHost`, plus its test project, which needs the same RID to reference a RID'd exe. A new project copying this document rather than the props file would have inherited nothing.
@@ -162,44 +169,89 @@ repairs state, it does not grant machine-wide reach on its own).
> after every sweep; the flag and the button run the same rule. The uninstall hook is still the updater's
> (P4, Velopack).
## The alpha line (owner decisions D73–D76, 2026-10-31)
## The main line's .NET (owner decisions D94 and D95, 2026-11-11)
**The main line — the betas and the first release — builds on .NET 11 since 0.1.0-beta.37**: SDK `11.0.100` (the band
`global.json` pins), runtime and Windows Desktop `11.0.0`, C# 15 by `LangVersion latest`, and the three packages that are
serviced with the runtime (`Microsoft.Extensions.Hosting`, `System.Diagnostics.EventLog`, `Microsoft.Data.Sqlite`) at
`11.0.0`. .NET 11 shipped on 2026-11-10; the owner moved the line the next day.
- **What decided it, and what did not.** `20_OPEN_QUESTIONS` §G held a decision table written on 2026-10-31, before any
evidence: move only after (a) two alphas on the released runtime had passed the gate and an owner's run of a hooked
title, (b) the owner's word on Windows 10, and (c) something .NET 11 brings being wanted. **The owner set (a) aside** —
no alpha had run on the released runtime; three had run on Release Candidate 3 (alpha.1 to alpha.3, the gate green
each time) — decided (b) as D95, below, and for (c) asked for C# 15 in the same release (D96). What stands in for (a)
is this release's own gate on 11.0.100 (`spike-notes` §28) and owner-only item 34.
- **Windows 10 (D95).** Microsoft supports .NET 11 on Windows 11, and on Windows 10 only in its LTSC and IoT editions.
FrameLedger does not block Windows 10 22H2: `SupportedOSPlatformVersion` stays `10.0.19045.0`, so CA1416 still holds
the code to it, and nothing checks the Windows version at start. It is **untested** there — it always was
(`ACCESSIBILITY.md`, `08_UI` §Accessibility) — and now **unsupported** too; NFR-8, `LIMITATIONS.md`, the README and the
guide say so.
- **The processor.** .NET 11 needs x86-64-v2 — SSE4.2 and POPCNT among its instruction sets — and does not start without
it ("The current CPU is missing one or more of the baseline instruction sets"). Its ReadyToRun code targets x86-64-v3
(AVX2 among others); Microsoft says a processor that lacks those may pay for it at start-up, where that code is
compiled again. Every processor Windows 11 accepts has x86-64-v2.
- **What .NET 11 changes in the software**, measured or read: `IHost.RunAsync` rethrows the exception of a
`BackgroundService` that failed, so the Agent's `--serve` leaves through its crash report where on .NET 10 it exited 0
(`10_LOGGING` §Crash handling; `HostBackgroundFailureTests` holds it); `Microsoft.Data.Sqlite` 11 declares SQLitePCLRaw
3.0.5, whose engine is SourceGear's `SQLite` package; nine `Microsoft.Extensions.*` libraries are part of the runtime
and no longer packages beside it (`legal/licenses/nuget/`: 78 packages became 69).
- **C# 15 (D96)** is the language the main line compiles, and none of its constructs is used yet: the analyses do not
read them (`13_CI_CD` §C# 15 and the analyses, `spike-notes` §29). What beta.37 takes from .NET 11 is API.
- **Going back** is installing 0.1.0-beta.36 over a later build: the ledger's schema did not move with the runtime.
## The alpha line (owner decisions D73–D76, 2026-10-31; resting since D98, 2026-11-11)
The owner, 2026-10-31: an alpha build, "a place where new .NET versions are tried", with code updated where needed and a
warning that it is an experiment used at one's own risk. **The alpha line** is the source of a released beta built on a
prerelease .NET — .NET 11 RC3 first (SDK `11.0.100-rc.3.26522.148`, runtime and Windows Desktop
`11.0.0-rc.3.26522.148`) — from the long-lived `alpha` branch, published as `vX.Y.Z-alpha.N` GitHub pre-releases on
Velopack's `alpha` channel (`13_CI_CD` §The alpha line, `11_UPDATER` §Channels).
warning that it is an experiment used at one's own risk. **The alpha line** is the source of a released beta built to
try a newer .NET — at times a prerelease one — from the long-lived `alpha` branch, published as `vX.Y.Z-alpha.N` GitHub
pre-releases on Velopack's `alpha` channel (`13_CI_CD` §The alpha line, `11_UPDATER` §Channels). It tried .NET 11:
Release Candidate 3 in alpha.1 to alpha.3 (2026-10-31 to 2026-11-06), the release itself in alpha.4.
- **What differs from main, and nothing else** — `git diff main...alpha` is the whole difference: `RELEASE_LINE`
(`alpha`); `global.json` (the prerelease SDK, `allowPrerelease: true`); the TFM in `Directory.Build.props`
(`net11.0-windows10.0.22621.0` — `SupportedOSPlatformVersion` stays `10.0.19045.0`, so CA1416 still holds the code to
Windows 10 22H2 although Microsoft does not support .NET 11 on its Home and Pro editions); the three Microsoft packages
in `Directory.Packages.props` that are serviced with the runtime (`Microsoft.Extensions.Hosting`,
`System.Diagnostics.EventLog`, `Microsoft.Data.Sqlite`, each `11.0.0-rc.3.26522.148` — the last declares SQLitePCLRaw
3.0.5, SQLite 3.53.4, where main's 10.0.14 declares 2.1.13, SQLite 3.53.3); `legal/licenses/nuget/` regenerated for
that package set; and a fix only .NET 11 can take, if one is ever forced. C# is what `LangVersion latest` gives the
SDK — C# 15 on the 11 SDK — and the source uses none of its syntax: **main must keep building on .NET 10** (D76: retarget
only, no preview feature, no runtime-async).
- **The line rests since 0.1.0-alpha.4 (D98, 2026-11-11).** The main line moved to .NET 11 the day after it shipped
(§The main line's .NET), so alpha.4 is 0.1.0-beta.37's source on the same .NET 11.0.0 — published so that a copy of
alpha.3 leaves the release candidate by an update, since an alpha never updates to a beta. After it no alpha is cut
until a newer .NET has a preview to try (.NET 12's, expected in 2027). While the line rests, the `alpha` branch differs
from main in **`RELEASE_LINE` alone** (`git diff main alpha --name-only` prints that one name), and a resting alpha
gets no update: its anti-cheat list, which reaches a user only inside a release (`13_CI_CD` §rules-publish.yml), stays
the one it shipped with. Its notes, the guide and the Alpha notice tell its user to install the beta — **and the build
itself does**: `Directory.Build.props` stamps its `FlAlphaLineRests` property (true on main since beta.37) into every
assembly, `Shared.AlphaLine.Rests` reads it, and an alpha cut while it is true shows a banner that says the line is
paused and to install the beta, in place of the experimental-build one (`App/Services/AlphaTexts`; D97: the words
"a prerelease .NET" appear only where the runtime's own description is a prerelease's).
- **The switch — what `alpha` differs from main in while it tries a newer .NET, and nothing else:** `RELEASE_LINE`
(`alpha`); `global.json` (that SDK, with `allowPrerelease: true` when it is a prerelease); the TFM in
`Directory.Build.props` (`SupportedOSPlatformVersion` stays `10.0.19045.0`); the three Microsoft packages in
`Directory.Packages.props` that are serviced with the runtime (`Microsoft.Extensions.Hosting`,
`System.Diagnostics.EventLog`, `Microsoft.Data.Sqlite`); `legal/licenses/nuget/` regenerated for that package set and
the *Transitive* row of `legal/THIRD_PARTY_NOTICES.md`; `FlAlphaLineRests` set to `false` in `Directory.Build.props`;
and a fix only the newer .NET can take, if one is ever forced.
As it stood for .NET 11 RC3: SDK `11.0.100-rc.3.26522.148`, `net11.0-windows10.0.22621.0`, the three packages at
`11.0.0-rc.3.26522.148` (`spike-notes` §24). **No newer language syntax and no preview feature on `alpha`** (D76): the
source is main's, and main compiles with its own SDK.
- **Everything else lives on main**, dormant on the main line. `RELEASE_LINE` is read like `VERSION` and stamped into
every assembly as the `FrameLedger.ReleaseLine` metadata (`Directory.Build.props` refuses any value but `main` and
`alpha`); `Shared.BuildLine` reads it back for the App and the Agent, which compose with it rather than reading it
where they act, so a test hands either line to the code it tests. On the alpha line they act on it (D74): the Alpha
notice in the Legal Gate with its own tick box, the Agent's wait for it, the ALPHA badge and banner, the locked update
channel (`08_UI`), and the ledger's copy before an alpha first opens it (D73, `06_DATA_MODEL` §Migrations). What the
alpha's SDK asks of the source and .NET 10 can also take lands on main.
alpha's SDK asks of the source and main's can also take lands on main.
- **The native layer does not change.** MSVC builds it from the same source; only its build id differs (the alpha's
tag). The hook, the guard and the Vulkan layer in an alpha are the beta's.
- **Sync, and the cadence.** After each beta release a pull request merges main into `alpha` with a merge commit (the
repository allows them; the linear rule is main's) and the next alpha is cut from it — **one alpha per beta, never
several in a row**, because Velopack reads only the newest ten releases (`11_UPDATER` §Channels). A conflict in
`legal/licenses/nuget/` is resolved by regenerating it on `alpha` (`tools/license-gather.ps1`). An alpha older than the
ledger's schema refuses to open it (`LedgerSchemaException`), so a beta that adds a migration is followed by an alpha
soon.
- **Building it locally:** check out `alpha` with the prerelease SDK installed — `global.json` selects it — and run
`./build.ps1 check` as on main.
- **The next .NET build** (an RC, the GA on 2026-11-10, a servicing release) is one line of `global.json` and three
package versions on `alpha`; RC3's support ends at the GA.
- **Whether the main line follows** is not decided here: `20_OPEN_QUESTIONS` §G, *The main line's .NET*.
- **Sync, and the cadence while it tries a .NET.** After each beta release a pull request merges main into `alpha` with
a merge commit (the repository allows them; the linear rule is main's) and the next alpha is cut from it — **one alpha
per beta, never several in a row**, because Velopack reads only the newest ten releases (`11_UPDATER` §Channels). A
conflict in a switch file is resolved for the switch; `legal/licenses/nuget/` by regenerating it on `alpha`
(`tools/license-gather.ps1`). An alpha older than the ledger's schema refuses to open it (`LedgerSchemaException`), so
a beta that adds a migration is followed by an alpha soon. **The sync that ended the .NET 11 trial** (alpha.4) went
the other way round: every switch file conflicted and main's version was taken for each, leaving `RELEASE_LINE`.
- **Building it locally:** check out `alpha` with the SDK its `global.json` names installed and run `./build.ps1 check`
as on main.
- **Starting the next trial** (a preview of the next .NET): the switch above on `alpha`, the whole gate in a throwaway
worktree first (`spike-notes` §24 is the pattern, with its decision table written before the run), and the Alpha
notice's sentence naming the .NET.
- **Whether the main line follows a trial** is the owner's each time: `20_OPEN_QUESTIONS` §G, *The main line's .NET*,
holds the table and what happened to it for .NET 11.
## Version
@@ -355,9 +407,12 @@ would not fit).
13b. `tools/accuracy-check.ps1 -SelfTest`, then live — the accuracy block is ONE text (`legal/ACCURACY.md`) embedded verbatim in `README.md` and `legal/DISCLAIMER.md` §4; a copy that drifted, a copy with no markers, or an empty source is red. Four self-test cases, both directions (§S23-6, 2026-09-06)
13c. **`tools/notice-check.ps1 -SelfTest`, then live** (beta.12, owner decision D45, 2026-10-03) — every first-party source file (`.cs .cpp .h .inl .ps1 .xaml`, tracked or new, minus `src/native/third_party/`) opens with `SPDX-License-Identifier: GPL-3.0-only` and `Copyright (C) 2026 poli0981 - additional terms under GPLv3 section 7: see NOTICE` in its own comment syntax, because GPLv3 §7 wants additional terms named "in the relevant source files" (`NOTICE` carries them); the second line may name another holder and year in the same form (a contributor's own file).
Seven self-test cases, both directions; `./tools/notice-check.ps1 -Fix` writes a missing header, and `tools/resx-gen.ps1` generates the accessor with it
13d. **`tools/wiki-build.ps1 -SelfTest`, then `-Check`** (beta.17, owner decisions D53/D54, 2026-10-07) — assembles the GitHub Wiki from `wiki/`, `guide/` and `LIMITATIONS.md` into a temporary folder, as `wiki.yml` does at a release (`13_CI_CD` §wiki.yml), and fails on a link to a file, folder or section that does not exist, two sources for one page, a page that neither Home nor the sidebar leads to, a sidebar that lists the guide in another order than `guide/README.md`, a backticked repository path in a `wiki/` page that names nothing, and a version number written into a `wiki/` page (the footer says which release the wiki shows). Nine self-test cases, both directions
13d. **`tools/wiki-build.ps1 -SelfTest`, then `-Check`** (beta.17, owner decisions D53/D54, 2026-10-07) — assembles the GitHub Wiki from `wiki/`, `guide/`, `LIMITATIONS.md` and (beta.35) `ACCESSIBILITY.md` into a temporary folder, as `wiki.yml` does at a release (`13_CI_CD` §wiki.yml), and fails on a link to a file, folder or section that does not exist, two sources for one page, a page that neither Home nor the sidebar leads to, a sidebar that lists the guide in another order than `guide/README.md`, a backticked repository path in a `wiki/` page that names nothing, and a version number written into a `wiki/` page (the footer says which release the wiki shows). ~~Nine~~ Ten self-test cases since beta.35 (the statement's page, D88), both directions
14. `tools/changelog-check.ps1 -SelfTest` — nine cases, five expected RED. The live half needs a pull request's changed-file list and is supplied by `ci.yml`
14b. `tools/release-notes.ps1 -SelfTest` — five cases over a fixture changelog (a dated and an undated section, a missing one, a prefix that must not match, an empty one), and since beta.24 three over a temporary tree: a version with its own `docs/changelog/<version>.md` is read from it, one without from `CHANGELOG.md`, a prefix of a file name is not that file. The live half needs a tag's version and runs in `release.yml`, where a missing or empty section is a red release (P4 PR-5)
14e. **`tools/nvapi-readonly-check.ps1 -SelfTest`, then live** (owner decision D108, 2026-11-11) — every `NvAPI_*` call in `src/native/FrameLedger.NvapiBridge/src/fl_nvapi_bridge.cpp`, the one translation unit that links `nvapi64.lib` into a shipped binary, is on `tools/nvapi-allowed-calls.txt` and every name there is called; a name that sets, saves, restores, deletes, registers, resets, enables, disables or controls something is red **even when the list names it**, and so is one that returns an identifier (D104) — three reads the verb rule would refuse are exempt by name, each with its reason. Comments and string literals are taken out first, so a setter named in the comment that says it is never called is not a call. Seventeen self-test cases, both directions. Text only: it runs with or without the native build. `18_GPU_VENDOR_APIS` §Runtime policy's "never call a setter", which was a sentence until the bridge went from fourteen calls to forty-four
14g. **`tools/input-readonly-check.ps1 -SelfTest`, then live** (owner decisions D155–D163, 2026-11-14) — what FrameLedger asks Windows about controllers, the keyboard and the mouse. Every Windows function the input reader calls is declared in ONE file, `src/FrameLedger.Infrastructure/Input/Win32Input.cs`, and is on `tools/input-allowed-calls.txt`; every name there is declared; a name that sends to a controller (`XInputSetState`, any `HidD_*`), injects input (`SendInput`), reads a key (`GetAsyncKeyState`, `GetKeyState`, `GetKeyboardState`, `XInputGetKeystroke`) or installs a hook is red **even when the list names it**; no managed source outside `src/FrameLedger.Infrastructure/Input/` names an input function at all, and `NativeMethods.txt` asks CsWin32 for none; raw input is registered for joysticks and gamepads only — `InputReader.RegisteredUsages` is exactly usage page 1, usages 4 and 5, and the backend builds its registration from it and writes no usage as a number; and **no native source names an input function** — nothing loaded into a game reads input. Comments and string literals are taken out first. Seventeen self-test cases, both directions. Text only; it does not look at the tests, which may drive a real keyboard for the App's own shortcuts. `21_INPUT_DEVICES` §What is never read, which the Privacy Policy repeats to the user
14f. **`tools/overlay-fields-check.ps1 -SelfTest`, then live** (owner decision D116, 2026-11-12) — the same for the code loaded into a game: every member of a vendor's struct the Overlay reads for Advanced metrics is read in ONE header, `src/native/FrameLedger.Overlay/include/fl_adv_fields.h`, through parameters with fixed names, and is on `tools/overlay-allowed-fields.txt`; every name there is read; the parameters are not aliased and `->` reaches nothing but `desc` and the Overlay's own `region`; no other Overlay or Vulkan-layer source reads a listed member; **a camera's position, axes and matrices** (`cameraPos`, `cameraUp`, `cameraRight`, `cameraFwd`, the five matrices, `cameraPinholeOffset`, `cameraPosition`, `cameraForward`) **are named nowhere in those sources and not on the list, whatever the list says**; and neither `sl::Constants` nor `ffxDispatchDescFrameGenerationPrepareV2` is copied by value or measured with `sizeof`. Comments and string literals are taken out first. Nineteen self-test cases, both directions. Text only. It reads names, not types — which struct version carries a member is `17_HOOK_ENGINE` §Which struct version carries which field and ctest `fl_adv_fields`
14d. **`tools/docs-split-check.ps1 -SelfTest`, then live** (beta.24, owner decision D66) — `CHANGELOG.md` holds no released section beside `[Unreleased]`; each `docs/changelog/<version>.md` holds one `## ` heading, its own `## [<version>] - date`, and no relative link (the App renders release notes as `CHANGELOG.md`); `CHANGELOG.md` and the files under `docs/changelog/`, and `docs/HANDOFF.md` and the files under `docs/handoff/` — and since PR-5 `docs/20_OPEN_QUESTIONS.md` and `docs/open-questions/`, `docs/spike-notes.md` and `docs/spike-notes/` — name each other in both directions. Eleven self-test cases, both directions
14c. **`tools/buildid-check.ps1 -SelfTest`** (beta.18) — eight cases, both directions: a binary built from a dirty tree (`-dirty`, or the short form's `+`), one with no id, a missing one, and an expected id that is dirty or not an id are red. The live half needs the release's clean-checkout id and runs in `release.yml` over the published tree (§Release-time token substitution)
15. `tools/resx-audit.ps1` — ~~**skipped loudly; it does not exist, and no `.resx` file does either**~~ **built 2026-09-13 (P3 PR-2)**: `-SelfTest` first (11 fixture cases, both directions), then the live pass over every `Strings.resx` family under `src/` — key sets equal across en/vi/ja, `Safety_*` in `ja` marked for review or signed, `Strings.Designer.cs` current with `tools/resx-gen.ps1`. A tree with no family is red (`09_I18N` §Translation workflow)
+400 -26
View File
@@ -95,7 +95,7 @@ FrameLedger uses the **`poli0981/.github` ops repo** where its templates fit, an
- **Placeholder guard:** fails if any `{{` token survives in `README.md` or `legal/*.md` *(other than `{{RELEASE_DATE}}`, which `release.yml` substitutes — added 2026-10-04)*. Those are shipped, legally operative documents (FR-11 displays them in the first-run Legal Gate); an unsubstituted `{{DEVELOPER_NAME}}` in an EULA is not a cosmetic defect.
- ~~`permissions: contents: read`.~~ `permissions: contents: read, pull-requests: read` — the second for the Changelog step's list of a pull request's files, read-only *(corrected 2026-10-04)*. *Since 2026-10-25 per job:* the workflow grants nothing (`permissions: {}`), `check` holds those two and `sonarcloud` `contents: read` alone — a workflow-level grant reaches every job (SonarCloud `githubactions:S8264` on the first analysis from CI).
### `codeql.yml` — push, PR, weekly cron · **repo-local**
### `codeql.yml` — push, PR, weekly cron, `workflow_dispatch` (since 2026-11-11) · **repo-local**
> **Not a caller stub, which this heading said until 2026-08-06.** `codeql.yml`'s own header says
> the opposite: the ops repo has `codeql-mixed.yml`, but C++ needs manual build mode driven by the
@@ -105,6 +105,36 @@ FrameLedger uses the **`poli0981/.github` ops repo** where its templates fit, an
- **And `actions`, since 2026-10-25** (build mode `none`): CodeQL reads `.github/workflows` for injection from event data, untrusted checkouts and over-broad tokens — the class of finding SonarCloud raised once (`githubactions:S7631`, `wiki.yml`, beta.26). Its context, `analyze (actions, none)`, is not one of `main`'s required checks; adding it to them is the owner's setting.
- **Concurrency, since 2026-10-25:** a new push to a pull request cancels that pull request's run; a push to `main` and the weekly scan each run in a group of their own (the run id), so none is cancelled or left pending, and the default branch's alerts come from a finished analysis.
- `permissions: security-events: write, contents: read`.
- **On .NET 11 since beta.37 (D94) — measured on the pull request that moved the line (#376, run 42916491104, CodeQL
2.27.4).** CodeQL's documentation lists C# up to 14 and .NET up to 10, and `analyze (csharp, none)`, a required
context, had never met `net11.0`. With build mode `none` the extractor reads the source with its own compiler front
end and uses a `dotnet` only for references: it found the 11.0.100 SDK `setup-dotnet` had installed and used that (on
.NET 10 it downloaded 10.0.100 for itself), restored, and reported 363 resolved references and none unresolved, 9
types it could not resolve (main's last run on .NET 10, 42862718138: 23, in three `Windows.Win32` namespaces), and
**`CodeQL scanned 1220 out of 1220 C# files`** — main's line, word for word — in 4 min 5 s. The source it read used
no C# 15 construct. **What it reads of C# 15 is measured construct by construct before one is used** (D96): a
construct in whose file the extractor reports an error is not used on main until CodeQL reads it, and the same holds
for Qodana's engine (§Qodana).
*(Corrected the same day, two pull requests later: that run was an **overlay** — on a pull request the C# analysis
extracts the files it changed on top of the database cached from `main`, then a .NET 10 one — so its "1220 out of
1220" was `main`'s baseline figure and its other numbers the changed files'. The whole source on `net11.0` was first
read by `main`'s own run after the merge (42921492229, "Overlay mode not enabled"): 1,221 source files, 363
references resolved and none unresolved, 23 types unresolved in the three `Windows.Win32` namespaces — the figure it
had on .NET 10 — and "CodeQL scanned 1221 out of 1221 C# files", in 4 min 45 s. So CodeQL reads `net11.0` as it
read `net10.0`; and for a construct it has not met, a pull request's run is not the evidence — a
`workflow_dispatch` on the branch is, §C# 15 and the analyses.)*
### `qodana.yml` — push to `main`, pull requests into `main`, `workflow_dispatch` · **repo-local, built 2026-11-10 (D93); Qodana for .NET since 2026-11-11 (D99)**
- ~~One job, `qodana`, on `ubuntu-latest` — the only Linux job in this repository, because the linter is a Linux
container~~ Two jobs, on `windows-latest`, each a report of its own: **`native`** (*Qodana (native)*) has CMake
write a Visual Studio solution and reads the C++ of `src/native` with `qodana-native.yaml`'s settings; and
`managed` (*Qodana (managed)*): `JetBrains/qodana-action` (pinned to
`v2026.2.2`'s commit; Dependabot moves the pin) runs **Qodana for .NET** in native mode over `FrameLedger.slnx`
with `qodana.yaml`'s settings, after a `dotnet build`. No job here runs on Linux any more. §Qodana below says what
it reads, what was changed from JetBrains' template and where an inspection is switched off.
- **Advisory, with a gate of zero problems:** one problem turns the job red; it is not a required context.
- `permissions: contents: read, checks: write, pull-requests: write` — the annotations and the one comment; granted to
the job, the workflow grants none.
### `release.yml` — on tag `v*` · **repo-local, built 2026-09-14 (P4 PR-5)**
@@ -211,9 +241,9 @@ FrameLedger uses the **`poli0981/.github` ops repo** where its templates fit, an
### `wiki.yml` — after a successful Release run, and by hand · **repo-local, built 2026-10-07 (beta.17, D53/D54)**
- **What:** the GitHub Wiki is a copy of the repository's own pages — `wiki/` (Home, the sidebar, the developer pages)
and `guide/` (the user guide: the End-user part, the same pages the App shows) with `LIMITATIONS.md`.
`tools/wiki-build.ps1` names each page (a guide page after its first heading, `guide/README.md` as *User-guide*,
`LIMITATIONS.md` as *Limitations*), rewrites links between pages to page names and every other repository link to
and `guide/` (the user guide: the End-user part, the same pages the App shows) with `LIMITATIONS.md` and, since
beta.35 (D88), `ACCESSIBILITY.md`. `tools/wiki-build.ps1` names each page (a guide page after its first heading,
`guide/README.md` as *User-guide*, `LIMITATIONS.md` as *Limitations*, `ACCESSIBILITY.md` as *Accessibility*), rewrites links between pages to page names and every other repository link to
its address on GitHub **at the release's tag**, drops each page's first heading (the wiki prints the page name), and
writes a footer naming the tag. The same assembly is `build.ps1`'s `wiki-check` gate (`12_BUILD` §Local quality gate,
13d), so a dead link is red on the pull request, never on the published wiki.
@@ -235,28 +265,31 @@ FrameLedger uses the **`poli0981/.github` ops repo** where its templates fit, an
- `permissions: contents: write` — the repository's default workflow token is read-only (read from the API,
2026-10-07), and pushing to the wiki repository needs write.
## The alpha line — the `alpha` branch (owner decisions D73–D76, 2026-10-31)
## The alpha line — the `alpha` branch (owner decisions D73–D76, 2026-10-31; resting since D98, 2026-11-11)
What it is and what differs on the branch: `12_BUILD` §The alpha line. What CI and the release do with it:
What it is, what differs on the branch and why it rests since 0.1.0-alpha.4: `12_BUILD` §The alpha line. What CI and the
release do with it:
- **`ci.yml`** runs `check` on a push to `alpha` and on every pull request into it, under the branch's own `global.json`
— `setup-dotnet` installs the exact prerelease SDK it names whatever the `rollForward`. The `sonarcloud` job does not
run for the branch (the organisation's plan analyses main alone, §SonarCloud), and neither does the `Changelog` step:
what an alpha changes is written in its release's own file on main, never in an `[Unreleased]` entry on `alpha`, which
would collide with main's at every sync.
- **`codeql.yml`** stays on main: the source is main's, and CodeQL's extractor reads C# 14 (CodeQL 2.25.6, June 2026)
while the alpha's compiler is C# 15.
— `setup-dotnet` installs the exact SDK it names whatever the `rollForward`, a prerelease one included. The
`sonarcloud` job does not run for the branch (the organisation's plan analyses main alone, §SonarCloud), and neither
does the `Changelog` step: what an alpha changes is written in its release's own file on main, never in an
`[Unreleased]` entry on `alpha`, which would collide with main's at every sync.
- **`codeql.yml` and `qodana.yml`** stay on main: the source is main's, and it is analysed there. *(Until beta.37 the
reason given here was the alpha's compiler — C# 15, where CodeQL's extractor read C# 14. Main compiles C# 15 since
beta.37; what CodeQL reads of it is §`codeql.yml`'s to say.)*
- **The branch is not protected** (an owner's setting; none was asked for): pull requests into it are how changes land
(squash), and the sync from main is a pull request merged with a merge commit — main's linear rule is main's alone.
- **A release of the line**, in order: the alpha's notes on main (`docs/changelog/x.y.z-alpha.N.md` opening with the
warning, its row in `CHANGELOG.md`'s Releases table, a line in HANDOFF's START HERE) → the sync pull request into
`alpha` → a `release.yml` dispatch on `alpha` (`gh workflow run release.yml --ref alpha -f version=x.y.z-alpha.N`) →
the signed tag on `alpha`'s head → the run (§release.yml: the line's tag rule, `--channel alpha`). **No wiki publish**
(`wiki.yml` skips an `-alpha.` tag) and **no release line moves**: README, ACCURACY, the Disclaimer's copy and
LIMITATIONS keep naming the newest beta.
- **The cadence:** one alpha per beta, cut from it. Velopack's GitHub source reads the newest ten releases (one page,
`perPage` 10), so ten alphas in a row would hide every beta from the betas' own update check, and the other way round
(`11_UPDATER` §Channels).
(`wiki.yml` skips an `-alpha.` tag) and **no release line moves**: README, ACCURACY, the Disclaimer's copy,
LIMITATIONS and ACCESSIBILITY keep naming the newest beta.
- **The cadence while the line tries a .NET:** one alpha per beta, cut from it. Velopack's GitHub source reads the
newest ten releases (one page, `perPage` 10), so ten alphas in a row would hide every beta from the betas' own update
check, and the other way round (`11_UPDATER` §Channels). **While it rests** no alpha is cut, and that limit has a
consequence the alpha's notes state: once ten betas have followed alpha.4, a copy of alpha.3 no longer finds it.
## SonarCloud — from CI, `ci.yml`'s `sonarcloud` job (owner decision D68, 2026-10-25)
@@ -324,7 +357,10 @@ from CI.
writes into the checkout, since the scanner walks all of it (`build/`, with the sources CMake fetches there, `out/`,
every `bin/` and `obj/`, `TestResults/`) — SonarCloud also drops what `.gitignore` names (the first run: "63 files
ignored because of scm ignore settings"), and the patterns say it in the file rather than rely on that; the same
duplication exclusions; C++20; UTF-8; the OpenCover paths; the compilation database;
duplication exclusions *(and, since 2026-11-12, `src/FrameLedger.Domain/Metrics/HookFamilies.cs`: the Domain's
mirror of `FlHookFamily`, the same enum twice by design and held together by `MetricEnumMirrorTests`. The six
members 0.1.0-beta.45 added to both were 3.7 % duplication on new code against the gate's 3 %, and failed the gate
on the release's commit with no issue open)*; C++20; UTF-8; the OpenCover paths; the compilation database;
**`sonar.cs.roslyn.ignoreIssues`** — Roslynator, Meziantou, the VS threading and the .NET
analyzers already gate the build, warnings as errors, and what they report at a lower severity is this project's
choice, so only SonarCloud's own rules are imported; and **two rules ignored where they apply by design** — `S8417`
@@ -414,7 +450,7 @@ the comment below; the rest are fixed in the same train (beta.24 PR-2).
| `cpp:S935` `FlGetBuildId` returns no value | 1 | **False positive** | `FL_BUILD_ID` is defined by the build (`12_BUILD` §Version); the analysis compiles without it. |
| `cpp:S3519` out-of-bounds `memcpy` | 1 | **False positive** | `fl_patch_check.h` reads the prologue of a function this DLL hooked, under SEH, by design (`17_HOOK_ENGINE` §Unhooking). |
| `cpp:S836` garbage value returned | 1 | **False positive** | `hook-harness`'s `VtableOf` reads the vtable of a COM object it created itself. |
| `csharpsquid:S3869` `DangerousGetHandle` | 1 | **Accept** | `ParentDump` passes the dump file's handle to `MiniDumpWriteDump` inside the `using` that owns the stream, on the same thread, so it cannot be released during the call; CsWin32 generates no `SafeHandle` overload here (`NativeMethods.json`: `useSafeHandles: false`). |
| `csharpsquid:S3869` `DangerousGetHandle` | 1 | **Accept** | `ParentDump` passes the dump file's handle to `MiniDumpWriteDump` inside the `using` that owns the stream, on the same thread, so it cannot be released during the call; CsWin32 generates no `SafeHandle` overload here (`NativeMethods.json`: `useSafeHandles: false`). **Raised again 2026-11-12:** #398 moved that call into `TryWriteOnce` (the dump is now tried three times), and SonarCloud tracks an issue by where it is — the accepted one closed and a new one opened on the same expression, as a blocker on new code. Nothing about the call changed; it is to be accepted again (HANDOFF owner-only item 38 (j)). |
| `cpp:S1669`, `cpp:S6191`, `csharpsquid:S4036`, `githubactions:S7637`, `csharpsquid:S2190`, `S3427`, `S8949` | 35 | **Fixed** in beta.24 PR-2 — 5 of the 23 `S1669` were in the vendored Vulkan headers and leave with the exclusions | — |
**`cpp:S5813`, the 17 accepted, and why each string is terminated** (lines as of beta.24):
@@ -457,6 +493,341 @@ on `src/FrameLedger.Infrastructure/**/*` and `src/FrameLedger.Shared/**/*`, so n
stop reappearing on pull requests. *(Since D68 the `S8417` and `S6640` rows are `SonarQube.Analysis.xml`'s, in the
repository; `S6232`'s are not — a union read outside `LARGE_INTEGER` would be a finding — and its 24 stay marked.)*
## Qodana — `qodana.yml`, a workflow of its own (owner decisions D93, 2026-11-10; D99 and D100, 2026-11-11)
**Since 2026-11-11 the linter is Qodana for .NET, on the Ultimate licence (D99).** The owner took an Ultimate licence for
the Qodana Cloud organisation and asked for Qodana over every language. So the linter is **Qodana for .NET**
(`qodana-dotnet`: Rider's engine) where it was the Community one, in **native mode** on a Windows runner: no container,
the projects built and loaded with the SDK the job installs from `global.json`, and no `EnableWindowsTargeting`.
`qodana.yaml` names the linter and the mode (`withinDocker: false`); nothing is an action argument, so `qodana scan` on
a developer's PC reads the same file. The job is `managed` — *Qodana (managed)* in a pull request's checks: checkout,
`setup-dotnet`, the NuGet cache restored, `dotnet build -c Release` first — so that a build that fails is one red step
with the compiler's own message, not a report of unresolved symbols — then the action, pinned by commit. No MSVC step,
on purpose: it exports `Platform=x64`, which MSBuild reads as a solution platform. The linter's log says what the token
bought: "Qodana license plan: Ultimate", and the Qodana Cloud project it is linked to (`frameledger`). Still advisory,
still a gate of zero, still `main` and its pull requests alone. From here down to *As it was built* is the job as it
is; below that is D93's own account, which is now history.
**Two reports (D99).** Asked how to gather the C# and the C++, the owner chose two reports from the start, and made a
second Qodana Cloud project (`frameledger-cpp`, the `QODANA_TOKEN_NATIVE` secret). The managed report leaves
`src/native` out; the native one (*The native report*, below) reads nothing else, so every file is in exactly one.
**What is read, and by what.**
| What | Read by | |
|---|---|---|
| The C#, XAML and `.resx` of the thirteen projects | `managed` | ReSharper's engine, as under D93 |
| Markdown and XML around the solution | `managed` | the IntelliJ platform's own inspections — 23 and 10 results on the first run |
| JetBrains' security rules over the C# — pattern rules for a process start, an SQL text and the like | `managed` | new with this linter (its OpenGrep plugin) — 36 results on the first run, every one read; the files they were read in are listed in `qodana.yaml`, so a new one is reported |
| The C++ of `src/native` | `native` | ReSharper's C++ engine and the clang-tidy checks it bundles, over a Visual Studio solution CMake writes for it |
| JSON and YAML, the workflows among them | nothing reported | whether the profile inspects them and found nothing, or does not inspect them, was not established; CodeQL's `actions` language reads the workflows |
| PowerShell (`build.ps1`, `tools/`), CMake files, the HLSL shader | no linter of JetBrains' reads them here | SonarCloud reads the PowerShell |
**The first runs on this linter, and what was done (2026-11-11).** Run 42923079119 — the job with no settings beyond the
solution: **3,953 problems** in 92 inspections, and 3,607 of them **hints**: the lowest level an IDE has, which the
editor offers without underlining anything (this linter files them as `TYPO`, Qodana as *Low*). Meziantou's "name the
parameter" 1,084 times, a trailing comma 979, a XAML binding whose `DataContext` the engine cannot know 414, the MVVM
Toolkit's partial-property proposal 224, "invert `if`" 114 — and every other Roslyn analyzer message below the level
this build makes an error. **Hints are off as a severity**, one group in `qodana.yaml` (`severity:TYPO`): the Community
linter never reported them, and D93's rule is about warnings and suggestions. With that, run 42928814104: **347**.
| What | n | Done |
|---|---|---|
| `OverriddenWithSameValue` — a vi or ja value equal to the English one | 271 | off in `.editorconfig` (`[*.resx]`): product and technology names, units and format strings are not translated, and a ja safety string is the English verbatim until a reviewer signs it |
| `csharp_injection_rule-CommandInjection` — a process started with something that is not a literal | 35 | **read, one by one** (below): 11 places in `src`, 24 in tests; one place was hardened. The rule stays on, and the files read are left out of the plugin's inspection by path in `qodana.yaml` (*How the security rules are scoped*, below), so a new file that starts a process is reported |
| `MarkdownIncorrectTableFormatting` | 20 | off in `qodana.yaml`: Rider wants columns padded to one width; these tables are written to be diffed |
| `XmlHighlighting` in `src/FrameLedger.App/app.manifest` | 10 | that file ignored for that inspection: the linter has no schema for a Windows application manifest |
| `UnusedVariable.Compiler` and `RedundantUsingDirective` in `PrimaryDisplay.cs` | 6 | **the linter's own fault**, and the file is left as it is: resolving `PInvoke.EnumDisplaySettings(null, …)` — the `null`-to-`PCWSTR` conversion — the engine of this linter's build (QDNET-262.12299) fails an internal assertion (`IsStandardImplicitConversionFromExpression(conversion.Kind)`, in its `backend.log`), loses the method, and reports its three locals and the file's three usings as unused. The compiler builds the file with warnings as errors, and InspectCode 2026.2.3.1 and 2026.3.0 report nothing in it. Those two inspections are excluded for that one file in `qodana.yaml`, to be removed when a linter build reads it |
| `MarkdownUnresolvedHeaderReference` | 3 | off in `qodana.yaml`: all three resolve on GitHub, whose anchor rule the linter does not follow (`#display-mode--both-tiers-beta10`) |
| `csharp_injection_rule-SQLInjection` | 1 | the same: the one place assigns a constant command text and binds every value as a parameter, as every statement in this repository does; its file is listed with the others |
| `RedundantBoolCompare`, in a test | 1 | fixed — the expression also carried a clause its first half implied |
With those settled, run 42940926340 (2026-11-11, this pull request's branch, the whole solution): **no problem
reported** — the gate's zero, on this linter.
**How the security rules are scoped — by the plugin's one inspection, not by a rule.** Every result of the
OpenGrep plugin's several hundred rules is produced by one inspection, `OpenGrepInspection`; a rule's own id
(`csharp_injection_rule-CommandInjection`) is in the report and is honoured nowhere. Three runs measured it — a
profile entry's `ignore` (42932842462), an `exclude` path under the rule's id (42935752891), `enabled: false`
on the rule (42938295264): all 36 results stood each time, and the linter's log shows why — the plugin hands its
server every rule file whatever the profile says. What the plugin's inspection does honour is a path: an
`exclude` entry named `OpenGrepInspection` (its scopes are in the report's `log/effective.profile.xml`). So the
eleven files of `src` the results were read in, and `tests`, are listed there — and in the run above the
inspection was performed on 943 files, where it had been 1,308, and reported nothing: it is on, and it reads
what is not listed. The price is stated in `qodana.yaml`: a listed file is left out of **all** of the plugin's
rules, not only the one that fired. The plugin is also most of the job's time — about six of its fifteen
minutes.
**A second tripwire, in the gate: `ProcessStartSitesTests`.** Qodana is advisory, and its rule is a pattern for
`Process.Start`. The test sweeps `src` for every way a process is started here — `Process.Start`, a
`ProcessStartInfo`, `CreateProcessWithTokenW` through CsWin32, and the registration of the logon task — and fails
on a file that is not in its list, until that file has been read and listed. `Startup/UnelevatedProcess`, which
the rule never saw, is in it.
**Every place that starts a process, as read on 2026-11-11.** The rule is a pattern, not a data flow: it reports any
`Process.Start` whose argument is not a literal. What matters at each is what is started and where its path and
arguments come from.
| Place | What it starts | Where the path and the arguments come from |
|---|---|---|
| `App/Services/AgentLauncher` | `FrameLedger.Agent.exe` | the file beside the App (`AppContext.BaseDirectory`); arguments the App builds (`--serve`, the data folder) |
| `App/Services/AgentTool` | the same Agent, with one maintenance flag | a constant at each of its two callers (`--install-task`, the Settings page's flags) |
| `Infrastructure/Startup/AgentElevation` | the Agent, elevated (`runas`) | the Agent's own path; `--serve --for-user <this user's SID>` |
| `Infrastructure/Startup/LogonTask` | nothing itself — it registers the logon task, whose action Windows starts at sign-in | the Agent's path and constant arguments (`--serve`), into a Task Scheduler definition for this user, at the lowest privileges |
| `Infrastructure/Diagnostics/CrashDumpWriter` | this process's own executable, as the dumper | `Environment.ProcessPath`; an `ArgumentList`, no shell |
| `Infrastructure/Capture/ProcessLauncher` | the game | the executable the user added and enabled hooking for, and the launch arguments the user wrote for it; no shell |
| `App/Services/ShellStarter` | a game, directly or through its store (`steam://…`, `explorer.exe shell:AppsFolder\<AUMID>`) | the launch `GamePlayer` resolves from the record of a game the user added (D59, D80) |
| `App/Services/Explorer` | `explorer.exe` on a folder or a file | a quoted path of FrameLedger's own: the data folder, the logs, a bug report's zip, the licences folder |
| `App/Services/ShellUrlOpener` | the browser or the mail client | the project's own pages, or a Markdown link `MarkdownLinks` has already held to `http`, `https` and `mailto`. **Hardened here**: the opener now refuses any other scheme itself (`ShellUrlOpenerTests`) — the shell runs whatever it is handed, and a `file:` address is a program |
| `App/ViewModels/SettingsViewModel` | the browser | one constant address |
| `Infrastructure/Startup/UnelevatedProcess` | the Agent or a game with the shell's token, from an elevated process (`CreateProcessWithTokenW`) | its two callers above, `AgentLauncher` and `ProcessLauncher` — not a `Process.Start`, so the rule did not see it; the sweep does |
None takes a value from outside the PC or from a file FrameLedger did not write, except the game's own path and
arguments — which starting the game the user chose is.
**The native report — `qodana-native.yaml`, the job `native` (D99, 2026-11-11).** The C++ of `src/native`, read by
the same linter — Qodana for .NET, whose C++ is ReSharper C++ — in native mode on a Windows runner. It is the only
linter of JetBrains' that can: the native layer is MSVC-only, and the two C/C++ linters (`qodana-cpp`,
`qodana-clang`) are Linux images. It reads C++ through a Visual Studio solution, and `src/native` is a CMake tree, so
the job has CMake write one first: the preset **`x64-vs-analysis`** (generator "Visual Studio 18 2026", tests and
tools on) puts `FrameLedgerNative.slnx` under `build/native/x64-vs-analysis`, and the job builds from it the one
target whose output is a header (`fl_ac_floor_gen`, the blocklist floor). Nothing else is compiled, and nothing from
that tree ships. Then the action, with `--config qodana-native.yaml`: the solution, `RelWithDebInfo` and `x64` — the
flags the shipped binaries are built with — and, left out by path, what is not FrameLedger's code (the build tree,
`third_party`, the vendor stubs, a compiled shader as a byte array). Its own results folder, cache keys and artifact
(`qodana-report-native`), and the Qodana Cloud project `frameledger-cpp` through `QODANA_TOKEN_NATIVE`. It posts no
comment and no annotations — a second job of the same linter would overwrite the managed job's — so its result is the
job's own status and its report. Advisory, and a gate of zero, like the other.
**What it runs** is ReSharper's C++ inspections and, through them, the clang-tidy checks and clang diagnostics that
engine bundles. clang never builds this code — MSVC does, with `/W4 /WX` — so what clang says is read as advice. Every
one of them has a key in `.editorconfig` (`resharper_cpp_…_highlighting`; a clang-tidy check's is
`resharper_cpp_clang_tidy_<check>_highlighting`), so D100's rule holds without a second file: no `.clang-tidy`.
**Its first runs, and what was done.** Run 42923079119: **4,695 problems**, 4,040 of them hints, which are off as on
the managed side. Run 42928814104: **655** — **494 distinct**, because a source file compiled into three targets is
reported three times.
| What | distinct | Done |
|---|---|---|
| A choice this code made, off for the native tree: where the compiler pads a struct (46); MSVC's own keywords, `__try` among them (54); a function pointer got by name cast to its real type, through `void*` (55); `reinterpret_cast` where memory is read as something else (5); `std::cmp_less` proposed for comparisons that already cast (20); a `default:` under a full enum switch (3); a copied verdict (4); the moves of a type that owns a handle (11); a zeroed out-parameter enum (8); an enum's width (5); "can be made const / constexpr / static" (52); "possibly unused `#include`" (15) | 278 | off in `.editorconfig`, each with its reason. Of the fifteen includes two were unused and were removed; the others are a file including what it uses although another header already brings it |
| A newer idiom proposed for working code — a range-for, `std::ranges`, `any_of`, `starts_with`, a raw string, an `if` with an initialiser | 53 | off: the native code is written to be read beside the hook path, an index and a plain loop |
| **The engine misreading**: `std::atomic_ref<T>::fetch_or` as an expression without side effects (7); a Win32, D3D12 or ffx-api structure zero-initialised and then filled field by field as "always rewritten" and "assigned, never used" (15); a test of a `constexpr` function as a redundant argument (3) | 25 | off, with what the engine gets wrong written beside the keys |
| Fields nothing names, in a structure Windows defines and the SDK's headers do not give (`RTL_PROCESS_MODULE_INFORMATION`, `PROCESS_MEMORY_COUNTERS_EX2`) and in a test's stand-in for a Vulkan object | 25 | off for those four files: the fields are what puts the others at their offsets |
| `%lu` with an `NvU32`, in `fl-probe-nvapi` | 19 | off for that file: NVAPI's header makes `NvU32` an `unsigned int` under clang and an `unsigned long` under MSVC — the linter's compiler, not the build's |
| The probes, the harness and the tests: an unchecked `printf` (35), `atoi` on the harness's own arguments (20), a `&pointer` handed to a `void*` parameter (7), a printf-shaped helper (4), a fake handle built from a number (4), a test's fixtures (4), `layout_test`'s run-time checks of compile-time facts (3), a macro that pastes a name (1) | 78 | off for `tools` and `tests` |
| Code loaded into a game (below) | 7 | kept as it is, off for those files; one of them is dead code, recorded for the owner |
| **Fixed**: `sizeof(a) / sizeof(a[0])` → `std::size(a)` (4); a multiplication done in a narrower type than the one it is stored in or compared with (2: a constant in the guard's sources, a test); a member function, and a type alias, nothing used (3) | 9 | in the code; none changes what a shipped binary does |
With those settled, run 42942068024 — the first with these settings — reported **one** problem: `LogNoteW`, now by
ReSharper's own inspection where clang's had said it before (below). And run 42943094415 (2026-11-11, this pull
request's branch): **no problem reported** — the gate's zero, in the native report too.
**Code loaded into a game — the Overlay, the Vulkan layer, the shared-memory headers, the guard's rule matcher.**
D100: there only a defect is fixed, in a pull request of its own with the hook's overhead measured (rule 5,
`14_TESTING` §Hook overhead), and a proposal about style is not acted on. About a quarter of the 655 results were in
those files; nearly all of them are the tree-wide kinds above. **The report found no defect there.** What is
particular to them, read and kept: two `if` chains whose arms have the same body and are different cases, each with
its comment (`dllmain.cpp`); MinHook's `MH_ALL_HOOKS`, which is `NULL` by its header; a retired enumerator kept so
that its bit is not reused; `slSetTag`, deprecated by a header whose replacement no measured module exports;
`ResolveTolerance` clearing its out-parameter with `memset`, as `ResetRules` does — a `Tolerance` is flags whose
default initialisers are zero. And one thing that is not as it should be: **`LogNoteW` in `dllmain.cpp`, which nothing
calls** — dead code. It is not a defect, so D100 leaves its removal to the owner (HANDOFF); until then that one file
is not read for an unused function (`.editorconfig` says so, in a section to be removed with it).
**"Zero" has to mean "analysed".** A report with no problem cannot tell a clean tree from one the linter did not
read. So the settings were proved on a branch that puts back three of the things fixed here — the unused alias, the
unused member function, the narrow multiplication. Run 42942390715 reported exactly those three, each by the
inspection that had found it the first time (`CppTypeAliasNeverUsed` in `tools/fl-probe-interposer/main.cpp`;
`clang-diagnostic-unused-member-function` and `bugprone-implicit-widening-of-multiplication-result` in
`tests/ring_test.cpp`), and `LogNoteW`, and nothing else. The branch was deleted. Do the same before trusting a
zero after a change to the preset, the settings or the linter's version.
**What it does not read.** CMake files and PowerShell have no inspection in this linter (SonarCloud reads the
PowerShell); the HLSL shader and the `.def` file are not C++. CodeQL (`analyze (cpp, manual)`, a required check) and
SonarCloud go on reading the C++ as before: this is a third reader, and the only one of the three that is ReSharper's.
**Where an inspection is switched off: by the engine that raised it (D100, amending D93's "never in `qodana.yaml`").**
What Qodana reports is fixed, unless the inspection argues with a choice this project made; then it is off, with the
reason on the line above — in the place the engine that raised it reads:
- **ReSharper's** — the C#, the XAML, the `.resx`, and the C++ of the native report with the clang-tidy checks it
runs — in **`.editorconfig`** (`resharper_*_highlighting`), where Rider and ReSharper in the editor read it too:
sixty-four keys for the managed code, fifty-four lines for the C++, some of them for a folder or a file alone.
- **What has no key there** — in **`qodana.yaml`**: a whole severity (the hints) and the IntelliJ platform's Markdown
and XML inspections under `profile.inspections`; under `exclude`, by inspection and path, the files JetBrains'
security rules were read in and a file the linter itself fails on.
- **What is not FrameLedger's code** is left out by path (`exclude`): `SonarQube.Analysis.xml`'s list, for its reasons.
**What it does not use, and what happens if the licence lapses.** Quick-fixes (the linter would push commits — the job
has no `contents: write`, and a change nobody read is not wanted), code coverage in Qodana (the gate's coverage and
SonarCloud's already read the same reports) and FlexInspect are Ultimate features this job leaves alone. Without a
valid licence the linter does not start, so the job is red — advisory, never a blocked merge; the way back is
`linter: qodana-cdnet` with the job on `ubuntu-latest`, as it ran until beta.37.
**Running it on a developer's PC.** `qodana scan` in the repository root reads `qodana.yaml`, downloads the linter and
needs the project's token in `QODANA_TOKEN` — the Ultimate linter checks the licence. Without a token, InspectCode
still gives ReSharper's part (*As it was built*, below), with `--toolset-path` at the 11.0 SDK's `MSBuild.dll`; the
Markdown, XML and security results are the Rider linter's alone.
**As it was built — the Community linter, 2026-11-10 to beta.37 (D93).** Kept as written; where it says *is*, read *was*.
**D93 — Qodana reads the C# and the XAML from CI.** The owner made the project on Qodana Cloud, set the `QODANA_TOKEN`
secret (2026-11-10) and asked for Qodana in CI from the workflow JetBrains gives, changed where it had to be, with what
it reports fixed. Asked which licence the Qodana Cloud organisation holds, the owner answered: a 30-day trial. So the
linter is **Qodana Community for .NET** (`qodana-cdnet`) — ReSharper's InspectCode in a Linux container, which runs
under every licence — and not Qodana for .NET (`qodana-dotnet`, Rider), which needs Ultimate and would turn the job red
the day the trial ends. JetBrains calls the Community linter Early Access.
**How it runs.** `.github/workflows/qodana.yml`, on every pull request into `main` from this repository, every push to
`main`, and by hand: a full-depth checkout of the pull request's own commit, then `JetBrains/qodana-action` pinned to
the commit of `v2026.2.2` (the newest; the template names `v2026.1`, and every action that is not GitHub's own is pinned
by commit here). The action brings the Qodana CLI of its own version, the CLI the linter's image
(`jetbrains/qodana-cdnet:2026.2` — InspectCode 2026.2.3.1 on the image's .NET SDK 10.0.403), and the linter reads
`qodana.yaml`: the solution (`FrameLedger.slnx`), the configuration (Release), the gate. It builds the solution itself,
then inspects it — the whole solution every time (`pr-mode: false`), so a pull request and `main` are judged by one
list. The report goes to Qodana Cloud, a summary to the pull request as a comment and annotations, and the SARIF and
the linter's logs to the run as the `qodana-report` artifact.
**What had to change from the template, and why.**
- **`EnableWindowsTargeting=true`, passed to the container** (`env-vars-to-linter`). Every project here targets
`net10.0-windows`, and the .NET SDK on Linux refuses such a project unless that property is set. With it, the
container restores, builds and loads all thirteen projects — WPF's markup compiler and CsWin32's generator are managed
code — which the first run showed (below). It is the workflow's, not `Directory.Build.props`': no build of the product
is meant to run anywhere but Windows. JetBrains' own page says a project with an OS-specific target needs Qodana for
.NET in native mode; that is the Ultimate linter, and the Community one did the work.
- **`ubuntu-latest`, the one Linux job here** — the Community linter ships as a Linux image only.
- **Permissions:** `contents: read`, `checks: write` (the annotations), `pull-requests: write` (the comment). The
template's `contents: write` is for pushing quick-fixes, an Ultimate feature this job never uses.
- **`main` only.** The `alpha` branch is main's source under another release line (§The alpha line); main's is what
is analysed. *(Until beta.37 the reason was the alpha's prerelease .NET, which the image does not carry.)*
- **Interim since beta.37 (D94, 2026-11-11): the .NET 11 SDK is installed into the container first.** Main builds on
.NET 11 and the image carries the SDKs 8 to 10, so `qodana.yaml` has the `bootstrap` line JetBrains' page gives for
an SDK the image lacks — `dotnet-install.sh` with `global.json`, into `/usr/share/dotnet`. It lasts until the
Ultimate linter's pull request replaces this job with one on Windows, where `setup-dotnet` does it. Measured on its
first run (#376, run 42916491047): the script installed SDK 11.0.100 beside the image's own, InspectCode 2026.2.3.1
loaded and inspected all thirteen `net11.0-windows` projects, and it reported one problem — a redundant `!` in the
test that pull request added, fixed there. So the 63 keys and the zero held across the move.
- **Not a fork's pull request, nor Dependabot's:** neither gets the secret or a token that can write the comment.
- **Caches are saved from `main` only** (`cache-default-branch-only`); a pull request reads them.
**What it reads, and what it does not.** The C# of the thirteen projects, their XAML and `.resx`. **Not the C++:**
`src/native` is a CMake tree no solution names, and it is MSVC-only, which no Linux container compiles — CodeQL and
SonarCloud read it. The native DLLs are absent in the container, so the `.targets` that copy them warn and the build
goes on, as on a fresh clone (measured here the same day: `dotnet build` of a clean worktree, 15 warnings, 0 errors).
**The gate is zero, and it is advisory.** `qodana.yaml` sets `failureConditions.severityThresholds.any: 0`: one
problem and the job is red. `main`'s required contexts stay `check`, `analyze (csharp, none)` and
`analyze (cpp, manual)`, so a problem Qodana finds — or a Qodana Cloud outage — is a red job beside a green `check`,
never a blocked merge. That it can fail was shown before it was
relied on: run 42847369329 had the settings and none of the fixes, found 409 problems and ended red — "Detected 409
problems across all severities, fail threshold 0".
**Where an inspection is switched off: `.editorconfig`, never `qodana.yaml`.** InspectCode reads its severities from
`.editorconfig`'s `resharper_*_highlighting` keys, and so do Rider and ReSharper in the editor — one list for the job
and for the person at the keyboard, beside the analyzer severities already kept there, and testable without CI. Each
key has its reason on the line above it. The rule: **what Qodana reports is fixed, unless the inspection argues with a
choice this project made** — then it is off, and the reason says which choice. Sixty-three keys, in these groups:
| Group | Inspections | The choice it argues with |
|---|---|---|
| The compiler's missing-XML-comment warning | `CS1591` | already a suggestion in this file; InspectCode reports suggestions — 2,050 of the first run's 3,679 |
| Naming | `InconsistentNaming` | names follow this file's rule and the .NET analyzers'; ReSharper's scheme renames `OpenGL`, `XeSS`, `Presents5s` and every `Chart_*` resource key |
| Unwrapping a nullable | `ConvertTypeCheckPatternToNullCheck`, `PatternAlwaysMatches` | `x is long v` and `x is > 0 and int v` name the type, as `csharp_style_var_for_built_in_types = false` asks |
| What a guard is for | three `…AccordingToAPIContract` inspections | the compiler's nullable analysis is the gate; a dependency property, a JSON document and a WPF event hand over null whatever an annotation promises |
| Said on purpose | `RedundantExtendsListEntry`, `RedundantArgumentDefaultValue`, `RedundantEmptySwitchSection`, `MemberHidesStaticFromOuterClass`, `CompareOfFloatsByEqualityOperator` | a code-behind names its base; an argument repeats a default it relies on; `default: break;` says the rest is left alone; exact float comparisons are with a copied value or a sentinel |
| Not this code's to change | `UnusedParameterInPartialMethod`, `Xaml.StaticResourceNotResolved`, `Xaml.PossibleNullReferenceException`, `LocalizableElement` | CommunityToolkit's generated signatures; WPF UI's run-time resources (`PagesLoadTests` fails a wrong key); a binding through null shows nothing, by design; exception and log text is English (`09_I18N`), and `resx-audit` gates what a user reads |
| Records | `NotAccessedPositionalProperty.*` | a positional property is part of equality and of what the record prints |
| Style proposals | twenty, from `ConvertToPrimaryConstructor` to `UseAwaitUsing` | the style is this file, `dotnet format` and the analyzers; `await using` under CA2007 reads worse than what it replaces |
| Accessibility proposals | `MemberCanBePrivate.*`, `PropertyCanBeMadeInitOnly.*`, `AutoPropertyCanBeMadeGetOnly.*` | a member's accessibility is what its type offers, not a count of today's callers |
| "Never used" across the solution | eight `.Global` inspections | enum values mirror the native layout and stored columns, exception types carry CA1032's constructors, rows and messages are read by Dapper and System.Text.Json; the private kind (`.Local`) stays on — and what this group named on 2026-11-10 is listed below |
| In `tests/` only | eleven | an assertion lambda runs inside the `using` it captures from; a test that dereferences null fails, which is its job; a test double keeps the shape of what it stands for |
Five places carry a `// ReSharper disable once …` comment with its reason instead, where the inspection is right in
general and wrong once: the tolerance channel closed early in `CaptureSession`, the volatile read in
`SessionEventPublisher`, the awaited `Task.Run` in `CaptureOrchestrator`, the window read lazily in
`SessionSummaryOpener`, and a test row that is never on.
**The first run, and what was done (2026-11-10, PR #373).** Run 42842741359, the template's shape with no settings:
the container built the solution and inspected it in under six minutes, and reported **3,679 problems** — 858 warnings
and 2,821 suggestions, in 93 inspections. About 400 were fixed in the code, in 181 files, none changing what the
software does:
| What | n | Done |
|---|---|---|
| Redundant qualifiers, casts, `using` directives, `!` suppressions, `unsafe` modifiers, tuple names | 290 | removed, by script from the report's own positions; the build (warnings as errors, Release and Debug) and `dotnet format` held every one — one cast of a pair in `SqliteGameMerge` was needed after all, and stayed |
| `var` for a built-in type, a block body for a one-line method | 52 | this file's own style (`csharp_style_var_for_built_in_types`, `…expression_bodied_methods`), applied by `dotnet format` |
| XAML bindings the inspection could not resolve | 35 | every item template names its `DataType`; a binding that reached the page through `DataContext.ViewModel` with `AncestorType=Page` names the page's type and its `ViewModel` — the same object at run time, and now navigable in the editor (`ViewerModeTests` pins the new text) |
| Left over in the code | 12 | `LogsViewModel`'s unread `_strip` field and `PipeServer.Client`'s unread `_run`; two initialisers a constructor overwrote; `GameFileProbe.UnityDataFile`'s unused parameter; `SpokenTextAutomationPeer` no longer assumes `GetNameCore` returns a string; a redundant setter and two grid positions in XAML; a `$` on a string with no hole; `!Any(==)` as `All(!=)`; a `foreach` variable's type |
| In the tests | 14 | unused variables, lambda parameters and an accessor; a linear `Contains` on a set; three simplifications |
The rest was settled by the settings above — 2,050 of it the one inspection the project had already made a
suggestion. A suggestion to use `m[^1]` inside an expression tree was wrong (CS8790) and its inspection is off.
**The runs (PR #373).** 42842741359 — the template's shape, no settings: 3,679 problems, the job green (no gate yet).
42847369329 — the settings and the gate, none of the fixes: 409 problems, every one in an inspection left on, the job
red; so `.editorconfig`'s keys are read in the container as they are on Windows, and the 3,270 others were the
settings'. 42849331796 — the fixes: **0 problems**, the job green; and `main`'s first run, the merge of #373
(run 42852304026): 0 problems. A job takes about ten minutes.
**Named by the "never used" group and left as it is (2026-11-10).** Nothing in the solution, its tests included, uses
these; whether each is vocabulary, a piece not wired yet, or dead is the owner's to say, so none was removed:
`AtomicFileWrite` (the class and `PublishAsync`), `ExecutableMarkerScan` and `IExecutableMarkerScan.Scan`,
`FpsPresentation.GeneratedLine`, `IPageNavigator.GoBack`, `OperatorDisclosure.VersionOf` and `ProvenanceOf`,
`CrashReporterCatalog.AlwaysOn`, `ProcessWatcher.Tracked`, `LedgerWalCheckpoint.Complete`,
`LhmTelemetryOptions.EnableCpuAndMemory`, `MeasuredFacts.StreamlineNoneWithheldFrom`, `ILegalAcceptanceStore.ListAsync`
and `ITelemetryPoller.Dropped` (only their implementations are called), `DocumentBrowser.PageChanged` and
`UrlOpener`'s getter, the charts' `PercentilePlot`, `NoteText` and `TempsPlotView`, `GameDetailViewModel`'s
`TrendEmptyText`, `LatencyHeader` and `SensorsEmptyText`, `GamesViewModel.SortLabel`, `TriStateChipModel.IsNo`,
`AboutWindow`'s `AppTab` and `ThirdPartyTab`, `NvapiNgxWords.Degraded`, `IpcMessages`' `TelemetryUnavailable` and
`DbWriteFailed`, `IpcProtocol.Keepalive`, `ShmLayout.GuardTickDeadlineMs`, `HarnessOverlayLogSweep.Own`; and the
getters of 49 auto-properties, among them view-model texts no page binds and columns of stored rows. The list is
`UnusedMember.Global` and its siblings in run 42842741359's report.
**Running it on a developer's PC.** The same InspectCode is a .NET tool:
`dotnet tool install JetBrains.ReSharper.GlobalTools --version 2026.2.3.1 --tool-path <somewhere>`, then, after
`dotnet build FrameLedger.slnx -c Release`:
`jb inspectcode FrameLedger.slnx --no-build --format=Sarif --output=<file> --caches-home=<folder> --properties:Configuration=Release --toolset-path="C:\Program Files\dotnet\sdk\<11.0.x>\MSBuild.dll" --dotnetcore="C:\Program Files\dotnet\dotnet.exe"`.
Two things measured on the owner's PC: without `--toolset-path` it picked Visual Studio 18 Insiders' MSBuild, which
crashed at start ("No files to inspect were found"); and a `--caches-home` a failed run had written left one project
unloaded on the next (2,875 false compile errors, all in `Infrastructure.Tests`) — a fresh folder loaded all thirteen.
A clean run there reports what the container does, to within a handful of findings that differ between Windows and
Linux (3,679 in the container, the same inspections on Windows with counts a few apart).
**What it does not cover, and what would change it.** No C++ (above); no `alpha`; no pull request from a fork or from
Dependabot; no quick-fixes, coverage or licence audit (Ultimate). **If the organisation gets an Ultimate licence** —
JetBrains has a programme for open-source projects — Qodana for .NET runs natively on Windows: `linter: qodana-dotnet` in
`qodana.yaml`, `runs-on: windows-latest`, `args: --within-docker false`, .NET set up from `global.json` first, and
`EnableWindowsTargeting` dropped. Until then the trial's end changes nothing here: the token only names the project
the report is sent to.
## C# 15 and the analyses (owner decision D96, 2026-11-11)
The main line compiles C# 15 since beta.37 — `LangVersion latest` on the .NET 11 SDK — and the owner asked for its
constructs to be used. **None is used yet, and this section is why.** The rule the plan was approved with: a construct
is used on main only where (1) the build, its analyzers and `dotnet format` take it with no analyzer switched off,
(2) CodeQL extracts its file without an error, and (3) Qodana reports nothing that is only its engine not knowing the
syntax — the first two are required checks, and Qodana's gate is zero. Each construct was written once, in a file of its
own, and shown to each tool (`spike-notes` §29, the canary):
| Construct | Build and analyzers | ReSharper's engine — InspectCode 2026.2.3.1, which Qodana 2026.2 runs, and 2026.3.0 | Used on main |
|---|---|---|---|
| Collection expression arguments, `[with(…), …]` | taken | an error: "Feature 'collection expression arguments' is not available. Please use language version 'preview'." | no |
| Closed hierarchies, `closed` | taken | not parsed ("Cannot resolve symbol 'closed'") | no |
| Unions, `union` | CA1815, from the SDK's own rules, on every union | not parsed ("Cannot resolve symbol 'union'") | no |
| Labeled `break` / `continue` | taken; `dotnet format` moves the label one indent to the left | not parsed by 2026.2.3.1; "not available… language version 'preview'" in 2026.3.0 | no |
**CodeQL's reading of them is not established**: its documentation lists C# up to 14, and a pull request's C# analysis
is an overlay that prints no per-file result (§`codeql.yml`). A full analysis of a branch is a `workflow_dispatch` of
`codeql.yml` on it.
**What moves this:** a ReSharper engine that reads C# 15 as a released language, in the Qodana linter this repository
runs; then the canary again, and a dispatched CodeQL run. `spike-notes` §29 lists where each construct would go.
**What main does take from .NET 11:** its API — the `string` overloads that name the comparison of a `char` search, and
`Process.TryGetProcessById`. The two preview features (runtime-async, the updated memory-safety rules) stay out, as D76
kept them out of the alphas.
## Dependabot (`.github/dependabot.yml`) and the NuGet audit
```yaml
@@ -481,9 +852,10 @@ Central package management makes Dependabot PRs single-file diffs.
pulled or found compromised in its first week never becomes a pull request here. Security updates do not wait — the
option does not apply to them. The actions' minor and patch releases come as one pull request; a new major comes
alone, because it is the one that changes inputs.
- **No `dotnet-sdk` entry, on purpose:** `global.json` is a floor (10.0.100) with `rollForward: latestFeature`, so a
developer's machine and the runner already build with the newest 10.0 SDK they have (the runner installed 10.0.402
on 2026-10-20). Raising the floor is a decision about what a contributor must install, not an update.
- **No `dotnet-sdk` entry, on purpose:** `global.json` is a floor (11.0.100 since beta.37, D94; 10.0.100 before) with
`rollForward: latestFeature`, so a developer's machine and the runner already build with the newest SDK of that band
they have (the runner installed 10.0.402 on 2026-10-20, 10.0.404 on 2026-11-10). Raising the floor is a decision
about what a contributor must install, not an update — the owner's, as D94 was.
- **Dependabot has never proposed a NuGet library update here — measured 2026-10-25.** Every NuGet pull request it
opened, #66 (2026-08-09) through #314 (2026-10-18), updated analyzers only (Meziantou, Roslynator, the VS threading
analyzers), while Dapper 2.1.89 (published 2026-09-23), coverlet 10.1.0 (09-27), xunit.v3.mtp-off 4.0.2 (10-09) and
@@ -520,8 +892,8 @@ Central package management makes Dependabot PRs single-file diffs.
taken: FluentAssertions 8 (its licence), and the vendored native SDKs (NVAPI R615's licence file changed; the Vulkan
and Streamline headers follow the measured loader and titles).
- **The NuGet audit** (`Directory.Build.props`, written down 2026-10-25): `NuGetAudit` on, `NuGetAuditMode` `all`
(transitive packages too), `NuGetAuditLevel` `low` — the SDK's defaults for net10.0, stated so that a later SDK's
defaults cannot move them. With `TreatWarningsAsErrors` a vulnerable package is NU1901–NU1904 and the restore, so
(transitive packages too), `NuGetAuditLevel` `low` — the SDK's defaults for net10.0 when they were written down,
stated so that a later SDK's defaults cannot move them (and net11.0's did not). With `TreatWarningsAsErrors` a vulnerable package is NU1901–NU1904 and the restore, so
the gate, is red — what `SQLitePCLRaw` 2.1.11 did (`Directory.Packages.props`). `dotnet list FrameLedger.slnx
package --vulnerable --include-transitive` read none on 2026-10-25.
- **The repository's own alerts:** Dependabot alerts and security updates are enabled, and so are secret scanning and
@@ -544,8 +916,10 @@ Central package management makes Dependabot PRs single-file diffs.
**In the tree — done:**
- [x] Repo-local `ci.yml`, `codeql.yml`, `release.yml`, `rules-publish.yml` (all four repo-local by decision, §Workflows); ~~caller stubs pointing at `poli0981/.github`~~ — never the design, see the first paragraph *(and `wiki.yml` since beta.17, repo-local too: five)*
- [x] Issue forms `bug_report.yml`, `feature_request.yml`, `safety_gap.yml`; `ISSUE_TEMPLATE/config.yml` pointing a vulnerability at private reporting; PR template referencing CLAUDE.md's definition of done
- [x] Repo-local `ci.yml`, `codeql.yml`, `release.yml`, `rules-publish.yml` (all four repo-local by decision, §Workflows); ~~caller stubs pointing at `poli0981/.github`~~ — never the design, see the first paragraph *(and `wiki.yml` since beta.17, repo-local too: five; and `qodana.yml` since 2026-11-10, D93: six)*
- [x] Issue forms `bug_report.yml`, `feature_request.yml`, `safety_gap.yml`; `ISSUE_TEMPLATE/config.yml` pointing a vulnerability at private reporting; PR template referencing CLAUDE.md's definition of done *(and since 2026-11-09, beta.35, D88: `accessibility.yml`, the form `ACCESSIBILITY.md` sends a barrier to, and the PR template's user-interface section)*
- [x] The labels the forms apply exist *(2026-11-09: `accessibility`, and `safety` and `priority-high`, which `safety_gap.yml` had named since it was written (2026-08-02) while the repository had neither — GitHub applies only a label that exists, so no Safety gap report was ever labelled)*
- [x] `ACCESSIBILITY.md` — the accessibility statement: WCAG 2.2 AA through WCAG2ICT as a goal, never a claim; how a barrier is reported and handled (2026-11-09, D88)
- [x] `.github/dependabot.yml` (nuget + github-actions, weekly; a seven-day cooldown and the actions grouped since 2026-10-25 — and a NuGet library update it never proposed is `dotnet list package --outdated`'s to find, §Dependabot and the NuGet audit)
- [x] `SECURITY.md` — the private route for a vulnerability, the public route for a safety gap, and the honest response intent (2026-09-16)
- [x] A release rehearsal: `release.yml` `workflow_dispatch` runs every step but the Release itself (§release.yml)
+46 -3
View File
@@ -28,9 +28,10 @@ The anti-cheat guard is the one component where a bug can cost someone an accoun
- **Ring buffer:** SPSC correctness under a hammering writer thread; seqlock torn-read detection; wrap and overwrite-oldest semantics; power-of-two capacity assumptions. Include the **one-full-lap case**: a reader stalled exactly `capacity` records must not validate a different frame as unchanged (this is what the never-reset `seq` counter defends, `07_IPC` §Protocol rules).
- **Drop accounting:** the reader-side computation (`writeIndex - readIndex > capacity`) reports the right count and resumes at the right index. The writer has no drop counter to test — by design.
- **Torn records produce gap markers,** not silently shortened streams; a golden test asserts the gap does not appear as a stutter.
- **Record and header layout:** `static_assert(sizeof(...) == 64)` and `offsetof` assertions for `FlFrameRecord` **and all ~~three~~ four header structs** (`FlShmHandshake`, `FlWriterState`, `FlControlBlock`, and `FlDisplayState` since layout 4, beta.10) — plus D33's `FlTolerance`, mirrored the same way *(corrected 2026-10-04)* — the Agent reads across the process boundary in every one of them. Both halves exist as of 2026-08-05: the native side is `layout_test.cpp` plus the `static_assert`s (ctest `fl_shm_layout`), and the runtime offset dump from `tools/fl-layout-dump` is consumed by `ShmLayoutMirrorTests`, which walks the field list in **both directions** and additionally asserts blittability — `Marshal.SizeOf` and every offset stay correct for a mirror that uses `[MarshalAs(ByValTStr)] string`, while `Unsafe.SizeOf` drops from 64 to 40 and the struct can no longer be read from a mapped view.
- **Record and header layout:** `static_assert(sizeof(...) == 64)` and `offsetof` assertions for `FlFrameRecord` **and all ~~three~~ ~~four~~ five header structs** *(five since layout 5, 2026-11-11: `FlAdvancedState`, 1472 bytes; what reads and writes it is tested at four levels since 0.1.0-beta.40 — `fl_adv_fields` the Overlay's header alone, `fl_guard` `[advanced]` the injected Overlay with the word off and on, `InGameAdvancedSamplerTests` / `CaptureSessionTests` / `SessionRecorderTests` what the Agent makes of a publication, and `AgentEndToEndTests.WhatTheGameTellsFsrIsStoredOnlyByASessionThatRecordsAdvancedMetrics` the shipped Agent, the real Overlay and the harness: two sessions of one target, the first storing nothing of it)* (`FlShmHandshake`, `FlWriterState`, `FlControlBlock`, and `FlDisplayState` since layout 4, beta.10) — plus D33's `FlTolerance`, mirrored the same way *(corrected 2026-10-04)* — the Agent reads across the process boundary in every one of them. Both halves exist as of 2026-08-05: the native side is `layout_test.cpp` plus the `static_assert`s (ctest `fl_shm_layout`), and the runtime offset dump from `tools/fl-layout-dump` is consumed by `ShmLayoutMirrorTests`, which walks the field list in **both directions** and additionally asserts blittability — `Marshal.SizeOf` and every offset stay correct for a mirror that uses `[MarshalAs(ByValTStr)] string`, while `Unsafe.SizeOf` drops from 64 to 40 and the struct can no longer be read from a mapped view.
- **Fault policy:** injected SEH faults in a fake hook body → counter increments, original still called, self-disable at 3.
- **Hook index verification, by behaviour.** A vtable slot carries no identity, so "check that slot 8 *is* `Present`" is not a question the runtime can answer — the earlier wording here required something unimplementable. What is answerable: patch the slot, call the method, see whether the detour ran (`17_HOOK_ENGINE` §Getting vtable addresses; proved for slots 8, 13 and 22 in `spike-notes.md` §H4). Runs headless on CI via WARP and `CreateSwapChainForComposition`. A deliberate mismatch aborts installation instead of patching.
- **A run that was cut short is not a pass** (2026-11-13). ctest calls a test passed when its process exits with 0, and on that day the development PC's security software removed a freshly linked `fl_guard_test.exe` while the gate ran: ctest printed `fl_guard ... Passed 0.66 sec` for a suite that takes three minutes, and the gate went red only because the next test runs the same executable and could not find it. So **every native test now names the line it ends on** — `fl_test_completes(<test> <regex>…)`, or `fl_catch2_test_completes(<test>…)` for a Catch2 executable (its `All tests passed (` line, or its summary row when a case was skipped, and never a row that names a failure) — and ctest requires that line (`PASS_REGULAR_EXPRESSION`; the helpers and the reasons are in `src/native/CMakeLists.txt`, above `enable_testing()`). Three things hold it: **the configure fails** when a test in `tests/` or `tools/` was registered without one (tried: a test added bare stops CMake with its name); **four canary tests** print what a Catch2 run would — nothing, its last line, a row with a skipped case, a row with a failure beside "passed" — through the same helper, and the two that must be red are `WILL_FAIL`; and the tests that were passing on a line printed in the *middle* of their run (`fl_nvapi_bridge`, `fl_process_stats`, `fl_nvapi_probe`, `fl_signer_probe`) now need that line and, after it, their last one. For `fl_nvapi_probe` this also closes what `fl_nvapi_bridge` was fixed for on 2026-09-25: a pass expression makes ctest ignore the exit code, so a run that printed its branch and then `FAILED` passed. `fl_signer_probe` accepts either ending — its criterion was always that the answers are printed. A `SKIP_RETURN_CODE` still wins over a missing line. **Not covered**: the managed tests (`dotnet test` reports a host that died) and a process that is stopped *after* printing its last line.
- **Every probe must be shown to fail.** A probe that ends in an unconditional assertion is green by construction and its ctest is decorative — `fl_proxy_swapchain` shipped in exactly that state. Before a probe counts as a regression net, break the thing it watches and watch it go red.
## Managed unit tests
@@ -145,7 +146,7 @@ own work item on the STA thread and puts back — and with the keys a parameter,
## Hook overhead measurement (NFR-1, per release)
1. `hook-harness` in a fixed workload, present-rate uncapped: 3 runs hooked vs 3 unhooked → mean present-call cost delta must be ≤ 1 µs (measured with QPC around the call site in an instrumented harness build).
2. Real game, fixed 10-minute scene, capture ON vs OFF, 3 runs each: **game's own Avg FPS delta ≤ 0.5%**, Agent CPU ≤ 1% of a core, Agent RSS ≤ 150 MB, Overlay resident ≤ 8 MB.
2. Real game, fixed 10-minute scene, capture ON vs OFF, 3 runs each: **game's own Avg FPS delta ≤ 0.5%**, Agent CPU ~~≤ 1%~~ ≤ 2.5 % of a core *(NFR-2 as D117 amended it, 2026-11-11; `tools/agent-cost.ps1` measures the unhooked half without a game)*, Agent RSS ≤ 150 MB, Overlay resident ≤ 8 MB.
> **NOT MEASURED. The runbook exists as of 2026-09-10 (P2 PR-G); the run is the owner's.**
> `tools/fps-impact-runbook.ps1` sequences the six launches, holds consent granted for the ON leg and
@@ -158,6 +159,43 @@ own work item on the STA thread and puts back — and with the keys a parameter,
> Until the run happens, the honest state of this line is *unmeasured*, not *passing*.
3. Results recorded in the release notes ("measured overhead this release: …").
**What a patch costs when it goes in and comes out (since 0.1.0-beta.46) is not the per-present number above**, and
is measured where it happens: `fl_guard`'s Streamline `[advanced]` case prints the microseconds of each install
batch and of the stop's one restore (the Overlay's `PATCH_BATCH` lines), and the time from the unhook request to
`UNHOOKED`; it asserts that every lazily installed patch is under a batch line, that the stop is one, and that the
restore stays under a second — the bound the one-patch-at-a-time stop did not meet on the development PC. In a real
game the same lines are in `logs\overlay-<pid>-….log` — read there on 2026-11-13 for five games: eighteen install
batches, 81–262 ms for sixteen of them and 0.61 s and 1.26 s for two (`17_HOOK_ENGINE` §Unhooking, *Batches*); and
on 2026-11-14 two restores of twenty-nine patches each, 83 ms and 86 ms, when the owner stopped the part inside a
running game.
**The call into the driver (train C, 0.1.0-beta.47) is tested in three places, each for what the
others cannot say.** `fl_reflex_summary` (Catch2) drives `fl_adv_fields.h`'s reader on reports written by hand: the
two newest frames held back, a frame summed once, a span the driver did not report, a span no frame can have, frame
ids that start again, and the ten words of a publication — and since 0.1.0-beta.48 the driver's two own times
(summed over the frames that gave one, each with its own count; zero and more than a second are not counted and
refuse nothing) and the bit a multiplier has — and since 0.1.0-beta.49 the counts for the native log: every place
a summed frame can be in, a span of a microsecond, an active time as long as the span, a refused frame, when a
batch is due, a batch closed by a change of state and the thirty-two such closings a request has, the six lines
and what each carries; a hidden `[.bench]` case is the measurement of one report's cost. `fl_watchdog_tick`
(Catch2, no thread and no clock) drives the grid the watchdog's periodic tick is on: sixty ticks each late and
each followed by work keep whole seconds, a wait is never rounded down, a tick woken early is not periodic and
moves nothing, a late tick is not followed by another within half a period. `fl_guard`'s nine `[latency]` cases
(the eighth reads the log: the first batch after ten reads, by number, and what is left when a flush is asked for
— against the stand-in's `kModeGpuShapes`; the ninth keeps the loader waking the watchdog without pause — a
thread that asks for `opengl32.dll` for as long as the target lives — and holds one publication a second, each
with a read, none within half a second of another: 35 720 publications in eight seconds on the code before it; a
hidden `[.measure]` case prints ninety reads' own lines) inject the Overlay
into a Direct3D 12 target that has loaded a stand-in `nvapi64.dll`, and read that module's call counts from outside
through its named section (`fl_stub_nvapi_shared.h`) — so "the driver was not called" is a count: not asked → not
even asked for the addresses; asked → about one read a second, sums equal to the stand-in's, every counted read a
received call; an error; a span not reported; the request off; a second session; the stop; a call of 50 ms (the
last one); a fault (counted once, the hooks stay); a module without the getters; no module (none is loaded); a
module beside the executable (refused, never asked); Direct3D 11 (not read). `fl_reflex_probe_stub` runs the same
header, `fl_reflex.h`, in a process that **enforces Control Flow Guard** — the injected cases do not: `hook-harness`
is not a CFG image. **None of them is a driver.** `fl-probe-reflex --real` is the measurement against the installed
one, run by hand (`spike-notes` §31), and a real game is owner-only item 45.
## Tier cross-validation (accuracy regression net) — **GONE, 2026-08-28, and this is the entry to read before planning P5**
~~Run the same game session with Tier 1 and Tier 2 simultaneously where possible (ETW does not conflict with hooking) and compare frametime distributions: Avg/1% Low must agree within **1%**. Divergence beyond that means one of the two paths is wrong and blocks the release. This is also the check that would have caught the original detection-accuracy problem.~~
@@ -201,7 +239,7 @@ that it was checked against anything.
| Axis | Values |
|---|---|
| OS | Win 10 22H2 VM · Win 11 dev machine |
| OS | Win 11 dev machine · a Win 10 22H2 VM where one exists — since beta.37 (D95) Windows 10 is not blocked and is neither tested nor supported, so a run there is information, not a gate |
| GPU vendor | NVIDIA (primary) · AMD or Intel if available, else document as untested |
| API | D3D11 · D3D12 · Vulkan (layer path) · OpenGL · **a 32-bit title, asserting it is correctly refused and recorded as Tier 2 with the reason** |
| Upscaler | DLSS SR · DLSS-G · DLSS-RR · FSR2/3 · XeSS · none |
@@ -209,9 +247,14 @@ that it was checked against anything.
| Mode | launch mode · attach mode · mid-session settings change |
| Safety | game with anti-cheat → toggle disabled · simulated late AC load → unhook · double-crash → auto-disable |
| Tier | forced Tier 2 (records duration, available sensors and the reason — and nothing else) · Tier 1 → Tier 2 degradation notice |
| Accessibility *(added 2026-11-09, beta.35, D88)* | **keyboard only** with the mouse disconnected — every page, menu and dialog, Ctrl+1…5, Esc · **Narrator** — names, values, headings, the notices as they appear · a **contrast theme** on and off (Aquatic, Desert) · Windows **animation effects** off · a large **display scale**. The owner's first pass: 2026-11-09, keyboard only and Narrator on 0.1.0-beta.34 and 0.1.0-alpha.3, no fault (`ACCESSIBILITY.md` §How it is checked; the automated half is `08_UI` §Accessibility) |
| Update | Velopack delta; update deferred while a game is hooked (FR-12) — *a delta needs a previous release, so this row is reachable from the second release, not the first* *(corrected 2026-10-04: no release has carried a delta — `release.yml` never fetches the previous release before `vpk pack`, `11_UPDATER` §Built — so the update leg is a full package; **since beta.13** a release carries a delta against the previous one, and the first
update that can use it is beta.12 → beta.13 — HANDOFF owner-only item 10)* |
## Release smoke
Clean-VM install → Legal Gate → Agent setup (unelevated) → add a game → consent dialog → hooked 2-min session → charts render → CSV opens in Excel → Vulkan layer registered/unregistered cleanly → uninstall removes layer registration, task, and asks about data.
*(Added 2026-11-09, beta.35, D88:)* and an **accessibility smoke**, the same build: the Legal Gate, the consent dialog and
the session summary reached and answered with the keyboard alone; Narrator reads the notice a refused game raises; the
window readable under one contrast theme. The release's owner-only item in `HANDOFF` lists what that release moved.
+44
View File
@@ -332,6 +332,50 @@ CI/CD with the native build (`release.yml` exists since P4 PR-5, 2026-09-14 —
**Total ≈ 5.5–6.5 weeks full-time** — roughly 1.5–2 weeks more than the ETW design, which is the honest price of the accuracy.
## After the controller train — the owner's order (2026-11-14)
Advanced metrics is closed (D154) and the controller feature is being built as 0.1.0-beta.50's train. Asked what
comes after, the owner was shown six directions and chose four of them, and added distribution. **This is the
order of work after the controller train; each begins with a plan of its own and the owner's approval.** Nothing
here is designed yet, and what is marked *to verify* is a lead, not a finding.
1. **Time actually played (AFK segmentation).** The v2 backlog's "AFK segmentation via `GetLastInputInfo`", moved
up: the controller train already stores when the keyboard, the mouse or a controller was in use, so the time
nobody was at the game can be told apart and left out of a session's averages. Timestamps of activity only —
never a key, as that train's Privacy text says.
2. **More from what is already stored — interface only, no hook.** The comparison of two stretches of one
session ("before and after I changed DLSS Quality to Balanced", v1.1 below), and a timeline for a game that
marks where the driver, the game's build or the settings changed, so a drop in frame rate has a date. *To
verify*: which of those changes the ledger already records for every session.
3. **Stutter from shader compilation (PSO).** `03_METRICS` defines *PSO stutter %* and `05_DETECTION` says no
writer exists. New hooks at pipeline creation: staged like Advanced metrics' train B — small releases, the
overhead measured, the owner's real games between them, never first on a game inside the user-mode
exception's trial (D38).
4. **The in-game overlay (FR-15).** v1.1's first line below: live frame rate, frame time, upscaler and
ray-tracing state drawn in the game, per game, off by default. The largest and the riskiest of the four — it
draws in the game's process — so it comes last, with a risk review of its own (rule 5, and what an
anti-cheat makes of a drawing module).
5. **Distribution: winget and the Microsoft Store.** The owner uploads to the Store and fills in its listing by
hand. FrameLedger has no product in Partner Center yet — the one there today is another program of the
owner's — so which type it is submitted as (*EXE or MSI app*, or *MSIX or PWA app*) is part of what the plan
settles. The repository's part is what the two channels need from a release. *To verify, before anything is promised*:
- whether the Store accepts an **unsigned** installer for an *EXE or MSI app* — the installer is unsigned
today (`CLAUDE.md`, Packaging), and if the Store wants a certificate that is a cost and a decision of the
owner's; whether an MSIX package is a real alternative for a program that loads a library into games and
registers a Vulkan layer;
- the installer's silent switch and a versioned, permanent download address for each release, which both
channels ask for; whether either takes a pre-release version;
- who updates an installation that came from winget or the Store — the App's own updater (Velopack) or the
channel — and that the two cannot fight;
- a winget manifest per release (address, SHA-256, switches) and whether `release.yml` writes it or the owner
submits it;
- rule 8 is untouched: a channel that delivers the installer is not a network call of the App's.
**Not chosen, and not refused**: a stabilising pass with a cross-check against another tool (the limits page
still says no reading has been checked against one), and export for CapFrameX and for video overlays. Two things
only the owner can settle stay open beside them: what leaving beta takes, and code signing — which direction 5
may force.
## v1.1 — earned features (only after the data path is stable)
- **In-game overlay** (FR-15): we are already in the process; draw live FPS/frametime/upscaler/RT state. Opt-in per game, off by default. Borderless and fullscreen-flip both handled since we own the swapchain.
+63 -2
View File
@@ -6,7 +6,7 @@ Mandatory conventions for all XAML/C# in `FrameLedger.App`. WPF UI = lepoco's Fl
| Package (NuGet id) | Pin | Notes |
|---|---|---|
| `WPF-UI` | **= 4.3.0** (exact; bump deliberately) | Assembly/namespace is `Wpf.Ui`. Targets `net8.0-windows` → resolves cleanly for our `net10.0-windows10.0.22621.0` TFM (`12_BUILD` §Managed build; this row said `19041` until 2026-09-21, which is no TFM this repository ever built — `Directory.Build.props` has been `22621` with a 19045 floor throughout) |
| `WPF-UI` | **= 4.3.0** (exact; bump deliberately) | Assembly/namespace is `Wpf.Ui`. Targets `net8.0-windows` → resolves cleanly for our `net11.0-windows10.0.22621.0` TFM — `net10.0-…` until beta.36 — (`12_BUILD` §Managed build; this row said `19041` until 2026-09-21, which is no TFM this repository ever built — `Directory.Build.props` has been `22621` with a 19045 floor throughout) |
| `WPF-UI.Abstractions` | ~~transitive~~ **= 4.3.0**, referenced directly (corrected 2026-10-04) | Navigation abstractions (`INavigationViewPageProvider`) |
| `WPF-UI.DependencyInjection` | match `WPF-UI` | DI glue (`services.AddNavigationViewPageProvider()`); verify exact id/version at scaffold |
| `WPF-UI.Tray` | **not used** | Tray stays on H.NotifyIcon.Wpf (CLAUDE.md stack table) |
@@ -62,6 +62,7 @@ private static readonly IHost _host = Host.CreateDefaultBuilder()
Theme rules:
- Manual Light/Dark: `ApplicationThemeManager.Apply(ApplicationTheme.Dark)`.
- System: `SystemThemeWatcher.Watch(mainWindow)` — call it in the window's `Loaded` handler (needs an HWND), and unwatch when switching to manual.
- A Windows contrast theme (beta.35, D91): wins over all three while it is on, and is never left to the watcher, which would apply WPF UI's own high-contrast files (§Gotchas). `Services/ThemePlan.For` is the table, `Services/WpfThemeApplier` applies it.
- Persist the choice in `settings`; re-apply before the main window shows to avoid a flash. **Measured 2026-09-13: this is a correctness rule, not a cosmetic one** — `ApplicationThemeManager.Apply(theme, backdrop, updateAccent: true)` is what registers the `SystemAccentColor*` resources the control templates `StaticResource` at first Measure; a `FluentWindow` shown before any `Apply` throws `XamlParseException: Cannot find resource named 'SystemAccentColorPrimary'`. The shell applies the theme in its constructor before `InitializeComponent()` and again from `Loaded` for the watcher.
- Subscribe `ApplicationThemeManager.Changed` once, centrally, to re-theme ScottPlot (below).
@@ -115,6 +116,14 @@ Code-behind wires services once: `navigationService.SetNavigationControl(RootNav
- Navigate **only** via `INavigationService.Navigate(typeof(SomePage))` — never manipulate `Frame` directly, never `new` a Page.
- Pages implement `INavigableView<TViewModel>` with `ViewModel` injected via constructor DI; `DataContext = this` in the page constructor (WPF UI template idiom) so bindings read `ViewModel.*`.
- Page state: pages are Transient; anything that must survive navigation lives in the ViewModel-backing services/DB, not the Page.
**Except the library, since beta.34** (owner request 2026-11-06, "Games stalls 0.5–1.5 s"; `spike-notes` §27): the window
keeps the `GamesPage` it built — `Services.WindowPageCache`, the window's own `INavigationViewPageProvider`, set after
`INavigationService.SetNavigationControl` (which hands the NavigationView the container's provider) — and its view model
reads the library again each time the page is shown (`GamesViewModel.Attach`). Still Transient in the container; one
instance per window, so a language change's new window builds its own in the new language. **Not**
`NavigationViewItem.NavigationCacheMode`: WPF UI 4.3.0 reads it only when the NavigationView has neither a service
provider nor a page provider (`NavigationView.Navigation.cs`, `GetNavigationItemInstance`), and this App sets one — the
attribute would have changed nothing.
## Control mapping (always prefer these)
@@ -164,6 +173,22 @@ Native `Menu`, `TabControl`, `ComboBox`, `Slider`~~, `ListView`~~ are fine — t
WCAG contrast, black in both themes. The chart reads `ChartTheme.Current` at every render; the WPF chip in the list
(`Controls/ChangeBadge`) follows a switch through the same resource reference as `ArtTint` and `FgMixBar`
(`ChartPalette.ForBackground`). No colour in XAML.
- Third exception (beta.35, D91): **a Windows contrast theme's colours**, read from `SystemColors` in one place
(`Services/ContrastPalette.FromSystem`) and laid over the theme by `Services/ContrastOverlay` — never written in XAML. A
colour the App works out itself reads `Services.ContrastTheme.Active` first (the charts, the badges, the artwork band);
XAML that must change with it binds `{Binding Path=(services:ContrastTheme.IsOn)}`, a static property WPF hears through
its `StaticPropertyChanged` event.
- **`ui:Button Appearance="Danger"` / `Info` / `Success` / `Caution` / `Dark` / `Light` paint fixed colours** — values in
WPF UI 4.3.0's button style (`#F44336` for Danger), not resources, so no theme and no contrast theme reaches them, under
the theme's button text (white on the red in the dark theme, 3.68:1). Danger is the one the App uses: its implicit
`ui:Button` style (`Styles/FrameLedger.xaml`) puts the theme's dark text on the red, and the theme's own button under a
contrast theme — message boxes' buttons included. The other five are refused
(`ContrastThemeTests.NoButtonOrBadgeTakesAnAppearanceThatPaintsAFixedColour`).
- **A snackbar's appearances are fixed colours too** (`ISnackbarService.Show`): Success, Caution, Info, Danger, Primary,
Dark and Light are values under white or black text — Success 2.78:1, Caution 2.16:1 in both themes. Only Secondary is
the theme's: `Services/SnackbarStrip`, the one place a snackbar is shown, uses it for every message.
- **A `ui:Badge` with a BOUND palette appearance** is the same hole as a literal one, and no XAML scan of literals sees it
(the agent's pill, beta.35: `AgentPill` now). A bound badge appearance is refused by the same test.
## FlowDocument (rendered documents, beta.12)
@@ -207,7 +232,12 @@ var result = await box.ShowDialogAsync(); // Wpf.Ui.Controls.MessageBoxResult.Pr
⚠ Name clash with `System.Windows.MessageBox` — in App code, `using MessageBox = Wpf.Ui.Controls.MessageBox;` ~~and ban the System one (BannedSymbols/analyzer note)~~ *(corrected 2026-10-04: nothing bans it. Three paths use the Win32 box on purpose — the crash dialog, the startup-failure dialog and the uninstall question — because they run when the shell's dialog host cannot be trusted. Use the WPF UI one everywhere else.)*
- Rich in-flow dialogs: `IContentDialogService` (host wired in MainWindow — a `ui:ContentDialogHost`, since 4.3.0 marks `SetDialogHost(ContentPresenter)` obsolete).
- Transient in-app: `ISnackbarService.Show(title, message, ControlAppearance.Success, new SymbolIcon(SymbolRegular.Checkmark24), TimeSpan.FromSeconds(4))`.
- Transient in-app: ~~`ISnackbarService.Show(title, message, ControlAppearance.Success, new SymbolIcon(SymbolRegular.Checkmark24), TimeSpan.FromSeconds(4))`~~
`IMessageStrip` (`Info` / `Success` / `Warn`) — `Services/SnackbarStrip` is the one caller of `ISnackbarService.Show`, and
says each message to a screen reader (beta.35, D89). A view model never holds `ISnackbarService`.
- A change a screen reader must hear goes through `IAnnouncer` (`Services/UiaAnnouncer`, UI Automation's notification
event). **`AutomationProperties.LiveSetting` alone says nothing in WPF**: the app must raise `LiveRegionChanged` itself,
and no XAML of the App's sets one (beta.35).
- System/tray notifications: H.NotifyIcon only (08_UI §Notifications policy).
## ScottPlot theme sync
@@ -224,6 +254,15 @@ On startup and on `ApplicationThemeManager.Changed`: for every live plot set fig
> `Add.HorizontalSpan` for the ribbon, `Add.Bars` over `Statistics.Histogram.WithBinCount`, `Add.ScatterLine`
> for the percentile curve, a series' `Axes.YAxis = plot.Axes.Right` for the sensor overlay.
>
> **Accessible charts, beta.35 (D90).** Every plot in XAML is `charts:ChartPlot` (a `WpfPlot` subclass; ScottPlot's
> default style and input stay) with `AutomationProperties.Name` and `HelpText` from `A11y_Chart_*` / `A11y_ChartData_*`.
> `ChartTheme.Attach` adds `Charts/ChartKeys` (arrows and A marked handled after ScottPlot's class handler has acted — an
> instance handler runs after it) and `ResetOnDoubleClick` (`DoubleLeftClickBenchmark(false)` and a `DoubleClickResponse` on
> the left button that autoscales). A series takes `ChartPatterns.Line` / `Points` / `Hatch` by its index — a new series
> without one is red in `NoChartTellsTwoSeriesApartByColourAlone`. Under a Windows contrast theme `ChartTheme.Current` is
> `ChartPalette.ForContrast(ContrastTheme.Active)` — the theme's window under every plot, its text, link, highlight,
> button text and disabled text for the rest — and the patterns carry the series apart (D91).
>
> **The trend's change markers, beta.30 (D70):** `Charts/ChangeMarkerPlottable`, an `IPlottable` + `IRenderLast` added
> with `Add.Plottable` (which assigns no axes: the chart sets `Axes.XAxis`/`YAxis` itself), draws the dashed lines in
> `Render` — clipped to the data area — and the numbered badges in `RenderLast`, after the panels and unclipped, in a
@@ -244,6 +283,13 @@ On startup and on `ApplicationThemeManager.Changed`: for every live plot set fig
- [ ] **`WpfPlot.DisplayScale` is read once, in the constructor**, from `VisualTreeHelper.GetDpi` before the control has
a window. Read a pointer with `GetPlotPixelPosition(e)` (the element's DPI now); ScottPlot's pixels are device
pixels while no `Plot.ScaleFactor` is set, and none is.
- [ ] **WPF UI 4.3.0's high-contrast theme cannot be applied as it is** (beta.35, D91). `ApplicationTheme.HighContrast` has
no dictionary (the watcher picks HC1/HC2/HCBlack/HCWhite by Windows' theme name); each of those holds 84 `#FFFF0000`
placeholders among 97 colours and 17 brushes (`SystemFillColor*Brush`, `MenuBar*`) that throw when the file is read
alone — inside the App their forward references fall back to the application's resources and read as the Fluent
theme's fills; their colours are Windows 11's defaults, not the user's; their text on the accent is the window text.
Never let `SystemThemeWatcher` watch while a contrast theme is on: it would apply them. The App reads them for each
key's role only (§Theming & brushes; `08_UI` §Contrast).
- [ ] `SystemThemeWatcher.Watch` after HWND exists (`Loaded`), not in the constructor — but `ApplicationThemeManager.Apply(…, updateAccent: true)` BEFORE the first window's XAML loads, or the accent resources are missing at first Measure (§App bootstrap).
- [ ] Menu row lives **outside** `ui:TitleBar` (its area is the drag region; a Menu inside becomes undraggable/unclickable territory).
- [ ] Don't set `AllowsTransparency`/`WindowStyle` on `FluentWindow` — it manages its own chrome.
@@ -258,6 +304,21 @@ On startup and on `ApplicationThemeManager.Changed`: for every live plot set fig
bar with no exit, which is exactly how the safety notices shipped (2026-09-14). `Styles/FrameLedger.xaml` carries
the upstream template plus one presenter column; `PagesLoadTests.TheInfoBarTemplateRendersItsContent` renders both
templates and turns red when upstream gains a presenter, which is when the copy comes out.
- [ ] **What a screen reader is given is the UI Automation tree, and WPF builds it with four holes** (beta.35, D88;
`PagesLoadTests.Automation` sweeps every surface for each). (1) **A control with no automation peer** — a plain `Control`
or `ContentControl`, WPF UI's `Image`, `Card`, `Badge`, `InfoBar` (measured on .NET 10.0.14) — is skipped by the tree; if
it is also focusable (a `Control` is, by default) it is a Tab stop Narrator says nothing about. Give a display-only
control a peer (`Controls/SpokenTextAutomationPeer`) and take it out of the tab order; `ui:CardControl` has a peer with an
empty name even with a string `Header` — the App's implicit style takes it out of the tab order. (2) **A list's,
drop-down's or grid row's item is named by its `ToString()`**: a record's is its printed form (`Choice { Value = dark,
Label = Dark }`), a class's its full type name. Every item type overrides it (`Services/SpokenText` builds the words).
(3) **A `TextBlock` written straight into a `ControlTemplate` is no control element** (`TextBlockAutomationPeer`: not when
its templated parent is something other than a `ContentPresenter`): draw a template's text through a `ContentPresenter`
(the App's InfoBar template), or give the control a peer that says it. (4) **A `DataGridTemplateColumn`'s cell has no text
of its own**, so WPF names it `Item: <row>, Column Display Index: n`: name it with `controls:CellName.Path` (a cell style
based on `DefaultDataGridCellStyle`, the one WPF UI's grid gives every cell). Also: a plain `ItemsControl` is focusable
(the App's implicit style says no), and WPF UI's `CardExpander` leaves its expand button (`ExpanderToggleButton`)
unnamed (`Controls/ExpanderToggleName`, set by the App's implicit CardExpander style).
- [ ] **WPF UI 4.3.0 keys its DataGrid parts' styles and defines no implicit one** (`DefaultDataGridRowStyle`,
`DefaultDataGridCellStyle`, `DefaultDataGridColumnHeaderStyle`, handed to the grid by its implicit `DataGrid` style's
`RowStyle`/`CellStyle`/`ColumnHeaderStyle` setters). A local `RowStyle` replaces WPF UI's; based on
+331 -16
View File
@@ -172,6 +172,81 @@ Every hook must be listed here with a purpose. Anything not on this list is not
| ✅ `opengl32.dll!wglSwapBuffers` (2026-09-06, P1 item 4) | The OpenGL present hook (§Presentation). A system module, not a vendor one, so — like the row below — it is deliberately **not** in `FL_HOOK_INVENTORY` and `hookinventory-check` does not cover it; listed here so the table above and these two rows together are the whole inventory |
| ✅ `kernelbase.dll!LoadLibraryExW` (2026-09-06, P1 item 1) | The wake for lazily loaded runtimes and the in-process anti-cheat stop — §DLL entry step 3 above carries the mechanism. **Rule 4:** it reads one argument the API was given (`flags`) and one property of a handle the loader just returned to *its own caller* (the module file name); nothing of the game's. **~~It is the only hook on a system module, and~~ Like `wglSwapBuffers` above it, it is deliberately not in `FL_HOOK_INVENTORY` (corrected 2026-10-04):** that table is *vendor* symbols resolved by name and gated by `hookinventory-check` against `vendor-exports.json`, and `LoadLibraryExW` is neither vendor nor uncertain. It is listed here instead so the table above and this row together are the whole inventory. It **never** installs, uninstalls or patches anything itself (§H2) |
### Calls the Overlay makes into a driver — not hooks, and the only ones
Everything else on this page learns what it learns from the arguments of a call the game made. This is the one
place the Overlay **makes a call of its own** into code that is neither the game's nor FrameLedger's, and it is
listed here so that nobody takes it for a hook or adds a second one without the same ruling
(`CLAUDE.md` rule 4's note of 2026-11-13, owner decisions D132, D136, D138–D141).
| Call | When, where | Rule 4 | What bounds it |
|---|---|---|---|
| **NVIDIA's `NvAPI_D3D_GetLatency` and `NvAPI_D3D_GetSleepStatus`** — the driver's two read-only latency getters (Advanced metrics train C; `fl_reflex.h`; **since 0.1.0-beta.47**) | **Only while a session asks for Advanced metrics** (D138: no switch of its own), only for a Direct3D 12 swap chain whose device is the one the Overlay holds, **about once a second, on the presenting thread, after that present's record is published**. The watchdog finds the addresses once for the process and arms one read at a time; a present pays one relaxed load. The addresses come from an `nvapi64.dll` **the game had already loaded** — asked for by name, never loaded — through its `nvapi_QueryInterface` and the two interface ids of the vendored `nvapi_interface.h`. The Overlay links no NVAPI | The owner's ruling (D132): a driver's read-only getter called with the device behind the swap chain the Overlay hooks reads the driver's report, not game memory, and writes nothing. Getters only — never a setter, a marker or a sleep (`tools/nvapi-readonly-check.ps1` holds which ids code loaded into a game may name: these two). What is read of the two structs is on `tools/overlay-allowed-fields.txt` like every other vendor member | **An `nvapi64.dll` in the executable's own folder is refused** (D140): the driver's is in Windows' directories. **A fault** in a getter is caught by a guard of its own, counted once in `faultCount`, and ends the reads for the process. **One call of 2 ms or more** ends them too (D141). **A call that has not returned after two seconds** is logged, flushed and published as stalled — nothing can stop it, and the presenting thread it holds is the game's. Each is until the game exits |
**Where the addresses are, measured** (`fl-probe-reflex --real`, driver 32.0.16.2114): System32's `nvapi64.dll` is a
shim and answers the two ids with addresses inside **`nvapi64_impl.dll`**, in the driver's own store. So an address
is accepted when it is executable code of an image the process has loaded, and **the Overlay keeps a reference on
that image as well as on `nvapi64.dll`** — two modules the game loaded, never released, for the reason
`ResolveScoped` gives: a presenting thread may be about to call. `nvapi64_impl.dll` is a Control Flow Guard image;
the probe, which enforces CFG, has made the calls from the header that ships, so on this driver the getters are
valid targets. **Not excluded**: a game that enforces CFG on a driver the probe has not been run against — a target
that were not valid is a fast-fail no guard catches.
**In a real game** (the owner, 2026-11-14, Phantom Blade Zero on 0.1.0-beta.47, three sessions, frame generation
×3 and off): the module was found four to five seconds after the attach — `nvapi64.dll` in System32, the getters in
`nvapi64_impl.dll`, as on the probe — a read was made about every second for 344 s, 52 s and 1 119 s, **the slowest
call took 24, 20 and 45 µs**, and no latch fired: no error, no slow call, no fault, no stall. **Cyberpunk 2077 the
same morning**, two sessions (frame generation ×2 and off, 260 s and 581 s): found 4.8 s and 2.6 s after the attach,
the slowest call 34 and 65 µs, no latch. Two Direct3D 11 games and one Direct3D 12 game with no `nvapi64.dll` were
left alone, each with its one line. Two games read, one driver.
**Two more members of the same report, and the multipliers seen (0.1.0-beta.48; owner decisions D142, D144).** In
Phantom Blade Zero the span between the driver's two GPU marks was longer than the frame's whole: it includes the
card's waiting. NVIDIA's header has the figure that does not — `gpuActiveRenderTimeUs`, "excluding the idles in
between" — and beside it `gpuFrameTimeUs`, one frame's GPU end to the next's. Both are read from the report the
getter already filled: **no call is added, and no interface id** (`tools/nvapi-readonly-check.ps1` is unchanged;
the two names are on `tools/overlay-allowed-fields.txt`). The buffer is the Overlay's own, so there is no question
of a member lying past what a game allocated; an older driver leaves both zero, and zero is read as *not given* —
each time has its own count, and neither decides which frames the seven spans are made of. The sleep status'
frame-generation multiplier is now also kept as the set of values a session saw (a session that generated ×2 and
ended in a menu read 1). **One defect fixed with it**: a read that ended after a session's last tick was taken by
the next session as its first point; a new request now drops it. **Measured** (the header's own function, the
Overlay's flags, a report of 62 summed frames, 2 million reports × 3, twice): 393–396 ns before, 429–434 ns after
— **+36 ns a read**, once a second, and after the call's time has been taken, so it cannot reach the 2 ms latch.
The present path is unchanged: one relaxed load.
**Since 0.1.0-beta.49 each summed frame is also counted for the native log** (owner decisions D148, D151;
§Native logging, `LATENCY_FRAMES`). The owner's ten sessions on beta.48 (HANDOFF, item 46) gave both of the
driver's own times in every frame report of both games — and in Cyberpunk 2077 without frame generation the mean
active time was **longer** than the mean span between the two GPU marks (3.17 against 2.79 ms), which the two
means cannot explain: the span's is over every summed frame, one of no length included, the active time's over
the frames that gave one. So the reader places each summed frame by its own two numbers (`CountGpuTimes`), after
the frame is accepted: which frames are summed does not change, and the seven spans, the two times and their
counts are what they were. **No new call, no further member of the report** — the two GPU marks and
`gpuActiveRenderTimeUs` were read already — and nothing of it leaves the process: the watchdog adds each read's
counts to a batch and writes the batch to the log. **Measured** (the same instrument: 2 million reports × 3,
twice): 424–435 ns before, 490–499 ns after — **+65 ns a read**, once a second, after the call's time has been
taken. `fl_reflex_summary_test "[.bench]"` is that measurement in the tree. The present path is unchanged.
**Since 0.1.0-beta.49 the read is on a grid of one second** (owner decision D150; `fl_watchdog_tick.h`). The
watchdog used to wait a second *after* each tick's work, and a tick the LoadLibrary detour woke was a tick like
any other: it took what the last read left, published it and — less than a second having passed since the last
arm — armed nothing, so the next publication carried no read. The Agent samples on a period of exactly one second
since 0.1.0-beta.48, and its sessions showed the holes: 0.4 to 16 % of a session's samples without a report
(HANDOFF, item 46). Now the watchdog's **periodic** tick is due at a deadline by the performance counter, and the
next deadline is one period after the one that was due, not after the moment the tick ran; a tick woken early
does everything it did — the request is followed and the hooks a session asks for go in at once — **except take
the driver read and publish region 5**, which are the periodic tick's. One read is armed at every periodic tick
(`g_latArmedQpc` and its "a second since the last arm" are gone). A periodic tick that ran more than half a
second late is not followed by another on the grid's next mark: the grid starts again from it, so **two reads are
never less than half a second apart** — "about once a second" still, and no longer "at most". Nothing on a game
thread changed: the presenting thread makes the read when it finds the gate armed, as before. **Measured** on `hook-harness` with the stand-in module, by the batches' own lines (`fl_guard_test "[.measure]"`): ten reads in 9 996 ms, the next twenty in 20 002, the next sixty in 59 997, each batch with as many periodic ticks as reads; and with a thread asking the loader for a module without pause — 18 240, 36 427 and 143 424 early wakes in the three batches — 9 997, 20 000 and 59 999 ms for the same reads. On the code before it the second case published 35 720 times in eight seconds, with nine reads. **Not measured in a game**, and not the whole of the holes: the Agent overwrote a reading that waited for a sample, which is its own change (`07_IPC` §Region 5).
**The gate.** One word, four states — idle, armed, reading, off (`dllmain.cpp`, above `ReadLatencyOnce`). Only the
watchdog arms; a presenting thread takes *armed* to *reading* by compare-exchange and owns the 15.4 KB report
buffer (static storage, never a game thread's stack) and the read's result until it stores *idle* or *off*; the
watchdog takes the result only when nobody is reading. No lock, and no second read while one is under way.
### Watchdog observations — not hooks, and listed here so nobody mistakes them for one
| Observation | Mechanism | Rule 4 | Why |
@@ -194,12 +269,17 @@ Every hook must be listed here with a purpose. Anything not on this list is not
|---|---|
| NGX: `NVSDK_NGX_D3D11/D3D12/VULKAN_CreateFeature`, `EvaluateFeature`, `ReleaseFeature` | Which NGX feature is *actually created and evaluated per frame*: SuperSampling (DLSS), RayReconstruction (DLSS-D), FrameGeneration (DLSS-G) |
| NGX parameter accessors (`NVSDK_NGX_Parameter_SetI/GetI/SetUI`) — **exported by `sl.common.dll` only; see below** | `Width`/`Height` (render) vs `OutWidth`/`OutHeight` (output), `PerfQualityValue` (quality preset), sharpness |
| ✅ Streamline: `slEvaluateFeature` · ⏳ `slInit`, `slSetFeatureLoaded`, `slSetConstants`, `slGetFeatureRequirements` | Feature set actually active when the game goes through SL rather than NGX directly (`kFeatureDLSS`, `DLSS_G`, `DLSS_RR`, `Reflex`, `NIS`). **Built 2026-08-09**, identity only: one MinHook detour on `sl.interposer.dll!slEvaluateFeature`, resolved **module-scoped** and installed lazily by the watchdog. Sets `FL_MEASURED_UPSCALER` + `FL_HOOK_UPSCALER_IDENTITY`; decodes `kFeatureDLSS`/`kFeatureNIS` and reports `FL_UPSCALER_UNKNOWN` for anything else. **Never `FL_UPSCALER_NONE`** — a Streamline-only writer cannot see FFX, XeSS or NGX-direct, and `NONE` is the only one of the three states that may be aggregated as a negative |
| ✅ Streamline: `slEvaluateFeature` · ~~⏳ `slInit`~~ *(not hooked, and not to be: owner decision D129, 2026-11-12 — its row below says why)*, ~~`slSetFeatureLoaded`~~ *(its own row below since beta.45)*, ~~`slSetConstants`~~ *(its own row below since beta.43)*, ~~`slGetFeatureRequirements`~~ *(its own row below since beta.44, with `slGetFeatureVersion`)* | Feature set actually active when the game goes through SL rather than NGX directly (`kFeatureDLSS`, `DLSS_G`, `DLSS_RR`, `Reflex`, `NIS`; since beta.34 `kFeatureDLSS_NR` — DLSS 5's neural rendering, its own fact bit and family `FL_HOOK_SL_NEURAL_RENDERING`, D84; rule 4: the `feature` argument of the API hooked, nothing else). **Built 2026-08-09**, identity only: one MinHook detour on `sl.interposer.dll!slEvaluateFeature`, resolved **module-scoped** and installed lazily by the watchdog. Sets `FL_MEASURED_UPSCALER` + `FL_HOOK_UPSCALER_IDENTITY`; decodes `kFeatureDLSS`/`kFeatureNIS` and reports `FL_UPSCALER_UNKNOWN` for anything else. **Never `FL_UPSCALER_NONE`** — a Streamline-only writer cannot see FFX, XeSS or NGX-direct, and `NONE` is the only one of the three states that may be aggregated as a negative. **Since 0.1.0-beta.40 (Advanced metrics train B, D111), and only while the Agent asks** (`FlControlBlock.advancedRequested`): the `sl::DLSSOptions` a title chains into `inputs` — the struct the quality mode is read from — is read further: `outputWidth`/`outputHeight`, `preExposure`, `exposureScale`, `colorBuffersHDR`, the five presets of version 1, and where the struct's own `structVersion` proves them present `ultraQualityPreset` and `useAutoExposure` (≥ 2) and `alphaUpscalingEnabled` (≥ 3) — §Which struct version carries which field. Rule 4: more members of the same argument of the same call; `fl_adv_fields.h` is the one place, `tools/overlay-fields-check.ps1` holds it. No new patch site. **Measured** (the header's own functions, `/O2 /guard:cf /Qspectre /MT`, 20 M calls × 3): +9.5–11.7 ns per evaluation that chains the options while asked; with the word at 0, one relaxed load and a branch not taken — below what the bench resolves. Most titles set these through `slDLSSSetOptions`, which is not an export, and read N/A (train D) |
| ✅ Streamline: `slSetConstants` (0.1.0-beta.43, Advanced metrics train B; owner decisions D111–D116, D122) | **What the game tells Streamline about a frame**, once per frame and viewport, read from the `sl::Constants` it passes — rule 4: an argument of the call hooked, and of it only the members named here. **Read:** `jitterOffset`, `mvecScale`, `cameraNear`, `cameraFar`, `cameraFOV`, `cameraAspectRatio` and the seven `Boolean`s (`depthInverted`, `cameraMotionIncluded`, `motionVectors3D`, `reset`, `orthographicProjection`, `motionVectorsDilated`, `motionVectorsJittered`). **Never read:** the five matrices (bytes 32..351 of the struct), `cameraPinholeOffset`, and the camera's position and three axes (bytes 368..423) — `tools/overlay-fields-check.ps1` refuses their names anywhere in code loaded into a game, and ctest `fl_adv_fields` publishes two structs that differ in every one of those bytes and gets the same bytes out. **The one row that is not installed unless asked for** (D111): `InstallByFamily` returns before it resolves the symbol in a process the Agent has never asked for Advanced metrics — resolving takes a reference on the module — so with the option off the export's first bytes are the file's (`fl_guard` `[advanced]` reads them from the target). Once asked, the patch goes in on that watchdog tick and stays for the life of the process; the detour forwards without looking while the request is off, and after a stop (`MayObserve`). **Every call** checks the struct's GUID and a `structVersion` of at least 1 before any member is read, and the viewport handle's the same way; every member read is one version 1 carries, so a version-1 struct laid against unreadable memory is read whole without a fault (the same ctest). **Summaries are the first viewport's**: a title with two viewports sets constants for each every frame, and "the last value" across them would flip twice a frame; the call count is of every viewport. `FLT_MAX` is Streamline's "not set" here and is kept nowhere (D119's plane at infinity is FidelityFX's). Family `FL_HOOK_SL_CONSTANTS` (bit 13), published with the patch: it entitles no field of a frame record, and tells "hooked, never called" from "not hooked". **Measured overhead** (the detour's body on the headers, the Overlay's flags, 20 million calls a run, three runs): not asked **+0.57 / +0.60 / +0.58 ns**; asked, the summarised viewport **+19.61 / +19.54 / +19.79 ns**; asked, another viewport +5.83 / +6.01 / +5.62 ns. |
| ✅ Streamline: `slGetFeatureVersion` and `slGetFeatureRequirements` (0.1.0-beta.44, Advanced metrics train B's third step; owner decisions D111–D116, D122) | **What Streamline answers when the game asks about a feature** — which version it is, and what it requires. Two rows, two patches, two detours of one shape: **the original runs first, exactly once, on every path, with both arguments forwarded untouched**, and the struct the game passed is read only when the original said `eOk` — a refusal leaves the struct as the game made it, which is not an answer. Rule 4: an argument of the call hooked, read after the call, and of it only the members named here. **Read from `sl::FeatureVersion`:** `versionSL`, `versionNGX`. **Read from `sl::FeatureRequirements`:** `flags`, `maxNumCPUThreads`, `maxNumViewports`, `numRequiredTags`, `osVersionDetected`, `osVersionRequired`, `driverVersionDetected`, `driverVersionRequired`. **Never read:** `requiredTags` — the pointer the count belongs to — and everything after `driverVersionRequired`: the Vulkan queue counts and the pointers to the names of extensions and features. **No pointer of the game's is followed** (`tools/overlay-fields-check.ps1`, parameters `fv` and `fr`; ctest `fl_adv_fields`: two answers that differ only in those bytes publish the same region, and an answer cut off where the Vulkan members begin, laid against unreadable memory, is read without a fault; the injected `fl_guard` case's stub hands back wild pointers and no fault is counted). **Installed only where asked for** (D111), as `slSetConstants` is: `InstallByFamily` returns before it resolves either symbol in a process the Agent has never asked for Advanced metrics; once in, a patch stays for the life of the process and its detour forwards without looking while the request is off. **They are the first two rows of the table**: a game asks these while it starts, the watchdog patches rows in the table's order, and each patch stops the game's threads for a moment (0.1–0.25 s apart in the owner's logs of 2026-11-12, where `slSetConstants` was the fifth Streamline patch in, 1.9–7.8 s after the attach). *(Since 0.1.0-beta.46 the rows a tick finds go in together, under one stop of the threads — §Unhooking, Batches — so the order no longer decides which row waits behind which; it is still the order of their log lines.)* In a process never asked they are passed over and the order changes nothing. ~~**Most sessions will still have missed the calls** — an attach after the game's start-up is the usual case — and the session then says the call was hooked and the game did not ask again.~~ *(Measured 2026-11-12, the owner's eight sessions on beta.44: both calls were read in six of six Streamline sessions on five games — Unreal's Streamline plugin asks again after the attach, Cyberpunk 2077 asks again for requirements and not for versions. A game that asks only while starting still reads "hooked, not asked again".)* Three conditions before a read, as for `slSetConstants`: the Agent is asking, observing has not stopped, and the struct carries its own GUID and a version of at least 1 on this call (a struct with another GUID is not kept: the injected case sends one). Every member read is version 1's (§Which struct version carries which field). **Twelve slots** in region 5 (`FlAdvancedState.features`, layout 5 as it was), one per feature in the order first asked about; the last answer about a feature is the one kept; a thirteenth feature is not kept. **A new request does not empty them**: what version a feature is and what it requires are facts about the process, and a second session of a game that kept running would otherwise have nothing where the first had the answer — the one group `Reset` leaves, with its bit of the mask. Streamline 1.x is refused before anything is resolved (`SpeaksStreamline2`). Families `FL_HOOK_SL_FEATURE_VERSION` (bit 15) and `FL_HOOK_SL_FEATURE_REQUIREMENTS` (bit 16), each published with its patch; neither entitles a field of a frame record. **Measured overhead** (the detours' bodies on the headers, the Overlay's flags, 20 million calls a run, three runs, eight features asked about in turn): the version — not asked **+0.58 / +0.61 / +0.58 ns**, asked **+14.82 / +14.54 / +14.47 ns**; the requirements — not asked **+0.58 / +0.63 / +0.61 ns**, asked **+20.88 / +21.01 / +20.96 ns**. A game makes these calls a handful of times. |
| ✅ Streamline: `slIsFeatureSupported`, `slIsFeatureLoaded`, `slSetFeatureLoaded` (0.1.0-beta.45; owner decisions D129–D131) | **Whether a feature is supported, whether it is loaded, and the game switching it on and off.** Three rows, three patches. Rule 4: bare arguments and results of the calls hooked, and no vendor struct at all — the feature's id; the `sl::Result` `slIsFeatureSupported` returned, **every result**, since a refusal is the answer to that question (the original runs first, both arguments forwarded untouched); the `bool` `slIsFeatureLoaded` wrote, read only when it returned `eOk`; the `bool` the game passed to `slSetFeatureLoaded`, noted before the call. **The adapter `slIsFeatureSupported` is asked about is never read** (D131): `sl::AdapterInfo` holds the adapter's LUID or a Vulkan physical device; the detour forwards the reference and `tools/overlay-fields-check.ps1` refuses the names of its members (`deviceLUID`, `deviceLUIDSizeInBytes`, `vkPhysicalDevice`) anywhere in the Overlay and the Vulkan layer — the fixture's adapter carries a wild pointer. **Installed only where asked for** (D111), as the four rows before them. `slIsFeatureSupported` and `slIsFeatureLoaded` follow the two rows of beta.44 at the head of the table; `slSetFeatureLoaded` follows `slSetConstants`. **Kept in word 1 of the feature's slot** (region 5, layout 5 as it was): bit 2 and the result in bits 8..15, bits 3 and 4, bits 5 and 6 — the last two a session's, cleared by a new request; the rest the process's, kept (`fl_adv_fields.h` `kFeatureSessionBits`). A refused feature therefore has a slot now, with nothing in it but that word. The declarations of the three functions are the same from Streamline v2.0.0 to v2.14.16, and `sl::Result` is the same forty codes in the same order at all sixteen tags (`include/sl_result.h`, `05a3c7e3889c` … `8a844b68ff55`), so a code means the same whatever interposer answered. Families `FL_HOOK_SL_FEATURE_SUPPORTED` (bit 17), `FL_HOOK_SL_FEATURE_LOADED` (18), `FL_HOOK_SL_SET_FEATURE_LOADED` (19), each published with its patch. **Measured overhead** (the detours' bodies on the headers, the Overlay's flags, 20 million calls a run, three runs, eight features in turn): `slIsFeatureSupported` — not asked **+1.50 / +1.48 / +1.50 ns**, asked **+11.97 / +11.98 / +12.08 ns**; `slIsFeatureLoaded` — **+1.79 / +1.77 / +1.89 ns**, **+12.35 / +12.31 / +12.33 ns**; `slSetFeatureLoaded` — **+1.44 / +1.47 / +1.80 ns**, **+11.28 / +11.54 / +11.45 ns**. **`slInit` is not a row and will not be** (D129): a game calls it once, before the Overlay is in the process or before a patch can land — launch mode waits for the presentation runtime before it injects, nothing is ever patched on a game's thread (§H2), and a row that exists only for Advanced metrics waits for the Agent's request. **A stop takes about 160 ms for every patch it restores** (measured on the fixture, 0.1.0-beta.45: each `MH_DisableHook` snapshots the system's threads): twelve patches took 1.9 s, fifteen 2.4 s, and the watchdog's wait for the stop's log lines, 2 s until then, is now a quarter of a second for every patch the registry can hold. **Since 0.1.0-beta.46 (D134) a stop restores its patches under one stop of the threads**: eighteen in 76–90 ms on the fixture, where one by one they took 1.37–1.56 s (§Unhooking, Batches). |
| ✅ Streamline: `slAllocateResources`, `slFreeResources`, `slShutdown` (0.1.0-beta.45; owner decision D130) | **The game allocating and freeing a feature's resources, and shutting Streamline down — counted, and nothing else.** Three rows, three patches. Rule 4: the feature's id is the only argument looked at; the command buffer and the viewport of the two resource calls are forwarded and never read (the fixture's command buffer is a wild pointer), and `slShutdown` has no argument. **`slFreeResources` and `slShutdown` are on a game's way out**: each detour notes the call first and then calls the original, exactly once, and runs nothing of ours after it returns; the fixture goes on calling the stub after `slShutdown`, through every detour. **Installed only where asked for** (D111); the last three Streamline rows of the table. **Kept** in word 1 of the feature's slot as two four-bit counts — allocations in bits 16..19, frees in 20..23, each stopping at 15 and never carrying into its neighbour — and, for `slShutdown`, in `FlAdvancedState.slShutdowns` (@308, the first of three reserved words; layout 5 as it was) with `FL_ADV_VALID_SL_SHUTDOWN` (bit 6 of the mask). All three are a session's: a new request clears them. Families `FL_HOOK_SL_ALLOCATE_RESOURCES` (bit 20), `FL_HOOK_SL_FREE_RESOURCES` (21), `FL_HOOK_SL_SHUTDOWN` (22). The three declarations are the same from Streamline v2.0.0 to v2.14.16. **Measured overhead** (the detours' bodies on the headers, the Overlay's flags, 20 million calls a run, three runs): `slAllocateResources` — not asked **+0.70 / +0.72 / +0.80 ns**, asked **+10.43 / +10.53 / +10.74 ns**; `slFreeResources` — **+0.61 / +0.84 / +0.58 ns**, **+10.33 / +10.57 / +10.44 ns**; `slShutdown` — not asked within the bench's noise (+0.06 / −0.12 / +0.04 ns), asked **+3.39 / +3.37 / +3.52 ns**. Inventory 23 → 26 rows; a Streamline title with Advanced metrics asked for now carries thirteen patches on the interposer, and a stop restores each in about 160 ms (the row above). **In real games (2026-11-13: six sessions, five games)** `slAllocateResources` was not called once after the attach, `slFreeResources` once in one game (Phantom Blade Zero, frame generation's) and `slShutdown` once in one (Dying Light: The Beast); no session counted a fault, and the two games that fault while closing did so at the offsets they had before these hooks. Installing the thirteen interposer patches took 0.7 s to 2.8 s of a game's start, 57 to 215 ms a patch. The owner's word on both (2026-11-13): `slAllocateResources` keeps its patch and is looked at again with more sessions (D135); installs and restores are to be batched, one stop of the threads for a batch (D134, 0.1.0-beta.46). |
| ✅ Streamline: `slSetTag` — **and, since 2026-09-05, every tag's TYPE** (`slTagCensus`; a HUD-less / UI tag marks the present `FL_FG_DLSS_G`, rule 4: the same argument) | The **global** resource tags, where a title states the size of the buffer it is upscaling *from*: `kBufferTypeScalingInputColor`'s extent → `renderW/H` — **or, since 2026-09-05, the size the tagged `sl::Resource` declares when the extent is zero** (`TagSize` in `fl_sl_inputs.h`, shared by all three tag routes; GUID-checked; a Resource declaring no size is still the honest unknown). **Built 2026-08-15** (`FL_HOOK_UPSCALER_PARAMS`), and it is an ALTERNATIVE to the local tags in `slEvaluateFeature`'s `inputs`, not a layer over them — `sl_core_api.h:258` says global and local tags "do NOT interact", so a title using one yields nothing from the other and both are read. Published only on a frame where an evaluation was also seen, because a tag is viewport state that outlives any one frame. This row was MISSING from this table while the hook shipped, which is the gap the banner above now names |
| ✅ Streamline: `slSetTagForFrame` — **and every tag's TYPE, as above** | The same global tags through **Streamline 2.8's frame-based entry point**, which deprecates `slSetTag`: `(const FrameToken&, const ViewportHandle&, const ResourceTag*, uint32_t, CommandBuffer*)` — the token first, then exactly `slSetTag`'s list, handed to the same tag walk. **Built 2026-09-04** (`FL_HOOK_UPSCALER_PARAMS`, its own trampoline; the family is now **published when whole** — once every tag export the loaded interposer has is patched — because publishing on the first row entitled records the second was not yet producing, 38 of 41 on the fixture: the ffx leaves' #110 defect again, fixed the same way). Measured need: Dying Light: The Beast (SL 2.8.0) published `Dlss` identity on every batch and the params bit on none across three captures, because a 2.8 title that tags per frame never calls the deprecated export. The token is not read; the frame count stays with `slGetNewFrameToken`. The installer picks this detour by SYMBOL (`kSymbolSlSetTagForFrame`, bound to the row by `static_assert`), not by family, since the two rows share a family bit and differ in signature |
| ✅ **FidelityFX ffx-api: `ffxDispatch` on the FOUR modules a game calls** — `amd_fidelityfx_dx12.dll` (SDK 1.1.x monolith), `amd_fidelityfx_upscaler_dx12.dll`, `amd_fidelityfx_framegeneration_dx12.dll`, **and `amd_fidelityfx_loader_dx12.dll`** (added the evening of 2026-09-04: three loader-shipping titles at FSR produced zero dispatches at any leaf export, so the loader is where their calls arrive) · ⏳ FSR 3.0 host route `ffx_fsr3_x64.dll!ffxFsr3ContextDispatchUpscale` / `ffxFsr3DispatchFrameGeneration` · ~~`ffxFsr2ContextCreate`~~ · ~~`ffxCreateContext`~~ | **Built 2026-09-04** (`HANDOFF` item 7c). One observer behind three trampolines reads the DESCRIPTOR the title passed — head type first, body only after the type matched a constant from the vendored MIT headers (`third_party/fidelityfx/`, tag v2.3.0): **UPSCALE** → identity (`FSR3` from the monolith, `FSR_UNVERSIONED` from the 2.x upscaler DLL, which hosts FSR 3.1 and FSR 4 behind one type) + `renderSize` → `renderW/H` (quality and sharpness `0xFF`: the dispatch carries neither — *since beta.25 the quality byte is the mode the `ffxQuery` row below names for the dispatched size, `0xFF` when none does*) + one COUNT; **PREPARE / PREPARE_V2** → `frameID` counted on a *new* index only — this vendor's `slGetNewFrameToken`, the `fgEvaluations` producer once a title has ever prepared, with the UPSCALE count as the fallback for a title that never does; **FRAMEGENERATION** → `fgMode = FSR_FG` on that present. Rule 4: the argument of the API we hooked, never `context`, `pNext`, a command list or a resource. **Why all four:** a UE5 title ships the two leaves with no loader and calls their exports; a loader-shipping title calls the loader's export and the signed loader reaches its providers through an object, not the leaves' exports (measured: Dying Light: The Beast, KCD2, Wukong). The morning's "never the loader" rested on the export tables alone and was reversed by the run. What guards against a loader that DID re-enter a leaf's export is the K = 1 control with all four hooked and the consumer's `frames/upscale-drained`, both of which read 2× in that case. `ffxCreateContext` is deliberately NOT a row — nothing it carries fills a field the dispatch does not, and its honest family would collide with `slEvaluateFeature`'s under the installer's equality binding. The family is compound (`IDENTITY \| PARAMS \| FG_EVALUATIONS`), pinned by `static_assert`s that also bind the leaf table to the rows both ways and refuse the loader as a row. Streamline keeps precedence for the count once it has ever issued a token, so every §S31-validated title is byte-identical. Second count: the consumer prints `frames/upscale-drained`, the AMD `tokens/batch`. **FSR 3.0 — BUILT 2026-09-05 as a fifth AMD target:** `ffx_fsr3_x64.dll!ffxFsr3ContextDispatchUpscale` (Cyberpunk's copy; named export, its own descriptor, vendored from tag **`fsr3-v3.0.4`** — the tag whose twelve `ffxFsr3*` declarations match the shipped module, not the `v1.1.4` first named). Rule 4: the detour reads `renderSize` of the descriptor the title passed and nothing else — never the context, a resource or the command list; the offset is pinned to a literal (1256) so a re-vendoring that moved it fails to compile. Same family and publish point as the four rows (`PublishFfxFamilyIfWhole` waits for a loaded host too), keyed by module in the installer; identity `FSR3` as a fact. `ffxFsr3DispatchFrameGeneration` is deliberately not a row (§H11 carries the reversal condition); the `ffx_fsr3upscaler_x64.dll` leaf is the pre-committed alternative if the facade reads silent (row B2) — never both, or `frames/upscale-drained` reads 2.00 |
| ✅ **FidelityFX ffx-api: `ffxQuery` on the same four modules** (beta.25, owner decision D65) — `amd_fidelityfx_dx12.dll`, `amd_fidelityfx_upscaler_dx12.dll`, `amd_fidelityfx_framegeneration_dx12.dll`, `amd_fidelityfx_loader_dx12.dll` | **Built 2026-10-19.** The mode an FSR title runs at reaches the vendor only as an argument of `ffxQuery` — "what render size (`…GETRENDERRESOLUTIONFROMQUALITYMODE`, `0x00010003`) or ratio (`…GETUPSCALERATIOFROMQUALITYMODE`, `0x00010002`) does this mode give at this display size?" — asked before the title sizes its targets; the per-frame dispatch carries the size alone. The detour calls the original FIRST, exactly once, and only when it returned `FFX_API_RETURN_OK` reads the descriptor's type, display size and mode and the values the vendor wrote through the title's own out pointers, into a five-slot table (one per `FfxApiUpscaleQualityMode`); the `ffxDispatch` UPSCALE arm names the ONE mode whose kept answer fits the dispatched `renderSize` (and `upscaleSize`, where stated) as `0x10 + mode` in `upscalerQuality`, else `0xFF` (`03_METRICS` §Upscaling). **Rule 4:** the arguments of the API we hooked — the descriptor the title passed and the out values of that same call — never `context`, a resource or anything the title did not hand the hooked export; nothing of the title's is written, and every read is under the hook's SEH guard. Both query layouts and type ids are pinned to literals by `static_assert`s, identical at SDK v1.1.4 and v2.3.0. **Its own family**, `FL_HOOK_UPSCALER_QUALITY` (bit 11 of `hooksInstalledMask`, mirrored in `FlHookFamily` and `HookFamilies`): `InstallByFamily` binds a row by equality, and the query entitles no measured field by itself — the byte rides under `FL_MEASURED_UPSCALER_PARAMS` from the dispatch arm — so the bit says only that the query was hooked, which tells "never asked" from "could not have seen it ask". Published like the dispatch family, once every loaded leaf is patched (`PublishFfxQueryFamilyIfWhole`); the FSR 3.0 host has no `ffxQuery` and no row (`static_assert`). A loader that passes a query on to a leaf's export refreshes one kept answer twice; nothing is counted, so nothing is counted twice. **Measured overhead** (a copy of the two bodies, `/O2 /guard:cf /Qspectre /MT`, 20 M calls × 3 runs, every slot holding both a size and a ratio — the worst case): 7.6–8.2 ns added per UPSCALE dispatch, i.e. per application frame, and 1.6 ns per query beside the trampoline; the per-present path gains one load. Harness: `--ffx-query sizes|ratios`; `fl_guard` "FSR's quality mode". |
| ✅ **FidelityFX ffx-api: `ffxDispatch` on the FOUR modules a game calls** — `amd_fidelityfx_dx12.dll` (SDK 1.1.x monolith), `amd_fidelityfx_upscaler_dx12.dll`, `amd_fidelityfx_framegeneration_dx12.dll`, **and `amd_fidelityfx_loader_dx12.dll`** (added the evening of 2026-09-04: three loader-shipping titles at FSR produced zero dispatches at any leaf export, so the loader is where their calls arrive) · ⏳ FSR 3.0 host route `ffx_fsr3_x64.dll!ffxFsr3ContextDispatchUpscale` / `ffxFsr3DispatchFrameGeneration` · ~~`ffxFsr2ContextCreate`~~ · ~~`ffxCreateContext`~~ | **Built 2026-09-04** (`HANDOFF` item 7c). One observer behind three trampolines reads the DESCRIPTOR the title passed — head type first, body only after the type matched a constant from the vendored MIT headers (`third_party/fidelityfx/`, tag v2.3.0): **UPSCALE** → identity (`FSR3` from the monolith, `FSR_UNVERSIONED` from the 2.x upscaler DLL, which hosts FSR 3.1 and FSR 4 behind one type) + `renderSize` → `renderW/H` (quality and sharpness `0xFF`: the dispatch carries neither — *since beta.25 the quality byte is the mode the `ffxQuery` row below names for the dispatched size, `0xFF` when none does*) + one COUNT *(since 0.1.0-beta.43, owner decision D125: **counted at the first source that saw an UPSCALE and at no other** — `g_ffxUpscaleOwner`, `fl::adv::Owns`. Cyberpunk 2077's loader 2.3.0 hands a dispatch on to the leaf's export — its frame-generation dispatch types are logged on the loader and on the leaf — so an UPSCALE through such a loader reaches two hooked modules, and one count per module read K = 1 as 2.0 on the fixture bound that way (`--ffx-topology 2x-export`: 60 presents, 120 counted). No title run so far upscales through one; it is a precaution. Measured on the header with the Overlay's flags: +0.02 to +0.03 ns on the module that is counted, about 4 ns less on the forwarded copy)*; **PREPARE / PREPARE_V2** → `frameID` counted on a *new* index only — this vendor's `slGetNewFrameToken`, the `fgEvaluations` producer once a title has ever prepared, with the UPSCALE count as the fallback for a title that never does; **FRAMEGENERATION** → `fgMode = FSR_FG` on that present. Rule 4: the argument of the API we hooked, never `context`, `pNext`, a command list or a resource. **Why all four:** a UE5 title ships the two leaves with no loader and calls their exports; a loader-shipping title calls the loader's export and the signed loader reaches its providers through an object, not the leaves' exports (measured: Dying Light: The Beast, KCD2, Wukong). The morning's "never the loader" rested on the export tables alone and was reversed by the run. What guards against a loader that DID re-enter a leaf's export is the K = 1 control with all four hooked and the consumer's `frames/upscale-drained`, both of which read 2× in that case. `ffxCreateContext` is deliberately NOT a row — nothing it carries fills a field the dispatch does not, and its honest family would collide with `slEvaluateFeature`'s under the installer's equality binding. The family is compound (`IDENTITY \| PARAMS \| FG_EVALUATIONS`), pinned by `static_assert`s that also bind the leaf table to the rows both ways and refuse the loader as a row. Streamline keeps precedence for the count once it has ever issued a token, so every §S31-validated title is byte-identical. Second count: the consumer prints `frames/upscale-drained`, the AMD `tokens/batch`. **FSR 3.0 — BUILT 2026-09-05 as a fifth AMD target:** `ffx_fsr3_x64.dll!ffxFsr3ContextDispatchUpscale` (Cyberpunk's copy; named export, its own descriptor, vendored from tag **`fsr3-v3.0.4`** — the tag whose twelve `ffxFsr3*` declarations match the shipped module, not the `v1.1.4` first named). Rule 4: the detour reads `renderSize` of the descriptor the title passed and nothing else — never the context, a resource or the command list; the offset is pinned to a literal (1256) so a re-vendoring that moved it fails to compile. Same family and publish point as the four rows (`PublishFfxFamilyIfWhole` waits for a loaded host too), keyed by module in the installer; identity `FSR3` as a fact. `ffxFsr3DispatchFrameGeneration` is deliberately not a row (§H11 carries the reversal condition); the `ffx_fsr3upscaler_x64.dll` leaf is the pre-committed alternative if the facade reads silent (row B2) — never both, or `frames/upscale-drained` reads 2.00. **Since 0.1.0-beta.40 (Advanced metrics train B, D111), and only while the Agent asks:** the UPSCALE arm reads the rest of `ffxDispatchDescUpscale` — `enableSharpening`, `sharpness`, `jitterOffset`, `preExposure`, `cameraNear`, `cameraFar`, `cameraFovAngleVertical`, `flags`, `reset`, `frameTimeDelta` — into process-local summaries (an "ever seen" mask, a largest magnitude, a lowest and highest, a count and a sum; one module's dispatches are counted, so a dispatch that reaches the loader and a leaf is one). Never on the FSR 3.0 host's legacy descriptor, which is vouched Line truncated
| ✅ **FidelityFX ffx-api: `ffxQuery` on the same four modules** (beta.25, owner decision D65) — `amd_fidelityfx_dx12.dll`, `amd_fidelityfx_upscaler_dx12.dll`, `amd_fidelityfx_framegeneration_dx12.dll`, `amd_fidelityfx_loader_dx12.dll` | **Built 2026-10-19.** The mode an FSR title runs at reaches the vendor only as an argument of `ffxQuery` — "what render size (`…GETRENDERRESOLUTIONFROMQUALITYMODE`, `0x00010003`) or ratio (`…GETUPSCALERATIOFROMQUALITYMODE`, `0x00010002`) does this mode give at this display size?" — asked before the title sizes its targets; the per-frame dispatch carries the size alone. The detour calls the original FIRST, exactly once, and only when it returned `FFX_API_RETURN_OK` reads the descriptor's type, display size and mode and the values the vendor wrote through the title's own out pointers, into a five-slot table (one per `FfxApiUpscaleQualityMode`); the `ffxDispatch` UPSCALE arm names the ONE mode whose kept answer fits the dispatched `renderSize` (and `upscaleSize`, where stated) as `0x10 + mode` in `upscalerQuality`, else `0xFF` (`03_METRICS` §Upscaling). **Rule 4:** the arguments of the API we hooked — the descriptor the title passed and the out values of that same call — never `context`, a resource or anything the title did not hand the hooked export; nothing of the title's is written, and every read is under the hook's SEH guard. Both query layouts and type ids are pinned to literals by `static_assert`s, identical at SDK v1.1.4 and v2.3.0. **Its own family**, `FL_HOOK_UPSCALER_QUALITY` (bit 11 of `hooksInstalledMask`, mirrored in `FlHookFamily` and `HookFamilies`): `InstallByFamily` binds a row by equality, and the query entitles no measured field by itself — the byte rides under `FL_MEASURED_UPSCALER_PARAMS` from the dispatch arm — so the bit says only that the query was hooked, which tells "never asked" from "could not have seen it ask". Published like the dispatch family, once every loaded leaf is patched (`PublishFfxQueryFamilyIfWhole`); the FSR 3.0 host has no `ffxQuery` and no row (`static_assert`). A loader that passes a query on to a leaf's export refreshes one kept answer twice; nothing is counted, so nothing is counted twice. **Measured overhead** (a copy of the two bodies, `/O2 /guard:cf /Qspectre /MT`, 20 M calls × 3 runs, every slot holding both a size and a ratio — the worst case): 7.6–8.2 ns added per UPSCALE dispatch, i.e. per application frame, and 1.6 ns per query beside the trampoline; the per-present path gains one load. Harness: `--ffx-query sizes|ratios`; `fl_guard` "FSR's quality mode". **Since 0.1.0-beta.40 (Advanced metrics train B, D111), and only while the Agent asks:** when the query the vendor answered OK is the upscaler's memory — `FFX_API_QUERY_DESC_TYPE_UPSCALE_GPU_MEMORY_USAGE` (`0x00010008`, ffx-api 1.1.2) or `…_V2` (`0x00010009`, 2.0) — the `FfxApiEffectMemoryUsage` it wrote through the title's own out-pointer is kept (total and aliasable bytes), one module's queries counted. Only a title that asks is answered: FrameLedger never calls `ffxQuery` itself. The frame-generation memory queries (`0x00020007`, `0x0002000b`) are not read. +4.3–6.1 ns per such query while asked. Harness: `--ffx-memory-query v1|v2`; `fl_guard` "region 5 stays empty until the Agent asks". |
| ~~XeSS: `xessD3D12CreateContext`, `xessD3D12Init`, `xessD3D12Execute`~~ **Closed at the census by licence, 2026-09-04** (`20_OPEN_QUESTIONS` §Scope decisions, `18_GPU_VENDOR_APIS` §Intel XeSS SDK): no hook on any `libxess*.dll` export *(corrected 2026-10-04)* | ~~`outputResolution`, `qualitySetting`, XeSS version; `xess_fg` variants for XeFG~~ |
| ✅ **FidelityFX ffx-api: `ffxConfigure` on the same four modules** (0.1.0-beta.43, Advanced metrics train B; owner decisions D111–D116, D122) | **How the game configures FSR frame generation** — a title calls `ffxConfigure` with a frame-generation descriptor, usually every frame. Rule 4: an argument of the call hooked, and of it only what is named here. **Read**, and only when the header's type is `FFX_API_CONFIGURE_DESC_TYPE_FRAMEGENERATION` (`0x20002`): `frameGenerationEnabled`, `allowAsyncWorkloads`, `flags`, `onlyPresentGenerated`, `generationRect`. **Never read:** `swapChain`, the two callbacks and their contexts, `HUDLessColor`, `frameID`; and of every other configure type — key-value pairs, the distortion field, the camera info — nothing but the header's type. ctest `fl_adv_fields` publishes two descriptors that differ in all of those and gets the same bytes out, and hands the detour a key-value descriptor, which is shorter, without a read past it. **Installed only where asked for** (D111), as `slSetConstants` is: `InstallByFamily` returns before it resolves anything in a process never asked for Advanced metrics, so with the option off the export's first bytes in the target are the file's (`fl_guard` `[advanced]`). Published when every ffx-api module the process loaded is patched (`FL_HOOK_FFX_CONFIGURE`, bit 14), as the dispatch and query families are. **Counted at the first module that saw a frame-generation configure** — an owner slot of its own: the fixture's loader has always handed `ffxConfigure` on to the leaf's export, and Cyberpunk 2077's loader does the same with dispatches, so one call reaches two hooked modules (the injected case reads one configure a frame through the loader, not two). The rectangle is stored in one word, sixteen bits a side, so a publication never carries half of one rectangle beside half of the next; a side that is negative or past fifteen bits makes it four `-1`. The descriptor's type is kept for the log whatever it is (`FFX_CONFIGURE`, §Native logging). The detour consults `MayObserve()` first. **Measured overhead** (the detour's body on the headers, the Overlay's flags, 20 million calls a run, three runs): not asked **+1.72 / +1.68 / +1.78 ns** (the type kept for the log); asked, the counted module **+15.02 / +15.55 / +15.11 ns**; asked, the forwarded copy +2.70 / +2.81 / +2.65 ns; asked, another configure type +2.43 / +2.55 / +2.47 ns. |
| ✅ **Application frames per present** (`fgEvaluations`) — **`sl.interposer.dll!slGetNewFrameToken` since 2026-09-03** | Native vs Displayed frame counts at Tier 1 — see `03_METRICS` §Frame Generation. **The producer is the frame-token export**: Streamline hands a title one token per application frame and the title must ask; the detour counts distinct tokens between two presents (pointer changes, nothing dereferenced), `FL_HOOK_FG_EVALUATIONS` is its family, and it installs on every Streamline 2 title including the ones that evaluate nothing through `slEvaluateFeature`. Rule 4: the out-parameter of an API we hooked. Premise and its gate in `03_METRICS`; pre-committed table in `20_OPEN_QUESTIONS` §S31. *The row as it stood before:* ~~NGX/SL/FFX/XeSS FG feature evaluations per present. **Built 2026-08-15 for Streamline, and MEASURED TO YIELD NOTHING ON THAT ROUTE**~~ — `slEvaluateFeature(kFeatureDLSS_G)` is never called by Cyberpunk 2077 (0 across ~14,000 batches, four FG settings), so this row installs, is honest, and counts zero; see `docs/HANDOFF.md` item 3. Built on the detour that was already there: `kFeatureDLSS_G` contributes to a saturating COUNT in the same word the feature bits live in (`fl_sl_seen.h`) rather than to a bit, because a bit collapses several evaluations between two presents into one — and under multi-frame generation that is the common case, not the edge. The row's family is therefore compound (`FL_HOOK_UPSCALER_IDENTITY \| FL_HOOK_FG_EVALUATIONS`), pinned by a `static_assert` because `hookinventory-check` reads that column as an opaque identifier. **Counts EVALUATIONS, not generated frames** — the 2026-08-14 owner ruling; the multiplier lives in `sl::DLSSGOptions`, which is only reachable by the route §2b refused. ~~Non-NVIDIA FG vendors are **deferred with a written rationale**: `amd_fidelityfx_framegeneration_dx12.dll` 3.1.5 exports only the five generic `ffx*` entry points, identical to five sibling modules, so identity is in the arguments and we have no headers~~ — **AMD has a producer since 2026-09-04** (the FidelityFX row above: the PREPARE dispatch's `frameID`, or the UPSCALE count on a title that never prepares, feeds `fgEvaluations` when Streamline has never issued a token); Intel stays at the census (`20_OPEN_QUESTIONS` §Scope decisions, by licence); the FSR 3.0 host DLL's UPSCALE (`ffxFsr3ContextDispatchUpscale`, since 2026-09-05) counts like an ffx-api UPSCALE when neither Streamline nor a PREPARE has spoken |
> `GetFrameStatistics().PresentCount` is **not** in this inventory and must not be
@@ -300,6 +380,72 @@ So:
> **`third_party/streamline/README.md` asserted this pass existed for five days
> before it did**, which is why it is described here rather than only there.
### Which struct version carries which field (owner decision D115, 2026-11-11)
**A field of a vendor struct is read only where it is proven to be there.** A game hands a hook a struct of the
version it was built against; a member a later version added lies past the end of what an older game allocated, and
reading it is reading game memory outside the argument (rule 4) — not merely a wrong value. So every read of a tail
member is gated on the struct's own version, and the gate comes from this table, not from the one header in the tree
— which declares only the newest shape, and for `sl::DLSSOptions` says nothing about which version added what.
Established 2026-11-11 from the upstream tags, through GitHub's API: each tag's header was fetched by blob and the
struct's declarations — type, name and order — compared across tags. The members' own types were checked at the
oldest and the newest tag (`enum Boolean : char`, `enum class DLSSMode : uint32_t`, `enum class DLSSPreset :
uint32_t`, `float2`/`float3`/`float4x4`; no `#pragma pack` in either).
**Streamline** (`NVIDIA-RTX/Streamline`; tags v2.0.0, v2.1.0, v2.2.0, v2.4.0, v2.4.15, v2.7.2, v2.7.30, v2.7.32,
v2.8.0, v2.9.0, v2.10.0, v2.10.3, v2.11.1, v2.12.0, v2.14.1, v2.14.16):
| Struct | `structVersion` | First tag (blob of the header) | Members that version adds, in order |
|---|---|---|---|
| `sl::DLSSOptions` (`include/sl_dlss.h`) | 1 | v2.0.0 (`7de7411e507d`) | `mode`, `outputWidth`, `outputHeight`, `sharpness`, `preExposure`, `exposureScale`, `colorBuffersHDR`, `indicatorInvertAxisX`, `indicatorInvertAxisY`, `dlaaPreset`, `qualityPreset`, `balancedPreset`, `performancePreset`, `ultraPerformancePreset` |
| | 2 | v2.1.0 (`60882f60e6f4`) | `ultraQualityPreset`, `useAutoExposure` |
| | 3 | v2.4.0 (`27d352db364f`) | `alphaUpscalingEnabled` |
| `sl::Constants` (`include/sl_consts.h`) | 1 | v2.0.0 (`db6f37adcf66`) | every member through `motionVectorsJittered` — the matrices, `jitterOffset`, `mvecScale`, `cameraPinholeOffset`, the camera's position and axes, `cameraNear`, `cameraFar`, `cameraFOV`, `cameraAspectRatio`, `motionVectorsInvalidValue`, the seven `Boolean`s |
| | 2 | v2.4.0 (`bba15338a623`) | `minRelativeLinearDepthObjectSeparation` |
| `sl::FeatureVersion` (`include/sl.h`; `include/sl_core_types.h` from v2.7.2) | 1 | v2.0.0 (`fc2d6ee39fea`) | `versionSL`, `versionNGX` |
| `sl::FeatureRequirements` (the same headers) | 1 | v2.0.0 (`fc2d6ee39fea`) | `flags`, `maxNumCPUThreads`, `maxNumViewports`, `numRequiredTags`, `requiredTags`, `osVersionDetected`, `osVersionRequired`, `driverVersionDetected`, `driverVersionRequired`, then ten Vulkan members from `vkNumComputeQueuesRequired` to `vkFeatures13` |
| | 2 | v2.1.0 (`79037e0974d2`) | `vkNumOpticalFlowQueuesRequired` |
`sl::DLSSOptions` keeps those seventeen declarations, same types and order, through v2.14.16 (`9a875bd4f843`); v2.8.0
only re-spelled the deprecated `sharpness`. `sl::Constants` is unchanged from v2.4.0 through v2.14.16
(`15589e577df9`). Streamline 1.x is refused before any read (`SpeaksStreamline2`). **Since 0.1.0-beta.43 the
Overlay reads `sl::Constants`, and only members of version 1** — nothing of version 2 — so no version test beyond
"at least 1" is needed, and the read ends before `minRelativeLinearDepthObjectSeparation`, the member version 2
added.
**The two answers about a feature** (established 2026-11-12, the same way, over the same sixteen tags):
`sl::FeatureVersion` is two members at every tag, and `sl::FeatureRequirements` is the nineteen of v2.0.0 plus the
one v2.1.0 added, unchanged through v2.14.16 (`a255f5c5e99a`). `sl::Version` is three `uint32_t` — `major`, `minor`,
`build` — in `include/sl_version.h` from v2.0.0 (`deec7cd04363`) to v2.14.16 (`dc3041ae8b44`), and the two functions'
declarations are the same at both ends: `sl::Result slGetFeatureVersion(sl::Feature, sl::FeatureVersion&)` and
`sl::Result slGetFeatureRequirements(sl::Feature, sl::FeatureRequirements&)`. **Since 0.1.0-beta.44 the Overlay reads
both, and only members of version 1**: the read of the requirements ends at `driverVersionRequired`, before the first
Vulkan member, so no version test beyond "at least 1" is needed.
**FidelityFX** (`GPUOpen-LibrariesAndSDKs/FidelityFX-SDK`; tags v1.1.0 — FSR 3.1.0, the first ffx-api — v1.1.1,
v1.1.2, v1.1.3, v1.1.4, v2.0.0, v2.1.0, v2.2.0, v2.3.0, v2.4.0). The descriptor header carries a `type` and no size
or version, so the question is whether the declaration ever changed:
| Struct | Tags | Result |
|---|---|---|
| `ffxDispatchDescUpscale` | v1.1.0 (`68860479bb94`) … v2.4.0 (`9d4a3e5f6c60`) | **one declaration in all ten**: twenty-three members, `flags` the last |
| `ffxConfigureDescFrameGeneration` | v1.1.0 (`0ecdd4b4e59b`) … v2.4.0 (`823a3b4720f1`) | **one declaration in all ten**: thirteen members |
| `FfxApiResource`, `FfxApiResourceDescription`, `FfxApiFloatCoords2D`, `FfxApiDimensions2D`, `FfxApiRect2D` | v1.1.0 … v2.4.0 | one declaration each in all ten |
| `ffxQueryDescUpscaleGetGPUMemoryUsage` (`0x00010008`), `FfxApiEffectMemoryUsage` | from v1.1.2 | one declaration since; a game that sends the type knows the struct |
| `ffxQueryDescUpscaleGetGPUMemoryUsageV2` (`0x00010009`) | from v2.0.0 | one declaration since |
**What follows for a hook** (Advanced metrics, train B):
- `sl::DLSSOptions`: a member is read at `structVersion >=` the row that adds it — never "at least 1" for the tail.
- `sl::Constants`: everything this project may read is version 1's. The camera's position, axes and matrices are
never read at any version (rule 4; the catalog's `PlayerState`).
- `ffxDispatchDescUpscale`, `ffxConfigureDescFrameGeneration`: every member may be read on the ffx-api route, since
no ffx-api release has a shorter one. `ffxDispatchDescFrameGenerationPrepareV2`'s camera arrays are never read.
- **Not established, so nothing more is read there**: the FSR 3.0 host's legacy descriptors (`fsr3-v3.0.x`) beyond
the `renderSize` the existing hook pins; any struct of a Streamline or FidelityFX release newer than the last tag
above — a re-vendoring re-runs this comparison (`third_party/*/README.md` §Checklist).
### Ray tracing
| Hook | Yields |
|---|---|
@@ -334,7 +480,7 @@ So:
| Hook / call | Yields |
|---|---|
| ~~`IDXGIAdapter3::QueryVideoMemoryInfo` (we **call** it, not hook it, once per second from our own thread)~~ | ~~**Per-process VRAM** `CurrentUsage` + `Budget` — previously impossible; now exact~~ **Never built; superseded 2026-10-03 (D43):** the game's memory is read from OUTSIDE the game by the Agent (`18_GPU_VENDOR_APIS` §The game process's memory), so nothing in the Overlay calls it. Only the budget would need it, and that stays open (`20_OPEN_QUESTIONS` §M11) |
| NVAPI Reflex: hook `NvAPI_D3D_SetSleepMode`, `NvAPI_D3D_SetLatencyMarker`; call `NvAPI_D3D_GetLatency` | Reflex on/off + PC latency breakdown. Declarations come from the vendored MIT NVAPI headers (`18_GPU_VENDOR_APIS` §L3); the Overlay links them like any other header. NVIDIA-only — the whole block compiles out to a no-op capability flag elsewhere |
| NVAPI Reflex: ~~hook `NvAPI_D3D_SetSleepMode`, `NvAPI_D3D_SetLatencyMarker`~~; call `NvAPI_D3D_GetLatency` and `NvAPI_D3D_GetSleepStatus` *(the two hooks cannot be: neither function is an export — `20_OPEN_QUESTIONS` §M14. **The calls are made since 0.1.0-beta.47** — §Calls the Overlay makes into a driver, above; owner decisions D132, D136)* | Reflex on/off + PC latency breakdown. Declarations come from the vendored MIT NVAPI headers (`18_GPU_VENDOR_APIS` §L3); the Overlay links them like any other header. NVIDIA-only — the whole block compiles out to a no-op capability flag elsewhere |
### Explicitly not hooked
Input APIs, file I/O, network, memory allocators, window messages beyond what the overlay needs, anything game-specific. Rule 4 in CLAUDE.md.
@@ -511,9 +657,11 @@ static_assert(sizeof(FlFrameRecord) == 64);
> v3 makes it safe by CONSTRUCTION. Every enum's 0 is now `NOT_REPORTED`, and
> `rtFlags`' polarity is flipped so its bits mean *observed*.
>
> Four answers items 4/6/7 owe had no home and now do: DLSS-SR **and** Ray
> Reconstruction concurrently (`featureFlags`, since RR was a mutually exclusive
> `upscaler` value), `upscalerSharpness`, and — in `FlWriterState`, because they
> Four answers items 4/6/7 owe had no home and now do: DLSS **with** Ray
> Reconstruction (`featureFlags`, since RR was a mutually exclusive
> `upscaler` value — *corrected 2026-11-14: this said "DLSS-SR and Ray
> Reconstruction concurrently"; RR runs instead of super resolution, it is still
> DLSS*), `upscalerSharpness`, and — in `FlWriterState`, because they
> are session facts and not per-frame — the device **RT tier** without which
> `03_METRICS`' definite RT `No` had no producer at all, plus `rtStateObjectsCreated`
> and `rasterPsoCreated`.
@@ -571,7 +719,7 @@ Field notes — each of these was a defect in an earlier revision:
- **SPSC lock-free ring**, capacity a power of two (default 8192 records = 512 KB), overwrite-oldest.
- Writer/reader seqlock protocol, including the rule that the payload write must
**exclude** the `seq` field and that `seq` is never reset: `07_IPC` §Protocol rules.
- Header: ~~three~~ **four** separate 64-byte lines before the ring *(corrected 2026-10-04: since layout 4, 2026-09-27,
- Header: ~~three~~ **four** separate 64-byte lines before the ring — and, since layout 5 (2026-11-11), region 5's twenty-three, which the watchdog alone writes and only while the Agent asks (`07_IPC` §Region 5) — *(corrected 2026-10-04: since layout 4, 2026-09-27,
beta.10, `FlDisplayState` at 0xC0 — Overlay-written at most every 500 ms; the ring starts at 0x100)* — `FlShmHandshake` (write-once),
`FlWriterState` (Overlay-written: `writeIndex`, `status`, `apiMask`, `faultCount`),
and `FlControlBlock` (**Agent**-written: `pauseRequested`, `unhookRequested`,
@@ -579,7 +727,7 @@ Field notes — each of these was a defect in an earlier revision:
header — they are written by the other process, and mixing them into an
Overlay-written line reintroduces exactly the false sharing the split exists to
prevent. Layout is normative in `07_IPC` §A + B.
- **The hot path performs: one QPC read, a few field reads from cached state, one 60-byte store, two relaxed atomic stores and two compiler fences.** No syscall, no allocation, no lock, no logging. Target ≤ 1 µs per present. *(Added 2026-10-04: the summary leaves out the per-present `IDXGISwapChain::GetLastPresentCount` call (§H5, the `dxgiUnseen` count) and the display sample every 500 ms at most (region 4).)*
- **The hot path performs: one QPC read, a few field reads from cached state, one 60-byte store, two relaxed atomic stores and two compiler fences.** No syscall, no allocation, no lock, no logging. Target ≤ 1 µs per present. *(Added 2026-10-04: the summary leaves out the per-present `IDXGISwapChain::GetLastPresentCount` call (§H5, the `dxgiUnseen` count) and the display sample every 500 ms at most (region 4).)* *(Added 2026-11-13, train C, 0.1.0-beta.47: after the store, one relaxed load of the latency gate; and, while a session asks for Advanced metrics and the watchdog armed one, a call into the driver about once a second — §Calls the Overlay makes into a driver. That call is the one thing on this path that is not bounded by FrameLedger's own code: 0.5–2.6 µs measured against the installed driver, and a call of 2 ms or more is the last.)*
- Per-frame mutable state (current upscaler, render res, dispatch counts) lives in a small struct updated by the feature hooks and *read* by the present hook; counters reset at present.
## The watchdog thread — the only thread we add to the game
@@ -643,8 +791,9 @@ Every hook body is wrapped:
```
- Any SEH exception inside our code ⇒ increment `faultCount`, record which hook, **return control to the original function** so the game continues.
- 3 faults total ⇒ set `status = self_disabled`, ~~uninstall all hooks (`MH_DisableHook(MH_ALL_HOOKS)`)~~ restore each patch's prologue only while its bytes are still our jump (compare-and-restore per patch through `StopObserving`, §Unhooking, §H7 — `MH_ALL_HOOKS` would have removed an overlay chained after us; corrected 2026-09-27), stop writing, stay dormant. The Agent sees the flag and finalizes the session as `degraded`.
- 3 faults total ⇒ set `status = self_disabled`, ~~uninstall all hooks (`MH_DisableHook(MH_ALL_HOOKS)`)~~ restore each patch's prologue only while its bytes are still our jump (compare-and-restore per patch through `StopObserving`, §Unhooking, §H7 — `MH_ALL_HOOKS` would have removed an overlay chained after us; corrected 2026-09-27; since 0.1.0-beta.46 the patches that are still ours come out together, under one stop of the threads — §Unhooking, Batches), stop writing, stay dormant. The Agent sees the flag and finalizes the session as `degraded`.
- Never `__try` around the call to the original function — only around *our* code, so we never mask a game bug as ours or vice versa.
- **One exception, since train C (2026-11-13): the two driver getters the Overlay calls itself** (§Calls the Overlay makes into a driver) are under a guard of their own, `fl_reflex.h`'s, with the same filter. They are not "the original function" — the game did not make these calls, we did, so a fault in them is ours to answer for. It is counted once through `NoteFault`, and the read is switched off for the process in the same breath, so that one fault cannot become the three that self-disable. **A caught fault does not give back a lock the driver held**: the guard is containment, not safety, which is why nothing is called again after it.
- The trampoline call to the original is always executed exactly once, on every path, including error paths.
## Unhooking
@@ -663,6 +812,94 @@ Every hook body is wrapped:
> foreign one in ctest `fl_unhook_inline` (the vtable form stays in `fl_unhook_preserves_foreign`), and
> against the injected Overlay in ctest `fl_guard` `[log]`, where every patch reads `UNHOOK_RESTORED`
> because the harness is alone in its process.
>
> **The registry refuses what it cannot hold, and a patch is registered before it can run (0.1.0-beta.42,
> `fl_patch_registry.h`).** The registry was an array and a counter, and two things were wrong with it — found by
> reading the code for the plan of the next two hooks, not by a failure. **Full meant silent:** `RegisterPatch`
> wrote its entry only while it fitted, after `PatchTarget` had already enabled the hook, so a 49th patch would
> have been live and unknown to the stop, with no line anywhere. (Twenty patch sites exist; the rule no longer
> depends on that count being right, and a `static_assert` now checks the capacity against the inventory's own row
> count plus the seven sites that are not rows.) **Enabled before registered:** `PatchTarget` enabled the hook,
> re-read `g_observing`, and registered last; a stop between the re-read and the registration walked a table
> that did not hold the patch yet, and the patch stayed in after "stopped". The DXGI three and the loader detour
> registered after enabling too, with no re-read at all.
>
> One sequence now, `fl::patch::Install`: **take a slot — no slot, no patch, nothing created — create, enable,
> commit the slot, then re-read the flag.** The stop clears the flag and then walks the committed slots. Each side
> writes and then reads what the other wrote, sequentially consistent, so at least one sees the other: the walk
> restores the patch, or the installer finds the flag clear and takes its own patch out (and retires the slot,
> so the walk does not report it as somebody else's). A slot whose patch failed is free again; one that was ever
> live is never reused — the walk reads it without a lock. Nothing is created once observing is off. A patch
> that did not go in is logged once per target and reason, `PATCH_FAILED a=<reason>` (`2` stopped, `3` registry
> full, `4` create failed, `5` enable failed, `6` the stop arrived meanwhile); until now every such failure was
> silent. The three DXGI patches are still enabled together and keep the same rules by hand; when any step of
> theirs fails, whatever a partial enable left in is taken out — it used to stay, unregistered, with init
> reporting failure. And two stores that could overwrite a stop's reason no longer do: `READY` is written only
> over `INIT`, and init's failure path stores nothing. No hook body changed. ctest `fl_unhook_inline` drives
> the shipped sequence against real MinHook patches: a full registry leaves the target's bytes untouched and
> MinHook unasked; a stop arranged between the enable and the re-read ends with the function unpatched; a failed
> patch gives its slot back; a slot is invisible to the walk until committed.
>
> **Batches, 0.1.0-beta.46 (owner decision D134): one stop of the game's threads for a group of patches.** MinHook
> writes a jump with every other thread of the process stopped — a snapshot of the system's threads, then a
> suspend, a context read and a resume for each of the process's — and `MH_EnableHook` / `MH_DisableHook` for one
> target do all of that for one jump. The owner's logs of 2026-11-13 put a number on it: the thirteen Streamline
> patches took 0.74 s to 2.8 s of a game's start, 57 to 215 ms each, and a stop restored its patches the same way.
> MinHook 1.3.4 can queue — `MH_QueueEnableHook` / `MH_QueueDisableHook` mark a hook, `MH_ApplyQueued` executes
> every mark under one stop — and `fl_patch_registry.h` now has the two sequences that use it:
> `fl::patch::Batch` (`Add`: slot and create, nothing marked; `Commit`: enable them all) and
> `fl::patch::RestoreAll`, which is the stop's walk. `InstallUpscalerHooks` queues every row a tick finds
> something for and commits once, before the families that are published "when whole"; `InstallRtHooks` batches
> its two; `StopObserving` calls `RestoreAll`. A patch's `HOOK_INSTALLED` line, family bit and latch still come in
> that order and only once it is in. The loader detour, OpenGL and the three DXGI patches of init are as they were.
>
> **What the queue makes dangerous, and what the code therefore never does** (the plan was reviewed against
> `hook.c` before a line was written, and its first form was refused). `MH_ApplyQueued` applies **every** hook
> whose mark differs from its state — the process has one table, not one per caller. So (1) **no mark is ever left
> standing**: a hook that failed to enable has its mark removed, or the next apply — the stop's own — would enable
> it; (2) **only a patch that is still ours is ever marked for a restore** (`StillOurs`, as before; §H7), and
> nothing is marked "all" — the refused draft marked everything and then un-marked what it must not restore, and
> un-marking a hook that is already out marks it for *enabling*: the stop would have put a patch back in after the
> stop; (3) **mark, apply and commit happen inside a gate** (`MinHookEngine::BeginQueue`, a slim lock taken
> exclusively) so the watchdog's apply cannot execute the stop's marks nor the stop's the watchdog's. The gate is
> held for one apply and never across a symbol lookup, which takes the loader's lock. It is the one lock in the
> Overlay, and it is on no hook body's path: the only way from a hook to it is the stop itself, which already
> waited for MinHook's own lock and a stop of the threads for every patch. Its owner re-enters it without
> waiting. **The verdict on an install is the engine's, not the bytes'**: a foreign overlay patching over us a
> moment after the apply would read as "not in", and treating that as a failure would mark a restore nobody had
> compared. `MH_ApplyQueued` stops at the first hook that fails and says only that one did; the ones after it
> are then enabled — or, in the stop, restored — one at a time, so one bad patch still costs one patch.
> The write-then-read pair is unchanged: the stop clears the flag and then walks; `Commit` commits its slots and
> then re-reads the flag, and takes out — compared first — what it put in when the flag is clear. A stop that
> lands before the commit enables nothing.
>
> **Measured** on `hook-harness`'s Streamline target, the development PC, three runs before and after (`fl_guard`
> `[advanced]` prints them): the nine Advanced-only patches went in one by one over 0.60 s / 0.70 s / 0.60 s
> from the first log line to the ninth, 76–88 ms each; as one batch they take **81 / 90 / 78 ms**. The stop restored
> eighteen patches in 1.37 s / 1.56 s / 1.37 s; under one apply, **77 / 90 / 76 ms** — and from the Agent's request
> to `UNHOOKED`, 1.27–1.48 s became 94–110 ms. One stop of the threads costs what it cost for one patch, whether it
> covers four, nine or eighteen. A new log line says so in the field: `PATCH_BATCH install` / `restore`, `a` the
> microseconds, `b` the patches. ~~**Not measured:** a real game — how long one stop of the threads is with a few
> hundred of them, and whether any title minds patches arriving together (owner-only item 44).~~ **In real games**
> (the owner's five sessions of 2026-11-13 on 0.1.0-beta.46 — Hogwarts Legacy, Phantom Blade Zero, Black Myth:
> Wukong, Cyberpunk 2077, Clair Obscur: Expedition 33; eighteen install batches of two to sixteen patches): every
> batch went in whole, with the hooks beta.45 installed and no `PATCH_FAILED`; **sixteen batches cost 81–262 ms
> whatever they held**, which is what one patch cost in the same game before. Streamline's thirteen, first line to
> last on beta.45 against the one batch now: Phantom Blade Zero 0.74–0.85 s → 88 ms, Cyberpunk 2077 1.18 s → 93 ms
> (with three more patches in the batch), Black Myth: Wukong 1.65 s → 188 ms (twelve), Hogwarts Legacy 2.80 s →
> **1.26 s**. That last one and a batch of three in Cyberpunk 2077 (**0.61 s**) are not explained: the cost does
> not follow the count — sixteen patches in 93 ms, three in 609 ms — and **the log cannot tell one long stop from an
> apply that failed part-way and was finished one patch at a time**, because a batch writes a line only for a patch
> that fails for good. **Since 0.1.0-beta.47 the line says which** (owner decision D137): `install` or `restore` is a batch that went under its one apply; `install singly` or `restore singly` is one whose apply said it had failed, and the high half of `b` then counts the patches handed to MinHook one at a time (`b=0x0003000d`: thirteen in, three of them singly) — attempts, not stops: MinHook makes no stop for a patch the failed apply had already done. So a long `a=` on a plain `install` line is one long stop of the threads. ~~**Still not measured: the restore in a real game**~~ **The restore in a real game** (the owner, 2026-11-14,
> Phantom Blade Zero, twice): twenty-nine patches out under one apply of **83 ms and 86 ms**, every one restored,
> none declined, and the game went on for minutes and closed cleanly (owner-only item 44 (d)). One more thing the
> queue changes for the better: MinHook's single-target disable does not move a thread that is suspended inside
> the patched bytes (`hook.c` passes the enable action to its thread fix-up in both directions); the queued apply
> does. ctest `fl_unhook_inline` drives `Batch` and `RestoreAll` against real MinHook (three in under one apply,
> a duplicate target refused without holding the others up, a foreign patch left byte for byte) and against a
> model of MinHook's table for what real MinHook cannot be made to do on demand: an apply that fails halfway or
> not at all, a stop inside the apply, a committed patch that is already out — the refused draft's failure — and
> that no mark is ever standing or made outside the gate.
~~`FlRequestUnhook()` (or the control flag) ⇒ `MH_DisableHook(MH_ALL_HOOKS)`, restore vtable entries **only where they are still ours**, flush the ring, set status.~~ *(Corrected 2026-10-04: both reach `StopObserving` — compare-and-restore per patch as the note above says, then the status; there is no `MH_ALL_HOOKS`, no vtable entry and no ring flush. The watchdog flushes the native log on its way out.)* **The DLL is not `FreeLibrary`'d from the live process** — a thread could still be inside a trampoline. It goes dormant and unloads with the process. This is deliberate and documented.
@@ -691,6 +928,12 @@ before it. And when the slot has changed, going dormant costs nothing: we stay
loaded anyway, and our detour remains in someone else's chain doing nothing
harmful.
> **Said plainly, 2026-11-13: for the inline patches that re-check has never been made.** MinHook owns the write
> window — it stops the threads, changes the protection and writes — and `StillOurs` is asked before it is called,
> not inside. Between the two a foreign overlay could patch over us, on a thread MinHook has not stopped yet. The
> batches of 0.1.0-beta.46 neither close that window nor widen it: the comparison and the write are apart by one
> stop of the threads, as they were for each patch, once instead of once a patch.
Verified by `hook-harness --probe-unhook` (ctest `fl_unhook_preserves_foreign`),
which simulates the foreign hooker rather than depending on RTSS being
installed, so it is deterministic and runs on CI. Both halves are asserted: we
@@ -702,13 +945,41 @@ compare-and-restore that never restores is not a fix.
No logging in hook bodies. A small fixed-size in-memory ring of structured events (hook installed, symbol missing, fault, unhook) is flushed to `logs\overlay-<pid>-*.log` by the init thread at session end, on unhook, or on Agent request — never mid-frame.
> **Built 2026-09-06 (P1 item 4), as specified, with the details a bug report needs.** A ring of 256 ×
> 64-byte events (`qpc`, code, two operands, a 40-char name) appended with one `fetch_add` from init and
> ~~64-byte events (`qpc`, code, two operands, a 40-char name)~~ 96-byte events (`qpc`, code, two operands, a
> 72-char name — 64 bytes and 40 characters until beta.42) appended with one `fetch_add` from init and
> watchdog paths and from a hook's **SEH handler** — `NoteFault` records the exception code and the hook's
> `__func__` — never from a hook body. Codes: `RING_CREATED`, `PRESENT_HOOKS`, `HOOK_INSTALLED` (one per
> patch, with the target address), `SYMBOL_MISSING` (once per pair, and only when the module IS mapped),
> `FAULT`, `STOP` (the reason), `UNHOOK_RESTORED` / `UNHOOK_DECLINED` (one per patch), `LOADER_DETOUR`,
> `SUPERVISION_LOST`, and since beta.24 `STOP_FLUSHED` (the last line after a stop: `a=1` when the stop had finished
> writing). **Three flush points, none on a present path:** the end of `InitThread` (so the
> `SUPERVISION_LOST`, since beta.24 `STOP_FLUSHED` (the last line after a stop: `a=1` when the stop had finished
> writing), and since beta.42 `PATCH_FAILED` (a patch that did not go in, once per target and reason — §Unhooking)
> and `FFX_DISPATCH`, `FFX_QUERY`, `UPSCALE_ROUTE`, `RUNTIME_CENSUS` (the note at the end of this section), and
> since beta.43 `FFX_CONFIGURE` — each configure type a hooked ffx module was handed, in a process where
> `ffxConfigure` is hooked at all (one asked for Advanced metrics) — and since beta.46 `PATCH_BATCH`: one line an
> apply, `install` or `restore`, `a` the microseconds it took and `b` the patches it covered (§Unhooking, Batches).
> **`LATENCY` (train C, 0.1.0-beta.47)**: one line each time the state of the driver read changes — `a` the
> `FlLatencyState`, the text what it means — and, when the module is found, two more with the end of the path
> `nvapi64.dll` was loaded from and of the image its getters are in. Written by the watchdog; a fault in a getter
> is a `FAULT` line from the presenting thread, as any fault is.
> **`LATENCY_FRAMES` (0.1.0-beta.49, owner decisions D148, D151)**: the frames the driver reads summed, counted
> for this log and nowhere else — nothing of it is published, stored or shown. Each summed frame is placed by its
> own two numbers, the span between its two GPU marks and the driver's active render time: no start mark, a span
> of no length, an active time with no span, a span with no active time, and — where a frame has both — *active
> longer than the span* or *active within the span*, each with its count and both sums in microseconds. No
> threshold: a span of one microsecond is a span. **A batch is seven lines and belongs to one state of the driver's
> sleep status** — its low-latency switch, its "the game sleeps" switch and its frame-generation multiplier —
> because the counts mean nothing apart from it: a read in another state closes the batch first (thirty-two such
> closings a request, after which a batch runs to its length and its first line says what it held). A batch is
> also written after a request's tenth read, its thirtieth and every sixtieth from there, when the Agent asks for
> a flush, at a new request and at a stop. `a` is printed in decimal and `b` in hex: *reads* and the switches
> seen on, off (`<<8`) and the multipliers seen (`<<16`); *frames summed* and refused; *GPU span absent* and of no
> length; *active time with no span* and span with no active time; then the two heaps, `a` their frames and `b`
> their span sum (`<<32`) and active sum; and the watchdog's own line — the *milliseconds the batch spans* and its
> ticks in them, on time (`<<32`) and woken early: a read a second is a thousand milliseconds a read and as many
> periodic ticks as reads (D150). **A game that exits by itself takes the batch in hand with it** — under a
> minute of reads. About seven lines a minute while a session with *Advanced metrics* reads the driver; none
> otherwise.
> **Three flush points — four since beta.42, the same note —, none on a present path:** the end of `InitThread` (so the
> file exists from the start of the session and a crash mid-session still leaves it), the watchdog when
> `FlControlBlock.logFlushRequested` changes (a counter the Agent bumps at session end; nothing clears it),
> and the watchdog on its way out after any stop. Appended, never recreated; events overwritten before a
@@ -719,15 +990,59 @@ No logging in hook bodies. A small fixed-size in-memory ring of structured event
> it reads 0. When the stop ran on a presenting thread — `MayObserve`'s one-frame safety stop — the watchdog's one
> final flush could run while that thread was still writing, and the lines after it (the kernelbase loader detour is
> the last patch) reached the ring after the last flush ever made. The stop now says when its last line is in, and
> wakes the watchdog; the final flush waits for that, bounded at 2 s, then writes `STOP_FLUSHED`. And each slot has a
> wakes the watchdog; the final flush waits for that, bounded at 2 s *(a quarter of a second for every patch the registry can hold since beta.45; the stop is one apply since beta.46 and the bound is for its fallback)*, then writes `STOP_FLUSHED`. And each slot has a
> commit word written after its fields: `LogNote` claims a slot with one `fetch_add` and fills it afterwards, so a
> flush on another thread could print a slot still being filled. A flush now prints only slots whose writer has
> finished — an ordinary flush stops at the first that is not, for the next one; the final flush names it in a `#`
> line. Nothing on a present path changed. Path: `%LOCALAPPDATA%\FrameLedger\logs\overlay-<pid>-<yyyymmdd-hhmmss>.log`,
> resolved by `SHGetKnownFolderPath` (§S21's rule, not the environment). No allocation: static ring,
> static 64 KiB format buffer, static path. The CaptureHost prints the file and its `FAULT` / `STOP` /
> `UNHOOK_*` / `SYMBOL_MISSING` / `SUPERVISION_LOST` lines at the end of every session
> `UNHOOK_*` / `SYMBOL_MISSING` / `PATCH_FAILED` / `SUPERVISION_LOST` lines at the end of every session
> (`Consume/OverlayLog.cs`).
>
> **The log outlives the game, and says what each ffx module was handed (0.1.0-beta.42, owner decision D121).**
> Cyberpunk 2077, set to FSR upscaling with FSR frame generation, was recorded with frame generation measured and
> the upscaler "unknown": PREPARE and FRAMEGENERATION dispatches arrived and no UPSCALE ever did. The log could
> not say why. **It ended half a second into the session** — written at the end of init, then only when the Agent
> asked or a stop ran, and a game that exits does neither: the watchdog dies with the process. And **it could not
> have said much**: the two one-target rows logged their symbol and every other row the word `hook`, so four ffx
> modules were one indistinguishable line each. Four changes, none of them a new patch site:
>
> 1. **A fourth flush point: the watchdog, on any tick that finds something new in the ring** (`FlushLogIfNew`:
> one load on an uneventful tick, no file opened). Events are few and bounded — installs, first sights, three
> faults at most — so this is a handful of appends a session. The init thread's flush moved to **before** the
> watchdog is created: `FlushLog`'s state belongs to one thread at a time, and with the watchdog flushing on its
> first tick the old order could run two at once.
> 2. **Every install line names its module and its symbol** — `amd_fidelityfx_dx12.dll!ffxDispatch` — built at
> compile time from the inventory row's own two spellings (`fl::inventory::NameOfRow`), so no second list of
> names exists. The same name is on the `UNHOOK_*` and `PATCH_FAILED` lines. The event's text grew from 40 to
> 72 characters for it (an event is 96 bytes; the ring 24 KiB, static), and `PatchTarget` lost its default
> name, so a new call site cannot log `hook`.
> 3. **`FFX_DISPATCH` / `FFX_QUERY`: each descriptor type a hooked ffx-api module was handed, once**
> (`fl_ffx_types.h`). Text = the module, `b` = the type (`0x10001` UPSCALE, `0x20004` / `0x2000c` PREPARE,
> `0x20003` FRAMEGENERATION, …); `a` ≠ 0 with `b=0` says what did not fit (`1` a ninth distinct type, `2` a
> type of zero). The detour keeps the value of `type` — the member it reads already to choose an arm — in the
> first free of eight slots per module and entry point; the watchdog writes it. A query's type is kept whatever
> the vendor answered. `UPSCALE_ROUTE` names each source an UPSCALE came from, once (the FSR 3.0 host has no
> descriptor type, so this is the only line that says its export was called), and `RUNTIME_CENSUS` repeats the
> census word (`b`) whenever it gains a bit — both from words the detours kept before, at no cost to a hook.
> 4. **`PATCH_FAILED`** (§Unhooking) is the fourth: a patch that did not go in used to leave no line at all.
>
> **This is a diagnostic, not a measurement, and it is not behind the Advanced-metrics switch** — like
> `HOOK_INSTALLED`, it reaches the log file and nothing else: no record, no region of the shared memory, no session.
> Rule 4: the only thing read is `type`, of a descriptor handed to a hooked call. Rule 5: no allocation, no lock,
> no logging in the body; a type already kept costs loads and no write.
>
> **Measured overhead** (on `fl_ffx_types.h` itself with the Overlay's compile flags — `/O2 /GS /guard:cf
> /Qspectre`, 50 million calls a run, three runs, i7-14700KF): a frame's three dispatch types in turn, **+3.8 ns a
> dispatch** (3.79 / 3.87 / 3.81); a type in the first slot +0.9 to +1.0 ns; in the eighth, the worst steady
> state, +8.1 to +8.4 ns. A new type costs one compare-exchange, once per type and module.
>
> ctest `fl_ffx_types` (the header, including eight threads meeting the same six types at once: each kept exactly
> once) and `fl_guard` `[log]`: a harness on the 1.1.x monolith, generating frames and asking a ratio query, is
> supervised and told **nothing else** — no flush request, no stop — and its log holds the two install lines by
> module and symbol, the three dispatch types and the query type once each, the route and the census; the process
> is then killed and the file is as it was, one header, no `STOP`.
> **Tests and the real logs folder (2026-09-15).** The injection tests use the shipped Overlay, so their harness
> logs land in the user's real `logs\` too: measured that day, 2506 overlay logs, **2492 of them hook-harness's**,
@@ -753,7 +1068,7 @@ verification a dev-machine-only affair before:
- **WARP** (`D3D_DRIVER_TYPE_WARP`) — a software rasteriser, so no adapter is required.
- **`CreateSwapChainForComposition`** — no `HWND`, so no dependency on a window station or an interactive session.
Current modes: `--probe-vtable` (§H4, ctest `fl_vtable_indices`), `--probe-proxy` (§H5, ctest `fl_proxy_swapchain`), `--probe-unhook` (§H7, ctest `fl_unhook_preserves_foreign`), `--probe-cost` (NFR-1, measurement only), **`--probe-frames`** (ctest `fl_frame_identity` — what counts as a frame, against `GetLastPresentCount`), **`--probe-d3d12`** (ctest `fl_d3d12_acquisition` — device → command queue → swapchain), `--present N`, `--hold N`, **`--real`** and **`--plus-ui K``**, **`--vulkan --hold-presenting N`** (P1 item 3, real loader, exit 77 = cannot run here) and **`--opengl --hold-presenting N`** (P1 item 4, gdi32 `SwapBuffers` on a hidden window, Microsoft's generic renderer — runs on CI). *(Added 2026-10-04, each with its own ctest: `--probe-d3d12-vtable`, `--probe-dxr`, `--probe-dxr-inputs`, `--probe-upscaler-resolve`, `--probe-sl-abi`, `--probe-ffx-resolve`, `--probe-sl-inputs`, `--probe-sl-seen`, `--probe-dxgi-count`; and the `--hold-presenting-*` injection fixtures with `--load` / `--load-after-ms`.)*
Current modes: `--probe-vtable` (§H4, ctest `fl_vtable_indices`), `--probe-proxy` (§H5, ctest `fl_proxy_swapchain`), `--probe-unhook` (§H7, ctest `fl_unhook_preserves_foreign`), `--probe-cost` (NFR-1, measurement only), **`--probe-frames`** (ctest `fl_frame_identity` — what counts as a frame, against `GetLastPresentCount`), **`--probe-d3d12`** (ctest `fl_d3d12_acquisition` — device → command queue → swapchain), `--present N`, `--hold N`, **`--real`** and **`--plus-ui K``**, **`--vulkan --hold-presenting N`** (P1 item 3, real loader, exit 77 = cannot run here) and **`--opengl --hold-presenting N`** (P1 item 4, gdi32 `SwapBuffers` on a hidden window, Microsoft's generic renderer — runs on CI). *(Added 2026-10-04, each with its own ctest: `--probe-d3d12-vtable`, `--probe-dxr`, `--probe-dxr-inputs`, `--probe-upscaler-resolve`, `--probe-sl-abi`, `--probe-ffx-resolve`, `--probe-sl-inputs`, `--probe-sl-seen`, `--probe-dxgi-count`; and the `--hold-presenting-*` injection fixtures with `--load` / `--load-after-ms`.)* *(Added 2026-11-12, Advanced metrics train B: `--probe-adv-fields`, ctest `fl_adv_fields` — `fl_adv_fields.h` alone: a struct of a lower version is not read past its end, the vendors' sentinels never leave as values, the summaries, the publication, a new request starting afresh; and `--advanced` / `--ffx-memory-query v1|v2`, which make the `--hold-presenting-upscaled` and `--hold-presenting-ffx` fixtures send what Advanced metrics reads.)*
> **Every present here used to carry `DXGI_PRESENT_TEST`, which submits nothing.** Measured: 500 of them leave `GetLastPresentCount` at 0 while 37 real presents move it by 37. "N presents → N records" was therefore satisfiable only by a writer counting non-frames. `--real` issues real presents; the test-present path is kept as a named mode because it is what an alt-tabbed title actually runs.
+646 -2
View File
@@ -145,7 +145,7 @@ L1 gives no temperatures on its own (unless the D3DKMT probe pans out). That is
`LibreHardwareMonitorLib` ≥ 0.9.6, consumed as an **unmodified NuGet package**. It already implements per-vendor GPU sensor access internally — which means the vendor-interop licensing problem is one LHM has solved upstream, under a licence that works for us. That is the entire reason this layer exists.
- `Computer` opened with `IsGpuEnabled` always; ~~`IsCpuEnabled` + `IsMemoryEnabled` only when the Agent is elevated and PawnIO is available~~ nothing else, ever — the CPU temperature uses a separate `Computer` (below) (corrected 2026-10-04).
- `Computer` opened with `IsGpuEnabled` always; ~~`IsCpuEnabled` + `IsMemoryEnabled` only when the Agent is elevated and PawnIO is available~~ nothing else, ever — the CPU temperature uses a separate `Computer` (below) (corrected 2026-10-04). *(2026-11-11: still true of THIS layer's `Computer`. Advanced metrics opens a third one, of its own — the graphics card, and as administrator with PawnIO the processor, memory and motherboard groups; D103, §Advanced metrics. Storage, the power supply, controllers, the battery and the network are opened by nothing.)*
- Poll on a dedicated thread: `computer.Accept(updateVisitor)` then read mapped sensors. Never faster than 500 ms; default 1000 ms.
- Sensor mapping by `SensorType` + name heuristics per vendor (`GPU Core`, `GPU Hot Spot`, `GPU Memory`, `GPU Package Power`, …), kept in `SensorMap.cs` with unit tests against captured sensor-tree fixtures.
- **P0 verification items:** (a) which fields LHM actually returns per vendor on real hardware — fill the capability matrix below — **NVIDIA filled 2026-09-03, AMD/Intel deferred (§R5/§R6)**; (b) ~~whether GPU-only usage works **without** elevation and without PawnIO (expected yes, since GPU sensors go through user-mode vendor paths, but confirm — it determines whether the default unelevated Agent has temperatures at all)~~ — **confirmed 2026-09-03, §M5 row R1**: eight fields unelevated, the same eight elevated, PawnIO never opened; (c) ~~confirm LHM's sources are not marked with MPL-2.0 Exhibit B~~ — **done, 2026-08-02: clear.** No source file applies the notice; the only repository hit is the `LICENSE` template itself, and every file we depend on carries the permissive Exhibit A. Method and evidence in `docs/spike-notes.md` §0. Re-check on every version bump, and check *source headers*, never the LICENSE file — MPL-2.0's own text contains Exhibit B as a template, so grepping the licence finds it in every MPL project ever published.
@@ -204,6 +204,8 @@ Still resolve **entry points defensively at runtime**: `nvapi64.dll` may be abse
> **Not in this build, stated so nobody infers it from the table below:** VRAM (`FlNvSample.vram` is
> reserved and its bit is never set — see the Memory row), power, hotspot temperature, per-domain
> utilisation beyond the GPU domain, and anything Reflex (§M8: a hook fact, not a poller field).
> *(2026-11-11: still true of `FlNvSample`, the 1 Hz telemetry layer's struct. The memory figures and the other
> three utilisation domains are read since ABI 3 — below — into structs of their own, for Advanced metrics only.)*
>
> **ABI 2 (2026-09-25, beta.8): the driver profile, read-only.** The owner asked whether the NVIDIA App overrides a
> game's DLSS (frame generation, ray reconstruction, the model preset). The NVIDIA App writes those overrides into the
@@ -222,6 +224,39 @@ Still resolve **entry points defensively at runtime**: `nvapi64.dll` may be abse
> ctest `fl_nvapi_bridge` (argument refusals, DEGRADED before init, the global profile of the test's own executable on
> a GPU machine) and `NvapiBridgeMirrorTests` (`FlNvDriverProfileSize`, ABI 2).
>
> **ABI 3 (2026-11-11, owner decision D101): what "Advanced metrics" reads.** Five reads, each a struct with its
> size first and a `present` mask, and each call behind it a query: `FlNvReadFacts` (what is true of the card whatever
> is running — its type, bus, PCI ids, VBIOS version, architecture, core counts, the base and boost clocks, the
> sensor's range, `NvAPI_GPU_GetMemoryInfoEx`'s four sizes, the frame-buffer sizes, the driver's package and base
> branch, NVAPI's interface version, the chipset), `FlNvReadLive` (the same memory call's free figure with evictions
> and promotions, the frame-buffer, video and bus utilisation, the video clock, the P-state, the board and
> power-supply temperatures where the card has them, the throttle word, the encoder's session count and averages),
> `FlNvReadDisplays` (per connected display: HDR formats and luminances, variable-refresh support, the adaptive-sync
> flags), `FlNvSessionDisplay` (the display a GDI device name names: its variable-refresh state and the colour space
> the driver reports for one process) and `FlNvProfileSettings` (every setting of the global profile, or only what a
> game's own profile sets — a game no profile names is answered `GLOBAL` with nothing listed, never with the PC's
> settings). No struct of ABI 2 changed size; `FL_NV_FIELD_VRAM` stays clear.
>
> **No identifier has an entry point** (D104): not `NvAPI_GPU_GetBoardInfo`, `NvAPI_GPU_GetUUID`, an EDID read, or
> `NvAPI_GPU_GetEncoderSessionsInfo`, which lists WHICH programs encode. A display is passed on by its place in the
> driver's list; NVAPI's display id is a handle of this boot and is neither stored nor shown.
>
> **Read-only is now a gate** (D108): `tools/nvapi-readonly-check.ps1` holds every `NvAPI_*` call of the bridge's
> source to `tools/nvapi-allowed-calls.txt` in both directions — forty-four names — and refuses a name that sets,
> saves, restores, registers or controls something, or that returns an identifier, **even when the list names it**.
> Managed: `INvapiAdvancedBridge` (a port of its own beside `INvapiBridge`; `NativeNvapiBridge` implements both, and a
> DLL that predates ABI 3 answers `Unavailable` rather than throwing) and `NvapiAdvancedReader`, which turns the five
> structs into the catalog's facts and readings — every `nvapi.` entry whose reach is `Agent`, and no other, held by
> `NvapiAdvancedReaderTests` in both directions. **Nothing composes the reader yet**: no Agent code calls it.
>
> **Measured on the dev box the day it was written** (RTX 5080, a driver of base branch `R620`, ctest `fl_nvapi_bridge`'s
> AVAILABLE branch — `20_OPEN_QUESTIONS` §M12): all twenty facts answer (`present = 0xFFFFF`: discrete, desktop, PCI
> Express, 84 ray-tracing and 336 tensor cores, base 2295 / boost 2670 MHz, 16,303 MiB dedicated, a DCH Game Ready
> driver); eight of the ten live groups answer (`0x33F`) — **the board and power-supply temperatures do not exist on
> this card**, as the memory temperature does not, and read N/A; the one display answers all four groups; the global
> profile lists 24 settings. One thing the header had wrong: it misspells `NV_ENCODER_STATISTICS_VER` as
> `NNV_ENCODER_STATISTICS_VER`, so the bridge names `NV_ENCODER_STATISTICS_VER1`.
>
> **The ctest could not fail on a GPU machine until the same day.** `fl_nvapi_bridge` was registered with
> `PASS_REGULAR_EXPRESSION` alone, which makes ctest ignore the exit code: the AVAILABLE branch's reference-counting
> check (two references, two releases, then a read expected to work) failed on every machine with an NVIDIA GPU while
@@ -234,7 +269,7 @@ Still resolve **entry points defensively at runtime**: `nvapi64.dll` may be abse
| Temperatures | `NvAPI_GPU_GetThermalSettings` (+ extended thermal query for hotspot/memory where available). **Bridge (2026-09-10):** the public enumeration only — core is the `GPU` target; memory is the `MEMORY` target when the driver lists one, and on the RTX 5080 it does not (bit clear → `null`); hotspot is not read |
| Utilisation | `NvAPI_GPU_GetDynamicPstatesInfoEx` (GPU / FB / VID / BUS domains). **Bridge (2026-09-10):** the GPU domain only, as `loadPct` |
| Clocks | `NvAPI_GPU_GetAllClockFrequencies` — **bridge (2026-09-10):** current graphics and memory clocks |
| Memory | `NvAPI_GPU_GetMemoryInfoEx` — **not `NvAPI_GPU_GetMemoryInfo`**, which this table named until 2026-08-05. The vendored headers mark it `__nvapi_deprecated_function` ("deprecated in release 520"), so under `/W4 /WX` a call to it fails the native build. **Neither is called (2026-09-10):** the vendored `nvapi.h` (`cd6918f6`) does not declare `GetMemoryInfoEx` at all — the deprecation message names a function the same header set does not ship — so the bridge's `vram` field is **reserved**, its bit is never set, and VRAM stays L1's PDH adapter counter. Taking a newer header set is the way to change that, not a call to the deprecated one |
| Memory | `NvAPI_GPU_GetMemoryInfoEx` — **not `NvAPI_GPU_GetMemoryInfo`**, which this table named until 2026-08-05. The vendored headers mark it `__nvapi_deprecated_function` ("deprecated in release 520"), so under `/W4 /WX` a call to it fails the native build. **Neither is called (2026-09-10):** ~~the vendored `nvapi.h` (`cd6918f6`) does not declare `GetMemoryInfoEx` at all — the deprecation message names a function the same header set does not ship~~ — so the bridge's `vram` field is **reserved**, its bit is never set, and VRAM stays L1's PDH adapter counter. ~~Taking a newer header set is the way to change that, not a call to the deprecated one~~ **Corrected 2026-11-11: the vendored set does declare it** — `nvapi_lite_common.h:600`, which `nvapi.h` includes, with `NV_GPU_MEMORY_INFO_EX_V1` above it and its id in `nvapi_interface.h`; the search that found nothing looked in `nvapi.h` alone. It was not called because of that sentence, from 2026-09-10 until now. The bridge calls it from ABI 3 for the advanced memory rows (§Advanced metrics); `FlNvSample.vram` stays reserved, so which layer fills the base `vram_adapter` series does not change |
| **Throttle reasons** | `NvAPI_GPU_GetPerfDecreaseInfo` — not available from L1 or L2. **Bridge (2026-09-10):** the raw reason mask, and 0 is a measurement (no reason), not a clear bit |
| Driver version | `NvAPI_SYS_GetDriverAndBranchVersion` — **bridge (2026-09-10):** `FlNvDriverVersion`, surfaced as `616.64 (r616_41)` |
| **NGX per-process state** | `NvAPI_NGX_GetNGXOverrideState` (R570+) — the driver's feedback word per DLSS feature for a PID, read OUT OF PROCESS by `fl-probe-nvapi --ngx-state` and, since 2026-09-06, by the capture host beside each module snapshot — **in-process through the bridge since 2026-09-10** (`NvapiNgxStateProbe`; the spawn is retired and the words, the merge and the printed block are unchanged). Measured: the SR word's `CREATED \| EVALUATE` is set without an NVIDIA-app override; the FG word follows DLSS-G once created (`CREATED \| EVALUATE` at ×4, corrected the same day — the first reading was blind because it was early), the multiplier bits stay clear; `scalingRatio` is 0 even under an override. Identity only, SR and FG (`03_METRICS` §Upscaling) — and, since beta.24 (D63), the mode an NVIDIA App override set where the word says it applied; its `performanceMode` is in NGX's own numbering (Silent Hill 2, 2026-10-13: the App's Ultra Performance, profile value 5, reported as 3). *(Beta.27: never read after the game has gone — that reading, the last, had cleared `CREATED \| EVALUATE` on every stored session — and the two bits stand once a reading carried them: `03_METRICS` §The driver-reported rung.)* |
@@ -493,3 +528,612 @@ alone could never catch.
> (`NvapiTelemetrySource` over `FrameLedger.NvapiBridge.dll`, §L3). §M5 in particular (do LHM GPU
> sensors work unelevated, without PawnIO?) decided whether the default unelevated
> Agent has temperatures at all — **it does, on NVIDIA; row R1 above.**
## Advanced metrics (owner decisions D101–D108, 2026-11-11)
The owner asked for a setting, *"Thông số nâng cao"* (Advanced metrics), with a navigation item of its own, that shows
as much as the Streamline and NVAPI headers and PawnIO make available — numbers, tables, charts — **with each value
assigned to what it belongs to: one game session, or the PC**. Asked while planning, the owner chose: everything
that can be reached inside a game as well, in stages, the route refused on 2026-08-09 re-examined; stored in full with
the session; off by default, and off means nothing more is read; and the page never asks for administrator rights
itself. This section is the design those answers gave, the decisions taken with it, and the inventory.
**Nothing here is built by the pull request that wrote it** except the catalog
(`FrameLedger.Application.Advanced.AdvancedMetricCatalog`) and this section. Each later pull request of the train
states here what it built.
### The decisions
- **D101 — the option.** `telemetry.advanced`, off by default, read by the Agent at each session's start. **Off means
not read**: no additional call to NVAPI or LibreHardwareMonitor, nothing more stored, and the navigation item is
hidden. A session recorded while it was off shows N/A for everything here — it is not filled in afterwards.
- **D102 — where it is stored.** In two tables of their own (schema 0023), not in new columns of `sessions`: every
read of a `sessions` row, the JSON export and the bug bundle carry the whole row, and a driver profile does not
belong in a bug bundle. `machine_profiles` holds the facts of the PC, one row per distinct set, and a session points
to the row that was true when it ran — the trend's hardware-change markers and `hardware_snapshots` are untouched.
`session_advanced` holds one session's facts, its series' statistics and their directory. The series themselves go
to `sensor_blobs` under names that start with `adv.`, **with NaN for a tick that has no reading**: the fixed series
store −1 there and their readers drop every negative value, which would silently eat a −12 V rail.
- **D103 — LibreHardwareMonitor's other groups.** CPU, memory and motherboard (its sensor chip and embedded
controller) are opened **only when the Agent already runs as administrator and PawnIO is installed** — the existing
admin mode (D34); the page says what is missing and links to that setting. **Refused**: the power-supply and
fan-controller groups, which make the library search every USB HID device. **Deferred**: the battery group — and
**storage**. The plan had storage in, temperature only and internal drives only, with one outcome written down
before any run: *if the group cannot be opened without touching removable drives, it stays closed*. It cannot: the
library's `StorageGroup` (0.9.6, read 2026-11-11) lists every drive through DiskInfoToolkit's `ReloadStorages` with
no filter by bus, and subscribes to drives being plugged in afterwards. So `lhm.storage.*` is `Deferred`, a drive's
temperature reads N/A, and `LhmGroups` — the only way this code can name a group to open — has no member for it
(`20_OPEN_QUESTIONS` §M13). This amends §L2's "nothing else, ever" from the pull request that opened the three.
- **D104 — identifiers are never read.** The card's board serial number and unique id, a display's EDID, and the list
of programs that are encoding. They are named in the inventory so the page can say so; a test holds that no stored
JSON has their keys. This is the only part of "everything" left out on purpose.
- **D105 — the navigation item** is the last of the pane's items; Ctrl+1 to Ctrl+5 keep their pages and **Ctrl+6**
opens it only while the option is on.
- **D106 — the `.partial` file skips what it does not know.** A chunk of an unknown type whose checksum is right is
passed over; until now it ended the read, so a build one release older lost everything after the first new chunk.
- **D107 — with no game running**, the Agent reads sensors only while the page is open, through one reader shared
with the sessions, closed again some seconds after the last use. A PC idling with FrameLedger in the tray reads
nothing (NFR-2's "~0% idle").
- **D108 — read-only, by a gate.** §Runtime policy's "never call a setter" becomes a build check: every `NvAPI_*`
call in the bridge's source is compared with a committed list, in both directions, and a name that changes
something or returns an identifier is refused whatever the list says. An entry point that both reads and writes by
a command argument — the colour and info-frame controls — is refused like a writer: the check reads names, not
arguments, and nothing on the list is one. *(Built with ABI 3: `tools/nvapi-readonly-check.ps1`.)*
### The reader (built 2026-11-11; composed by the Agent since the switch)
`Application.Advanced.IAdvancedHardware` is the port: `Acquire()` keeps the reader running until the handle is let
go, `Latest` is the last tick without waiting, `ReadSessionFacts` reads one session's facts on the caller's thread.
`Infrastructure.Telemetry.AdvancedHardwareHub` implements it — **one reader for the whole Agent, on a thread of its
own (`fl-advanced`), running only while somebody holds it** and for a linger after, so that two sessions in a row,
or a session ending as the page opens, do not close and reopen the sensors (D107). The Agent records several
sessions at once, each with a poller; a reader per session would open the motherboard's sensor chip as many times.
What runs on that thread is `AdvancedHardwareReader`: NVAPI through `NvapiAdvancedReader`, and
`LhmAdvancedReader` over a `Computer` of its own, opened with `LhmAdvancedReader.GroupsFor(elevated, pawnIo)` —
`LhmGroups`, an enum that has a member for the four groups this project may open and for no other. Each sensor is a
member of a catalog family by the device at the top of its tree and its type (`LhmAdvancedMap.FamilyOf`); its series
is named from the library's own identifier (`/lpc/nct6798d/0/voltage/3` → `adv.lhm.board.voltage.lpc_nct6798d_0_3`)
and labelled with the hardware's words. A sensor with no family is not read.
- **A source that is not there is a state** (`AdvancedSources`): no NVIDIA driver, not administrator, PawnIO
missing. Never an exception, and never a value in its place.
- **A device that throws twice is left alone** until the reader opens again; the others go on. The library's own
traversal stops at the first exception, so each top-level device is refreshed in a guard of its own.
- **NVAPI answering with no field, twice running, after it had answered**: released and initialised again, once
(`20_OPEN_QUESTIONS` §M12).
- **The cadence was set by a measurement, before anything composed the reader.** `FrameLedger.CaptureHost
probe-advanced` runs this reader on the caller's own rights and prints what it read and what it cost. On the dev
box (RTX 5080, unelevated, 2026-11-11): 35 facts and 58 readings a tick — 17 from NVAPI, 41 from the card's sensors;
**one NVAPI live read costs 0.4 ms of processor time; one refresh of LibreHardwareMonitor's graphics-card group
costs 14 ms (62 ms elapsed)**. Once a second that refresh alone is 1.4% of one core, above NFR-2's 1% for the
whole Agent. So NVAPI is read every tick and **the sensor library every fifth** (0.28%), and on the ticks between a
sensor has no reading — its last value is not repeated. The same refresh is what the L2 telemetry layer does once
a second in every session: that cost is not this feature's, and is `20_OPEN_QUESTIONS` §M16 — measured in a real
Agent on 2026-11-11 (1.93 % of one core off, 2.04 % on) and answered by amending NFR-2 to 2.5 % (D117).
### From the reader to the ledger (built 2026-11-11; fed since the switch)
- **The poller attaches, it does not read.** `TelemetryPoller` takes an optional `IAdvancedHardware`, holds it from
`Start` to `Dispose`, and at each poll takes `Latest` — a read of one reference. A snapshot it has not taken yet
rides on that poll's `TelemetrySample.Advanced`; one it has already taken is not attached again, because a
reading is of one tick. Null — the option off — composes nothing, and a sample is what it always was.
- **The `.partial`** carries those readings in a chunk of their own after the sensor chunk (type 9) and the
session's context in another (type 10), so a recovered session keeps both; with the option off neither is
written. An unknown chunk type is now skipped rather than the end of the file (D106). `06_DATA_MODEL` §The
`.partial` file.
- **The finalizer** (`AdvancedSessionBuilder`, both tiers) turns the samples' readings into one series each,
aligned to `t_ms` with NaN for a tick with none, and builds the session's row: facts, sources, statistics,
directory. Schema 0023 and the rule for NaN are `06_DATA_MODEL` §Blob encoding and §Migrations.
### The switch (built 2026-11-11, Advanced PR-5)
`telemetry.advanced` (`SettingsRegistry.TelemetryAdvanced`, Boolean, default `0`, read by the Agent) becomes
`RecorderOptions.AdvancedMetrics` through `SettingsRecorderPolicy`, at each session's start like the other recording
settings; exactly `1` is on. From there:
- **The process has one sensor library** *(2026-11-12)*. One *reader* was never the whole of it: a session's L2
source and its CPU-temperature reader each open a LibreHardwareMonitor `Computer` of their own beside the
Advanced reader's, and the library keeps part of its state — the driver it reads the processor through, its bus
mutexes — per process. `Computer.Close()` lets go of that part whoever else is still reading. The first sessions
recorded as administrator with PawnIO showed it (`20_OPEN_QUESTIONS` §M13): a processor device that faulted out
some seconds after each session's sources closed, and one session whose reader could not open at all. Every
`Computer` is now opened and closed through `Infrastructure.Telemetry.LhmLibrary` (`LhmComputerAdapter` is its
user): opens and closes never run at once, and a `Computer` closed while another is in use is **parked** — left
open, read by nobody, handed to the next user of the same groups — until the last user lets go, when all are
closed. The reads are not serialised: the readers keep their own threads, so a sensor that hangs in one still
holds up neither the other nor the pipe.
- **The Agent has one reader** — `AgentServices` registers `AdvancedHardwareHub` as `IAdvancedHardware`, a tick a
second and a linger of ten — and nothing starts it but a holder. `SessionRecorder` holds it for a session's length
**only when the option is on**, and `AgentRecording` gives the session's `TelemetryPoller` the reader on the same
condition. Off: not held, not asked, no chunk 9 or 10, no row (`SessionRecorderTests.WithAdvancedMetricsOff…`).
- **A session's facts are read once, off the capture loop.** Fifteen seconds after the game's process is known — by
then it has a window — or at the session's end if that comes first, `IAdvancedHardware.ReadSessionFacts` runs on a
pool thread: asking the driver for a profile takes some hundred milliseconds. The finalize waits at most five
seconds for a read still in flight and stores the session without its facts rather than hold it. A read that
throws costs the facts, never the session.
- **Which display is the game's** comes from Windows, not from the game: `WindowGeometryProbe.MonitorDeviceOf` — the
monitor the process's largest visible window is on, as its GDI device name — which the reader hands to
`FlNvSessionDisplay`. No window, or a display another adapter drives: the session's display facts are absent.
- **The context reaches the crash file as it changes** (chunk 10), from `IAdvancedHardware.Latest` — a read of one
reference on the capture loop's task — so a recovered session keeps its machine facts, its sources and its
directory of series.
- **A session recorded with the option on always stores a row**, if only one that says every source is `Unknown`:
no row means "the option was off when this was recorded", which the page (PR-7) shows as such.
What elevation adds changed with this: an Agent running as administrator with PawnIO reads the processor, memory
and motherboard groups in a session with the option on. The texts that state what the admin mode adds say so —
Disclaimer 2.15 §5 and §5A (the owner's answer of 2026-11-11), the admin mode's disclosure (`agent-admin-dialog/2`),
Settings, the first run, the guide and `LIMITATIONS.md` — and Privacy Policy 2.15 names what is read and stored.
### The page's feed (built 2026-11-11, Advanced PR-6; shown by the page since PR-7)
With no game running the Agent reads sensors **only while the page is open** (D107), and nothing tells the Agent
that a page opened or closed:
- **`GetAdvancedMachine`** (`07_IPC` §Messages) is answered by `Application.Ipc.AdvancedMachineResponder` from
`IAdvancedHardware.Latest` alone. The first request acquires the shared reader and each one renews a five-second
lease; a lease after the last request the hold is let go, and the reader closes its sources ten seconds later
unless a session holds it. A page that closes, an App that exits and a pipe that drops all look the same: no
request. With the option off the answer is `Error AdvancedOff`, and nothing is held.
- **`App.Services.AdvancedMachineFeed`** asks once a second while a page holds `Watch()`, and keeps the facts (asked
for once, by their hash), **the last value of every series with the tick it was read on** — a sensor read every
fifth tick is "1400, as of then", never shown as new — five minutes of each series for the live charts, and the
sources' states. It stores nothing. Its states are the page's own words for why there is nothing to show: waiting
for the first read, off, no agent (a viewer never has one), an older agent.
### The page (built 2026-11-11, Advanced PR-7)
`08_UI` §Advanced is the screen. What this document's model becomes there:
- **A machine fact** is under *This PC* ▸ *About this PC* while the Agent answers, and under *Session* as it was
when that session was recorded (`machine_profiles`, one row per distinct set).
- **A session fact** is under *Session* only.
- **A series** is a row of *Right now* and a line of *Over time* while live — never stored then — and a row of the
session's statistics and a line of its chart once a session stored it.
- **An entry that is not read** is on *Sources* with its reason, in the same words as §Left out, by name.
`ISessionRepository.ListAdvancedAsync` lists the sessions that have a row; `SessionSeriesLoader.LoadAdvancedAsync`
decodes their series against `t_ms`, leaving a NaN tick out and keeping a negative reading.
### What belongs to what
| Scope | What it is | Stored | Shown |
|---|---|---|---|
| `MachineFact` | true of the PC whatever is running: the card's firmware and architecture, the driver's package, what a display can do | `machine_profiles`, once per distinct set; a session points to its row | This PC |
| `SessionFact` | true of one session: the game's display state, the settings the driver's profile for that game sets | `session_advanced.facts_json` | Session |
| `Series`, during a session | a sensor of the PC sampled while the game ran | `sensor_blobs` (`adv.*`), statistics in `session_advanced.stats_json` | Session |
| `Series`, with no session | the same sensor, read because the page is open | nowhere | This PC |
A third tab, **Sources**, lists every row of the inventory with its state on this PC — read, or N/A with the reason:
not an NVIDIA card, the Agent is not administrator, PawnIO is not installed, the option was off when the session was
recorded, the hook is not built, the route is refused, a vendor's licence. A PC without an NVIDIA card shows N/A for
every `nvapi.` row; nothing is estimated in its place (FR-4.9).
### The trains
| Train | Release | What | Its gate |
|---|---|---|---|
| A | 0.1.0-beta.39 | read by the Agent only — NVAPI and LibreHardwareMonitor — storage, the page. **Nothing changes in code loaded into a game** | this section |
| B | 0.1.0-beta.40, ~~.41, .42~~ ~~.42, .43~~ .42, .43, .44 *(three releases since D113, 2026-11-11; one was planned. Renumbered 2026-11-12, D118: 0.1.0-beta.41 is a fix release for the sensor library, not part of the train. **Four since D122, the same day**: the second step is beta.42 — no new patch site — and beta.43, the two new hooks; the third step is beta.44)* | inside the game: more fields of the Streamline and FidelityFX calls already hooked, and new hooks on functions `sl.interposer.dll` **exports** (`17_HOOK_ENGINE` §Hook inventory's ⏳ rows); shared-memory layout 5 — §Train B below | a plan of its own; overhead measured; each release after the first waits for the owner's run of the one before |
| B, a fourth step | 0.1.0-beta.45 *(D130, 2026-11-12)* | six more of Streamline's core calls — is a feature supported, is it loaded, the game switching it, its resources allocated and freed, Streamline shut down | hooks installed only where asked for, as in the second and third steps; `slInit` is not among them (D129) |
| C | ~~0.1.0-beta.41~~ ~~0.1.0-beta.43~~ ~~0.1.0-beta.44~~ ~~0.1.0-beta.45~~ after 0.1.0-beta.45 *(D118, D122, D130)* | Reflex latency per frame | a spike — written beside beta.45, a document only — then the owner's go or no |
| D | ~~0.1.0-beta.42~~ ~~0.1.0-beta.44~~ ~~0.1.0-beta.45~~ ~~0.1.0-beta.46~~ after train C *(D118, D122, D130)* | the Streamline feature state behind `slGetFeatureFunction` — DLSS frame generation's, DLSS's, Reflex's | a spike, then the owner's go or no |
Rules 4 and 5, the hook inventory, the measured overhead and the anti-cheat guard bind B, C and D exactly as they bind
every hook. `sl::Constants` is read without its camera position, axes and matrices: where the camera is, is where the
player is (rule 4). NGX stays closed by its licence (`17_HOOK_ENGINE` §NGX).
**Closed 2026-11-14 (owner decision D154): train D is not built, and the refusal stands.** The ten entries it would have read — what a game sets and asks through Streamline's feature functions — have reach `RefusedRoute`: seeing those calls takes handing the game a function of FrameLedger's in place of Streamline's, and a game that fetched its functions before the attach would never call it. What follows is the plan as it stood. **Train D re-examines a refusal; it does not lift it.** `handoff/history/p0-queue-1-6.md` refused, on five grounds,
hooking `slGetFeatureFunction` and patching the function it returns. The route to try is a different one:
`slGetFeatureFunction` is itself an export (`vendor-exports.json`), so its hook is an ordinary inventory row, and
what it hands back could be a wrapper for a fixed list of names instead of a patch at an address learned at run time
— which would answer grounds 1 to 4, and turn ground 5 into "measured only when injected at launch". Two things it
raises are the owner's: replacing the pointer the game receives is a **write**, which rule 4 forbids as written; and
a wrapper must outlive the Overlay's own unhook. If the spike cannot answer all of it, the route stays refused and
the page says so.
### Train B (planned 2026-11-11, owner decisions D110–D116)
What the catalog marks `HookedCall` and `NewHook`. Planned the day train A shipped, the owner's decisions asked while
planning, the design reviewed against the code before it was approved.
- **D111 — off means off inside the game too.** The Agent writes one word of the control block
(`advancedRequested`, ~~0 or 1~~ 0 or that session's request, unconditionally at attach) when the session records Advanced metrics. The Overlay
reads a new field, and installs a new hook, only while it is set; with the option off it hooks and reads exactly
what 0.1.0-beta.39's does.
**Built 2026-11-12 (0.1.0-beta.40, PR-3), with one thing the plan did not have:** "on" is not a 1 but a
*request* — a value no session of that ring used before. A game outlives a session, and the Overlay's summaries
with it; a second session of a game that kept running would have read the first one's field of view and counts.
The watchdog starts its summaries afresh at a request it has not seen and names the request in every publication;
the Agent reads only what answers its own (`07_IPC` §Region 5). The detours consult one process-local flag, the
watchdog is the region's one writer, and `fl_guard` shows the off half on a target that sends everything: every
byte of the region zero until asked, untouched after an "off", a count that starts again at the next request.
**The Agent and the App, 2026-11-12 (PR-4).** A session that records Advanced metrics says so through
`ICaptureObserver.WantsInGameAdvanced`; `CaptureSession` writes the word either way as the ring is attached, before
its first guard result, and from then on carries each publication of region 5 on `CaptureProgress.InGameAdvanced`
— one instance a publication, the same one on the ten ticks between two. `Application.Capture.InGameAdvancedSampler`
turns a publication into what a session stores: the session facts of the thirteen `Game` entries below (nineteen since 0.1.0-beta.43, twenty-one since 0.1.0-beta.44, twenty-three since 0.1.0-beta.45), in the
vendor's own words (a preset is its letter, a member its name), and two series —
`adv.ffx.upscale.frame_time_delta`, the mean of one second's dispatches, and `adv.ffx.memory_usage`, a point for
each second the game asked. The facts are one list until a field they are made of moves, so the recorder's context
is rewritten when a range widens and not once a second; the readings ride the newest telemetry sample of the tick.
**No schema change**: the facts go into `session_advanced.facts_json` beside the driver's, the series into
`sensor_blobs` under their `adv.` names, the `.partial`'s chunks 9 and 10 carry both. `AdvancedSources.InGame`
(optional; absent in every row 0.1.0-beta.39 wrote) says what became of those entries in the session: `Reading`,
`NothingSeen` — measured inside the game, and the game made none of the calls where they can be seen — or
`NotHooked`. The page says it in the session's sources line and, on Sources, that such an entry is read inside a
game FrameLedger measures when the game makes the call. The aliasable part of FSR's memory is published and not
shown: the catalog has one memory entry.
**The second step, 2026-11-12 (0.1.0-beta.43, PR-4).** Six more entries have reach `Game`: `sl.constants.jitter`,
`.motion_vectors`, `.depth`, `.projection`, `.resets` and `ffx.framegen.config`, from the two calls hooked only
for Advanced metrics (`17_HOOK_ENGINE` §Hook inventory, `slSetConstants` and `ffxConfigure`). The sampler makes
them from members region 5 already had (layout 5 as it was; schema 23 as it was): a switch is `on`, `off` or
`on_and_off` from the two words of what was seen, the flags are AMD's own names, the rectangle is the last one
the game gave and is left out where the capture side could not state it, the field of view is a range (D114).
**A field of view over π is said as given (owner decision D128, 0.1.0-beta.44).** Both vendors' headers say
radians; Cyberpunk 2077 gives Streamline 51–90, and the row read `51.0055 – 90 rad (2922.4° – 5156.6°)`. Such a
value is now the number alone, with a second row (`cameraFOV unit`, token `not_radians`) that says what it is
not; it is never converted, which would store this project's guess at the unit as the game's word. A
frame-generation configure that set no flag has a `flags` row of `none`, as an upscale does.
**A count is not part of what makes the facts new** — `sl.constants.resets` grows all session, and a context
rewritten on each would be one a second — so `InGameAdvancedSampler.Conclude` brings it up to the last
publication once, when the session ends. `AdvancedSources.StreamlineConstants` and `.FsrFrameGeneration`
(`AdvancedInGameCall`; null in every row written before, and in a session not measured inside the game) say for
each call which of three things happened: `Read`; `NeverCalled` — the capture side published the hook's family
(`FlHookFamily.SlConstants`, `.FfxConfigure` in the writer state's mask) and no publication carried the group; or
`NotHooked` — the family was never published, which is a game that did not load the library, a Streamline 1.x
interposer, or a patch that was refused. The page says it in one more sentence of the session's sources line.
**The third step, 2026-11-12 (0.1.0-beta.44).** Two more entries have reach `Game`: `sl.feature.versions` and
`sl.feature.requirements`, from what Streamline answers `slGetFeatureVersion` and `slGetFeatureRequirements`
(`17_HOOK_ENGINE` §Hook inventory). The sampler makes them from region 5's twelve feature slots (layout 5 and
schema 23 as they were): rows named `<feature> · <member>`, the feature by NVIDIA's constant (`kFeatureDLSS_G`;
`feature 4242` for an id this build has no name for), a version as `major.minor.build`, the requirement flags by
their enumerators. A version of all zeros is the vendor's "none" and has no row — a feature with no NGX model, a
requirement nobody stated — except `versionSL`; a count of 0 is a count. One bit of the mask covers both
questions, so which was answered is told by the slots (`InGameAdvancedSampler.FeatureCall`), and
`AdvancedSources.StreamlineFeatureVersion` and `.StreamlineFeatureRequirements` say for each: `Read`;
`NeverCalled` — hooked, and the game did not ask after that, ~~**the usual case, because a game asks while it
starts**~~ *(not the usual case: measured 2026-11-12, six of six Streamline sessions on five games read both, bar
Cyberpunk 2077's version question)*, and the page words it as "did not ask again"; or `NotHooked`. The slots are not emptied by a new request, so a second
session of a game that kept running stores what the first one read. `sl.init.preferences` stays unread (D112).
**The fourth step, 2026-11-12 (0.1.0-beta.45).** Two more entries have reach `Game`, from six calls
(`17_HOOK_ENGINE` §Hook inventory): `sl.feature.support` — `<feature> · slIsFeatureSupported`, the `sl::Result`
by NVIDIA's enumerator (`eOk`, `eErrorFeatureNotSupported`; `result 40` for a number past the forty this build
names, `result 255` for a code a byte could not carry), and `<feature> · slIsFeatureLoaded`, `on` or `off` — and
`sl.feature.lifecycle` — `<feature> · slSetFeatureLoaded` (`on`, `off`, `on_and_off`: what the game asked for
during the session, not the state it left), `· slAllocateResources` and `· slFreeResources` (a count; `≥ 15`
where the four bits stopped), and a row `slShutdown` with the number of shutdowns. All of it is word 1 of the
twelve feature slots and one word that was reserved (layout 5 and schema 23 as they were). A feature with nothing
to say has no row and a fact with no row is not stored. `AdvancedSources` has one member a call
(`StreamlineFeatureSupported`, `…Loaded`, `StreamlineSetFeatureLoaded`, `StreamlineAllocateResources`,
`StreamlineFreeResources`, `StreamlineShutdown`), `Read` / `NeverCalled` / `NotHooked` as before and null in a
row written before; the page says the six in one sentence. The switching and the counts are a session's — the
capture side clears them at a new request — so, unlike the answers, a second session does not inherit them.
`sl.init.preferences` now has reach `BeforeAttach` (D129): not a later release, never.
**Measured 2026-11-13** (six Streamline sessions on five games; `HANDOFF` item 43): `slIsFeatureSupported` was
read in every one — refusals among the answers, `eErrorFeatureMissing` for ImGUI, NIS and DeepDVC —
`slIsFeatureLoaded` in three games, `slSetFeatureLoaded` in two (every feature off and on), `slFreeResources`
and `slShutdown` once in one game each, and `slAllocateResources` in none.
- **D113 — three releases, because a new patch site cannot be proven in CI.** A crash in a game under the user-mode
anti-cheat exception's trial ends that exception for good (D38), and every release already means restarting the
game. **beta.40**: more fields of calls *already* hooked — `sl::DLSSOptions` where the game chains it into
`slEvaluateFeature`, the FSR upscale dispatch, the FSR memory query — shared-memory layout 5, the Agent and the
page; **no new patch site**. **~~beta.41~~ beta.42**: `slSetConstants` and `ffxConfigure`. **~~beta.42~~
beta.43**: `slGetFeatureVersion` and `slGetFeatureRequirements` — the owner's words: a release of their own, still
hooked. Each starts when the owner says the one before passed its owner-only item.
- **D122 — the second step is two releases** (owner decision 2026-11-12, after the owner's sessions on beta.40
and beta.41 and a review of the draft against the code). **0.1.0-beta.42 adds no patch site**: the session's
upscaler label (D120, `03_METRICS` §Upscaling), a patch registry that refuses what it cannot register, a native
log that survives the game's exit and says which descriptor types each ffx module saw — because Cyberpunk 2077,
set to FSR upscaling, was named "unknown" and nothing in the field could say why (D121) — and FSR's infinite depth
plane (D119). **0.1.0-beta.43 is `slSetConstants` and `ffxConfigure`**, begun only when that log was read. The
third step (`slGetFeatureVersion`, `slGetFeatureRequirements`) is beta.44, train C beta.45, train D beta.46.
- **D118 — 0.1.0-beta.41 is a fix release, and the train's numbers move by one** (owner decision 2026-11-12). The
owner's first evening with beta.40 showed the sensor library being closed under its other users (the paragraph on
`LhmLibrary` below; `20_OPEN_QUESTIONS` §M13); told that the fix would take the number D113 had given the train's
second release, the owner chose to release it at once. So: second release beta.42, third beta.43, train C
beta.44, train D beta.45. Text written before that day says beta.41 and beta.42 for the second and third — in
the released changelogs and the moved HANDOFF sections, which are not edited, and in the comments of `fl_shm.h`
and `ShmLayout.cs`, which the release that fills each member rewrites (they were left so that the code loaded
into a game stayed 0.1.0-beta.40's, line for line, in a release that fixes something else).
- **D112 — `slInit` is not hooked in train B.** A game calls it before FrameLedger attaches, so its preferences
would almost never be seen. If it is ever hooked: `applicationId`, `projectId` and the paths are never read.
- **D129 — `slInit` is not hooked at all** (owner decision 2026-11-12, asked to reopen D112 and shown what was
found). No mode of the shipped product is in the game in time: *Play ▸ Start directly* does not create the game
suspended — it waits for the presentation runtime, runs the guard, then injects; nothing is ever patched on a
game's thread, so a hook lands some hundred milliseconds after its module appears (`17_HOOK_ENGINE`, the loader
lock); and a hook that exists only for Advanced metrics waits for the Agent's request, which comes after the
attach. A hook on it would be a patch that reads nothing. `sl.init.preferences` says so in the catalog. Most of
what it would have told is read another way: which features are loaded (`slIsFeatureLoaded`), and whether a
newer plugin than the game's own is in use (the versions of 0.1.0-beta.44).
- **D130 — six more core calls in one release, 0.1.0-beta.45**: `slIsFeatureSupported`, `slIsFeatureLoaded`,
`slSetFeatureLoaded`, `slAllocateResources`, `slFreeResources`, `slShutdown`. Trains C and D come after it; train
C's spike is written beside it, as a document.
- **D131 — the adapter a game asks `slIsFeatureSupported` about is never read**: `sl::AdapterInfo` holds its LUID
or a Vulkan physical device. The feature's id and the result are all that is kept, and a build check refuses the
names of the struct's members in the code loaded into a game.
- **D114 — the field of view is a session's range**, lowest to highest, never a series: a series would say when the
player aimed.
- **D115 — a tail member is read only where it is proven to be there** (`17_HOOK_ENGINE` §Which struct version
carries which field): the table from the upstream tags, not the one header in the tree.
- **D116 — "never the camera" is a build check**, as "never a setter" is (D108): `tools/overlay-fields-check.ps1` —
every read of a vendor struct in one header, each allowed member named and used, the camera's names refused
anywhere in the Overlay and the Vulkan layer, no copy of `sl::Constants` by value. *(Built 2026-11-12:
twenty-eight members on `tools/overlay-allowed-fields.txt`; `12_BUILD` gate 14f. Fifty-six since 0.1.0-beta.44,
with the two answers about a feature.)*
- **What will read N/A, and says why.** Many titles set their DLSS options through `slDLSSSetOptions`, which is not
an export (train D), so `sl.dlss.*` is filled only for a title that passes its options with the evaluate call.
`slGetFeatureVersion` and `slGetFeatureRequirements` are calls a game makes while starting.
- **Privacy Policy 2.19** at beta.45 (approved with that release's plan): 2.18 named two of Streamline's answers
about a feature; whether it is supported and loaded, the game switching it, allocating and freeing its resources
and shutting Streamline down are in neither list, so they are named — with two more things never read: what
identifies the graphics card a game hands over with its question, and the content of a graphics resource.
- **Privacy Policy 2.18** at beta.44 (approved with that release's plan): 2.17 named two further calls and what
they give; Streamline's answers about a feature are in neither list, so they are named, with what of the answer
is never read — the lists it points to.
- **Privacy Policy 2.21 and Disclaimer 2.17** at beta.48 (wording approved with the plan of 2026-11-14, D145):
the Privacy Policy names the two further averages a session stores of the driver's report and that the
frame-generation multiplier is every value the session saw; the Disclaimer changes only with the accuracy
statement it carries, which now says two real games were read, that the stages overlap, and that the graphics
card's stage includes its waiting. No new call: the paragraph on the call the component makes is as it was.
- **Privacy Policy 2.20 and Disclaimer 2.16** at beta.47, train C (wording approved with the plan of
2026-11-13): the first thing read inside a game that is **not an argument of a call the game made**. The Privacy
Policy names what is stored — how long each stage of a frame took, as the driver reports it, and the driver's
low-latency state; never a time of day, never an input's moment. The Disclaimer drops "API arguments and nothing
else" for what is true, and §2 gains the paragraph on the one call the component makes itself: where, how often,
on which thread, what it never calls, what ends it, and that a call which never returned could not be stopped.
- **Privacy Policy 2.17** at beta.43 (approved with that release's plan): 2.16 listed DLSS's and FSR's upscaling
parameters by name, and Streamline's per-frame constants and the frame-generation configuration are in neither
list, so they are named — and §4 says the two calls are intercepted only once a session with the option on has
started in the game.
- **Privacy Policy 2.16** at beta.40, once for the three releases: with Advanced metrics on, for a game FrameLedger
injects into, the parameters the game gives its upscaler and frame generation through the intercepted calls are
*stored* — sizes, presets, exposure, sharpening, the jitter's, the depth's and the field of view's ranges, flags,
versions — and never the camera's position, axes or matrices.
### Left out, by name
Present in the vendored `nvapi.h`, readable from the Agent, and not in the inventory — each is either not a
measurement of how a game runs or exists on professional cards only: ECC status and errors, NVLINK, illumination
(the card's lighting zones), licensable features, virtualization and GSP information, Quadro status, VR-ready data,
the IRQ and bus slot numbers, logical GPUs and SLI state, HDCP support, scan-out configuration, the colour and
info-frame controls. `NvAPI_DISP_GetTiming` calculates a timing from an input and reads nothing.
`NV_GET_ADAPTIVE_SYNC_DATA`'s last-flip count and timestamp mean nothing sampled once a second.
`NvAPI_GPU_ClientRegisterForUtilizationSampleUpdates` registers a callback, which is not a read.
### The inventory
One row per entry of `AdvancedMetricCatalog`, in its order; `AdvancedMetricCatalogTests` fails when this table and
the catalog differ in either direction. **Reach** is where a value can be read from, or why it is not read: `Agent`
(the Agent's own process), `AgentElevated` (the same, as administrator with PawnIO), `Game` (read inside a game
FrameLedger measures, from the arguments of a call the Overlay hooks, while the session records Advanced metrics —
since 0.1.0-beta.40), `HookedCall` (an argument of a call the Overlay already hooks, not read yet), `NewHook` (a hook
that is not built — Train B; no entry carries it since 0.1.0-beta.45), `BeforeAttach` (a call the game makes once,
before FrameLedger is inside it, so no hook could read it — D129), `Spike` (Trains C and D; no entry since D154), `Refused`, `Deferred`, `Identifier` (D104), `PlayerState`
(rule 4), `Licence`, `NotInHeaders`. ~~Only the first two are read by a build of Train A.~~ The first three are read
(`AdvancedMetric.IsRead`).
**NVAPI, the graphics card: facts**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.gpu.type` | Integrated or discrete graphics | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.system_type` | Laptop or desktop, as the driver sees the PC | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.external` | The card is an external one | MachineFact | Nvapi | Agent | Flag |
| `nvapi.gpu.bus_type` | The bus the card sits on | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.pci_ids` | PCI device, subsystem and revision ids - the model, not the unit | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.vbios_version` | Video BIOS version | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.architecture` | Chip architecture, implementation and revision | MachineFact | Nvapi | Agent | Text |
| `nvapi.gpu.core_count` | Graphics cores | MachineFact | Nvapi | Agent | Count |
| `nvapi.gpu.rt_cores` | Ray-tracing cores | MachineFact | Nvapi | Agent | Count |
| `nvapi.gpu.tensor_cores` | Tensor cores | MachineFact | Nvapi | Agent | Count |
| `nvapi.gpu.ram_bus_width` | Width of the card's memory bus | MachineFact | Nvapi | Agent | Bits |
| `nvapi.gpu.clock_base_core` | Base core clock | MachineFact | Nvapi | Agent | Megahertz |
| `nvapi.gpu.clock_base_memory` | Base memory clock | MachineFact | Nvapi | Agent | Megahertz |
| `nvapi.gpu.clock_boost_core` | Boost core clock | MachineFact | Nvapi | Agent | Megahertz |
| `nvapi.gpu.clock_boost_memory` | Boost memory clock | MachineFact | Nvapi | Agent | Megahertz |
| `nvapi.gpu.temp_default_max` | The core sensor's default maximum temperature | MachineFact | Nvapi | Agent | Celsius |
**NVAPI, the card's memory**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.vram.dedicated` | Dedicated video memory on the card | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.dedicated_available` | Dedicated video memory available for surfaces | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.system` | System memory the driver reserved at load | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.shared` | Shared system memory the driver may commit | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.framebuffer_physical` | Physical frame buffer size | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.framebuffer_virtual` | Virtual frame buffer size | MachineFact | Nvapi | Agent | Mebibytes |
| `nvapi.vram.available` | Dedicated video memory available now, all programs | Series | Nvapi | Agent | Mebibytes |
| `nvapi.vram.in_use` | Dedicated video memory in use now, all programs: available for surfaces less available now | Series | Nvapi | Agent | Mebibytes |
| `nvapi.vram.evictions_size` | Memory released by evictions since the driver loaded | Series | Nvapi | Agent | Mebibytes |
| `nvapi.vram.eviction_count` | Evictions from dedicated video memory since the driver loaded | Series | Nvapi | Agent | Count |
| `nvapi.vram.promotions_size` | Memory moved in by promotions since the driver loaded | Series | Nvapi | Agent | Mebibytes |
| `nvapi.vram.promotion_count` | Promotions into dedicated video memory since the driver loaded | Series | Nvapi | Agent | Count |
**NVAPI, the card over time**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.load.framebuffer` | Frame-buffer (memory controller) utilisation | Series | Nvapi | Agent | Percent |
| `nvapi.load.video` | Video-engine utilisation | Series | Nvapi | Agent | Percent |
| `nvapi.load.bus` | Bus-interface utilisation | Series | Nvapi | Agent | Percent |
| `nvapi.clock.video` | Video clock | Series | Nvapi | Agent | Megahertz |
| `nvapi.pstate` | The card's current performance state (P-state id) | Series | Nvapi | Agent | Count |
| `nvapi.temp.board` | Board ambient temperature, where the card has the sensor | Series | Nvapi | Agent | Celsius |
| `nvapi.temp.power_supply` | The card's power-supply temperature, where it has the sensor | Series | Nvapi | Agent | Celsius |
| `nvapi.limit.ac_battery` | Held back by a switch to battery on the tick: 100 or 0 | Series | Nvapi | Agent | Percent |
| `nvapi.limit.api` | Held back at a program's request on the tick: 100 or 0 | Series | Nvapi | Agent | Percent |
| `nvapi.limit.insufficient_power` | Held back by a missing power connector on the tick: 100 or 0 | Series | Nvapi | Agent | Percent |
| `nvapi.encoder.sessions` | Active hardware-encoder sessions, all programs | Series | Nvapi | Agent | Count |
| `nvapi.encoder.fps` | Average encoding rate of the active sessions | Series | Nvapi | Agent | FramesPerSecond |
| `nvapi.encoder.latency` | Average encoding latency | Series | Nvapi | Agent | Microseconds |
**NVAPI, the driver**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.driver.package` | Game Ready, Studio or an RTX enterprise branch | MachineFact | Nvapi | Agent | Text |
| `nvapi.driver.dch` | The driver is a DCH package | MachineFact | Nvapi | Agent | Flag |
| `nvapi.driver.base_branch` | The driver's base branch | MachineFact | Nvapi | Agent | Text |
| `nvapi.driver.interface_version` | NVAPI's own interface version | MachineFact | Nvapi | Agent | Text |
| `nvapi.driver.chipset` | Chipset vendor and name, as the driver reports them | MachineFact | Nvapi | Agent | Text |
| `nvapi.driver.global_profile` | Every setting of the driver's global profile, read-only | MachineFact | Nvapi | Agent | Table |
| `nvapi.session.driver_profile` | The settings the driver's profile for this game sets itself, read-only | SessionFact | Nvapi | Agent | Table |
**NVAPI, displays**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.display.hdr_formats` | HDR formats a display accepts: HDR10, HDR10+, HDR10+ Gaming, Dolby Vision, EDR | MachineFact, one per device | Nvapi | Agent | Text |
| `nvapi.display.luminance_max` | A display's maximum luminance for HDR content | MachineFact, one per device | Nvapi | Agent | Nits |
| `nvapi.display.luminance_min` | Its minimum luminance | MachineFact, one per device | Nvapi | Agent | Nits |
| `nvapi.display.luminance_max_frame_average` | Its maximum frame-average light level | MachineFact, one per device | Nvapi | Agent | Nits |
| `nvapi.display.vrr_support` | Variable refresh: supported, G-SYNC module or adaptive sync, ULMB, usable at the current mode | MachineFact, one per device | Nvapi | Agent | Text |
| `nvapi.display.adaptive_sync_off` | Adaptive sync is switched off on the display | MachineFact, one per device | Nvapi | Agent | Flag |
| `nvapi.display.max_frame_interval` | The longest frame interval adaptive sync is set to hold | MachineFact, one per device | Nvapi | Agent | Microseconds |
| `nvapi.session.vrr_enabled` | Variable refresh was active on the game's display | SessionFact | Nvapi | Agent | Flag |
| `nvapi.session.vrr_requested` | Variable refresh was requested for it | SessionFact | Nvapi | Agent | Flag |
| `nvapi.session.vrr_possible` | Variable refresh was possible on it | SessionFact | Nvapi | Agent | Flag |
| `nvapi.session.vrr_display_mode` | The display itself was in variable-refresh mode | SessionFact | Nvapi | Agent | Flag |
| `nvapi.session.source_colorspace` | The colour space the game presented in, as the driver reports it for the process | SessionFact | Nvapi | Agent | Text |
**LibreHardwareMonitor, the graphics card (every vendor, unelevated)**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `lhm.gpu.temperature` | Every temperature sensor of the card | Series, one per device | Lhm | Agent | Celsius |
| `lhm.gpu.load` | Every load of the card: its engines, its bus, its memory controller | Series, one per device | Lhm | Agent | Percent |
| `lhm.gpu.memory` | The card's memory counters: total, free, used, Direct3D dedicated and shared | Series, one per device | Lhm | Agent | Mebibytes |
| `lhm.gpu.clock` | Every clock of the card | Series, one per device | Lhm | Agent | Megahertz |
| `lhm.gpu.power` | Every power reading of the card | Series, one per device | Lhm | Agent | Watts |
| `lhm.gpu.fan` | Every fan of the card | Series, one per device | Lhm | Agent | Rpm |
| `lhm.gpu.fan_control` | Each fan's duty | Series, one per device | Lhm | Agent | Percent |
| `lhm.gpu.voltage` | Every voltage of the card | Series, one per device | Lhm | Agent | Volts |
| `lhm.gpu.current` | Every current of the card | Series, one per device | Lhm | Agent | Amperes |
| `lhm.gpu.throughput` | PCI Express traffic to and from the card | Series, one per device | Lhm | Agent | BytesPerSecond |
**LibreHardwareMonitor through PawnIO (the Agent as administrator, D103)**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `lhm.cpu.temperature` | Processor temperatures: the package and each core | Series, one per device | Lhm | AgentElevated | Celsius |
| `lhm.cpu.load` | Processor load: the total and each core | Series, one per device | Lhm | AgentElevated | Percent |
| `lhm.cpu.clock` | Each core's clock, and the bus clock | Series, one per device | Lhm | AgentElevated | Megahertz |
| `lhm.cpu.power` | Processor power: the package, the cores | Series, one per device | Lhm | AgentElevated | Watts |
| `lhm.cpu.voltage` | Processor voltages | Series, one per device | Lhm | AgentElevated | Volts |
| `lhm.board.model` | The motherboard's maker and model | MachineFact | Lhm | AgentElevated | Text |
| `lhm.board.sensor_chip` | The board's sensor chip | MachineFact | Lhm | AgentElevated | Text |
| `lhm.board.temperature` | The board's temperature sensors | Series, one per device | Lhm | AgentElevated | Celsius |
| `lhm.board.fan` | Each fan header's speed | Series, one per device | Lhm | AgentElevated | Rpm |
| `lhm.board.fan_control` | Each fan header's duty | Series, one per device | Lhm | AgentElevated | Percent |
| `lhm.board.voltage` | The board's voltage rails | Series, one per device | Lhm | AgentElevated | Volts |
| `lhm.memory.modules` | The memory modules: size, speed, maker's part name - never a serial number | MachineFact | Lhm | AgentElevated | Table |
| `lhm.memory.load` | Memory in use: physical and virtual | Series, one per device | Lhm | AgentElevated | Percent |
| `lhm.memory.temperature` | Each memory module's temperature, where it has a sensor | Series, one per device | Lhm | AgentElevated | Celsius |
**What the Agent could reach and does not**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `lhm.psu` | Power-supply sensors: reading them makes the library search every USB device | MachineFact | Lhm | Refused | Table |
| `lhm.controller` | Fan-controller sensors: the same search of USB devices | MachineFact | Lhm | Refused | Table |
| `lhm.battery` | Battery sensors | MachineFact | Lhm | Deferred | Table |
| `lhm.storage.models` | The drives' model names: the library cannot list drives without touching removable ones | MachineFact | Lhm | Deferred | Table |
| `lhm.storage.temperature` | Each drive's temperature: the same, so the storage group is not opened | Series, one per device | Lhm | Deferred | Celsius |
| `nvapi.gpu.board_serial` | The card's board serial number | MachineFact | Nvapi | Identifier | Text |
| `nvapi.gpu.uuid` | The card's unique id | MachineFact | Nvapi | Identifier | Text |
| `nvapi.display.edid` | A display's EDID block, which carries its serial number | MachineFact, one per device | Nvapi | Identifier | Text |
| `nvapi.encoder.session_processes` | Which programs are encoding | Series | Nvapi | Identifier | Table |
| `nvapi.gpu.power` | Card power through NVAPI (LibreHardwareMonitor reads it) | Series | Nvapi | NotInHeaders | Watts |
| `nvapi.gpu.fan_duty` | Fan duty through NVAPI (LibreHardwareMonitor reads it) | Series | Nvapi | NotInHeaders | Percent |
| `nvapi.gpu.voltage` | Card voltage through NVAPI (LibreHardwareMonitor reads it) | Series | Nvapi | NotInHeaders | Volts |
| `nvapi.gpu.ram_type` | The card's memory type | MachineFact | Nvapi | NotInHeaders | Text |
**Inside the game: calls the Overlay already hooks**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `sl.dlss.output_size` | The output size the game gave DLSS | SessionFact | Streamline | Game | Text |
| `sl.dlss.presets` | The DLSS model preset per quality mode | SessionFact | Streamline | Game | Table |
| `sl.dlss.hdr_input` | The game told DLSS its colour buffers are HDR | SessionFact | Streamline | Game | Flag |
| `sl.dlss.auto_exposure` | DLSS auto-exposure | SessionFact | Streamline | Game | Flag |
| `sl.dlss.alpha_upscaling` | DLSS upscales the alpha channel | SessionFact | Streamline | Game | Flag |
| `sl.dlss.exposure` | Pre-exposure and exposure scale given to DLSS | SessionFact | Streamline | Game | Table |
| `ffx.upscale.sharpening` | FSR sharpening: on or off, and its strength | SessionFact | FidelityFx | Game | Table |
| `ffx.upscale.jitter` | The largest jitter offset the game gave FSR | SessionFact | FidelityFx | Game | Table |
| `ffx.upscale.exposure` | Pre-exposure given to FSR | SessionFact | FidelityFx | Game | Text |
| `ffx.upscale.depth` | Near and far planes and the range of the field of view given to FSR | SessionFact | FidelityFx | Game | Table |
| `ffx.upscale.flags` | The flags of the game's FSR dispatch | SessionFact | FidelityFx | Game | Text |
| `ffx.upscale.frame_time_delta` | The frame time the game reported to FSR | Series | FidelityFx | Game | Milliseconds |
| `ffx.memory_usage` | FSR's own video memory, when the game asks it | Series | FidelityFx | Game | Mebibytes |
**Inside the game: hooks that are not built**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `sl.constants.jitter` | The jitter offsets the game gave Streamline | SessionFact | Streamline | Game | Table |
| `sl.constants.motion_vectors` | Motion-vector scale, and whether they are 3D, dilated, jittered | SessionFact | Streamline | Game | Table |
| `sl.constants.depth` | Near and far planes, and whether depth is inverted | SessionFact | Streamline | Game | Table |
| `sl.constants.projection` | Field of view, aspect ratio, orthographic or not | SessionFact | Streamline | Game | Table |
| `sl.constants.resets` | How often the game reset the upscaler's history | SessionFact | Streamline | Game | Count |
| `ffx.framegen.config` | The game's FSR frame-generation configuration | SessionFact | FidelityFx | Game | Table |
| `sl.feature.versions` | Each Streamline feature's version and its NGX model version | SessionFact | Streamline | Game | Table |
| `sl.feature.requirements` | What each Streamline feature requires: operating system, driver, buffers | SessionFact | Streamline | Game | Table |
| `sl.feature.support` | Whether Streamline says each feature is supported, and loaded | SessionFact | Streamline | Game | Table |
| `sl.feature.lifecycle` | Each feature switched on or off, its resources allocated and freed, Streamline shut down | SessionFact | Streamline | Game | Table |
| `sl.init.preferences` | The preferences the game started Streamline with | SessionFact | Streamline | BeforeAttach | Text |
| `sl.constants.camera` | Camera position, axes and matrices | SessionFact | Streamline | PlayerState | Text |
**Inside the game: what the graphics driver reports, asked for by the part inside the game itself** *(train C,
owner decisions D132, D136, D138 — `17_HOOK_ENGINE` §Calls the Overlay makes into a driver; since 0.1.0-beta.47)*
Not an argument of a hooked call: the Overlay calls the driver's two read-only latency getters about once a second,
only while a session records Advanced metrics, only for a Direct3D 12 game that had already loaded the driver's
`nvapi64.dll`. The report becomes seven series — `adv.nvapi.reflex.frame_report.` and `simulation`,
`render_submit`, `present`, `driver`, `os_queue`, `gpu_render`, `total` — each point the mean, in microseconds, of
the frames one read summed; the sleep status becomes one table of the driver's five switches as seen over the
session and, since 0.1.0-beta.48 (D144), every frame-generation multiplier it gave (it kept the last one, and a
session that ended in a menu read 1). **Since 0.1.0-beta.48 (D142) two more series, an entry of their own** —
`adv.nvapi.reflex.gpu_times.` and `active_render`, `frame_time` — the driver's `gpuActiveRenderTimeUs` and
`gpuFrameTimeUs`, each the mean over the frames of that read that gave one. They are the driver's own figures, not
spans between two of its marks: in Phantom Blade Zero `gpu_render` (the two GPU marks, idle time included) was
longer than `total`, which is why its label now says what it includes (D143). **Reworded 2026-11-14 (D152)**:
NVIDIA's header defines the active time as that span "excluding the idles in between", and in Cyberpunk 2077
without frame generation the active time is the longer of the two on nine frames in ten (HANDOFF, item 47) — so
the labels now say what each is (*the driver's own figure*; *between the driver's start and end marks*) and no
longer that one is the other less its idle time. The numbers and the means are unchanged. An entry of their own keeps the
seven spans within what one chart draws as lines. What the *game* asked of Reflex (`nvapi.reflex.mode`) is an argument of a function no module exports:
reading it takes a patch at an address learned at run time, which is refused.
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `nvapi.reflex.mode` | Reflex: low-latency mode, boost, the frame-rate limit | SessionFact | Nvapi | RefusedRoute | Text |
| `nvapi.reflex.sleep_status` | What the driver's low-latency state is | SessionFact | Nvapi | GameDriver | Table |
| `nvapi.reflex.frame_report` | Latency per frame: simulation, render submit, present, driver, OS queue, GPU render | Series, one per device | Nvapi | GameDriver | Microseconds |
| `nvapi.reflex.gpu_times` | The graphics card's time per frame, as the driver works it out: active render, frame to frame | Series, one per device | Nvapi | GameDriver | Microseconds |
**Inside the game: Streamline's feature functions — not read (D154, 2026-11-14)**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `sl.dlss.optimal_settings` | The render size and sharpness DLSS recommends | SessionFact | Streamline | RefusedRoute | Text |
| `sl.dlss.vram_estimate` | DLSS's own estimate of its video memory | Series | Streamline | RefusedRoute | Mebibytes |
| `sl.dlssg.options` | DLSS frame generation: mode, frames to generate | SessionFact | Streamline | RefusedRoute | Text |
| `sl.dlssg.state` | DLSS frame generation: status, the most frames it can generate | SessionFact | Streamline | RefusedRoute | Text |
| `sl.dlssg.frames_presented` | Frames DLSS frame generation presented, by its own count | Series | Streamline | RefusedRoute | Count |
| `sl.dlssg.vram_estimate` | DLSS frame generation's estimate of its video memory | Series | Streamline | RefusedRoute | Mebibytes |
| `sl.dlssd.options` | Ray reconstruction's options | SessionFact | Streamline | RefusedRoute | Text |
| `sl.nis.options` | NVIDIA Image Scaling's options | SessionFact | Streamline | RefusedRoute | Text |
| `sl.deepdvc.options` | RTX Dynamic Vibrance's options | SessionFact | Streamline | RefusedRoute | Text |
| `sl.reflex.state` | Reflex through Streamline: availability and its frame reports | SessionFact | Streamline | RefusedRoute | Table |
| `sl.pcl.markers` | PC-latency markers the game sets | Series | Streamline | RefusedRoute | Count |
**Closed by a licence**
| Id | What it is | Scope | Source | Reach | Unit |
|---|---|---|---|---|---|
| `ngx.parameters` | NGX parameters: the quality value, the DLSS hints | SessionFact | Ngx | Licence | Table |
| `xess.options` | XeSS's options | SessionFact | Xess | Licence | Table |
| `sl.nvperf` | Streamline's performance counters (NSight Perf SDK licence) | Series | Streamline | Licence | Table |
+56 -2
View File
@@ -60,10 +60,30 @@ Implemented in `FrameLedger.Injector` and reached from managed code through a th
> at once), waits `kScanRetryPauseMs` (250 ms), takes a **fresh** scan set and scans it whole. A member still unreadable
> then refuses `kProcessUnreadable`, as before; one that has gone is not in the new set; a target that has gone leaves
> no set, which refuses `kProcessTreeUnavailable`. **`ERROR_ACCESS_DENIED` — a protected process — is never retried**,
> and no other answer of any source is. The sentence above stands: every failure still means REFUSE; what changed is
> and no other answer of any source is *(one more since D123, below)*. The sentence above stands: every failure still means REFUSE; what changed is
> how many times "still loading" is asked before it counts as one. ctest `fl_guard` `[D48]` holds both directions,
> against fakes and against the real source (a process created suspended, a pid with no process).
>
> **Amended 2026-11-12 (owner decision D123): a module list that changed while it was read is read again, at most
> three times — never a pass.** The module check lists a process's modules and then asks each one's name. A game at
> its launch loads and unloads modules for seconds, and a module that was in the list and could not be named a
> moment later made the whole answer "incomplete": a refusal before injection, or — 0.5 s after it, at the first
> supervision scan — a safety unhook with no frame recorded, the game then running on unrecorded. **Five such ends
> in one PC's Agent logs in a week** (Phantom Blade Zero three times on 2026-11-05, Crown & Coin on 11-10,
> Cyberpunk 2077 on 11-12), and twice in this repository's own gate. Asked with that, the owner chose: read it
> again. Such a list is now `Collected::kUnsettled`; the pass finishes — **a blocklisted module among those that
> were read still refuses at once** — waits `kScanRetryPauseMs`, takes a fresh scan set and reads it whole, up to
> `kMaxUnsettledRereads` (3) times: four reads and 750 ms of pauses at most. A list still unsettled on the last
> read refuses `kModuleScanFailed` ("a module list kept changing while it was read"). **What did not change:** a
> list too long for the guard's buffer is still `kIncomplete` and refuses at once (a second read would be as
> long); a protected process still refuses at once; D48's re-read is still one for the whole check; every read is
> of the whole scan set, and the read that allows is one in which every member was read whole and none held a
> finding. A failure to read is still never a pass — what changed is how many times a list that is moving is
> asked before it counts as unreadable. The launch poll's presentation-runtime probe treats an unsettled list as
> "ask again" rather than deciding on half of one. ctest `fl_guard` `[D123]`: against fakes, every direction
> above; against the real source, a process that loads and frees a module in a loop — read some 3,600 times in
> three seconds, about one read in four is unsettled and none is anything else.
>
> **Launch mode (2026-09-06, P1 item 2) does not weaken this.** `FlGuardedInjectWhenReady` polls
> the same enumeration to learn *when* the launched target has mapped a presentation runtime, and a
> failing enumeration there means "not yet" — the poll continues. That is not a pass: the poll
@@ -1027,7 +1047,7 @@ injected or layered**.
If `OpenProcess` with the needed rights fails (protected process, higher integrity), do **not** escalate creatively. Report "cannot attach" and offer Tier-2. Never attempt to acquire privileges beyond running the Agent elevated at the user's explicit request.
> **The explicit request exists since beta.10 (owner decision D34, 2026-09-27): the admin mode.** An option off by default,
> turned on through a versioned disclosure (`Safety_AdminMode_*`, `agent-admin-dialog/1`) that states what it adds, what it
> turned on through a versioned disclosure (`Safety_AdminMode_*`, `agent-admin-dialog/2` *(`/1` until 2026-11-11, when Advanced metrics widened what the option adds — D101)*) that states what it adds, what it
> does not change, how Windows asks and the risk of a per-user install; then **Windows asks at every start of the Agent**
> (`01_ARCHITECTURE` ADR-9). What it does NOT do, each held by code rather than by this paragraph:
>
@@ -1214,6 +1234,20 @@ outlives a rules change, a game update and *Change executable*; §What a finding
hooked session's early abnormal end; the policy counted them, and two turned hooking off. It cannot tell either from
the hang it exists for, so it stands as written; whether a witness — the same executable started again within
seconds — may exempt a hand-off is the owner's decision, not taken. Users are told in `LIMITATIONS.md` §Play.
- **The count starts again when the user switches hooking on (owner decision D126, 0.1.0-beta.44).** Until then
`hook_crash_count` only ever grew: Lies of P had two such ends on 2026-10-19/20 (the hand-off above), its user
switched hooking on again, and on 2026-11-12 one more early end — the game's own crash, by the owner, after a shader
failed to compile — switched it off at once, under a reason that says *twice*. The consent grant, the only
statement that turns hooking on, now sets the count to 0 (`SqliteGameConsentStore`). The policy still only
disables, a block is still never cleared, and the user-mode exception's trial (D38) keeps its own record, which
nothing sets back.
- **A fault while the game is closing is not seen by a hooked session (owner decision D127: written down, not
changed in 0.1.0-beta.44).** Hogwarts Legacy faults in the driver's `_nvngx.dll` at exit — six times in Windows'
application log since 2026-11-06, hooked or not. The one unhooked session reads `crashed`; the five hooked ones
read `normal` with exit code 0: the witness is looked for when the session is saved, which for a hooked session
is before Windows has written the event. Looking again later would make every session of such a game `crashed`,
and under a user-mode exception that ends the exception for good for a fault no hook caused — so it is a decision
of its own, not a fix to slip in beside two new hooks.
- **What reads as a crash changed in beta.14, in both directions** (`04_CAPTURE` §Crash & exit classification): a
Windows report that is not a crash (a memory-leak or hang report) no longer makes one, and an engine's crash reporter
started by a game that then left does — Unreal's crash ended with exit code 3 and read `normal`. Both feed this policy,
@@ -1231,6 +1265,26 @@ outlives a rules change, a game update and *Change executable*; §What a finding
> the user. Cautious mode returns with the overlay, when there is something
> for it to switch off.
## What the Agent reads about input, from outside every game (owner decisions D155–D159)
Since 0.1.0-beta.50 the Agent can say what a session was played with (`21_INPUT_DEVICES`). It is on this page
because of what it must never become:
- **It is read from Windows, by the Agent. Nothing is added to a game**: no hook, no injected code, nothing in
`17_HOOK_ENGINE` §Hook inventory. Rule 4 is untouched, and a game the guard refuses is read like any other —
nothing about it touches the game's process.
- **No hook of any kind.** `SetWindowsHookEx` stays forbidden everywhere (`tools/chokepoint-check.ps1`); keyboard
and mouse are known by one call that returns a time (`GetLastInputInfo`), and raw input is registered for
joysticks and gamepads only — never for a keyboard or a mouse. A tool that hooks or registers for the keyboard is
what security software, and some anti-cheat, look for; this is not one.
- **No device is opened and nothing is sent to one.** A handle of FrameLedger's on a controller could make a game's
own exclusive open fail; an output or feature report could change how a controller reports to the game. Neither
is done.
- **A committed list holds it**: `tools/input-readonly-check.ps1` (`12_BUILD` step 14g).
Whether any anti-cheat objects to a background program that receives controllers' reports is not known; none is
documented to. It reads nothing of, and does nothing to, the game.
## Honest limits to document to users
The Disclaimer states these explicitly, and the UI consent dialog echoes them:
+19 -2
View File
@@ -1549,6 +1549,16 @@ probe never showed a deadlock.
> real-game leg — a title that pulls Streamline or D3D12 in lazily, hooked through the wake — which the
> `loader detour:` report line will show as a non-zero wake count with the family present from the first
> bucket.
>
> **0.1.0-beta.46 (owner decision D134): one suspension for a batch, and the rule is as it was.** The rows a
> watchdog tick finds are enabled under one `MH_ApplyQueued` — one stop of the threads where there was one a row
> (`17_HOOK_ENGINE` §Unhooking, Batches). Still only on the watchdog, never from the detour and never on a game's
> thread; the gate around the queue is not held across a symbol lookup, so it is never held with the loader's lock
> wanted. What a stop of the threads costs in a real game was read from the owner's logs of 2026-11-13: 57 to
> 215 ms, thirteen times over for a Streamline title until this release. **On beta.46, the same evening** (five
> games, eighteen batches): 81–262 ms a batch for sixteen of them, and two that took 0.61 s and 1.26 s while their
> game was loading — one stop where there were three and thirteen, and still a second of every thread of Hogwarts
> Legacy standing still. Not explained (`17_HOOK_ENGINE` §Unhooking, Batches).
### H11 ◐ · XeFG and FSR3-FG identity — **AMD built 2026-09-04 through the ffx-api leaves, complete 2026-09-06; Intel closed by licence; FSR 3.0 host route built 2026-09-05**
@@ -1860,6 +1870,12 @@ Moved verbatim, in this order.
| M9 ✅ | **Closed 2026-08-02 by a decision.** The old file/module detection does not exist in this repository and the owner confirmed there is no copy elsewhere, so **"build a minimal static-hint detector as the measurement baseline" is now P0 item 3** (`15_ROADMAP`). It needs no guard and no injection. Until it exists, ADR-7's headline claim stays out of the README rather than being asserted unmeasured |
| M10 ❓ | **§CPU — is the CPU temperature this build reads the right sensor, and does it read at all on the owner's machine?** `LhmCpuTemperatureReader` (2026-09-21) opens a CPU-only LibreHardwareMonitor `Computer` when the Agent is elevated AND `PawnIo.IsInstalled`, takes the package sensor (`CPU Package` / `Core (Tctl/Tdie)`) and falls back to the hottest core. It was written against LHM's sensor names and a fake `ILhmComputer`; **no elevated run with PawnIO has been taken.** | **The measurement is one session:** Agent elevated, PawnIO installed, a five-minute capture with HWiNFO or Ryzen Master open beside it. Pre-committed: `max_cpu_temp` within 3 °C of the reference's package peak → keep; a core sensor picked where a package sensor exists, or a reading tens of degrees off (a distance-to-TjMax sensor mistaken for a temperature) → fix `Pick` and re-run; nothing read at all → `cpu_temp` stays N/A and `legal/ACCURACY.md` says processor temperature is not measured. CPU **load** and memory are not in question: `TheRealCountersAnswerOnThisMachine` reads them on every gate run. |
| M11 🅓 | **The game's VRAM budget and `budget_exceeded_pct` (FR-4.5's second half), 2026-10-03.** D43 reads the game's usage from outside (the `GPU Process Memory` counters); the budget the OS grants the process is what DXGI's `QueryVideoMemoryInfo` returns to the *calling* process (§M7), so only an in-process call — the Overlay's background thread, which today only checks for a stop — or `D3DKMTQueryStatistics`' `PROCESS_SEGMENT_GROUP` query (documented "Reserved for system use") could read it. **Deferred by decision:** neither is built; `vram_budget_exceeded_pct` stays NULL and the budget N/A, and the in-process route would change what the Disclaimer says the Overlay's thread does. Reopening it needs that decision, not a measurement. | — |
| M12 ❓ | **Advanced metrics — which NVAPI rows answer on real hardware (D101, 2026-11-11).** `18_GPU_VENDOR_APIS` §The inventory lists sixty `nvapi.` rows whose reach is `Agent`. They are declared in the vendored headers; **none has been called on a GPU**, and CI has no driver, so it never takes that branch. Unknown per row: whether the RTX 5080 on the current driver answers or returns `NVAPI_NOT_SUPPORTED` (the memory-temperature target already does not exist on that card). | **The measurement is the bridge's probe on the owner's PC**, then the page itself (the release's owner-only item). Pre-committed: a call that does not answer → that row is N/A with NVAPI's own status as the reason, and nothing is put in its place; a call that hangs or faults twice → the row is off for the session (§Runtime policy); a value that contradicts the vendor's own tool → the row is withheld until it is explained. **Half answered the same day, by the bridge's own test on the dev box** (ctest `fl_nvapi_bridge`, AVAILABLE branch; RTX 5080): every fact answers (`present = 0xFFFFF`), eight of ten live groups answer (`0x33F` — no board and no power-supply temperature sensor on this card: N/A, by the first pre-committed outcome), the display answers its four groups, the global profile lists 24 settings; figures in `18_GPU_VENDOR_APIS` §L3. **Still open**: the same values on the page beside the vendor's own tool, and the session rows (`nvapi.session.*`), which need a game. **And one thing seen once and not explained**: in one test run of about a dozen that day, `FlNvReadLive` came back with no field at all although `FlNvReadFacts`, a moment earlier in the same process, had answered everything; nine further runs did not show it. A handle the driver invalidated is the usual cause of that shape, and nothing here re-enumerates after `FlNvInit`. Pre-committed for the reader that owns the thread: two readings in a row with no field → release and initialise the bridge again, once, and say so in the log; if that does not bring the fields back the layer is off for the session, as any layer that fails twice is. |
| M13 ❓ | **Advanced metrics — LibreHardwareMonitor's CPU, memory, motherboard and storage groups, elevated, with PawnIO (D103, 2026-11-11).** Never opened by any build: §M10's elevated run has not been taken either, and this is a larger surface — a sensor chip on the LPC bus, the embedded controller, SPD over SMBus, SMART. Unknown: whether each group opens on the owner's board, what a refresh costs, whether opening the storage group touches removable drives (this PC once had a failing USB drive), and whether any group enables a privilege (`PrivilegeFootprintTests`). | **One elevated run with the probe verb, before the option exists.** Pre-committed: the storage group cannot be opened without touching removable drives → `lhm.storage.*` stays N/A in beta.39 and its reach reads `Deferred`; a node that throws twice → skipped for the session; the Agent above 1% of one core or 150 MB with the option on (NFR-2) → the slow groups are read less often or dropped before the release; a privilege enabled by a group → that group is not opened, and the owner is asked. **Two of the four answered 2026-11-11, before the option exists.** (1) **Storage: not opened, by the first outcome.** LibreHardwareMonitor 0.9.6's `StorageGroup` lists every drive through DiskInfoToolkit's `ReloadStorages` with no filter by bus and follows drives plugged in later — it cannot be opened without touching removable drives. `lhm.storage.*` reads `Deferred`, and `LhmGroups` has no member for it. (2) **Cost, unelevated, measured with `probe-advanced`** on the dev box: one refresh of the graphics-card group alone is 14 ms of processor time (62 ms elapsed) — 1.4% of one core at one a second, above NFR-2 — and one NVAPI read 0.4 ms. By the third outcome the sensor library is read every fifth tick (0.28%). **Still open, and the owner's**: the same probe as administrator with PawnIO — whether the processor, memory and motherboard groups open on this board, what they cost, and `PrivilegeFootprintTests`' new case on a token that holds the privileges (CI's does; the dev box's unelevated run passes by construction). **The first run was taken on 2026-11-12, by the owner, in ordinary use** (0.1.0-beta.40, the Agent as administrator with PawnIO, an i7-14700KF on a board with an IT8689E chip, an RTX 5080), and read here from a copy of the ledger and the Agent's log, not from a probe. **The groups open and read**: nine of ten sessions stored 119 processor series (a clock and a load per thread, temperatures, powers, voltages), 14 memory series and 21 motherboard series (six fans, six temperatures, nine voltages) beside the card's 41 — 195 from the sensor library in all. **No value was compared with another tool**; that is still the owner's. **Two faults, one cause** (`Infrastructure.Telemetry.LhmLibrary`): the Agent held up to three `Computer`s at once, and the library keeps the driver it reads the processor through and its bus mutexes per process — closing one `Computer` let go of them under the others. After seven sessions the Advanced reader's processor device threw `NullReferenceException` twice and was left alone (the page lost the processor until the reader reopened); once a session's start met another instance's close, the reader's `Open` threw `ObjectDisposedException`, and that session stored 17 series instead of 212. **Fixed 2026-11-12**: opens and closes are serialised and a `Computer` is parked, not closed, while another is in use. The cause is read from the log's timing and was not reproduced — it needs this elevation — so **whether the faults are gone is the owner's next sessions to show**: no `advanced: a device is left alone` and no `did not open` in the Agent's log. **First evidence, 2026-11-12, 0.1.0-beta.41, the same PC:** seven sessions between 06:38 and 07:01 with no `advanced:` line at all, each with the option on storing 212 or 213 series; on beta.40 that morning the processor device faulted out after each of six sessions (04:34–05:07) — thirteen of sixteen over the night. The reader failing to open was seen once in sixteen, so this is not yet enough to call that one gone. Still open: any value beside another tool. |
| M16 ✅ | **The L2 telemetry layer's cost — the first number, and it is above NFR-2 (found 2026-11-11, measuring something else).** `LhmTelemetrySource` refreshes LibreHardwareMonitor's graphics-card group once a second in every recorded session. `probe-advanced` measured that refresh, through the same library and the same group, at **14 ms of processor time** on the dev box (RTX 5080, 41 mapped sensors, about twenty of them the Direct3D engine nodes the library enumerates per card) — 1.4% of one core, where NFR-2 allows the whole Agent 1% during capture and `14_TESTING` §Hook overhead's Agent figure has never been measured. This is a measurement of the call, not yet of a session: L2's own poller was not instrumented. | **Measure L2 in a session** (the Agent's processor time over a ten-minute capture, telemetry at 1000 ms and at 2000 ms). Pre-committed: at or under 1% → this row closes with the number; above → the owner chooses between a slower default for `telemetry.interval_ms`, reading the sensors the layer keeps from NVAPI on NVIDIA cards (L3 already has most, at 0.4 ms), or amending NFR-2 to what is measured. Not decided here, and nothing about L2 changes in the Advanced-metrics train. **Measured in a real Agent, 2026-11-11 (owner decision D110), and answered (D117).** `tools/agent-cost.ps1` runs the shipped `FrameLedger.Agent.exe --console capture` against a copy of hook-harness in a scratch data folder — a session that is not hooked, so the telemetry poller runs as in every session and nothing drains frames — and reads the Agent process's own processor time at exit; the cost per second is the slope between a 45 s and a 165 s session, two runs of each. Dev box (RTX 5080, not elevated, layers `l1+lhm+nvapi`): **19.3 ms of processor time per second of session = 1.93 % of one core with Advanced metrics off** (2,312 and 2,312 ms over 45 s; 4,594 and 4,656 ms over 165 s), **20.4 ms = 2.04 % with it on** (2,984 and 3,000; 5,547 and 5,344); peak working set 113 MiB and 119 MiB. A first pass with 30 s and 90 s sessions gave 1.56 % and 2.21 % — its two short runs differed by 670 ms, which is why the lengths were raised. So the probe's number holds in a session: the Agent's steady cost is about twice NFR-2's 1 %, nearly all of it the layer every session has always used, and the new option adds about a tenth of a percent. **The owner's answer: NFR-2 is amended to ≤ 2.5 % of one core** — the limit had never been measured, the sensors keep their one reading a second, and nothing in the code changes. Not measured: a hooked session (the drain's cost on top), a session with the Agent elevated and PawnIO's groups open, another PC. |
| M14 ❓ | **Reflex latency per frame — is there a route within rules 4 and 5? (Train C, 2026-11-11.)** `17_HOOK_ENGINE` §Memory / latency names hooks on `NvAPI_D3D_SetSleepMode` / `SetLatencyMarker` and a call to `NvAPI_D3D_GetLatency`. NVAPI resolves all three through `nvapi_QueryInterface` by id — they are not exports, so a hook on them is a patch at an address learned at run time, the class refused on 2026-08-09. The route to try first needs no hook: the Overlay's own background thread **calls** `GetLatency` with the device of the swap chain it already hooks, once a second (the call returns the last 64 frames' reports). Unknown: whether calling into the driver from that thread is safe on every device kind, what it costs, what it returns when the game never enabled Reflex, and what it means for the Disclaimer's sentence about that thread. | **A spike on `hook-harness` and a stub first, then one real title.** Pre-committed: no route without a patch at a run-time address or a loader-lock call on the game's thread → Reflex stays N/A and the page says why; overhead above NFR-1 → the same. The owner says go or no on the spike's result. **2026-11-12 — the reading that comes before that spike is `spike-notes` §30** (owner decision D130: a document first, beside 0.1.0-beta.45; nothing was run). It finds: a hook on NVAPI's latency functions is the refused class, and a hook on Streamline's Reflex and PCL functions is train D's wrapper (§M15) — and would miss a game that fetched them before the attach, as `slInit` is missed (D129); the Overlay calling `slReflexGetState` itself breaks the thread contract the vendor's header states; **what remains is the route named in this row**, the Overlay calling two driver getters with the game's device, and a no-injection ETW route that gives the game-side markers only and makes the game post messages to its own window. Corrected there: `nvapi64.dll` exports **two** functions on the dev box (`nvapi_Direct_GetMethod` beside `nvapi_QueryInterface`), not one. The go or no is the owner's, and a go needs the owner's ruling that calling a driver getter with the game's device is inside rule 4. **2026-11-13 — the owner answered (D132, D133).** Route D goes to the spike this row names, on `hook-harness` and a stub `nvapi64.dll`, with the ruling: a driver's read-only getter called with the device of the swap chain the Overlay hooks is inside rule 4 (`CLAUDE.md`, the rule's note of that day). No real game before the owner has seen the result, and nothing shipped changes for the spike (`spike-notes` §31, written after 0.1.0-beta.46). Route C is **refused**; the ETW route **stays closed**, and §G is not reopened. **2026-11-13 — the harness spike is done (`spike-notes` §31), and none of its stop conditions fired.** With a stand-in module: nothing is called or loaded when `nvapi64.dll` is absent, a missing getter is refused whole, an error and a report of nothing read as what they are, and a call the driver holds for seconds stops only the thread that made it — so not the watchdog. With the installed driver (32.0.16.2114, RTX 5080), in the probe's own process and with the probe setting the markers a game would: no fault before `NvAPI_Initialize`, without markers, or across 38 million reads beside a marker-setting thread; a call costs half a microsecond to a microsecond and a half; the report is readable from markers alone, in microseconds; and the Direct3D 12 **device** — the object the Overlay already holds — is what the driver takes. What only a game can say is unanswered: a game's own device, a real report, another driver. **The owner's next word is go or no for one real title**; a go is a plan of its own first. **2026-11-13, the same evening — go (D136)**: a plan for one title, Cyberpunk 2077; no code before the plan is approved, and no release carrying it before a stop while a game runs has been exercised on 0.1.0-beta.46 (owner-only item 44 (d)). **2026-11-13, the plan and four more answers (D138–D141).** The read has **no switch of its own**: it is part of Advanced metrics, so after the release it reaches every Direct3D 12 game that loaded the driver's `nvapi64.dll`, for everyone who switched Advanced metrics on — "one title" is the order of the owner's first runs, not a limit in code (D138). The probe ran a leg with a sleep mode first (D139; `spike-notes` §31, *The leg with a sleep*): **the getters do not wait behind `NvAPI_D3D_Sleep`** — thirty-one million reads beside a thread that slept 97 % of the time, none of a millisecond; a read after a sleeping frame costs about 2.6 µs. An `nvapi64.dll` loaded from the game's own folder is refused with the reason (D140). One call of 2 ms or more switches the read off until the game exits (D141). The plan's design, as reviewed against the source, is in [`handoff/history/0.1.0-beta.46.md`](handoff/history/0.1.0-beta.46.md) §The train after 0.1.0-beta.46. **2026-11-13, night — written; released as 0.1.0Line truncated
| M17 | **What the input reader does on a real PC with a real controller — unmeasured (2026-11-14, owner decision D157: built without a probe).** Written and tested against a Windows whose answers the tests write, and run on the developer's PC with no controller connected. Open, each read from the counts a session stores (`InputHealth`) and the Agent's log: (a) **whether XInput answers a program with no window in front** — Microsoft's `XInputEnable` page says the system switches controller input by focus; the reader counts the slices in which a device's own reports changed and XInput said nothing, and falls back to the reports after ten; (b) whether raw input delivers, in the background, the reports of an Xbox pad on USB, on the wireless adapter and on Bluetooth, and of a DualSense on USB and on Bluetooth; (c) whether XInput returns a battery for an Xbox pad on Bluetooth (reported not to); (d) whether a DualSense's reports are laid out as `Domain.Input.DualSenseReport` reads them, and whether its battery is there on Bluetooth with Steam running; (e) whether a controller's own input moves `GetLastInputInfo`'s time (if it does, a controller reads as *both*); (f) **the reader's cost with a controller that sends several hundred reports a second**, against NFR-2's 2.5 % of one core, of which about 2 % is already used — `tools/agent-cost.ps1` measures it with no controller only. **Measured 2026-11-14, the real Agent, no controller connected**: in four sessions of 30 and 90 s the reader took 277 and 834 slices, Windows accepted the registration, XInput's library and the time of the last input were there, and the reader's thread used 15–62 ms of processor time — under 0.2 % of one core in the worst of them. The whole Agent, per second of session: 2.06 % of one core with nothing optional, 1.54 % with input read, 1.61 % with Advanced metrics — three numbers inside the tool's own run-to-run variation (the two 90 s medians with and without input differ by 48 ms), so **the reader's cost is not distinguishable from nothing with no controller connected, and unmeasured with one**; (g) an unelevated Agent beside a game that runs as administrator. Owner-only item 48 (`HANDOFF`) |
| M15 🅓 | **Decided 2026-11-14 (D154): not built. The owner closed Advanced metrics with the refusal standing; the ten entries are listed as not read, with the reason.** ~~**The Streamline state behind `slGetFeatureFunction` — re-examined, not reopened (Train D, 2026-11-11).**~~ The owner asked for the refusal in `handoff/history/p0-queue-1-6.md` to be looked at again. `18_GPU_VENDOR_APIS` §The trains states the route to try (hook the export, hand back a wrapper for a fixed list of names) and what it raises: the wrapper is a write to what the game receives, which rule 4 forbids as written; a wrapper must outlive the Overlay's unhook; a game that fetched its functions before the hook existed is never seen. | **A written design against each of the five grounds, then a spike.** Pre-committed: any ground unanswered → the route stays refused, and `sl.dlssg.*`, `sl.reflex.*` and the rest of that block read N/A with "refused" as the reason. Proceeding needs the owner's amendment of rule 4, asked with the rule quoted. |
| M10 ✅ | ~~PDH `\GPU Engine(*)\Utilization Percentage` summed across engines does not reproduce the Task Manager figure the doc invokes. Decide what we actually report and label it accordingly~~ **Decided 2026-09-09 (PR-E1): L1 does not read the engine counters at all.** `LoadPct` is L2's vendor-reported `GPU Core` load (measured ✓ on NVIDIA, §M5 row R1) and is labelled *GPU load (vendor)* wherever it is shown; on a machine without L2 the field is N/A, never a sum that is not the number the user expects. What L1 does read from PDH is one counter, `\GPU Adapter Memory(luid_…_phys_0)\Dedicated Usage`, bound by LUID — the figure Task Manager labels "Dedicated GPU memory", measured 1301 MB at an idle desktop on the RTX 5080. Reopening this needs a measurement of `max(engine)` against Task Manager's own number, not a decision |
#### §M5 — the decision table, pre-committed 2026-09-03 BEFORE the first run
@@ -1913,10 +1929,11 @@ Each of these is referenced by an existing doc but specified nowhere.
| ~~**Threading model**~~ | ~~Which component owns which thread, UI-thread rules, and how the 1 Hz telemetry poller, 10 Hz drain and pipe reader interact~~ **Answered 2026-09-09 (P2 PR-C): `04_CAPTURE` §Threading model** — four threads, one owner per shared object, the pipe reader named as one more producer. This row was not struck when the section was written; struck 2026-09-13. What is still unspecified is the UI thread's own rules, which is `08_UI`'s in P3 | `04_CAPTURE`, `18_GPU_VENDOR_APIS` |
| **Pipe/shm security** | Stated as policy ("DACL granting the current user's SID") but not as implementable SDDL, nor who creates the objects with what rights. The shm half is built (`fl_shm_host.h`, the Overlay creates the section under the current user's SID); the pipe half is **built 2026-09-13 (P3 PR-1, HANDOFF §P3 decision D11)**: `D:P(A;;GA;;;<user>)(A;;GA;;;BA)` (with the owner `O:<user>` stated since 2026-09-17, `07_IPC` §C) handed to `CreateNamedPipe` with `PIPE_REJECT_REMOTE_CLIENTS`, and the client's token user compared by impersonation on its first frame (`Infrastructure.Ipc.PipeAccessControl`, pinned by `PipeServerClientTests` including the refused stranger). **Row closed** | `07_IPC` |
| ~~**Legal doc versioning**~~ | ~~FR-11 re-shows documents "when a document version increments". Nothing says where the current version lives, who writes `legal_acceptance.version`, or how the markdown is rendered in-app~~ **Answered 2026-09-14 (P3 PR-9):** the version is the document's own `**Version:**` line, read from the copy embedded in the App at build (`App.Services.LegalDocuments`); the App writes `legal_acceptance.version` on the one Accept and re-shows any document whose embedded version differs from its row; ~~the markdown is shown as text with authoring comments stripped, not rendered~~ the markdown is shown with authoring comments stripped and, since beta.12, rendered (`08_UI` §Menu, *Documents are rendered*; §First-run flow built note) | FR-11, `06_DATA_MODEL` |
| **Accessibility / DPI** | NFR-9 states requirements; no design anywhere | NFR-9, `08_UI` |
| ~~**Accessibility / DPI**~~ | ~~NFR-9 states requirements; no design anywhere~~ **Answered 2026-09-15 (P4 PR-8): `08_UI` §Accessibility** — what each clause of NFR-9 means here and the check that holds it; this row was not struck then. Since 2026-11-09 (beta.35, D88) the users' half is `ACCESSIBILITY.md`, and NFR-9 names screen readers and contrast themes too. **Row closed** | NFR-9, `08_UI` |
| **Uninstall / data deletion** | One Velopack clause. No in-app "delete all my data", despite the privacy position | `12_BUILD`, `legal/PRIVACY_POLICY.md` |
| ~~**Cover art**~~ | ~~Appears in the schema, the data directory and the UI; nothing says where it comes from or the licence position on downloaded store art~~ **Answered 2026-10-14 (beta.23, owner decisions D60/D61):** from the user's own image (a copy in `covers\`) or what the Steam and GOG GALAXY apps already keep on this PC — nothing is downloaded, so no store art is redistributed or fetched; the art stays its owners' (`legal/TRADEMARKS.md`), the Privacy Policy 2.11 names the files read. `08_UI` §Games › Game detail. **Row closed** | `06_DATA_MODEL`, `05_DETECTION` |
| **The main line's .NET** *(added 2026-10-31, D73–D76)* | **When does the main line — the betas and the first release — leave .NET 10 LTS, and for what?** The alpha line (`12_BUILD` §The alpha line) builds the same source on .NET 11 to gather the evidence; it decides nothing by existing. What is known: .NET 11 is STS, supported 24 months to about 2028-11, no later than .NET 10 LTS (2028-11-14), so moving buys no support time; Microsoft's supported-OS list for .NET 11 has Windows 10 only in its LTSC and IoT editions, against NFR-8's Windows 10 22H2 floor; C# 15's unions and closed hierarchies would suit the tri-states and refusal reasons, but only once the main line compiles C# 15. **Decision table, written before the evidence:** move to .NET 11 only if (a) at least two alphas on the GA runtime pass the gate and an owner's run of a hooked title with no runtime-specific defect, (b) the owner accepts dropping NFR-8's floor to Windows 11 or ships knowingly unsupported on Windows 10, and (c) something .NET 11 brings is wanted — a measured gain or a language feature; otherwise stay on .NET 10 and take .NET 12 LTS (expected 2027-11) the same way, through the alpha line first | `12_BUILD`, `13_CI_CD`, `02_SPEC` NFR-8, CLAUDE.md pinned stack |
| ~~**The main line's .NET**~~ *(added 2026-10-31, D73–D76; decided 2026-11-11)* | **When does the main line — the betas and the first release — leave .NET 10 LTS, and for what?** The alpha line (`12_BUILD` §The alpha line) builds the same source on .NET 11 to gather the evidence; it decides nothing by existing. What is known: .NET 11 is STS, supported 24 months to about 2028-11, no later than .NET 10 LTS (2028-11-14), so moving buys no support time; Microsoft's supported-OS list for .NET 11 has Windows 10 only in its LTSC and IoT editions, against NFR-8's Windows 10 22H2 floor; C# 15's unions and closed hierarchies would suit the tri-states and refusal reasons, but only once the main line compiles C# 15. **Decision table, written before the evidence:** move to .NET 11 only if (a) at least two alphas on the GA runtime pass the gate and an owner's run of a hooked title with no runtime-specific defect, (b) the owner accepts dropping NFR-8's floor to Windows 11 or ships knowingly unsupported on Windows 10, and (c) something .NET 11 brings is wanted — a measured gain or a language feature; otherwise stay on .NET 10 and take .NET 12 LTS (expected 2027-11) the same way, through the alpha line first. **Decided 2026-11-11 (owner decisions D94–D96), the day after .NET 11 shipped: the main line moves with 0.1.0-beta.37.** Of the table: (a) was **set aside by the owner, not met** — no alpha had run on the released runtime; alpha.1 to alpha.3 ran on Release Candidate 3 with the gate green each time, and no owner's run of a hooked title on an alpha is recorded in this repository; (b) Windows 10 22H2 is not blocked and is stated as untested and unsupported (D95, NFR-8 amended); (c) C# 15 is used from the same release (D96). What stood in for (a): the whole gate on SDK 11.0.100 (`spike-notes` §28) and owner-only item 34. .NET 12 is taken the way this row described — through the alpha line first — and whether the main line follows is asked again then. **Row closed** | `12_BUILD`, `13_CI_CD`, `02_SPEC` NFR-8, CLAUDE.md pinned stack |
| **C# 15 on the main line** *(added 2026-11-11, beta.37, D96)* | **When can the main line use C# 15's constructs?** It compiles them since beta.37 and uses none: ReSharper's engine — InspectCode 2026.2.3.1, which Qodana 2026.2 runs, and 2026.3.0 — parses neither `closed` nor `union` and calls collection expression arguments and labeled jumps preview features, so each file that used one would be an error on Qodana's gate of zero; a union also draws CA1815 from the SDK's own rules; and what CodeQL (documented up to C# 14) extracts of them is not established, a pull request's analysis being an overlay. **Answered by:** the canary of `spike-notes` §29 run again on the engine the Qodana linter then carries, and a `workflow_dispatch` of `codeql.yml` on its branch. The survey of where each construct would go is in that note | `13_CI_CD` §C# 15 and the analyses, CLAUDE.md pinned stack |
| **Xbox app / Microsoft Store titles** *(added 2026-11-03, beta.33, D77–D82)* | GDK titles are imported, played and measured since beta.33 (`spike-notes` §25: full trust, the user's token, the session's object namespace, the image path the `Content` folder's); what is NOT answered: **(1) title lists know Steam ids only (D81).** `blockedStoreIds` has no `store` value for a Microsoft Store product id or a package family name, and the guard has no reader for one, so an Xbox copy of a game listed there by its Steam id is matched by its executables, modules, files, drivers and services but not by that entry. Keying one would take a schema value (`additionalProperties: false` there is a safety control) and a reader under D31's rule — listed refuses, none is not applicable, unreadable refuses — fed by `MicrosoftGame.config`'s `StoreId`, which the import already reads. The owner chose not to, for now. **(2) A Vulkan GDK title is not measured:** the layer is enabled by `FRAMELEDGER_ENABLE_VK_LAYER=1`, which only a launch FrameLedger makes can set, and a package is started by Windows. **(3) A package that keeps write virtualization on** (all 11 measured turn it off) would put the Overlay's log in its `LocalCache`, out of the bug report's reach; capture itself never reads the log. **(4) Unmeasured:** a start from inside the Xbox app (GameLaunchHelper's parent there — the scan set's next process), any GDK title with anti-cheat, the ring from a packaged process (the owner's first hooked session) | `02_SPEC` §Out of scope, `05_DETECTION`, `19_SAFETY` check 3, `LIMITATIONS.md` |
| **Tier-2 mechanism** *(was: PresentMon distribution)* | **By what mechanism does a shipped build reach a no-injection MEASUREMENT?** Nothing names one. Since 2026-08-28 the ladder is two rungs and Tier 2 is explicitly not a measurement — duration, sensors and the reason — so this row is no longer about degraded fidelity but about whether that capability returns at all. **It now owns TWO casualties, not one:** the frame-generation oracle §S31 spent four attempts looking for, and `14_TESTING`'s Tier-1 accuracy cross-check, struck the same day because it required a second instrument measuring the same frames. **One mechanism would restore both**, which is the strongest argument for costing it. Constraints on any answer: `README`'s capture-tier table and its **safety** row *"always a way out"*, `legal/DISCLAIMER.md` §73 and `legal/EULA.md` §33; ETW needs elevation whatever the tool; and `15_ROADMAP` has no v2 fallback either | `README`, `legal/DISCLAIMER.md`, `legal/EULA.md`, `04_CAPTURE`, `14_TESTING`, `15_ROADMAP` |
+168
View File
@@ -0,0 +1,168 @@
# 21 — Input devices: controllers, and keyboard-and-mouse activity
What FrameLedger reads about game controllers and about the keyboard and mouse, from where, and what it never reads.
The numbers a session makes of it are in [`03_METRICS.md`](03_METRICS.md) §Input; the columns are in
[`06_DATA_MODEL.md`](06_DATA_MODEL.md).
> **Status, 2026-11-14 — the Agent records it; the App does not show it yet.** A session recorded with the option
> on, after the terms were accepted, stores what it was played with: the model and the arithmetic
> (`FrameLedger.Domain.Input`, `FrameLedger.Application.Input`, `Application.Recording.InputSessionBuilder`), the
> storage (schema 0024, the `.partial` file's chunk 11 — `06_DATA_MODEL`), the reader (`Application.Input.InputReader`,
> `Infrastructure.Input`) and its composition into the Agent (`capture.input_devices`, `Application.Input.InputTerms`)
> are written and under test. The screens, the store's word and the input libraries come in the pull requests after
> this one, and this note is rewritten by each. **No report from a real controller has passed through any of this code**
> (owner decision D157): the developer has no controller, and what happens on a real one is owner-only item 48.
## The owner's request and decisions (2026-11-14)
The request: whether a game supports controllers, on the game's page; the share of a session played with a controller
and with the keyboard; how often a controller connected and disconnected in a session; the battery state that lasted
longest; switches between makers' controllers and each maker's share; stored, and drawn; Xbox controllers and the
DualSense first.
| | Decision |
|---|---|
| **D155** | Controllers are read **by the Agent, from outside the game**. No hook is added to a game and CLAUDE.md rule 4 is untouched. A session that is not hooked has the numbers as a hooked one does. |
| **D156** | The reading is **on unless switched off** (Settings ▸ Controllers and input, `capture.input_devices`), and **reads nothing until the Privacy Policy that describes it — 2.22 or later — has been accepted on the PC**. An Agent that the logon task started after an update records sessions without it until the App has been opened once. |
| **D157** | **Built without a probe.** Asked whether to run a probe tool with real controllers first, the owner chose to build straight. So every source says what it did (§What the reader says about itself), and each pull request and `LIMITATIONS.md` say what was verified against hand-made data only. |
| **D158** | Keyboard and mouse are **one class, known only by the time of the last input** (`GetLastInputInfo`). No Raw Input registration for a keyboard or a mouse, no hook: no key ever enters FrameLedger's process. |
| **D159** | **Nothing is ever sent to a controller**: no output report, no feature-report request, no `XInputSetState`. A DualSense on Bluetooth therefore has a battery level only when Steam or the game has already switched it to its long report (§DualSense). |
| **D160** | The numbers are **in the CSV/JSON export and in a bug bundle** — they are columns of `sessions`, as the display mode's are — and the Privacy Policy's §3A says so. |
| **D161** | **A session that is not hooked is: duration, hardware telemetry, the reason — and what the Agent reads from outside the game** (the game's memory since beta.12, controllers now). Every field measured inside the game still reads N/A. CLAUDE.md's sentence is amended with the pull request that makes it true. |
| **D162** | Also built: the number of switches between a controller and the keyboard and mouse, and the share of time with no input; marks on the frame-time chart where a controller connected, dropped or ran low; a battery's first and last level and its rate; the kind of connection; a note where Steam Input presented the controller as an Xbox pad; a trend metric, a row in Compare, and "usually played with" on the game's page. |
| **D163** | "Supports controllers" is **three pieces of evidence, never a verdict**: what Steam's own file on the disk says (no network: rule 8), which input libraries the game loaded, and how the game was actually played. From outside a game one can know that a controller was in use while the game was in front — not what the game did with it. |
## The switch, and the terms
`capture.input_devices` — Settings ▸ Recording ▸ *Controllers and input* — is **on unless switched off** (D156), and
is read at each session's start like every setting the Agent reads. On is not enough: the Agent is usually started
at logon, and after an update that is before the App has shown the Privacy Policy that describes this reading. So
`SettingsRecorderPolicy` gives a session the reader only when `legal_acceptance` holds the Privacy Policy at
`InputTerms.MinimumPrivacyVersion` (2.22) or later. Until then the session reads nothing of controllers or of the
keyboard and mouse, stores none of schema 0024's columns, and says so: the slot `input=terms_pending` in
`sessions.capture_notes` (`CaptureNotes.InputHeldForTerms`), and a line in its `.partial` file's notes. An operator's console with no acceptance store reads
nothing either. Off, or held back, a session is byte for byte what it was before the reading existed.
## What is read, and from where
Everything is a Windows API in a library Windows ships. **No third-party library is added.**
| Source | What it gives | Why this one |
|---|---|---|
| Raw Input (`RegisterRawInputDevices`, usage page 1, usages 4 and 5, `RIDEV_INPUTSINK \| RIDEV_DEVNOTIFY`; `GetRawInputBuffer`, `GetRawInputData`, `GetRawInputDeviceInfo`, `GetRawInputDeviceList`) | the reports game controllers send, while the Agent is in the background; a controller arriving and leaving; its vendor and product ids | the documented way to receive a HID controller's reports without being in front, and it opens no handle to the device |
| XInput 1.4 (`XInputGetState`, `XInputGetCapabilities`, `XInputGetBatteryInformation`) | each of its four pads: its state, wired or wireless, a battery in four levels | what an Xbox pad's battery is reported through |
| `GetLastInputInfo` | the time of the last keyboard or mouse input in the user's session | a time and nothing else (D158) |
| `GetForegroundWindow`, `GetWindowThreadProcessId` | whose window is in front | already used for the focus sample (`04_CAPTURE`) |
| Steam's `appcache\appinfo.vdf`, on the disk | whether the store lists controller support for a library game | a local file: no request is made (rule 8) |
**Rejected, with the reason** (the checklist of `18_GPU_VENDOR_APIS` §Checklist applies to any library):
- **Windows.Gaming.Input** — its readings stop when the program is not in front, which the Agent never is.
- **GameInput** — on a PC it is a NuGet package under Microsoft's licence terms with a redistributable to install.
- **Opening the HID device** (`CreateFile`) — a handle of FrameLedger's on a controller can make a game's, or a
remapper's, exclusive open of it fail. Raw Input needs none.
- **A sixth native library** — parsing bytes a device controls in C++, inside an Agent that may run as administrator.
The reader is C#, and what parses a report is `FrameLedger.Domain.Input`, with no pointer in it.
- **Windows' own HID parser** (`hid.dll`'s `HidP_*`), to read *any* controller generically. It would have let a
controller of a maker this build does not know count as *in use* — and nothing here could be tried on one. A
controller whose layout this build does not know is listed, with its connections and drops, and is never *in
use*: a guess would be a made-up number (§The reader).
## What is never read
Which button, where a stick or a trigger is, the motion sensors, the touchpad; which key, what was typed, where the
mouse is; a controller's serial number, its Bluetooth address, its device path. A report is looked at to answer one
question — *is anything on this pad in use?* — and, for a DualSense, to take its battery byte; the answer is kept and
the report is not. `InputSessionBuilderTests` hold the stored JSON to a list of keys, so a key that could carry an
identifier has to be added to a test first.
## Who a controller is
By its USB vendor and product ids only (`Domain.Input.ControllerCatalog`): Microsoft is the Xbox family, Sony
PlayStation, Nintendo Nintendo, anything else *other* — except that a pad Windows presents through XInput is in the
Xbox family whoever made it, because that is all Windows says of it. A model has a name only where the catalog lists
its ids; a wrong name would be worse than none. Two pads of one model on one kind of connection cannot be told apart
across a reconnect, and are not.
## XInput
A state is a pad *in use* when a button is down, a trigger is past `XINPUT_GAMEPAD_TRIGGER_THRESHOLD` or a stick is
outside its `XINPUT_GAMEPAD_*_THUMB_DEADZONE` (`Domain.Input.XInputPad`) — a stick held counts, one that drifts inside
the dead zone does not. Its battery is XInput's words: a wired pad, or four levels with no word for charging.
Two things about XInput are not established and are what item 48 is for: Microsoft's page for `XInputEnable` says
that on Windows 10 and later "game controller input is automatically enabled/disabled by the system based on the
application window focus", which may mean a program in the background reads a neutral state; and a pad on Bluetooth
is reported not to answer `XInputGetBatteryInformation`. XInput also has no documented way to say which device one of
its four pads is.
## DualSense
Sony's 054C:0CE6, and the Edge, 054C:0DF2 (`Domain.Input.DualSenseReport`). Three reports:
| Connection | Report | What it carries |
|---|---|---|
| USB | `0x01`, 64 bytes | sticks, triggers, buttons — and the battery: a level in tenths and what the battery is doing |
| Bluetooth, as connected | `0x01`, 10 bytes | sticks, buttons, triggers. **No battery** |
| Bluetooth, once switched | `0x31`, 78 bytes | the USB report's body, one byte later |
The pad sends the short report until some program asks it for a feature report or sends it an output report; from then
until it reconnects it sends the long one. **FrameLedger never asks** (D159): a game that reads the short report through
DirectInput would lose its input. So over Bluetooth the level is there when Steam or the game switched the pad, and is
not otherwise. Windows hands every report of a device over at the length of the device's longest, so the short report
arrives padded to 78 bytes with the id the USB report has: the connection, not the length, says which layout a report
`0x01` has.
The layout was written from the open descriptions of the protocol. No code was taken from the Linux kernel's driver,
which is under a licence this project cannot take code from.
## The reader
`Application.Input.InputReader` over `Infrastructure.Input.Win32InputBackend`, on the thread of
`Infrastructure.Input.InputDeviceHub` — named `fl-input`, and **there only while a session has a view of the hub**:
with no session there is no thread, no window and no registration.
**A slice, about ten times a second.** The reader takes what Windows has: the controllers that came and went, the
reports that arrived (in bulk — `GetRawInputBuffer` — not one message at a time: a DualSense sends some hundreds a
second, and the Agent's whole budget is 2.5 % of one core, NFR-2), the state of XInput's connected pads, the time of
the last keyboard or mouse input, and the process whose window is in front. From that it makes one `InputSlice`,
the same for every session that is running; each session's view counts it when a window of *its* game was in front
(`Application.Input.InputTickAccumulator`), and the telemetry poller takes the sum once a second.
**The window.** Raw input is delivered to a window. The reader's is a child of `HWND_MESSAGE`: never shown, in no
task bar, receiving no broadcast, answered by Windows' default procedure. It is registered for HID usage page 1,
usages 4 (joystick) and 5 (gamepad) — `InputReader.RegisteredUsages` — and for nothing else.
**What a controller is.** Each of XInput's pads that is connected is one; each controller Windows does *not* present
through XInput is one. A device Windows presents through XInput is also seen by raw input and is **not** a second
controller: it gives the pads their name where that cannot be mistaken — every XInput-class device a player could
hold has the same two ids — and is otherwise only the witness of the next paragraph. An empty XInput slot is asked
about once in twenty slices (XInput looks for a device each time) and at once when a controller arrives or leaves;
a connected pad's battery once in fifty.
**Where XInput says nothing.** If XInput returns a pad at rest for ever because the Agent's window is never in front
(§XInput), the device's own reports are the evidence: a slice in which an XInput-class device's reports *changed*
while no pad's packet number moved — in that slice or the one before — is counted (`XInputSilent`). After ten of
them, **and only if XInput has never once been seen to move**, such a slice counts the lowest connected pad as in
use. That is coarser than XInput's dead zones — a stick that drifts counts — and is used only where the better
source is silent; a session's stored counts say which it was.
**Which controllers are read for use.** XInput's pads, and a DualSense by its own reports. Any other controller —
a DualShock 4, a Switch Pro controller — is listed with its connections and drops and is never *in use*.
**Faults.** A slice that throws is counted and skipped; twenty in a row end the run with one line in the Agent's
log, and the sessions go on without readings. When a run ends, one line says what it did: slices, whether Windows
accepted the registration, reports received and not read, XInput's reads and silent slices, whether the time of the
last input could be read, the thread's processor time, faults. Nothing in it names a device.
**Held by a gate.** `tools/input-readonly-check.ps1` (`12_BUILD` step 14g): the reader's calls are a committed list,
none of which sends to a device, injects input or reads a key; no other managed source and no native source names
an input function.
## What the reader says about itself
`Application.Input.InputHealth`, carried by every reading and stored with the session: whether Windows accepted the
registration, how many reports came and how many could not be read, whether XInput's library was there and how many
states it returned, how often a controller's own reports said *in use* while XInput said nothing, whether the time of
the last input could be read, how many slices were taken and how much processor time the reader's thread used. Counts
and yes-or-no answers; nothing in it names a device.
+808 -62
View File
@@ -72,84 +72,154 @@ owner-only items stay below; the history of every train, the passed owner-only i
| `§beta.30` | [`handoff/history/0.1.0-beta.30.md`](handoff/history/0.1.0-beta.30.md) |
| `§beta.31` | [`handoff/history/0.1.0-beta.31.md`](handoff/history/0.1.0-beta.31.md) |
| `§beta.32` | [`handoff/history/0.1.0-beta.32.md`](handoff/history/0.1.0-beta.32.md) |
| `§beta.33` | [`handoff/history/0.1.0-beta.33.md`](handoff/history/0.1.0-beta.33.md) |
| `§beta.34` | [`handoff/history/0.1.0-beta.34.md`](handoff/history/0.1.0-beta.34.md) |
| `§beta.35` | [`handoff/history/0.1.0-beta.35.md`](handoff/history/0.1.0-beta.35.md) |
| `§beta.36` | [`handoff/history/0.1.0-beta.36.md`](handoff/history/0.1.0-beta.36.md) |
| `§beta.37` | [`handoff/history/0.1.0-beta.37.md`](handoff/history/0.1.0-beta.37.md) |
| `§beta.38`, `§NEXT` (the owner's request of 2026-11-11, the four trains) | [`handoff/history/0.1.0-beta.38.md`](handoff/history/0.1.0-beta.38.md) |
| `§beta.39`, train B's plan as approved (its `§NEXT`) | [`handoff/history/0.1.0-beta.39.md`](handoff/history/0.1.0-beta.39.md) |
| `§beta.40`, what the owner's first sessions with it showed, train B's traps | [`handoff/history/0.1.0-beta.40.md`](handoff/history/0.1.0-beta.40.md) |
| `§beta.41`, what the owner's sessions on beta.40 and beta.41 showed, the plan of 2026-11-12 (D119–D122) | [`handoff/history/0.1.0-beta.41.md`](handoff/history/0.1.0-beta.41.md) |
| `§beta.42`, what the owner's seven sessions on beta.42 showed (D121 answered, the guard's lost launches), beta.43's design as planned | [`handoff/history/0.1.0-beta.42.md`](handoff/history/0.1.0-beta.42.md) |
| `§beta.43`, what the owner's nine sessions on beta.43 showed, session by session (item 41), the traps of that train | [`handoff/history/0.1.0-beta.43.md`](handoff/history/0.1.0-beta.43.md) |
| `§beta.44`, what the owner's eight sessions on beta.44 showed, session by session (item 42), this release's facts, the traps of that train | [`handoff/history/0.1.0-beta.44.md`](handoff/history/0.1.0-beta.44.md) |
| `§beta.45`, what the owner's seven sessions on beta.45 showed, session by session (item 43), that release's facts and SonarCloud's gate on its commit, the traps of that train | [`handoff/history/0.1.0-beta.45.md`](handoff/history/0.1.0-beta.45.md) |
| `§beta.46`, what the owner's five sessions on beta.46 showed (item 44), the design of the train that followed as the plan's review left it (`§The train after 0.1.0-beta.46`), that release's facts, the traps of that train | [`handoff/history/0.1.0-beta.46.md`](handoff/history/0.1.0-beta.46.md) |
| `§beta.47`, what that release decided beyond D136–D141, what item 44's last part showed, that release's facts, the traps of that train, its NEXT as the owner's sessions of 2026-11-14 changed it | [`handoff/history/0.1.0-beta.47.md`](handoff/history/0.1.0-beta.47.md) |
| `§beta.48`, what that release decided beyond D142–D147, that release's facts, the traps of that train, its NEXT and its "Left open" as item 46 changed them | [`handoff/history/0.1.0-beta.48.md`](handoff/history/0.1.0-beta.48.md) |
| `§Owner-only`, an open item | [§Owner-only](#owner-only--no-pr-can-close-these) below |
| `§Owner-only`, a passed item (3, 12, 14–19) | [`handoff/owner-only-passed.md`](handoff/owner-only-passed.md) |
| `§Owner-only`, a passed item (3, 12, 14–19, 38, 39) | [`handoff/owner-only-passed.md`](handoff/owner-only-passed.md) |
| `§Traps`, `§Two corrections carried forward` | [`handoff/traps.md`](handoff/traps.md) |
| `Dnn` | [§Decisions index](#decisions-index) below |
---
## **START HERE** — 2026-11-03, 0.1.0-beta.33: games from the Xbox app and the Microsoft Store
## **START HERE** — 2026-11-14, 0.1.0-beta.49: item 46 read; the frames counted for the log, Reflex and frame generation together; a read a second on a grid; the Agent keeps the reading that waits
*Status is `docs/changelog/0.1.0-beta.33.md`; this is what no other file carries.* The owner: "Trong phiên này, ta sẽ tập
trung xử lý, đánh giá, phân tích cho các game thuộc MS Store/ Xbox App/Game Pass cho (các) phiên bản tiếp theo nhé. Quy
trình vẫn như mọi lần." — handle, evaluate and analyse the games of the Microsoft Store, the Xbox app and Game Pass for the
next versions, the usual train; the owner's games are in `C:\XboxGames` and `D:\xbox`. #351 (the library, the install
root, detection, the art, Privacy 2.13), #352 (Play, the guard's two refusals, one exe-path reader) and this release;
`0.1.0-alpha.2` follows (D82).
*Status is `docs/changelog/0.1.0-beta.49.md`; this is what no other file carries.* The owner ran item 46 on
beta.48 the afternoon it was released — ten sessions, then eight more — and said: check (#455, and item 46's last
paragraphs, below). The read works, the samples are a second apart, both games give both of the driver's own
times, the multipliers are listed, off means off, **and the owner's word: no hitch**. Two things were left, the
owner said *do both*, and four answers given while planning made the train (D148–D151): #456 (the Overlay counts
each summed frame for its log), #457 (the watchdog's periodic tick on a grid), #458 (the Agent keeps the reading
that waits; the guide on whose switches the table shows), and this release. **beta.48's own section, its traps
and its release facts are [`handoff/history/0.1.0-beta.48.md`](handoff/history/0.1.0-beta.48.md).**
**Decisions (asked while planning, do not re-ask):**
### What this release decided, beyond D148–D151
- **D77 — measure first, on the owner's titles**: the launch chain was measured by activating six of the eleven packages
with no FrameLedger hook, read-only, each game closed by its window and never killed, with the decision table committed
before the run (`spike-notes` §25, its two commits in #351).
- **D78 — GDK only**: an app-container (UWP) process is refused with a reason of its own, `TargetAppContainer`, Tier 2
(`19_SAFETY` §The target must be able to take the Overlay); `LIMITATIONS.md` says so.
- **D79 — the platform is "Xbox"** (id `xbox`), full scope: import, art, Play, Tier-1 capture, detection, docs and legal.
- **D80 — Play on an Xbox game always activates its package**, in both D59 modes (`08_UI` §Games): started by its path a
GDK game is never the game in its package — the run corrected the plan's premise that `CreateProcess` fails with error 5
(it hands the launch executable back to the package, so the game is not the launcher's descendant; an Unreal shipping
executable runs outside it). The Agent records such a game in attach mode.
- **D81 — title lists stay keyed by Steam id**: an Xbox copy is matched by everything but that entry; the gap is
`20_OPEN_QUESTIONS` §G *Xbox app / Microsoft Store titles*.
- **D82 — alpha.2 right after beta.33** (the owner runs `0.1.0-alpha.1`, and the alpha's channel is locked): on RC3; if
the train reaches the alpha after .NET 11's release on 2026-11-10, the standing rule applies (SDK 11.0.100, packages
11.0.0) — no new decision.
- **The counts are of each frame's own two numbers, with sums, and no threshold.** The draft counted spans of no
length; a sample whose mean span is half a microsecond says spans of one or two microseconds exist, and a count
of exact zeros would have read "the driver's" where it was "ours". So the frames that have both a span and an
active time are kept in two heaps — active longer than the span, or not — each with its count and both sums.
- **A batch's state is three things**: the driver's low-latency switch, its "the game sleeps" switch
(`bUseGameSleep`: NVAPI's "is `NvAPI_D3D_Sleep()` being called") and its frame-generation multiplier. The second
was added on the owner's word — Reflex goes with frame generation in the owner's games, and the low-latency
switch can be the NVIDIA App's override — and Clair Obscur's two sessions read exactly that: `bUseGameSleep`
off throughout without frame generation, on with ×3.
- **Only the driver read and the publication are the periodic tick's.** The request is still followed on every
tick, so the hooks a session asks for go in on the tick the loader woke — a game that has just loaded
Streamline calls it at once. A draft that moved the whole block would have missed those calls.
- **One read is armed at every periodic tick, and the grid is what keeps two apart**: half a second at the least,
after a tick that ran that late. "At most once a second" is no longer said anywhere.
- **A session's first publication may hold a moment's worth.** A request seen on a woken tick is published by the
periodic tick that follows, however soon. One test read a single publication as complete and had to stop.
- **The Agent keeps two sets, and an empty one does not wait.** A reading can stand at the sample after the one it
would have ridden — a second later. With the sensor interval above a second the oldest is still pushed out.
- **No legal text changed (D149)**, although a bug report carries the Overlay's logs and so these lines: the
owner was told and chose it. The Privacy Policy names the logs of the component loaded into games.
- **The plan was reviewed against the source by a second reader before the owner saw it, and sixteen things in
the draft were wrong** — among them the three above, and the whole of the Agent's half, which the draft did not
have.
The Privacy Policy 2.13's wording was approved with the plan (D36).
### Traps this train left
**Choices made here:**
- **A cause I offered and then counted away**: "the holes are the sensor library's fifth tick" — 25 of 134 fell on
one. The pattern that was there (46 of 134 exactly five samples apart) was the Agent's own attach rule, found
by reading `SessionRecorder`, not by looking harder at the Overlay.
- **A fixture that proved nothing**: the case for a woken watchdog first churned `sl.common.dll`, which wakes
nobody — it is in neither the inventory nor the census — and passed. It was caught only because the case
asserts, from the log, that the watchdog *was* woken (0 early ticks). **A test of a mechanism asserts that the
mechanism ran.** `opengl32.dll` is the wake that unloads nothing under a hook.
- **Qodana's native report reads the tests**: `static_cast<unsigned long long>(a * b)` is
`bugprone-implicit-widening-of-multiplication-result`. Widen an operand, not the product.
- **A here-document with an apostrophe in it failed again.** Edit scripts are files.
- **No edit while a gate runs still holds when CI asks for one**: the Qodana fix went into a second worktree of
the lower branch, and reached the upper one by a merge.
- **The item's own text said "nothing is lost by it" of the holes**, on a reading of the Overlay alone. A reading
was lost at every return to step. Say what was looked at.
- **A GDK package root is an install-root boundary**, for the guard's check 4 and for detection alike, and only where no
store boundary matched — so it can only widen what is walked (`19_SAFETY` §Pre-injection checks, amended 2026-11-03).
The guard gained `GamesInstalled\<X>` with it, and one case list read by both test suites
(`tests/fixtures/install-roots/cases.txt`) holds the two resolvers together.
- **`root_file`, a new detection signal**, answered from the install root's own listing, so the `xbox` platform rule —
first, because a GDK FINAL FANTASY ships Steam's DLL — is decidable even where a very large install's walk ran short.
- **The two refusals read the target once, before the primitive**, never at the 30 s re-scan; "could not tell" refuses as
`ProcessUnreadable`, and only enforcing flags count (`SetContextIpValidation` without strict mode is on every
CET-compatible process: `pwsh.exe` reads `0x105`).
- **`TargetResolver` reads the watcher's spelling** (`ProcessImagePath`) but keeps the guard's rights, so D21's
`TargetUnreadable` is unchanged.
### NEXT
**Measured on the way** (`spike-notes` §25): eleven GDK packages, full trust, the user's token, medium integrity, none of
the four mitigations; the image path and the loader's main module both in the `Content` spelling; every game `.exe`
refuses `FILE_READ_DATA`/`FILE_EXECUTE` outside its package, so no Unreal version and no *Runs as* for these titles;
`GetProcessMitigationPolicy` answers through `PROCESS_QUERY_LIMITED_INFORMATION`, and an unknown policy answers
`ERROR_INVALID_PARAMETER`.
1. ~~**The owner runs item 47**~~ **Run and read 2026-11-14** (item 47, *Run 2026-11-14*): the difference in
Cyberpunk 2077 is the driver's, the holes are gone from steady play, off means off, and the stop fired once.
2. **The two GPU labels were reworded on the owner's word** (D152, 2026-11-14; strings in three languages, the
guide — in `[Unreleased]`, not in a release yet). **D141's stop stays as it is, and nothing more of Reflex is read** (D153, the owner
the same evening: "it only has to tell whether the game has it or not; Kingdom Come: Deliverance II has no
Reflex in its settings" — so that session lost nothing to the stop). (f) and (g) are said: the App's *Low
Latency Mode* is off, and no hitch.
3. **`slAllocateResources`** keeps its patch (D135); looked at again when there are more sessions.
4. ~~**Train D**~~ **Advanced metrics is closed (D154, 2026-11-14): train D is not built.** The owner was shown
the catalog — 140 entries, 110 read — and what the ten entries of reach `Spike` hold (Streamline's own headers,
2.14.16: DLSS frame generation's options and state, the video-memory estimates, ray reconstruction's, Image
Scaling's and Dynamic Vibrance's options, the PCL markers), with what reading them takes: a function of ours
handed to the game in place of Streamline's (rule 4), which a game that fetched its functions before the
attach never calls. The owner: "Đóng luôn". The ten are `RefusedRoute` now, the guide no longer says a later
release reads more, and §M15 is decided. **What is next is the controller feature**, planned in a session of
its own.
**Traps met on the way:**
### Left open by this session
- **Malwarebytes quarantines `fl_guard_test.exe`** (`Malware.Heuristic.108256`, records in
`C:\ProgramData\Malwarebytes\MBAMService\RtpDetections`) on some builds and not others: ctest then reports `fl_guard` as
passed in 0.5 s with no output and `fl_prescan` with an empty command. Read `LastTest.log` for the assertion count before
believing a native pass; an exclusion for `E:\frameledger\build\` is the owner's to add (a security setting), and CI runs
both suites.
- **`build.ps1` finds the x64 clang-format; `Get-ChildItem … clang-format.exe` finds the ARM64 one first**, which will not
start on this PC.
- **SonarCloud's quality gate on #351 was 77.7 % new-code coverage** (≥ 80 % wanted) and one `cpp:S5813` (`wcslen`): both
closed in #352 — its own gate 95.4 %, main's after the merge 86.5 % on the code since beta.32 with no open issue. The
job is advisory, so a merge on green does not wait for it — read its verdict after each merge.
- **FluentAssertions' `OnlyContain` fails an empty collection**: a test of what this PC holds (`GamingRoots.OnThisPc()`)
passed here and failed on the hosted runner, which has no `.GamingRoot` (#352's first CI run) — assert that nothing
is wrong rather than that everything is right.
**Owner-only:** item 28 — the Xbox titles on real hardware.
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-14: 0.1.0-beta.49 was released 2026-11-14T11:12:44Z** (18:12 in the owner's time zone) from the signed
tag on `afe4984` (#459), by run 43587336452 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed, and the portable package hashed
here equals its line; the five native libraries in it carry `FL_BUILD_ID v0.1.0-beta.49`, none of them
`-dirty`. The delta from beta.48 is 1 100 211 bytes beside a full package of 110 803 443 — "Delta processed
0452 files. 0023 patched, 0429 unchanged". The wiki was published by run 43588972123. **Two rehearsals on the
release branch**: run 43580081466 at `1ab2809`, and — after a test-only fix reached the branch — run
43581734484 at `9f94356`, whose tree is the released one (a delta of 1 134 949 bytes, the same file counts).
**SonarCloud analysed `afe4984`: the quality gate passed and no issue is open on `main`. Qodana's two reports,
CodeQL and `check` are green on that commit.** On the way two advisory checks did find something, each in a
test of this train and each fixed before its pull request merged: Qodana's native report on #456 (two
products widened after the multiplication), SonarCloud on #457 (a line of the log handed to `printf`). **The
train stopped once for the owner**: two signed merges timed out at the passphrase prompt after #457; it went on
when the owner had unlocked the key. The tag's message reads `FrameLedger 0.1.0-beta.49` where the tags before
it read the bare version — nothing reads it, and a pushed tag is not moved.
- **Why Cyberpunk 2077's active render is longer than its span without frame generation** — item 47's to say. A
third session that afternoon repeated it (233: 3.40 against 3.13 ms at 292 of 296 samples) and a session of
menus did not (232: 0 of 126).
- **The Agent's half of the holes is read from the code and fits the ledger; it is not measured.** And with both
sides on a second a slip of phase can still leave a sample without a report: item 47 counts them.
- **A game that exits by itself loses the batch of counts in hand** — under a minute. Not changed: nothing runs
at a process's exit.
- **Where the third of a second of a poll goes is not established** (the history file of beta.48 has what is
known). Not looked at in this train.
- **With the sensor interval above one second, some readings of the driver's report are not stored** — the oldest
of the two that wait is pushed out. Said in the guide; not changed.
- **No test runs the whole path through the real Agent**: the injected tests stop at the shared memory, the
sampler's and the pending queue's start from it.
- **Two install batches of beta.46 took 0.61 s and 1.26 s and nothing explains them.** Item 46's ten sessions had none
above 191 ms.
- **The Japanese strings are unreviewed**, as every `ja` string is; this release added none.
- **(c) and (h) of item 43**, which only the owner sees, are still not exercised.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed).
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** item 47; one session each of Valor Mortis and Black Myth: Wukong with hooking off; what item 41
left unexercised; item 40 (d).
## Owner-only — no PR can close these
*The items that passed — 3, 12, 14–19 — moved verbatim to [`handoff/owner-only-passed.md`](handoff/owner-only-passed.md) on 2026-10-19 (D66); the numbers are kept, and a number is never reused.*
*The items that passed — 3, 12, 14–19, and 38 and 39 on 2026-11-12 — moved verbatim to [`handoff/owner-only-passed.md`](handoff/owner-only-passed.md) (the first eight on 2026-10-19, D66); the numbers are kept, and a number is never reused.*
1. **§S23-2 — branch protection.** `Rules / validate` is not a required status check on
`main`, so the gate that makes the anti-cheat blocklist un-removable is advisory.
@@ -249,7 +319,9 @@ refuses `FILE_READ_DATA`/`FILE_EXECUTE` outside its package, so no Unreal versio
setting*, the game's own library and no generation beyond what was observed. Record the names the override loads
under `%ProgramData%\NVIDIA\NGX` (the summary's *Loaded:* line) — they correct `05_DETECTION` §Shipped is not loaded.
(d) **DLSS 5** on the owner's DLSS 5 title: *DLSS 5 runtime loaded*. If it is not shown, record the module names it
loads (*Loaded:* line); `nvngx_dlssnr.dll` is then corrected in `CensusNames` and `05_DETECTION`.
loads (*Loaded:* line); `nvngx_dlssnr.dll` is then corrected in `CensusNames` and `05_DETECTION`. *(2026-11-06, `spike-notes` §26:
Hogwarts Legacy, Neural Rendering on, loaded `nvngx_dlssnr.dll` 310.9.16.0 from the game's folder — the name holds;
whether it ran is what beta.34 measures, item 30(a).)*
(e) **SonarCloud (D67):** bulk-mark per `13_CI_CD` §SonarCloud; optionally its *Ignore issues on multiple criteria*
rows; close PR #315. (f) The Legal Gate shows **Disclaimer 2.11** and **Privacy Policy 2.12** once. (g) The `ja`
strings marked `review`. *(2026-10-19, the owner's first runs — `spike-notes` §22: (a) the Dynamic context line
@@ -346,7 +418,600 @@ refuses `FILE_READ_DATA`/`FILE_EXECUTE` outside its package, so no Unreal versio
`Art_Source_Xbox`, `Import_NothingFound`, `FirstRun_Import_Body`, `Settings_GameArt_Body`, `Play_Xbox_Format`,
`Play_Xbox_NotStarted_Format`, `Play_PackagedGame_Format`, `Play_ToolTip_Store`, `Play_ToolTip_Direct`,
`Settings_PlayMode_Body`, `Guard_TargetAppContainer`, `Guard_TargetMitigation`. Then
`docs/LIST_GAME_TESTED.md` gains the Xbox titles from a ledger copy (D35), approved before the commit.
`docs/LIST_GAME_TESTED.md` gains the Xbox titles from a ledger copy (D35), approved before the commit. *(2026-11-06,
beta.34: done — inKONBINI and Valor Mortis, with the refresh the owner approved with the plan.)*
29. **0.1.0-alpha.2 on real hardware** (D73, D82). (a) **The update from alpha.1**: Help ▸ Check for updates offers
alpha.2 — the first update an alpha's updater has found — and *Restart to update* installs it; say whether it
downloaded the delta or the whole package. (b) The terms window shows the Privacy Policy 2.13 and the Alpha notice
with its box again, and Accept waits for the tick. (c)
`%LOCALAPPDATA%\FrameLedger\backups\ledger-before-0.1.0-alpha.2.db` exists beside alpha.1's, and opens. (d) Item 28's
Xbox checks, on the alpha.
30. **beta.34 on real hardware** (D83–D86). (a) **The update from beta.33** (or alpha.3's, item 31): the terms window
shows the Disclaimer 2.13 and the Privacy Policy 2.14 once. (b) **Hogwarts Legacy with Neural Rendering on**, one
hooked session → *DLSS 5 (neural rendering on N % of rendered frames)*; then **off** — `spike-notes` §26's table,
written before the run, decides (on and 0 %: neural rendering runs outside `slEvaluateFeature`, the line stays
*runtime loaded, not seen running*, and the session's `sl_tag_census` is recorded before anything is named).
(c) **Phantom Blade Zero** under the NVIDIA App's *Dynamic, up to ×3* → *DLSS 4.5 (Dynamic multi frame generation, set
in the NVIDIA App)*; the game's own ×3 with no override from the App → *DLSS 4*. (d) **Games** from any page: the
cards at once, a second visit at once; the Debug line `ui: page … shown … ms after its navigation`. (e) **Settings →
change the language** → the window comes back on Settings. (f) The filters on Games and Compare, the Compare columns
and selection bar, the selected page's label in the pane, in both themes. (g) The 34 `ja` strings this release adds,
all marked `review`: `Dlss_Basis_Dynamic`, `Dlss_NeuralRendering_NotSeen`, `Dlss_Feature_NeuralRendering`,
`Dlss_Basis_NeuralRendering_Format`, `Games_Filter_*`, `Games_Header_Filtered_Format`, `Filter_Clear`,
`Compare_Filter_*`, `Compare_Count_Format`, `Compare_Count_Filtered_Format`, `Compare_ClearSelection`,
`Compare_MidSession_Changed`, `A11y_Games_Filter_*`, `A11y_Compare_Filter_*`.
31. **0.1.0-alpha.3 on real hardware** (D73, D87). (a) **The update from alpha.2**: Help ▸ Check for updates offers
alpha.3 and *Restart to update* installs it; say whether it downloaded the delta or the whole package. (b) The terms
window shows the Disclaimer 2.13, the Privacy Policy 2.14 and the Alpha notice with its box again, and Accept waits for
the tick. (c) `%LOCALAPPDATA%\FrameLedger\backups\ledger-before-0.1.0-alpha.3.db` exists beside alpha.2's, and opens;
the ledger it left is at schema 22. (d) Item 30's checks, on the alpha.
32. **beta.35 on real hardware** (D88–D91). (a) **The update from beta.34** (or alpha.3's): no terms window, the ledger
unchanged. (b) **Narrator**: a safety notice is read as it appears; the agent's banner going and coming; Tools ▸
Database maintenance ▸ Check reads its result; a short message at the bottom is read; **H** moves between each page's
title and its sections; Settings ▸ Theme's options read their words; the Sessions table's cells read their values;
Tab to a chart says its name and where its numbers are, and its arrows, Ctrl+arrows and A move it without scrolling
the page; a double-click shows all of it. (c) **Contrast themes**: Aquatic and Desert, and one with an edited colour
(Windows 10 if at hand) — every page, a dialog, the title bar's agent pill, a short message, a chart; turned on and
off with FrameLedger open, it follows at once; Settings ▸ Theme shows its line. (d) **Ctrl+1 … Ctrl+5** put the
keyboard in the page and Narrator says its title. (e) **Animation effects off** (Settings ▸ Accessibility ▸ Visual
effects): no page slides in. (f) **Remove** on a game's page: the box opens over the window, its Cancel puts the focus
back on Remove; the light theme's agent pill and a *done* message read clearly. (g) Help ▸ Accessibility shows the
statement. (h) The 35 `ja` strings this release adds, all marked `review`: `Menu_Help_Accessibility`,
`Accessibility_Title`, `Settings_Theme_ContrastOn` and 32 `A11y_*` (the names, the chart names and keys, the
announcements).
33. **Qodana, and beta.36 on real hardware** (D93). (a) **The update from beta.35**: no terms window, the ledger
unchanged. (b) **The re-pointed bindings, by hand** — the tests pin their text, not a click: on Games, a card opens
its game and both Play buttons (card and list) start it, greyed in a viewer; on the Dashboard a recent session
opens; on a game's page a session row's *Open* opens its summary; in a session's summary a Yes/No/N/A chip opens the
override dialog; a safety notice's button dismisses it; Settings lists the hooked games with *Revoke*, the
exception rows with *Grant*/*Withdraw*, and the system information lines. (c) **Qodana Cloud**: the trial's end
changes nothing in CI (the Community linter needs no licence; the token names the project the report goes to). If
an Ultimate licence is wanted — JetBrains has a programme for open-source projects — `13_CI_CD` §Qodana says what
moves. (d) **`qodana` as a required check, or not**: advisory today, like `sonarcloud`. (e) **The members nothing
uses** (`13_CI_CD` §Qodana, 34 and 49 getters): say which are dead, and a pull request removes them.
34. **beta.37 on real hardware — the first build on .NET 11's release** (D94–D98). (a) **The update from beta.36**:
Help ▸ Check for updates offers beta.37 and *Restart to update* installs it — a download of about 70 MB, not a few;
the terms window shows the **EULA 1.7 and the Disclaimer 2.14** once and nothing else; the ledger stays at schema 22.
(b) **A hooked session on .NET 11.0.0**: one title you measure often, Tier 1 — frame times, the upscaler and frame
generation as before; the Agent's log opens with `release line main, .NET 11.0.0`, and Help ▸ About reads *Runs on
.NET 11.0.0*. This is the run `20_OPEN_QUESTIONS` §G's table asked for before the move and did not get (D94).
(c) **The Agent in admin mode and the logon task** still start it — the publish is self-contained on a new runtime.
(d) **Windows 10 22H2, if a PC or a VM with it is at hand**: does it install and start? No version has ever been run
there; D95 states it as untested, and one run would be the first evidence either way. (e) **The Japanese** of the
three alpha strings that changed or are new (`Alpha_Banner_Body_Format`, `Alpha_Banner_Body_Resting_Format`,
`About_Runtime_Alpha_Format`) and of the first-run summary's first line. (f) **Going back** — only if something is
wrong: install beta.36 over it; the ledger opens.
35. **alpha.4 on real hardware — the last alpha for now** (D98). (a) **The update from alpha.3**: Help ▸ Check for
updates offers alpha.4 and *Restart to update* installs it — a large download, the runtime moving from the release
candidate to 11.0.0. (b) The terms window shows the **EULA 1.7, the Disclaimer 2.14 and the Alpha notice 1.1** and
waits for the notice's box; `…\backups\ledger-before-0.1.0-alpha.4.db` exists beside the older copies, and opens;
the ledger stays at schema 22. (c) **The banner says the alpha line is paused** and to install the beta — not the
experimental-build warning — and Help ▸ About reads *Alpha build — runs on .NET 11.0.0*, with no "prerelease".
(d) **Then install beta.37 over it**: the installer asks to replace it; the beta opens the ledger, shows no Alpha
notice, and its update channel is no longer locked.
36. **beta.38, and the two Qodana reports** (D99, D100). (a) **The update from beta.37**: Help ▸ Check for updates
offers beta.38 — a small download — and *Restart to update* installs it; no terms window. (b) **Links still open**:
the documentation from the Help menu, a document's page on GitHub from its window, a link inside the user guide
and inside the update notes, and the source link in Help ▸ About each open in the browser, as before. (c) **Qodana Cloud**: the projects `frameledger` and
`frameledger-cpp` each show a report from `main` with no problem. A job's log says *Qodana license plan: Ultimate*
and the day the subscription is active until; when it lapses both jobs go red — advisory, never a blocked merge —
and `qodana.yaml` says the way back. (d) **`LogNoteW`**: the Overlay has one function nothing calls
(`src/native/FrameLedger.Overlay/src/dllmain.cpp`). Say whether it is removed — a pull request of its own, eleven
lines — or kept; either way the `.editorconfig` section that names it is then removed or reworded.
37. **beta.39, Advanced metrics on real hardware** (D101–D109). (a) **The update from beta.38**: the terms window
shows the **Privacy Policy 2.15** and the **Disclaimer 2.15** once; the ledger moves to schema 23 and every
earlier session opens as before. (b) **The switch**: Settings ▸ Recording ▸ *Advanced metrics* on — **Advanced**
appears as the last item on the left at once, **Ctrl+6** opens it; off — it goes, and Ctrl+6 does nothing.
(c) **This PC, on the RTX 5080**: the facts (firmware, cores, memory sizes, the display's HDR and variable
refresh), the *Right now* table beside the NVIDIA app or HWiNFO — the first comparison of any of these values
with another tool — and the chart moving; a fan's *Read at* a few seconds behind the memory's. Leave the page: the
Agent's log says the reader closes some fifteen seconds later, and opens again when the page does. (d) **A
session with a real game, the option on**: it is on *Session* with its statistics and its chart; its summary has
**Advanced metrics…** and the button opens it there; *variable refresh was active* and the colour space against
what the game and the display really did; the driver's settings for the game against the NVIDIA app. A session
recorded with the option off is not on the list. (e) **As administrator, with PawnIO installed** — the first run
there has ever been (`20_OPEN_QUESTIONS` §M10, §M13): the processor per core, the memory and the motherboard's
sensors appear on *This PC* and *Sources* says *Read*; watch for a pause or a hang when the page first opens,
and compare a few values with HWiNFO. Without it, *Sources* gives the reason in words. (f) **What it costs**: the
Agent's processor time and memory during a session with the option on and with it off (NFR-2: ~~1 %~~ 2.5 % of one
core since D117, 150 MB) — ~~and with that in hand, the answer to §M16~~ §M16 was measured without a game and
answered on 2026-11-11; what is left for a real game is the hooked session and the elevated one. (g) **The words**: the `vi` strings read as a Vietnamese
speaker would write them; the `ja` strings marked `review`. (h) **A copy of the ledger** (`--data-dir`) that holds
such a session: the page is there, opens on *Session*, and *This PC* says it is a copy.
40. **beta.42 — the log that answers D121, the label, the plane at infinity** (D119–D122). (a) **The update from
beta.41**: Help ▸ Check for updates, *Restart to update* with no game running; no terms window. (b) **Cyberpunk
2077 with FSR upscaling** (FSR 3 or 4, with or without FSR frame generation — as it was when the upscaler read
"unknown"): one session of a few minutes of play, then **quit the game from its own menu**. Nothing else is
needed: `%LOCALAPPDATA%\FrameLedger\logs\overlay-<pid>-….log` of that session is what is read. If the session
now names FSR, say so — nothing in this release should have changed that. (c) **A game that gave one depth
plane** — Lies of P, Clair Obscur or Dying Light: The Beast, FSR upscaling, **Advanced metrics on**: the
session's Advanced page shows both planes, one of them "∞". If it still shows one, the missing plane is not
`FLT_MAX` and the log of (b)'s kind says nothing about it — say so. (d) **A session that changes upscaler** —
Dying Light: The Beast as it was (DLSS evaluated, then FSR), or any game switched between DLSS and FSR in its
menu during one session: the summary names the one that ran longest, with that one's quality and render size,
and says *settings changed during session*. (e) **Any hooked game, ended by quitting it**: its overlay log has
lines after the first second — `HOOK_INSTALLED` with a module and a function, `RUNTIME_CENSUS` — and no
`PATCH_FAILED`; if there is one, its `a=` number is the reason.
**Four sessions were run on 2026-11-12 and read the same morning** (START HERE §What the owner's first four
sessions on beta.42 showed): (a) happened — the sessions carry the beta.42 build; (b) read, and D121 answered:
the session upscaled with FSR 2.1, built into the game; (c) Lies of P and Dying Light show `cameraNear = ∞`;
(e) all four logs are whole and none has a `PATCH_FAILED`. **(d) was not exercised**, and the owner has not
said the item passed.
41. **beta.43 — the first real games with the two new hooks, and the guard's re-read** (D123–D125). **Not first on
a game under the user-mode exception's trial** (D38): a crash there ends the exception for good. (a) **The
update from beta.42**: Help ▸ Check for updates, *Restart to update* with no game running; **the terms window
shows once, with Privacy Policy 2.17**. (b) **The guard**: over the games played, the Agent's log
(`logs\agent-<date>.log`) has no session that ended at a game's launch with `guard=ModuleScanFailed`. If one
does, its signal now says which kind — *kept changing while it was read* (four reads were not enough) or *came
back incomplete* — say which. (c) **A Streamline game with Advanced metrics on** (Dying Light: The Beast, or
Cyberpunk 2077 with DLSS): a session of a few minutes, quit from the game's menu. Its Advanced page shows the
Streamline rows — jitter, motion vectors, depth, projection, resets — and the line under the session says
*read*. **Read them for sense**: a field of view between about 0.3 and 2.5 rad, an aspect ratio near the
screen's, a jitter under 1, planes that are not both zero. Values that make no sense mean the game's struct is
not laid out as the one FrameLedger was built against: say so, with the numbers, and do not run further games.
(d) **FSR frame generation on** (Cyberpunk 2077 or Lies of P), Advanced metrics on: the page shows the
frame-generation rows — on, asynchronous or not, the flags, the rectangle — and the line says *read*. (e) **The
three answers of the line**: a game with neither library says *not hooked* for both; a game where it says
*hooked — the game did not make the call* for a library the game plainly uses is a finding, not a pass — say
which game. (f) **Off means off**: with Advanced metrics off, **start a game afresh** and record a session; it
has no Advanced button, and its `logs\overlay-<pid>-….log` has no `HOOK_INSTALLED` line naming `slSetConstants`
or `ffxConfigure`. (A game that was already running when a session with the option on was recorded keeps the
two patches until it exits — that is by design, and its log will have the lines.) (g) **Nothing else moved**:
no crash, no `PATCH_FAILED` in any overlay log, frame rates and the frame-generation factor as on beta.42 for
the same game and settings.
**What each outcome means, written before the run.** (c) or (d) empty with *not hooked*: the log of that
session says why — no `HOOK_INSTALLED` for the function (the library is a 1.x interposer or lacks the export)
or a `PATCH_FAILED` (its `a=` number is the reason). Empty with *hooked — the game did not make the call*: the
game reaches Streamline or FSR by a way that does not pass the export — beta.44's plan needs to know. A crash,
or a game that will not start: say so at once; *Disable all hooking* (Settings ▸ Capture) and switching Advanced
metrics off are the two ways to run without the hooks, and which of them cured it says which hook it was.
**Nine sessions were run on 2026-11-12 and read the same afternoon** ([`handoff/history/0.1.0-beta.43.md`](handoff/history/0.1.0-beta.43.md) §What the owner's nine
sessions on beta.43 showed): (a) happened — Privacy Policy 2.17 accepted once, at the update; (b) held — no
`guard=ModuleScanFailed` since the update; (c) read on Phantom Blade Zero, Cyberpunk 2077 and Hogwarts Legacy,
every value sane but one: **Cyberpunk 2077 gives its field of view in degrees** (51–90), and the page labels it
radians — the struct is laid out as expected, the label is wrong; (d) read on Lies of P, **not exercised through
the FidelityFX 2.3 loader**; (e) the three answers were each seen and none is a finding, **a game with neither
library was not run**; (f) held on Lies of P started afresh; (g) no `PATCH_FAILED`, no fault, frame rates as
before — Lies of P's 35 s end with exit code 3 is, by the owner, the game's own crash after a shader failed to
compile, and Hogwarts Legacy's fault at exit is in Windows' log six times since 2026-11-06, hooked or not.
**Judged passed 2026-11-12, on the owner's word to go on if it did**; the two parts in bold stay open.
42. **beta.44 — the two questions about a feature in real games, and what the nine sessions found, fixed**
(D126–D128). **Not first on a game under the user-mode exception's trial** (D38): a crash there ends the
exception for good. (a) **The update from beta.43**: Help ▸ Check for updates, *Restart to update* with no game
running; **the terms window shows once, with Privacy Policy 2.18**. (b) **Cyberpunk 2077 with Advanced metrics
on**, a session of a minute or two: its Advanced page shows the field of view as the game gives it —
`51.0055 – 90`, or whatever the game's setting makes it — with a line under it that says *Not radians*, and no
"rad" beside the number. The two sessions of 2026-11-12 keep `51.0055 – 90 rad (2922.4° – 5156.6°)`: they were
stored that way. (c) **Any Streamline game with Advanced metrics on** (Phantom Blade Zero, Hogwarts Legacy,
Cyberpunk 2077): the line under the session has one more sentence — *Which version each Streamline feature is:
…. What each requires: ….* — and for each it says *read*, or *hooked - the game asks this while it starts, and
did not ask again after FrameLedger attached*. **The second is the expected answer** for a game that was already
past its start-up when FrameLedger attached; it is not a failure. If it says *read*: the page lists features by
NVIDIA's names (`kFeatureDLSS`, `kFeatureDLSS_G`) with versions and requirements — check that the driver version
it shows as *detected* is the installed driver's, and that Streamline's version is the game's `sl.*.dll` file
version. (d) **Which games, if any, say *read***: nothing to set up — FrameLedger attaches as soon as it sees the
game — but say, for each Streamline game run, which of the two answers each call got. Hogwarts Legacy loaded its
Streamline library some six seconds after FrameLedger attached on 2026-11-12, so it is the likeliest to say
*read*; Phantom Blade Zero and Cyberpunk 2077 had it loaded already. (e) **A game with no Streamline** (Lies of P, Kingdom Come:
Deliverance II): both say *not hooked*. (f) **Off means off**: with Advanced metrics off, **start a game
afresh** and record a session; its `logs\overlay-<pid>-….log` has no `HOOK_INSTALLED` line naming
`slGetFeatureVersion`, `slGetFeatureRequirements`, `slSetConstants` or `ffxConfigure`. (g) **Nothing else
moved**: no crash, no `PATCH_FAILED` in any overlay log, frame rates and the frame-generation factor as on
beta.43 for the same game and settings. With Advanced metrics on, a Streamline game's log has
`slGetFeatureVersion` and `slGetFeatureRequirements` as its **first two** Streamline `HOOK_INSTALLED` lines.
(h) **The count of early ends** (D126) is not something to provoke: if a game whose hooking you switched back
on ends once within a minute of starting, hooking stays on; a second such end switches it off, as before.
**What each outcome means, written before the run.** (c) *not hooked* on a Streamline game: the log says why —
no `HOOK_INSTALLED` for the function (a 1.x interposer, or the export is missing) or a `PATCH_FAILED`. (c) and
(d) both *did not ask again* on every game: the calls are made before any attach can be in — then these two
hooks give nothing in attach mode, and whether they are worth their two patches is the owner's to say before
train C. *Read* with numbers that make no sense (a driver version that is not the installed one, a feature
named `feature <number>`): say so with the numbers — an id with no name only means NVIDIA added a feature this
build does not know. A crash, or a game that will not start: say so at once; *Disable all hooking*
(Settings ▸ Capture) and switching Advanced metrics off are the two ways to run without the hooks, and which
of them cured it says which hook it was.
**Eight sessions were run on 2026-11-12 and read the same evening** ([`handoff/history/0.1.0-beta.44.md`](handoff/history/0.1.0-beta.44.md) §What the owner's
eight sessions on beta.44 showed): (a) happened — Privacy Policy 2.18 accepted once; (b) held — Cyberpunk 2077 reads
`51.0055 – 90` and *not radians*; (c), (d) **both questions read on five games, six sessions of six** — the
expected answer was the other one — with the driver, Windows and NGX versions agreeing with the PC, and one
reading no file could give: frame generation's Streamline plugin is 2.14.16 on three games that ship an older
one; Cyberpunk's version question was *not asked again*; (e) held on Kingdom Come: Deliverance II; (g) no
`PATCH_FAILED`, no fault, the two rows first; (h) seen for real — Valor Mortis's second early end switched
hooking off, the owner switched it on, the count is 0. **(f) was not exercised.** Valor Mortis crashed 51 s
into a hooked session exactly as on beta.40, and Black Myth: Wukong faults in Steam's crash handler at every
exit: neither is explained, and neither began with this release. **Taken as passed on the owner's "update and
go on", 2026-11-12**; (f) and the two comparisons stay open.
43. **beta.45 — six more Streamline calls in real games** (D129–D131). **Not first on a game under the user-mode
exception's trial** (D38): a crash there ends the exception for good. (a) **The update from beta.44**: Help ▸
Check for updates, *Restart to update* with no game running; **the terms window shows once, with Privacy Policy
2.19**. (b) **Any Streamline game with Advanced metrics on** (Phantom Blade Zero, Clair Obscur, Valor Mortis,
Black Myth: Wukong, Cyberpunk 2077), a session of a few minutes that includes **changing a graphics setting
that touches frame generation or DLSS** if the game lets you: its Advanced page has two more tables —
*Whether Streamline says each feature is supported, and loaded* (`kFeatureDLSS_G · slIsFeatureSupported` =
`eOk`, `· slIsFeatureLoaded` = *On* or *Off*) and *Each feature switched on or off, its resources allocated and
freed, Streamline shut down* — and the line under the session ends with one more sentence: *Streamline's other
calls - read: …; hooked, the game did not make the call: …; not hooked: ….* Say, for each game, which calls
are in which group. (c) **Check one number you can know**: a feature the game has switched off in its own menu
and that reads *On* under `slIsFeatureLoaded`, or the reverse, is worth saying — "loaded" is Streamline's word
for the plugin being in, not for the feature being in use, and what a real game does here is not known.
(d) **A game with no Streamline** (Lies of P, Kingdom Come: Deliverance II): all six are in *not hooked*.
(e) **The way out**: close each game the way you usually do. **Black Myth: Wukong and Hogwarts Legacy already
fault while closing** (Steam's `crashhandler64.dll`, `_nvngx.dll`); what matters is whether anything is
*different* — a new fault in Windows' application log (Event Viewer ▸ Windows Logs ▸ Application, *Application
Error*) naming another module, a game that hangs on exit, a game that used to close cleanly and does not.
(f) **Off means off**: with Advanced metrics off, **start a game afresh** and record a session; its
`logs\overlay-<pid>-….log` has no `HOOK_INSTALLED` line naming any of the ten: `slSetConstants`, `ffxConfigure`,
`slGetFeatureVersion`, `slGetFeatureRequirements`, `slIsFeatureSupported`, `slIsFeatureLoaded`,
`slSetFeatureLoaded`, `slAllocateResources`, `slFreeResources`, `slShutdown`. (This is item 42 (f), which was
not run.) (g) **Nothing else moved**: no crash, no `PATCH_FAILED` in any overlay log, frame rates and the
frame-generation factor as on beta.44 for the same game and settings; with Advanced metrics on the log ends
with an `UNHOOK_RESTORED` line for each patch and no "the stop was still writing". (h) **The Advanced page's
catalog**: *The preferences the game started Streamline with* reads *Not read: the game makes this call once,
before FrameLedger is inside it*.
**What each outcome means, written before the run.** A call in *hooked, the game did not make the call* in
every game: the game does that before FrameLedger is in, or never — then that hook gives nothing in attach
mode, and whether it keeps its patch is the owner's to say. `slIsFeatureSupported` with a result that is not
`eOk` for a feature the game then uses: say which — the row is Streamline's answer to the game, and the game
may have asked about another graphics card. A count of `≥ 15`: the game allocates or frees every few frames;
the number stopped, nothing is wrong. *Not hooked* on a Streamline game: the log says why — no `HOOK_INSTALLED`
for the function (a 1.x interposer, or the export is missing) or a `PATCH_FAILED`. **A new fault on the way
out, a hang on exit, or a crash in play**: say so at once, with the game; switching Advanced metrics off and
starting the game afresh runs it without any of the ten hooks, and if that cures it the hook is one of the
six new ones or the four before — the overlay log's last `HOOK_INSTALLED` lines and Windows' fault entry say
which is nearer. A game that will not start at all: *Disable all hooking* (Settings ▸ Capture).
**Seven sessions were run on 2026-11-13 and read the same day** ([`handoff/history/0.1.0-beta.45.md`](handoff/history/0.1.0-beta.45.md) §What the owner's
seven sessions on beta.45 showed): (a) happened — Privacy Policy 2.19 accepted once; (b) five Streamline games, six sessions —
`slIsFeatureSupported` read in all, `slIsFeatureLoaded` in three games, `slSetFeatureLoaded` in two,
`slFreeResources` and `slShutdown` once in one game each, **`slAllocateResources` in none**; (d) held on Kingdom
Come: Deliverance II; (e) **nothing different**: Hogwarts Legacy and Black Myth: Wukong fault while closing in
the same module at the same offset as before, the other four games leave no fault; (g) no `PATCH_FAILED`, no
fault counted, rates as before — and no session was stopped while its game ran, so the part about
`UNHOOK_RESTORED` lines was not exercised and was worded wrongly. **(f) was not exercised; (c) and (h) are the
owner's to say.** Taken as passed on the owner's "check the logs and go on", 2026-11-13; (f) stays open.
`slAllocateResources` keeps its patch (D135, the same day).
44. **beta.46 — hooks in and out under one stop of the threads, in real games** (D134). **Not first on a game under
the user-mode exception's trial** (D38): a crash there ends the exception for good. (a) **The update from
beta.45**: Help ▸ Check for updates, *Restart to update* with no game running; **the terms window does not
show** (nothing new is read). (b) **Any Streamline game with Advanced metrics on** (Hogwarts Legacy first — its
thirteen patches took 2.8 s on beta.45 — then Phantom Blade Zero, Cyberpunk 2077, Dying Light: The Beast, Black
Myth: Wukong): play as usual and close the game. Its `logs\overlay-<pid>-….log` has `PATCH_BATCH install`
lines, and the `sl.interposer.dll` `HOOK_INSTALLED` lines above each carry **the same time** to the thousandth
of a second, not times 0.06–0.2 s apart. Say the `a=` of each `PATCH_BATCH` line: it is the microseconds the
game's threads were stopped for that batch. (c) **How the game's start feels**: no worse than on beta.45 — the
hitching while a game with Streamline starts should, if anything, be shorter. (d) **A stop while the game
runs**, which no session has exercised since beta.43: with a game running and being recorded, switch *Disable
all hooking* on in Settings ▸ Capture; the part inside the game stops at its next safety scan, within 30
seconds. The game must go on running without a freeze of seconds; the log then has `STOP`, an `UNHOOK_RESTORED`
line for every patch, one `PATCH_BATCH restore` line and `STOP_FLUSHED … a=1`. Switch it back off afterwards. (e) **A game with no
Streamline** (Kingdom Come: Deliverance II, Lies of P): its FSR rows are under `PATCH_BATCH` lines as well, and
the session reads as before. (f) **Nothing else moved**: no crash, no `PATCH_FAILED`, no `UNHOOK_DECLINED`
unless another overlay is running, the same hooks in `HOOK_INSTALLED` as on beta.45 for the same game, frame
rates and the frame-generation factor as before, the Advanced rows as before. (g) **Off means off** (item 43
(f), never run): with Advanced metrics off, start a game afresh; no `HOOK_INSTALLED` line names any of the ten
Advanced-only calls.
**What each outcome means, written before the run.** A `PATCH_BATCH install` whose `a=` is about what one
patch cost on beta.45 (60 000–220 000) is the expected answer: one stop of the threads. An `a=` of a second or
more: one stop is long in that game — say which game and the number; it is still one where there were thirteen,
but it is what would decide whether batches need a ceiling. `HOOK_INSTALLED` lines of one batch with different
times: the batch fell back to one patch at a time — the log then also has a `PATCH_FAILED` — say so with the
log. A game that hitches at its start as it did not on beta.45, or freezes at (d): say so at once, with the
game; 0.1.0-beta.45 is the build without this change. A hook that was in on beta.45 and is not on beta.46
(compare the `HOOK_INSTALLED` lines), or a session that reads *not hooked* where it read *read*: the batch lost
a patch — the log's `PATCH_FAILED` line names it and why. A crash in play or at exit that is new: say so at
once; *Disable all hooking* runs the game with nothing patched after the start.
**Five sessions were run on 2026-11-13 and read the same day** ([`handoff/history/0.1.0-beta.46.md`](handoff/history/0.1.0-beta.46.md)
§What the owner's five sessions on beta.46 showed): (a) held — the terms window did not show; (b) five Streamline games — every batch's lines carry one
time; `a=` is 81 417–261 920 for sixteen of the eighteen batches, **1 260 835 for Hogwarts Legacy's thirteen
and 608 700 for three of Cyberpunk 2077's**, neither explained — and the sentence above about a fallback is
wrong: a batch finished one patch at a time leaves no line unless a patch fails for good, so the log cannot say
which of the two a long `a=` was; (f) held — no crash in play, no `PATCH_FAILED`, no `UNHOOK_DECLINED`, the
same hooks as on beta.45, rates and factor as before, the two known faults on the way out at their old offsets.
**(d), (e) and (g) were not exercised; (c) is the owner's to say.** Not passed: (d) is the half of this release
that no real game has run.
**2026-11-14, three more sessions (203–205), read the same morning — (d) is still not exercised.** *Disable all
hooking* was switched on **before** the games were started, not while one was being recorded: Cyberpunk 2077
(06:34) and Phantom Blade Zero (06:38) were each recorded as `RefusedKillSwitch` — tier 2, nothing injected, no
overlay log for either process, which is FR-2.4 doing what it says and not what (d) asks. With the switch off
again, Phantom Blade Zero (06:46, 518 s) was hooked from its start to its exit: twenty-two patches under one
batch of 84 ms and ray tracing's two under one of 81 ms, ×3.0 at 89 → 268, no fault, a clean exit — and **no
`STOP`, no `UNHOOK_RESTORED`, no `PATCH_BATCH restore` line**, because nothing stopped it while it ran. Windows'
application log has no fault since 06:30. **How (d) is run**: start the game with the switch **off** and wait
until FrameLedger shows it recording; **then**, with the game still running, switch *Disable all hooking* on;
play on for half a minute; switch it off; close the game.
**2026-11-14, the second try (sessions 206 and 207, Phantom Blade Zero): (d) held, twice.** The game started with
the switch off and was being recorded; the switch was turned on while it ran. Each log has `STOP` at the next
safety scan, an `UNHOOK_RESTORED` line for every one of the twenty-nine patches at one time, one `PATCH_BATCH
restore` — `a=83309` and `a=85578`, `b=0x1d` — and `STOP_FLUSHED … a=1`; no `UNHOOK_DECLINED`. Both sessions
ended `KillSwitchEngaged`, tier 1, no fault, nothing dropped; the game ran on for three and four minutes and
closed with nothing in Windows' application log. Whether a hitch was felt at the stop is the owner's to say.
**Taken as passed for what 0.1.0-beta.47 was held on** — (a), (b), (d), (f). **(e) and (g) were never run**: they
are (f) and (g) of item 45.
45. **beta.47 — NVIDIA's latency report, read inside a game** (D132, D136, D138–D141). **Cyberpunk 2077 first**
(D136), **and not first on a game under the user-mode exception's trial** (D38): a crash there ends the exception for good.
This is the first release in which the part inside a game *makes a call of its own* — into the graphics driver,
about once a second, on the game's presenting thread — so read (a) before (b). (a) **The update**: Help ▸ Check
for updates, *Restart to update* with no game running; **the terms window shows once, with Privacy Policy 2.20
and Disclaimer 2.16** — the Disclaimer's §2 has a new paragraph, *One call it makes itself, into the graphics
driver*. Settings ▸ *Advanced metrics* says the same in its description. (b) **Cyberpunk 2077, Advanced metrics
on, frame generation OFF, Reflex on in the game's menu**, five minutes of play, then close the game as usual.
Its `logs\overlay-<pid>-….log` has `LATENCY` lines: *reading the driver's latency report*, and two more with
the end of two paths — `…\System32\nvapi64.dll` and `…\nvapi64_impl.dll`. On the Advanced page, the session
has a group **Latency**: the table *What the driver's low-latency state is* (five rows by NVIDIA's names, and
`fgMultiplier`) and, in the chart's list, *Latency per frame* with seven lines. The line under the session ends
with *NVIDIA's latency report, which FrameLedger asks the graphics driver for inside the game: read.* Say the
*total* line's usual value, and **whether the game felt any different with the option on** — a hitch once a
second is the thing to look for. (c) **The same with frame generation ON** (DLSS ×2 or more): the driver's
pacer is another thread, and this is the arrangement the probe could not make. (d) **Reflex switched off in the
game's menu**, if the game lets you: say what the line reads — *read* or *asked, and the driver reported no
frame*; both are possible and neither is wrong. (e) **Only after (b) and (c) are clean, a second game** that
uses Reflex through Streamline (Phantom Blade Zero, Clair Obscur, Black Myth: Wukong, Hogwarts Legacy). (f) **A
game that is not Direct3D 12 or has no NVIDIA library** (Kingdom Come: Deliverance II, Lies of P): the line
reads *not read*, with which of the two. (g) **Off means off**: with Advanced metrics off, start a game afresh;
its log has no `LATENCY` line. (h) **Nothing else moved**: no fault counted, frame rates and the
frame-generation factor as before, the game closes as it did.
**What each outcome means, written before the run.** *Read*, with a total of a few to a few tens of
milliseconds that moves with the frame rate: route D works in a real game; what comes next — more titles, the
session summary — is the owner's to say. *Asked, and the driver reported no frame* in a game whose Reflex is
on: the driver has a report for another object than the device FrameLedger holds; nothing is harmed, and the
title is not readable this way. *Not read — the game has not loaded nvapi64.dll* in Cyberpunk 2077: what its
files said did not happen at run time; another title is tried. *Stopped — a call took N µs*: **one call of a few
thousand microseconds, once, is most likely a pause of the game's thread on a busy PC** (`spike-notes` §31, *A
run on a busy PC*) — start the game again and see whether it repeats; if it repeats, or N is ten thousand or
more, the driver holds the call in that game and **route D stops for it** — say so with the number. *Stopped — a
call into the driver faulted*, or a fault counted in the session: **route D stops** (the commitment of §30) —
say so at once, with the log. **A game that freezes** with a `LATENCY` line saying a call has not returned, or
that freezes at all with the option on and not with it off: close it, say so at once; Advanced metrics off is
the way to play it, and route D ends. **A hitch once a second** with the option on that is not there with it
off: say so at once — the two-millisecond latch did not catch what a player can feel. **The game closing by
itself a second or two after it starts, only with the option on** (Windows' Application log: exception
`0xc0000409`): a Control Flow Guard refusal of the call — say so at once with the entry; that game enforces
what the probe could only check for this driver. Spans that make no sense — a part larger than the total, a
line that is always zero: say which; they are the driver's numbers and are shown as it reported them.
**Three sessions were run on 2026-11-14 and read the same morning — in Phantom Blade Zero, not Cyberpunk 2077**
(sessions 208–210, Direct3D 12, Streamline 2.14.1). **The read works in a real game.** (a) held: the terms were
accepted once, Disclaimer 2.16 and Privacy Policy 2.20. In all three the overlay log has `LATENCY … reading the
driver's latency report` four to five seconds after the attach, with `C:\WINDOWS\SYSTEM32\nvapi64.dll` and
`…\nvapi64_impl.dll` on the two lines after it, and **no later `LATENCY` line**: no error, no slow call, no
fault, no stall. Each session's line reads *read*, and the slowest call into the driver was **24, 20 and 45 µs**.
Frame generation on: session 208, ×3.0 at 85 → 256 — and the driver's own `fgMultiplier` read 3. Off: sessions
209 and 210, multiplier 1. Seven series in each, 227 / 24 / 823 points: 90 %, 62 % and 98 % of the Agent's own
once-a-second samples (the short session spent its first seconds before the module was found). With frame
generation ×3: total 23.4 ms, simulation 4.4, render submit 10.3, present 0.7, driver 10.9, OS queue 13.0, GPU
render 11.5 (means). The driver's switches: `bLowLatencyMode`, `bUseGameSleep` and `bFullscreenIFlip` *on at
times, off at others*; `bFsVrr` and `bCplVsyncOn` *off* — so in a game the low-latency switch does read on,
which it never did for the probe. No fault counted, nothing dropped, a normal exit each time, nothing about the
game in Windows' application log; rates as on beta.46 for the same game.
**One outcome the list above names: a part larger than the total.** *GPU render* reads 16.4 ms where the *total*
reads 5.4 ms (session 209), and about equals it in session 210 (16.0 against 15.9 ms). The span is the driver's
`gpuRenderStartTime` to `gpuRenderEndTime`, and in this game that start lies **before** the frame's own
simulation start: it reads like the time between two frames' GPU ends, idle time included — a reading, not a
fact; the driver has a field of its own for the active time (`gpuActiveRenderTimeUs`), which FrameLedger does
not read. It is shown as the driver reported it. The spans also overlap — render submit, driver and OS queue
together are longer than the total — so they are not parts that add up, and the page does not say they are.
**Not run**: Cyberpunk 2077, which (b) and (c) name; (d) Reflex off in the game's menu; (f) a game that is not
Direct3D 12 or has no NVIDIA library; (g) off means off. **Not said**: whether the game felt any different with
the option on. Not passed as written; what the three sessions were run for is answered.
**Five more sessions the same morning (211–215), read at 09:15: Cyberpunk 2077 twice, and three games the
read must leave alone.** (b), (c) **Cyberpunk 2077** (Streamline 2.14.1, DLSS): session 211 with frame
generation ×2.0 for 42 % of it (166 → 330) and menus for the rest, session 212 without, at 238 FPS. The module
was found 4.8 s and 2.6 s after the attach, in the same two files; each session's line reads *read*; **the
slowest call took 34 and 65 µs**; no later `LATENCY` line. 190 of 195 and 429 of 438 samples carry a report.
The total: 13.5 ms while frames were generated, 4.1 ms in the menus, 11.7 ms at 238 FPS without generation.
**GPU render is 2.7 ms there and never longer than the total** (0 of 429 samples; 1 of 190) — so what the
three sessions above showed is Phantom Blade Zero's, not the route's. `bLowLatencyMode` *on* throughout. The
table's `fgMultiplier` reads **1 in the session that held ×2 for 105 seconds**: the row keeps the last value
read, and that session ended in a menu — a row that lists what was seen would not mislead. D137's first
Cyberpunk batches: eight of them, 85–221 ms, every one `install` in one apply; beta.46's 0.61 s did not recur.
(f) **Not Direct3D 12**: Crown & Coin (213) and Psycho-Sleuth (214), both Direct3D 11 — one `LATENCY no
Direct3D 12 device` line each, nothing after it, the session's line reads *not read — not Direct3D 12*.
**Direct3D 12 with no NVIDIA library**: Dread Fields (215) — `the game has not loaded nvapi64.dll`, said once,
*not read*. The three ran at the same time and were recorded apart. (h) held in all five: no fault counted,
nothing dropped, normal exits, nothing in Windows' application log.
**A correction to the paragraph above**: the Agent's samples are **1.32 s apart, not one second** (the `t_ms`
series; so in every session looked at back to beta.42, while `sensor_blobs.hz` says 1.0 — not looked into),
so the shares are 227 of 258, 24 of 39 and 823 of 846 samples, and the gaps are the first seconds of a session.
**The owner's answer on the *GPU render* line (2026-11-14): read the driver's active time** — planned as one
train with what else is open. **Left of this item**: (d) Reflex off in the game's menu, (g) off means off, and
the owner's word on how the games felt; the owner runs what is left once, with that train's release.
**They are (d), (e) and (g) of item 46.**
46. **beta.48 — the driver's own GPU times, the multipliers seen, samples a second apart; and what item 45 left**
(D142–D146). **One batch**, as the owner asked: any Direct3D 12 games that use Reflex, as many as wanted, with
*Advanced metrics* on — **not first on a game under the user-mode exception's trial** (D38). Then say *check*;
the sessions are read from a copy of the ledger and the logs.
(a) **The update** shows the terms window once, with **Privacy Policy 2.21 and Disclaimer 2.17**.
(b) **Phantom Blade Zero and Cyberpunk 2077 again**, a few minutes of play each: on the Advanced page, under
**Latency**, a second chart — *the graphics card's time per frame, as the driver works it out* — with two lines,
*GPU active render* and *GPU frame time*; and the old line reads *GPU render, start to end (idle time included)*.
(c) **A session with both a menu and frame generation**: the table's `fgMultiplier` row lists both — *1, 2*, or
*1, 3* — not the last one.
(d) **Reflex switched off in a game's menu**, if a game lets you (item 45 (d)): say what the session's line reads.
(e) **Off means off** (item 45 (g)): switch *Advanced metrics* off, start a game afresh, play a minute, close
it; switch it back on afterwards.
(f) **Any other games of the batch** — Clair Obscur, Black Myth: Wukong, Hogwarts Legacy, Dread Fields, a
Direct3D 11 game: nothing to do but play and close them.
(g) **How the games felt** with the option on — a hitch once a second is the thing to look for; nobody but the
player can say it.
**Read from the ledger, nothing for the owner to do**: the samples of every session are one second apart, not
1.32 s, and `hz` says what they were.
**What each outcome means, written before the run.** *The two lines are there, and active render is shorter
than GPU render*: D142 did what it was for; in Phantom Blade Zero the reading of 2026-11-14 expects *GPU frame
time* near the 16 ms the old line showed, and *active render* well under it — **if active render is as long as
the old line there, that reading was wrong**, and the span is long for another reason: say the numbers. *The
second chart is not there, and the seven lines are*: the driver gave zero for both figures in every frame —
nothing is wrong with the read, and the texts already say a driver may give neither; say which game. *One line
and not the other*: the same, for one figure. *Active render longer than GPU render, start to end*: they are the
driver's numbers and are shown as reported — say which game and by how much. *`fgMultiplier` shows one value in
a session that had both*: the driver's status did not change with the menu in that game, or the mask is wrong —
say the game; the log's `LATENCY` lines and the frame-generation mix of the session tell the two apart. *Samples
still 1.3 s apart*: the period is not kept — the poll takes longer than the interval, or the change is not in
the build: stop and say so. *Samples a second apart and the Agent's processor use visibly higher*: a poll that
took a third of 1.32 s now takes a third of a second — say so; where that time goes is not established
(§Left open). *(d) reads "read"*: the driver reports frames with the game's Reflex off — both that and *asked,
and the driver reported no frame* are possible and neither is wrong. *(e) the log has a `LATENCY` line*: off
does not mean off — say so at once, with the log. **A fault, a freeze, a game closing by itself, a hitch every
second, or a session that says *stopped***: as item 45 says for each — say so at once.
**Run 2026-11-14, 12:49–14:12, and read the same afternoon** from a copy of the ledger and the logs: ten
sessions on beta.48 — Phantom Blade Zero seven times (217, 219, 220, 223–226), Cyberpunk 2077 three times (218,
221, 222). Every one tier 1, exit code 0, no fault counted, nothing dropped; nothing in Windows' application
log. (Session 216, an hour of Phantom Blade Zero, was still beta.47 — it ran while the release was cut — and
is 1.32 s apart like the ones before it.)
**The samples are a second apart**: the median gap is 996–1002 ms in all ten (session 223: 2 294 of 2 394 gaps
round to 1.00 s, the rest lie between 0.80 and 1.27 s), and `hz` reads 1.000–1.017.
(a) `legal_acceptance` holds Privacy Policy 2.21 and Disclaimer 2.17, accepted at 12:48:40.
(b) **Both figures were given by both games, in every frame report**: in the nine sessions with the option on,
every sample that has a *GPU render* span has both (3 834 of 3 834). **Phantom Blade Zero is what the reading of
2026-11-14 expected**: at 60 FPS (217) the span's median is 16.39 ms, *GPU frame time* 16.67 ms and *GPU active
render* **9.59 ms**; in the menu (224) 16.39 / 16.67 / **2.77**; at 91 FPS (225) 11.49 / 11.79 / 10.05; with
frame generation ×5 (223) 15.67 / 15.88 / 14.08. So in that game the span is the time from one frame to the
next less a third of a millisecond, whatever the card did in it, and active render is never longer than it
(0 of 3 406 samples). **Cyberpunk 2077 is the outcome written above as "active render longer than GPU render,
start to end"**: without frame generation it is longer at 157 of 158 samples (218: medians 3.17 against
2.79 ms, frame time 6.02) and at 48 of 48 (221: 2.03 against 1.82, frame time 2.37); with frame generation ×2
(222) it is shorter at all 222 (4.85 against 6.35, frame time 6.67). Shown as reported. **Why is not
established**, and two things could do it: the two are averages over different frames — the span over every
summed frame, one of no length included (the smallest mean span of a sample there is 0.5 µs, the smallest active
time 0.29 ms), active render over the frames that gave one — or the driver counts work the two marks do not
enclose. The ledger cannot tell them apart; a count of the frames whose span has no length would.
(c) `fgMultiplier` reads **1, 2** in Cyberpunk 2077 (222: ×2.0 for 69 % of it), **1, 5** in Phantom Blade Zero
(220: ×5.0 for 71 %, 30 → 151 FPS; 223: ×5.0 for 95 %, 63 → 316) and **1** in the seven sessions that generated
nothing.
(d) **Session 221** (Cyberpunk 2077, 59 s of menus at 421 FPS) is the only Cyberpunk session whose
`bLowLatencyMode` reads *on and off* — 218 and 222 read *on*, as 211 and 212 did — and is taken to be the one
where Reflex was switched off; the owner has not said which it was. Its line reads *read*, 48 of 57 samples
carry a report, every one from the first on, and the slowest call took 19 µs: the driver goes on reporting
frames.
(e) **Off means off, on a real game for the first time**: session 226 (`telemetry.advanced` 0) has no
`session_advanced` row and 19 series where the others have 240; its log has ten hooks, none of the ones
Advanced metrics adds, and **no `LATENCY` line**. The setting was still off when the ledger was copied.
(f) No other game was in the batch.
**The read itself**: one `reading the driver's latency report` line a session, 1.4–6.7 s after the attach, in
the same two files, and none after it; the slowest call of a session 18–123 µs; no latch. The install batches
took 75–191 ms, every one `install`.
**Samples with no report**: after a session's first report, 0.4–7.3 % of its samples carry none (1 of 223 in
222; 134 of 2 377 in 223, minutes of four in five there). They are not the sensor library's ticks (25 of the 134
fall on one, 24 just after one). What the code says, not measured: the watchdog ticks once a second and arms a
read when a second has passed since the last arm, so a tick that comes a moment early waits for the next one —
the reads are a second apart or two, the samples exactly one. Nothing is lost by it; a sample shows a gap.
(g) **The owner's answer, asked whether a game hitched once a second with the option on (2026-11-14): no
hitch.** What the frames say, in the sessions without frame
generation: a frame three times the median of the 120 before it and 8 ms longer comes 0.33 times a second
with the option on (225) and 0.32 with it off (226), and such frames do not gather at a period of one second
(vector strength 0.05–0.10 where chance gives 0.10–0.20; 217, 218 and 222 the same). Sessions with frame
generation cannot be judged that way. **Not seen in the ledger**: what the page looked like, and the Agent's
processor use now that a poll comes every second.
**Read, and (g) answered**: (a)–(g) each have their answer; ~~whether session 221 was (d) is inferred, and the
owner's to say~~ (d) is withdrawn in the next paragraph. The owner has not said the item is passed.
**Eight more sessions the same afternoon (227–234), read at once, and three things the owner said.** Still
beta.48: Cyberpunk 2077 three times (227 with the option off — no `LATENCY` line, no row; 232 mostly menus; 233
five minutes without frame generation), Phantom Blade Zero three times (228, 231, 234), Clair Obscur:
Expedition 33 twice (229 with frame generation ×3 for 79 % of it, 230 without). All tier 1, exit code 0, no
fault, nothing dropped; samples a second apart; slowest calls 10–29 µs; `fgMultiplier` *1, 3* in 229.
**The owner's words**: *Reflex is checked together with frame generation — without frame generation the game
does not switch Reflex on*; *a value that reads on and off is the NVIDIA App's override* (the App's page for
Phantom Blade Zero: **Low Latency Mode — Ultra**); and a session with Reflex off may be read. **So (d) above
is withdrawn as written**: session 221 was not "the one where Reflex was switched off in a menu", and nothing
here says what its `bLowLatencyMode` *on and off* was. What the sessions do read: **Clair Obscur without frame
generation has `bUseGameSleep` off throughout (230) and with ×3 on and off (229)** — the game's own Reflex
goes with its frame generation, as the owner said; **Phantom Blade Zero has `bLowLatencyMode` off throughout
in 234**; and in both the read is what it is with Reflex on — one `reading` line, a report at every read, the
two new figures in every one. **Cyberpunk 2077 without frame generation again has active render longer than
the span** (233: 3.40 against 3.13 ms at 292 of 296 samples), **and its menus do not** (232: 0 of 126; active
0.39 ms inside a span of 1.18).
**The holes are larger in short sessions**: 11.3 %, 16.2 % and 11.1 % of the samples after the first report
in 228, 229 and 231; 6.0 % in 230; 0.8 % in 232. **And "nothing is lost by it", above, was wrong**: it was
written on a reading of the Overlay alone. The Agent kept one set of readings until a sample carried it and
overwrote it — a publication that landed just after a sample left that sample without, and the set that had
waited was replaced at the next return to step (46 of session 223's 134 holes are exactly five samples
apart). The Overlay's part is the other half: its ticks were a second and their own work apart, and a tick the
loader woke published the read and armed none. Both are mended in 0.1.0-beta.49 (D150); neither is measured in
a game yet.
47. **beta.49 — whose the difference is in Cyberpunk 2077's GPU times; and the holes** (D148–D151). **One batch,
in pairs**: the same game with frame generation off and with it on, *Advanced metrics* on. Reflex is not
switched by hand — it goes with frame generation. **Not first on a game under the user-mode exception's
trial** (D38). Then say *check*; the logs and a copy of the ledger are read.
(a) **The update shows no terms window**: no legal text changed.
(b) **Cyberpunk 2077**: six minutes or more of play **without frame generation**, in a scene like session
218's; six minutes or more with **frame generation ×2**; two minutes of menus only.
(c) **Phantom Blade Zero**: six minutes or more without frame generation — the control — and six or more with.
(d) **About eight short sessions**, three minutes each, any Direct3D 12 game that uses Reflex: the step between
the game's side and the Agent's is different in every session, and one long session shows one of them.
(e) **One session with *Advanced metrics* off**; switch it back on afterwards.
(f) **What *Low Latency Mode* is set to for Cyberpunk 2077 in the NVIDIA App** — a word is enough.
(g) **How the games felt**, as always.
A game closed by itself loses its last batch of counts, under a minute: that is why six minutes and not two.
**What each outcome means, written before the run.** The log's `LATENCY_FRAMES` batches each carry the state
they belong to, so every one is read as *this state, these counts*.
*Cyberpunk without frame generation: nearly every frame that has both is in "active longer than span", and
their mean span is near the 2.8 ms the chart shows* — **the difference is the driver's**: it counts work the
two marks do not enclose. The numbers are shown as reported and the label "idle time included" is what is
wrong for that game: the owner decides its wording.
*"Active longer" is a minority whose spans are next to nothing, the rest have active within the span, and
spans of no length or of a microsecond are about as many as that minority* — **the difference is ours**: the
span's mean is pulled down by frames that have none. The owner decides whether the mean is taken over the
frames that have a span.
*Both heaps are large, with real spans* — both; the heaps' sums give the like-for-like figure, and it is said
with them.
*"GPU span absent" is not zero* — those reads published no GPU point at all, and were never in the comparison.
*Phantom Blade Zero shows many frames in "active longer"* — suspect the count before the driver: say so.
*The counts differ between a batch with `bUseGameSleep` on and one with it off, in the same game* — that is
the owner's point, measured: say the two side by side.
*The batches say a thousand milliseconds a read and as many periodic ticks as reads, and every session has
next to no sample without a report* — D150 did what it was for.
*The log says a thousand milliseconds a read and some sessions still lack a report at a steady share of their
samples* — the Agent's rule is not the whole of it: say the sessions and the share.
*The log says more than a thousand milliseconds a read* — the Overlay is still uneven; the line's tick counts
say whether the loader was waking it.
*(e) has a `LATENCY` or `LATENCY_FRAMES` line* — off does not mean off: say so at once, with the log.
**A fault, a freeze, a game closing by itself, a hitch every second, or a session that says *stopped***: as
item 45 says for each — say so at once.
**Run 2026-11-14 (sessions 236–245, ten on 0.1.0-beta.49), read the same day** from a copy of the ledger and
the Overlay's logs. All ten tier 1, exit code 0, no fault, nothing dropped. Phantom Blade Zero twice (236
without frame generation, 1 108 s; 237 with ×2, 811 s), Kingdom Come: Deliverance II (238), Clair Obscur (239),
Black Myth: Wukong (240), Lies of P (241, and 245 with the option off), Hogwarts Legacy (242), Dying Light: The
Beast (243), Cyberpunk 2077 (244). 239 and 240 ran at the same time, and so did 241, 242 and 243.
(a) **No new acceptance row** after the update: the terms window did not appear.
(b) **Cyberpunk 2077 was one session of 125 s without frame generation** (374 FPS); ×2 and the menus were not
run. It is enough for the question. 88 reads in one state — low-latency mode on, the game sleeping, ×1 — summed
5 394 frames: **4 917 in "active longer than span", mean span 1 571 µs, mean active 1 849 µs**; 96 within
(3 866 / 3 431 µs); 381 with a span of no length and an active time; none absent, none refused. **So the
difference is the driver's**: on frames that have a real span, its active time is 278 µs (18 %) longer than
the two marks are apart. A small part is ours as well — 7 % of the frames have a span of no length and pull the
span's mean down (the stored means are 1 892 and 2 138 µs). With frame generation on it is not measured on
this build (0.1.0-beta.48's means: never longer).
(c) **The count is not what is wrong**: in the six other sessions that reported (five games), **not one frame of 129 900** is
in "active longer", and no span is absent or of no length. Phantom Blade Zero: without frame generation
63 123 frames, span 16.4 ms, active 11.1 ms; with ×2 21 209 frames, span 33.2 ms, active 12.6 ms.
(d) **The holes**: in batches of twenty reads or more a read took 1 000.94 ms (236), 997.18 (237), 999.99 (242,
244), and up to 1 013 ms in the sessions that shared the PC with another game. Samples without a
report: 236 — 40 of 1 106, **39 of them in the first 46**, while the game started; 237 — 36 of 809, in the first 42 and
the last two; 244 — 9 of 123, in the first ten. Before, they were spread through a session (0.4–16 %). **D150 did what
it was for.** Where two or three games ran at once the holes are blocks in which the game did not present
(240: 103 of 477; 242: 80 of 402, 68 of them at its start; 243: 37 of 414; 239: 23 of 184): a read is taken on
the presenting thread, so a game that presents nothing is not read — Lies of P's log has sixty reads in 189 s.
(e) **Off means off**: session 245 has no `LATENCY` line and no Advanced row.
**The stop fired in a real game for the first time** (D141): Kingdom Come: Deliverance II, 12 s after the
attach, while the game was starting — the seventh read took 5 337 µs, and no further read was made; the
session says so (`Slow`), the game ran 287 s and closed normally. No frame had been reported in those seven
reads. **The owner: that game has no Reflex in its settings** — so there was no report to lose; the
session's line says *stopped* where, without the slow call, it would have said *the driver reported no
frame*. Elsewhere the slowest call was 22–235 µs. In Lies of P the driver reported no frame at all.
**What the states say of Reflex and frame generation** (each batch carries its own): low-latency mode was
**on at a multiplier of 1** in Cyberpunk 2077 (88 reads), Black Myth: Wukong (253), Hogwarts Legacy (299) and
Dying Light: The Beast (345); Phantom Blade Zero without frame generation sleeps through Reflex with
low-latency mode **off** (1 062 reads) and switches it on with ×2; Clair Obscur without frame generation has
both off and still reports its frames. The logs cannot say whether a game's menu or the NVIDIA App switched
the mode on. NVIDIA's own guide makes frame generation need Reflex and not the reverse
(`spike-notes` §30, *Added 2026-11-14*). Hogwarts Legacy gave a multiplier of 6 for seventeen reads.
(f), (g) **The owner, the same evening: "App reflex off, không khựng"** — *Low Latency Mode* is off in the
NVIDIA App, and no hitch. Read with (f) as it was asked, for Cyberpunk 2077: the low-latency mode its 88
reads show at a multiplier of 1 is then the game's own Reflex and not the App's override — Reflex on without
frame generation, as NVIDIA's guide allows. **Then shown, in two screenshots**: the NVIDIA App's *Program
Settings* for Cyberpunk 2077 have *Low Latency Mode* **Off**, set for that program (the rows that follow the
global setting read "Global - …"; this one does not); and the game's own *Video* page has **NVIDIA Reflex Low
Latency: On**, V-Sync off, borderless at 2560×1440, 378 FPS on the game's counter. So the driver's
`bLowLatencyMode` on, at a multiplier of 1, is the game's menu, seen from both sides. **The owner's**: the wording of the two GPU labels — *idle time excluded /
included* does not describe a game whose active time is the longer.
**Answered 2026-08-05, do not re-ask:** remove `gameguard` and keep `guard` (approved
over a red `Rules` gate, with the reasoning recorded in the merge commit); vendor NVAPI
headers **and** `amd64/nvapi64.lib`; Windows 11 is the **measurement scope** only —
@@ -442,7 +1107,88 @@ Where each owner decision is recorded — the file and its section, and a label
| D79 | [`05_DETECTION.md`](05_DETECTION.md) §Platform signatures & metadata | the platform is "Xbox" (`xbox`), in full: import, artwork, Play, capture, detection, docs and legal |
| D80 | [`08_UI.md`](08_UI.md) §Games | Play on an Xbox game always activates its package, in both modes; the Agent answers `PackagedGame` and records it in attach mode |
| D81 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §G | title lists stay keyed by Steam id; an Xbox copy is matched by everything but that entry |
| D82 | [`HANDOFF.md`](HANDOFF.md) §START HERE | alpha.2 right after beta.33; on RC3 unless the train reaches it after .NET 11 ships (then SDK 11.0.100, packages 11.0.0) |
| D82 | [`handoff/history/0.1.0-beta.33.md`](handoff/history/0.1.0-beta.33.md) | alpha.2 right after beta.33; on RC3 unless the train reaches it after .NET 11 ships (then SDK 11.0.100, packages 11.0.0) |
| D83 | [`05_DETECTION.md`](05_DETECTION.md) §DLSS generation | the NVIDIA App's Dynamic frame generation, reported applied by the driver in a session where DLSS Frame Generation was identified, is DLSS 4.5 (amends D64) |
| D84 | [`05_DETECTION.md`](05_DETECTION.md) §DLSS generation | DLSS 5 heads the DLSS line where its neural rendering was measured running (`kFeatureDLSS_NR`, `dlss_nr_pct`); loaded only reads "runtime loaded" (amends D64) |
| D85 | [`08_UI.md`](08_UI.md) §Games, §Compare | filters: Games — store, engine, state, installed, played; Compare — game, tier, frame generation (no period); Compare lists every session |
| D86 | [`08_UI.md`](08_UI.md) (the shell) | the left pane's selected item keeps its label: our copy of WPF UI 4.3.0's LeftFluent item template |
| D87 | [`handoff/history/0.1.0-beta.34.md`](handoff/history/0.1.0-beta.34.md) | alpha.3 right after beta.34, on RC3 (before .NET 11's release, 2026-11-10) |
| D88 | [`../ACCESSIBILITY.md`](../ACCESSIBILITY.md) | the accessibility statement at the root, under Help ▸ Accessibility and on the wiki: WCAG 2.2 AA through WCAG2ICT is a goal, never a claim; reports by severity, no promised times |
| D89 | [`08_UI.md`](08_UI.md) §Screen readers | what appears is said to a screen reader — notices, the agent's and the update's banners, the snackbars, a game page's notice, Database maintenance's results — while FrameLedger's window is active |
| D90 | [`08_UI.md`](08_UI.md) §Screen readers | page titles heading 1, section titles heading 2; every chart named with where its numbers are; no series told apart by colour alone |
| D91 | [`08_UI.md`](08_UI.md) §Contrast | Windows contrast themes are followed, whatever Settings ▸ Theme says |
| D92 | [`handoff/history/0.1.0-beta.35.md`](handoff/history/0.1.0-beta.35.md) | beta.35 in this train, no alpha; alpha.4 is the move to .NET 11's release (SDK 11.0.100, packages 11.0.0), a task of its own after 2026-11-10 |
| D93 | [`13_CI_CD.md`](13_CI_CD.md) §Qodana | Qodana Community for .NET from a workflow of its own (the organisation's licence is a 30-day trial, so not the Rider linter): a gate of zero problems, advisory; an inspection is switched off only in `.editorconfig`, with its reason |
| D94 | [`12_BUILD.md`](12_BUILD.md) §The main line's .NET | the main line moves to .NET 11 with 0.1.0-beta.37, the day after it shipped; condition (a) of `20_OPEN_QUESTIONS` §G's table — two alphas on the released runtime — set aside by the owner, not met |
| D95 | [`02_SPEC.md`](02_SPEC.md) NFR-8 | Windows 10 22H2 is not blocked, and is stated as untested and unsupported (Microsoft supports .NET 11 on Windows 11); a processor needs SSE4.2 and POPCNT |
| D96 | [`13_CI_CD.md`](13_CI_CD.md) §C# 15 and the analyses | C# 15 on the main line from beta.37, a construct only where the build, CodeQL and Qodana read it — none of the four yet (Qodana's engine reads none, 2026-11-11), so beta.37 takes .NET 11's API and no syntax; no preview feature |
| D97 | [`../legal/ALPHA_NOTICE.md`](../legal/ALPHA_NOTICE.md) | an alpha is "a beta's source built to try a newer version of .NET, at times a prerelease one": EULA 1.7, Disclaimer 2.14 (with a paragraph on Windows and the processor), Alpha notice 1.1, approved with the plan; the App says "a prerelease .NET" only where the runtime is one |
| D98 | [`12_BUILD.md`](12_BUILD.md) §The alpha line | alpha.4 is beta.37's source on .NET 11.0.0, and then the alpha line rests until a newer .NET has a preview; a resting alpha tells its user to install the beta |
| D99 | [`13_CI_CD.md`](13_CI_CD.md) §Qodana | Qodana for .NET on the Ultimate licence, in two reports from the start: the managed code to the Qodana Cloud project `frameledger`, the C++ to `frameledger-cpp` |
| D100 | [`13_CI_CD.md`](13_CI_CD.md) §Qodana | an inspection is switched off where the engine that raised it reads — ReSharper's in `.editorconfig`, the rest in `qodana.yaml` (amends D93); hints are off as a severity; in code loaded into a game only defects are fixed |
| D101 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | "Thông số nâng cao" — Advanced metrics: the option `telemetry.advanced`, off by default, and off means nothing more is read or stored and the navigation item is hidden; everything reachable inside a game follows in stages, four trains A–D |
| D102 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | stored in full with the session, in tables of their own (`machine_profiles`, `session_advanced`, schema 0023) and not in columns of `sessions`; series under `adv.` names with NaN, not −1, for a tick with no reading |
| D103 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | LibreHardwareMonitor's CPU, memory, motherboard and storage groups only when the Agent already runs as administrator with PawnIO — the page links to that setting and never asks; power-supply and fan-controller groups refused, battery deferred |
| D104 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | identifiers are never read: the card's serial number and unique id, a display's EDID, the list of programs that are encoding |
| D105 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | the navigation item is the pane's last; Ctrl+1–5 keep their pages and Ctrl+6 opens it only while the option is on |
| D106 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | a `.partial` chunk of an unknown type with a right checksum is skipped, where it ended the read |
| D107 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | with no game running the Agent reads sensors only while the page is open |
| D108 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics | "never call a setter" is a build gate: every `NvAPI_*` call of the bridge is held to a committed list |
| D110 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M16 | the L2 telemetry layer's cost is measured in a real Agent before anything is changed (`tools/agent-cost.ps1`); what to do about the number is asked after |
| D117 | [`02_SPEC.md`](02_SPEC.md) NFR-2 | asked with the measurement in hand (1.93 % off, 2.04 % on): NFR-2 is amended to ≤ 2.5 % of one core during capture; the sensors keep one reading a second and no code changes |
| D111 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | inside the game too, off means nothing more is read: the Overlay reads new fields and installs new hooks only while the Agent's control word says the session records Advanced metrics |
| D112 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | `slInit` is not hooked in train B; if ever, never `applicationId`, `projectId` or the paths |
| D113 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | train B is three releases: beta.40 fields of calls already hooked (no new patch site), beta.41 `slSetConstants` + `ffxConfigure`, beta.42 `slGetFeatureVersion` + `slGetFeatureRequirements`; each waits for the owner's run of the one before; C is beta.43, D beta.44 |
| D114 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | the field of view is stored as a session's range, never as a series |
| D115 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Which struct version carries which field | a tail member of a vendor struct is read only where the upstream tags prove it is there |
| D116 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | "never the camera" is a build gate: `tools/overlay-fields-check.ps1` |
| D118 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | told that the sensor-library fix would take the number D113 gave train B's second release, the owner chose to release it at once as 0.1.0-beta.41: the train's second and third releases are beta.42 and beta.43, train C beta.44, train D beta.45 |
| D119 | [`07_IPC.md`](07_IPC.md) §Region 5 | `FLT_MAX` (or +infinity) in an FSR depth plane is stored and shown as "infinite" — three engines pass it for the plane at infinity; for FSR only: for Streamline `FLT_MAX` is the "not set" sentinel and stays absent |
| D120 | [`03_METRICS.md`](03_METRICS.md) §Upscaling | a session's upscaler is the technology named on the most samples, and its quality and render size are read from that technology's samples — not the first one named, with the others' values beside it |
| D121 | [`HANDOFF.md`](HANDOFF.md) §NEXT | Cyberpunk 2077 was set to FSR upscaling and FrameLedger named none: a detection gap, to be investigated — which first needs a native log that survives the game's exit |
| D122 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | train B's second step is split: **beta.42** has no new patch site (D119–D121 and the patch registry's hardening); **beta.43** is the two new hooks, begun only after the owner has run Cyberpunk on beta.42 and its log was read. The third release is beta.44, train C beta.45, train D beta.46 |
| D123 | [`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) §Pre-injection checks, check 1 | a module list that changed while the guard read it is read again, at most three times, before it refuses — five real sessions on one PC in a week had ended as a safety unhook with no frame recorded; a list that never settles still refuses, and nothing else about what refuses changed |
| D124 | [`HANDOFF.md`](HANDOFF.md) §NEXT | the guard's re-read (D123) ships with 0.1.0-beta.43 and not as a release of its own — the train's numbers do not move a third time; until then a launch can still be lost that way on beta.42 |
| D125 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory | an FSR UPSCALE is counted at the first source that saw one and at no other — a precaution for a loader that hands a dispatch on to the leaf's export (Cyberpunk 2077's 2.3.0 does, for frame generation); no title run so far upscales through one |
| D126 | [`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) §Crash & stability safety | the count of early abnormal ends starts again when the user switches hooking on (the consent grant sets it to 0) — it was never set back, so a game once switched off was switched off again by every single such end, under a reason that says "twice" (Lies of P, 2026-11-12) |
| D127 | [`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) §Crash & stability safety | a fault while a game is closing, which a hooked session does not see (Hogwarts Legacy, the driver's `_nvngx.dll`, hooked or not), is written down and not changed in 0.1.0-beta.44 — seeing it would end a user-mode exception for a fault no hook caused; a decision of its own |
| D128 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | a field of view over π is shown as the game gave it, with a row that says it is not radians, and is never converted — Cyberpunk 2077 gives Streamline degrees; converting would store a guess at the unit as the game's word |
| D129 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | `slInit` is not hooked at all — asked to reopen D112 and shown that no shipped mode is in the game in time (launch mode waits for the presentation runtime, no patch is ever made on a game's thread, an Advanced-only hook waits for the Agent's request), the owner kept it out; a hook there would be a patch that reads nothing |
| D130 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B | six more Streamline core calls are hooked in one release, 0.1.0-beta.45 — `slIsFeatureSupported`, `slIsFeatureLoaded`, `slSetFeatureLoaded`, `slAllocateResources`, `slFreeResources`, `slShutdown`; trains C and D come after it, and train C's spike is written beside it as a document |
| D131 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory | the adapter a game asks `slIsFeatureSupported` about (`sl::AdapterInfo`: a LUID, or a Vulkan physical device) is never read — the feature's id and the result only; a build check refuses the struct's member names in code loaded into a game |
| D132 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14, `CLAUDE.md` rule 4's note of 2026-11-13 | Reflex route D goes to a spike on `hook-harness` and a stub first — and the ruling a go needed: calling a driver's read-only getter with the device of the swap chain the Overlay hooks is inside rule 4. No real game before the owner has seen the result |
| D133 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14, [`spike-notes/30-reflex-latency.md`](spike-notes/30-reflex-latency.md) | Reflex route C (the Overlay calling `slReflexGetState`) is refused; the no-injection ETW route stays closed, and §G with it |
| D134 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Unhooking, *Batches* | installs and restores are batched through MinHook's queue — one stop of the game's threads for a batch, not one a patch — in a pull request of its own, measured, released alone (0.1.0-beta.46) |
| D135 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory | `slAllocateResources` keeps its patch although no game called it in the first six sessions; looked at again when there are more |
| D136 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14, [`spike-notes/31-reflex-route-d-harness.md`](spike-notes/31-reflex-route-d-harness.md) §The question for the owner | Reflex route D goes on to one real title, Cyberpunk 2077 — a plan of its own first, no code before it is approved, and no release carrying it before item 44 (d) has been run |
| D137 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Unhooking, *Batches* | the `PATCH_BATCH` line says how a batch ended — one apply, or finished one patch at a time — after two batches in real games took 0.61 s and 1.26 s and the log could not say which; with the next release, not one of its own |
| D138 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14, [`handoff/history/0.1.0-beta.46.md`](handoff/history/0.1.0-beta.46.md) §The train after 0.1.0-beta.46 | the Reflex read has no switch of its own: it is part of Advanced metrics — so it reaches every Direct3D 12 game that loaded the driver's `nvapi64.dll`, for everyone who switched Advanced metrics on; "one title" is the order of the owner's first runs (item 45), not a limit in code |
| D139 | [`spike-notes/31-reflex-route-d-harness.md`](spike-notes/31-reflex-route-d-harness.md) §The leg with a sleep | before any Overlay code the probe runs a leg with `NvAPI_D3D_SetSleepMode` and `NvAPI_D3D_Sleep` — writes, in the probe's own process, on its own device, in no shipped binary |
| D140 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14 | an `nvapi64.dll` loaded from the game's own folder — a replacement, not the driver's — is refused, and the session says why |
| D141 | [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M14 | one call into the driver of 2 ms or more switches the Reflex read off until the game exits |
| D142 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Calls the Overlay makes into a driver | asked on 2026-11-14 with Phantom Blade Zero's *GPU render* longer than its total: read the driver's `gpuActiveRenderTimeUs`, and with it `gpuFrameTimeUs` — two more members of the report already fetched, no further call |
| D143 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §The inventory | the start-to-end *GPU render* line stays, stored and shown, under a label that says it includes the card's idle time |
| D144 | [`07_IPC.md`](07_IPC.md) §Region 5 | the driver's frame-generation multiplier is kept as every value a session saw, not the last one |
| D145 | [`../legal/ACCURACY.md`](../legal/ACCURACY.md) | the accuracy statement, the README and the guide say two real games were read on one driver (they said none); Privacy Policy 2.21 and Disclaimer 2.17 with it |
| D146 | [`06_DATA_MODEL.md`](06_DATA_MODEL.md) §sensor_blobs | the Agent's samples were 1.32 s apart under a 1 s setting (it waited the whole interval after its work): sample on a fixed period, and store the rate the samples had |
| D147 | [`03_METRICS.md`](03_METRICS.md) | the session summary's latency fields stay N/A: the driver's total is simulation start to GPU end, not input to display |
| D148 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Native logging | asked on 2026-11-14 with item 46's Cyberpunk 2077 numbers (active render longer than the span between the GPU marks): count the frames, and write the counts **only to the Overlay's native log** — no layout, schema, App or string change |
| D149 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Native logging | told that a bug report carries the five newest Overlay logs and so these lines: **no change to the legal texts** — the Privacy Policy names the logs of the component loaded into games, and the lines are counts and sums of microseconds |
| D150 | [`07_IPC.md`](07_IPC.md) §Region 5 | the samples that carry no report are mended **on both sides**: the Overlay reads and publishes on a grid of one second, and the Agent no longer overwrites a reading that waits for a sample |
| D151 | [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Native logging | the owner's word: Reflex is checked together with frame generation — without it the owner's games do not switch Reflex on, and a low-latency switch that reads on and off is the NVIDIA App's override. A batch of counts belongs to one state of the sleep status |
| D152 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §The inventory | the two GPU labels say what each figure is — *the driver's own figure*, *between the driver's start and end marks* — not "idle time excluded / included": item 47 counted the active time as the longer in Cyberpunk 2077. Strings only; numbers and means unchanged |
| D153 | [`spike-notes/30-reflex-latency.md`](spike-notes/30-reflex-latency.md) *Added 2026-11-14* | the owner's word: of Reflex it is enough to tell whether a game has it. Reflex and PCL leave train D (`sl.pcl.markers` stays refused), and D141's stop is not changed for a game's first seconds. My reading of one sentence; nothing shipped is removed by it |
| D154 | [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §The trains, [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) §M15 | Advanced metrics is closed: train D is not built, the ten Streamline entries that waited for a spike are not read (`RefusedRoute`), rule 4 is not amended for them. Nothing shipped is removed |
| D155 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | controllers are read by the Agent, from outside the game: no hook is added, rule 4 is untouched, and a session that is not hooked has the numbers too |
| D156 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | the reading is on unless switched off (`capture.input_devices`), and reads nothing until the Privacy Policy that describes it (2.22 or later) has been accepted on the PC |
| D157 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | built without a probe run on real controllers first ("Xây thẳng"): every source reports what it did, and what was verified against hand-made data only is said in each pull request |
| D158 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | keyboard and mouse are one class known only by `GetLastInputInfo`'s time: no Raw Input for them, no hook, no key in the process |
| D159 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §DualSense | nothing is ever sent to a controller; a DualSense on Bluetooth has a battery level only where Steam or the game switched it |
| D160 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | the numbers are columns of `sessions`, so they are in the CSV/JSON export and in a bug bundle, and the Privacy Policy's §3A says so |
| D161 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | a session that is not hooked is duration, hardware telemetry, the reason — and what the Agent reads from outside the game; CLAUDE.md's sentence is amended with the pull request that makes it true |
| D162 | [`03_METRICS.md`](03_METRICS.md) §Input | also built: switches of method and the share with no input, marks on the frame-time chart, a battery's first and last level and its rate, the kind of connection, the Steam Input note, a trend metric, a Compare row, "usually played with" |
| D163 | [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) §The owner's request and decisions | "supports controllers" is three pieces of evidence and never a verdict: Steam's own file on the disk (no network), the input libraries the game loaded, how the game was played |
| D109 | [`../legal/DISCLAIMER.md`](../legal/DISCLAIMER.md) §5, §5A | asked on 2026-11-11 with both sentences quoted: the Disclaimer goes to 2.15 with the Privacy Policy — PawnIO and the admin mode are no longer "for the CPU temperature" alone; with Advanced metrics on they also read the processor's, the memory's and the motherboard's sensors. The admin mode's disclosure follows (`agent-admin-dialog/2`) |
---
+46 -16
View File
@@ -1,43 +1,47 @@
# Games FrameLedger has been run with
*As of 2026-09-27 (0.1.0-beta.10). Built from a copy of the developer's own FrameLedger database and the repository's
*As of 2026-11-06 (0.1.0-beta.34). Built from a copy of the developer's own FrameLedger database and the repository's
measurement notes, and reviewed by the project's owner before it was published. It lists what happened on one machine;
it promises nothing about any other.*
*(Not re-read for 0.1.0-beta.11 or 0.1.0-beta.12 — noted 2026-10-04. Refreshing it needs a new copy of the database and
the owner's review; sessions since 2026-09-27, and the game-memory figures beta.12 adds, are not here.)*
*~~(Not re-read for 0.1.0-beta.11 or 0.1.0-beta.12 — noted 2026-10-04. Refreshing it needs a new copy of the database and
the owner's review; sessions since 2026-09-27, and the game-memory figures beta.12 adds, are not here.)~~ Refreshed
2026-11-06 from a copy of the database taken that day, and reviewed by the owner with the plan of 0.1.0-beta.34: the
sessions since 2026-09-27 are below, with the graphics memory a game used where it adds something.*
**The machine.** Every row below was run on one PC: an NVIDIA GeForce RTX 5080 (the only graphics card), an Intel Core
i7-14700KF, 32 GB of DDR5, Windows 11 (Insider builds 26300 → 29648) and a 2560×1440, 240 Hz monitor. **No AMD or Intel
i7-14700KF, 32 GB of DDR5, Windows 11 (Insider builds 26300 → 29648) and a 2560×1440, 240 Hz monitor; the graphics
driver went from 32.0.16.1714 to 32.0.16.2114 over the sessions in the database. **No AMD or Intel
graphics card and no second machine has been tested**, and ~~**no real Vulkan or OpenGL game has been measured** — every
measurement below is Direct3D 11 or 12~~ **one real Vulkan game has been measured (Everhex Demo, 2026-10-13) and no
OpenGL game** — every other measurement below is Direct3D 11 or 12 *(corrected 2026-10-14)*. What the tiers mean and what FrameLedger cannot do: [LIMITATIONS.md](../LIMITATIONS.md).
**Two periods.** From 2026-08-03 to 2026-09-06 the games were measured with the developer's capture tool, one run at a
time against the game's own settings menu (reports in [spike-notes.md](spike-notes.md)). From 2026-09-16 the shipped
app and agent recorded them as a user would; the database holds the sessions from 2026-09-22 on — earlier ones went
with the 2026-09-16 recovery and with library entries removed since.
app and agent recorded them as a user would; ~~the database holds the sessions from 2026-09-22 on~~ the copy of
2026-11-06 holds the sessions from 2026-10-04 on — earlier ones went with the 2026-09-16 recovery and with library
entries and sessions removed since; a row citing the database before 2026-10-04 was read from the copy of 2026-09-27.
## Measured — hooked (Tier 1)
| Game | Engine | API | What was measured | When | Evidence |
|---|---|---|---|---|---|
| Cyberpunk 2077 | — | D3D12 | DLSS Balanced 1485×835 → 2560×1440, exact against its settings file; DLSS Frame Generation off ×1.00, ×3 (61.4 → 183.5 FPS), ×4 (58.7 → 234.1); FSR 3 + FSR frame generation ×2 (74.3 → 148.6); ray tracing *Yes* with path tracing on and *No* with it off, as set | 2026-08-15 → 09-05, about 25 runs | [spike-notes §8, §9, §11](spike-notes.md) |
| Cyberpunk 2077 | — | D3D12 | DLSS Balanced 1485×835 → 2560×1440, exact against its settings file; DLSS Frame Generation off ×1.00, ×3 (61.4 → 183.5 FPS), ×4 (58.7 → 234.1); FSR 3 + FSR frame generation ×2 (74.3 → 148.6); ray tracing *Yes* with path tracing on and *No* with it off, as set; on 2026-10-20, ×2, ×3 and ×4 chosen in the game's menu within one session read as separate states, ×4.0 at 63.8 → 255.3 FPS, with ray tracing and Ray Reconstruction *Yes* | 2026-08-15 → 10-20, about 26 runs | [spike-notes §8, §9, §11, §23](spike-notes.md), database |
| Dying Light: The Beast | — | D3D12 | DLSS Frame Generation ×4: Native 70.5 → Displayed 282.1 FPS, counted through DXGI; FSR + FSR frame generation ×2 (70.5 → 140.9); ray tracing *Yes* | 2026-09-04 → 09-06 | [spike-notes §9](spike-notes.md), [03_METRICS §Frame Generation](03_METRICS.md) |
| Hell Is Us | Unreal Engine 5.5.4 | D3D12 | DLSS Frame Generation ×4 (76.5 → 306.2); DLSS shown as *driver-reported* (it bypasses Streamline); FSR + FSR frame generation ×2 (85.7 → 171.4); XeSS frame generation shown as Presented FPS only; ray tracing *No* | 2026-09-03 → 09-16 | [spike-notes §8, §9](spike-notes.md) |
| Black Myth: Wukong | Unreal Engine 5.0.0 | D3D12 | DLSS Frame Generation ×4 (55.4 → 221.5); ray tracing *Yes* (16.2 % of frames in the app's session); a session that changed state showed its steady state (×4.1) instead of an average | 2026-08-20 → 09-16 | [spike-notes §9, §14](spike-notes.md) |
| Clair Obscur: Expedition 33 | Unreal Engine 5.4.4 | D3D12 | DLSS + Frame Generation ×4 (54.9 → 218.4) and ×2 (76.0 → 149.8); frame generation off read *none*, with DXGI agreeing; ray tracing *No* | 2026-09-04 → 09-06 | [spike-notes §9](spike-notes.md) |
| Clair Obscur: Expedition 33 | Unreal Engine 5.4.4 | D3D12 | DLSS + Frame Generation ×4 (54.9 → 218.4) and ×2 (76.0 → 149.8); frame generation off read *none*, with DXGI agreeing; ray tracing *No*; the NVIDIA App's *Fixed ×6* read ×6.0 (81.8 → 490.9) on 2026-10-20 | 2026-09-04 → 10-20 | [spike-notes §9, §23](spike-notes.md), database |
| Rune Factory: Guardians of Azuma | Unreal Engine (version not in the executable) | D3D12 | DLSS Frame Generation ×4 (128.9 → 513.0); FSR frame generation compiled into the game counted ×2 but not named; ray tracing *No* (the game has none) | 2026-08-20 → 09-06 | [spike-notes §9](spike-notes.md) |
| Kingdom Come: Deliverance II | — | D3D12 | FSR (3.1 or 4) at 1505×847 → 2560×1440; frame generation *none* (the game ships none); ray tracing *No* | 2026-09-04 | [spike-notes §9](spike-notes.md) |
| Lies of P | Unreal Engine 4.27.2 | D3D12 | FSR 3 at 1506×848 → 2560×1440 with FSR frame generation ×2.04 (57.1 → 116.4) in the app; frame generation off read *none*; DLSS shown as *driver-reported*; ray tracing *No* | 2026-08-03 → 09-16 | [spike-notes §7, §9, §14](spike-notes.md) |
| Lies of P | Unreal Engine 4.27.2 | D3D12 | FSR 3 at 1506×848 → 2560×1440 with FSR frame generation ×2.04 (57.1 → 116.4) in the app; frame generation off read *none*; DLSS shown as *driver-reported*; ray tracing *No*; on 2026-10-19 and 10-20, FSR 3 at 1707×961 and at 1280×720 → 2560×1440 with FSR frame generation ×2.0 (54.6 → 109.6, 57.6 → 116.7) and no FSR quality mode — the game never asked for one while the hook was in | 2026-08-03 → 10-20 | [spike-notes §7, §9, §14, §22, §23](spike-notes.md), database |
| Onimusha: Way of the Sword (demo) | RE Engine | D3D12 | DLSS + DLSS Frame Generation: ×3.67 and ×3.93 with the capture tool, ×4.00 in the app's steady state (60 → 240 FPS), ×3.0 (60 → 180) in the app on 2026-09-22; ray tracing *Yes* | 2026-09-05 → 09-22 | [spike-notes §9, §14](spike-notes.md), database |
| Cronos: The New Dawn (demo) | Unreal Engine (version not in the executable) | D3D12 | DLSS with frame generation off (105 FPS, *none*); on ×2 (86.8 → 172.7), DLSS Frame Generation named; ray tracing *Yes* | 2026-09-06 → 09-16 | [spike-notes §9, §14](spike-notes.md) |
| Alan Wake 2 | — | D3D12 | Path tracing Ultra with DLSS and Ray Reconstruction: ray tracing *Yes* (96.8 % of frames), frame generation read ×1.00 (the game did not apply the setting that run). On 2026-09-04: DLSS with ray tracing *Yes* (99 % of frames), 92.5 FPS, frame generation *none* by count; then ray tracing off (*No*), DLSS, Presented 114.5 FPS. The quality preset was never read | 2026-08-04 → 09-04 | [spike-notes §8](spike-notes.md), run reports |
| SILENT HILL 2 | Unreal Engine 5.1 | D3D12 | DLSS Frame Generation identified, frames not counted (Presented 78 FPS); DLSS *driver-reported*; ray tracing *Yes* | 2026-09-25 | database |
| DARK SOULS III | FromSoftware | D3D11 | Presented 59.8 FPS; no upscaler, frame generation or ray tracing loaded | 2026-09-03 | [spike-notes §9](spike-notes.md) |
| Sekiro: Shadows Die Twice | FromSoftware | D3D11 | Presented 60 FPS (the game's cap), two sessions of 15 min and 2 h 18 min | 2026-09-26 | database |
| SILENT HILL 2 | Unreal Engine 5.1 | D3D12 | DLSS Frame Generation identified, frames not counted (Presented 78 FPS) on 2026-09-25; then under the NVIDIA App's *Fixed ×6* ×6.0 (53.3 → 320.0) and under its *Dynamic, up to ×6* states ×2.0, ×3.0 and ×6.1; DLSS *driver-reported*; ray tracing *Yes* | 2026-09-25 → 10-20 | database, [spike-notes §22, §23](spike-notes.md) |
| DARK SOULS III | FromSoftware | D3D11 | Presented 59.8 FPS; no upscaler, frame generation or ray tracing loaded; then 34 sessions at the game's 60 FPS cap, the longest 6 h 12 min, four of them ended by the game crashing | 2026-09-03 → 10-25 | [spike-notes §9](spike-notes.md), database |
| Sekiro: Shadows Die Twice | FromSoftware | D3D11 | Presented 60 FPS (the game's cap): two sessions of 15 min and 2 h 18 min, then three more of up to 1 h 6 min | 2026-09-26 → 10-29 | database |
| Yu-Gi-Oh! Master Duel | Unity 6000.0 | D3D11 | Presented 239 FPS across six sessions (up to 5 h 25 min) | 2026-09-22 → 09-25 | database |
| GIRLS' FRONTLINE 2: EXILIUM | Unity 2019.4 | D3D11 | Presented 53–101 FPS; the last three sessions under the user-mode anti-cheat exception (NetEase Yidun), one of them 3 h 9 min, all ended normally | 2026-09-21 → 09-27 | database, [HANDOFF](HANDOFF.md) |
| GIRLS' FRONTLINE 2: EXILIUM | Unity 2019.4 | D3D11 | Presented 53–101 FPS; the last three sessions under the user-mode anti-cheat exception (NetEase Yidun), one of them 3 h 9 min, all ended normally; ten more under the exception, 28–72 FPS, the longest 2 h 9 min, all ended normally | 2026-09-21 → 10-20 | database, [HANDOFF](HANDOFF.md) |
| Aniimo | Unity 2022.3 | D3D12 | DLSS at 1707×960 → 2560×1440, frame generation *none* (×1.00, 205 FPS), ray tracing *No* — then blocked (below) | 2026-09-22 | database |
| AbyssMemory | Unreal Engine 4.26.1 | D3D11 | Presented 55 FPS, one short session | 2026-09-22 | database |
| Umamusume: Pretty Derby | Unity 2022.3 | D3D11 | Presented 37 FPS, one session | 2026-09-27 | database |
@@ -45,6 +49,21 @@ with the 2026-09-16 recovery and with library entries removed since.
| Flower in Us | RPG Maker MV/MZ (NW.js) | D3D11 | Presented 59.8 FPS over three swap chains, after two refusals that found the game's browser process ambiguous (fixed) | 2026-09-03 → 09-04 | [spike-notes §9](spike-notes.md) |
| UMIGARI | Unreal Engine 5.5.4 | D3D12 | Presented 95–97 FPS, frame generation *none*, ray tracing *No* | 2026-09-05 → 09-06 | run reports only |
| Everhex Demo | — | Vulkan | Presented 239.7 FPS (1 % low 213.5), 14,955 frames in 69 s, through FrameLedger's Vulkan layer — the first real Vulkan game measured; started by Play set to *Start directly* | 2026-10-13 | database *(row added 2026-10-14 from a copy of the database)* |
| The Witcher 3: Wild Hunt — Remastered | — | D3D12 | DLSS 1707×960 → 2560×1440 with DLSS Frame Generation ×2.0 (118.8 → 237.4 FPS); ray tracing and Ray Reconstruction *Yes*; under the NVIDIA App's *Fixed ×6* it still presented ×2.0 — the setting is not the frames | 2026-10-04 → 10-20 | database, [spike-notes §23](spike-notes.md) |
| Phantom Blade Zero | Unreal Engine (version not in the executable) | D3D12 | DLSS *driver-reported*; DLSS Frame Generation under the NVIDIA App's *Dynamic* mode: up to ×6 held ×4.0 (59.5 → 235.6 FPS) and ×5.2 (44.3 → 231.2), up to ×3 held ×3.0 (58–76 → 175–227) for 75–94 % of each session longer than an hour; ray tracing *Yes*; thirteen sessions, the longest 6 h 29 min (5.26 million frames); up to 9.6 GB of graphics memory | 2026-11-01 → 11-05 | database, [spike-notes §26](spike-notes.md) |
| Hogwarts Legacy | Unreal Engine (version not in the executable) | D3D12 | With the game's *Neural Rendering* on, DLSS 5's runtime (`nvngx_dlssnr.dll` 310.9.16.0) loaded from the game's folder — the first game seen loading it; DLSS *driver-reported*; DLSS Frame Generation not counted (non-uniform, ×1.97 overall), Presented 140.2 FPS; ray tracing *Yes*, Ray Reconstruction *No* | 2026-11-06 | database, [spike-notes §26](spike-notes.md) |
| Valor Mortis (Xbox app) | Unreal Engine | D3D12 | DLSS Frame Generation ×6.0 (39.9 → 239.2 FPS), DLSS *driver-reported*, ray tracing *No*; recorded in attach mode | 2026-11-04 | database |
| inKONBINI (Xbox app) | Unity 6000.0 | D3D11 | Presented 205.9 FPS, one short session | 2026-11-03 | database |
| DARK SOULS II: Scholar of the First Sin | FromSoftware | D3D11 | The game's 60 FPS cap, nine sessions up to 5 h 29 min; one was unhooked by the safety check and recorded on | 2026-10-09 → 10-12 | database |
| Mithril | Unreal Engine (version not in the executable) | D3D12 | Presented 91.5–119.4 FPS, ray tracing *Yes*, four sessions up to 3 h 4 min | 2026-10-12 → 10-16 | database |
| KEEP IT ON HER | Unreal Engine 5.8 | D3D12 | Presented 59.5–108.8 FPS, ray tracing *Yes* | 2026-10-12 → 10-13 | database |
| WALTO | Unity 6000.3 | D3D12 | Presented 212.5 FPS over 2 h 41 min (2,057,022 frames) | 2026-10-08 | database |
| White Mask Mansion | Unity 6000.1 | D3D12 | Presented 286–309 FPS, five sessions | 2026-10-27 | database |
| LIMINAL: Within Demo | Unity 2022.3 | D3D11 | Presented 222.5–228.6 FPS | 2026-10-12 | database |
| Sleepytime Village | Unity 2021.3 | D3D11 | Presented 906–1,440 FPS (the game sets no cap) | 2026-10-25 → 10-26 | database |
| Scarlet Wolf | Unity 2021.3 | D3D11 | Presented 128.5 FPS | 2026-10-15 | database |
| March Out! Playtest | Unity 2022.3 | D3D11 | Presented 118.2 FPS | 2026-10-08 | database |
| A Space for the Unbound – Prologue (GOG) | Unity 2019.4 | D3D11 | Presented 61.0 FPS, started by FrameLedger | 2026-10-20 | database |
## Recorded without measurement (Tier 2)
@@ -52,8 +71,14 @@ with the 2026-09-16 recovery and with library entries removed since.
|---|---|---|---|
| HELLO, HELLO WORLD! | A 32-bit game (RPG Maker MV): refused by the guard once, and two sessions with hooking switched off — each recorded with its duration and sensors. *Its itch.io download puts malware in front of the game — do not run it (noted 2026-10-13, [spike-notes §20](spike-notes.md))* | 2026-09-23 | database |
| Umamusume: Pretty Derby | Three sessions played with hooking switched off for the game | 2026-09-27 | database |
| Sekiro: Shadows Die Twice | One session played with hooking switched off | 2026-09-25 | database |
| Sekiro: Shadows Die Twice | Two sessions played with hooking switched off | 2026-09-25, 10-07 | database |
| GIRLS' FRONTLINE 2: EXILIUM | One session whose process Windows would not let FrameLedger open, and two recovered after the agent itself stopped mid-session (*interrupted*) | 2026-09-22 → 09-23 | database |
| Bound Between Desks | Two sessions with hooking switched off (a Ren'Py 8.5 game) | 2026-10-15 → 10-20 | database |
| Phantom Blade Zero | Three sessions after a game update changed its executable on 2026-11-04, until hooking was consented to again on 11-05; one refused because a module list came back incomplete | 2026-11-04 → 11-05 | database |
| The Witcher 3: Wild Hunt — Remastered | Three sessions before hooking was consented to | 2026-10-13 → 10-20 | database |
| Mithril | Two sessions before hooking was consented to | 2026-10-16 | database |
| Hogwarts Legacy | One session with hooking switched off; Windows logged a crash of the game | 2026-11-06 | database |
| Valor Mortis, WALTO, Scarlet Wolf, Lies of P, A Space for the Unbound – Prologue | One session each with hooking switched off | 2026-10-08 → 11-04 | database |
## Refused by the anti-cheat guard
@@ -64,7 +89,10 @@ with the 2026-09-16 recovery and with library entries removed since.
| Aniimo | NetEase Yidun with a kernel driver (`NEPKernel.sys`) in the game's folder — never eligible for an exception | 2026-09-25 | [19_SAFETY](19_SAFETY_AND_ANTICHEAT.md) |
| Goose Goose Duck | Easy Anti-Cheat's service running, and its folder | 2026-08-04 | [spike-notes §13](spike-notes.md) |
| Neverness To Everness | Anti-Cheat Expert's kernel driver in the game's folder | 2026-08-04 | [spike-notes §13](spike-notes.md) |
| Wuthering Waves | Anti-Cheat Expert, kept beside the game's executable four folder levels in, where the check did not look before beta.21: a session with hooking on was injected and unhooked 37 seconds later, when the 30-second re-check met ACE's driver (`ACE-BASE.sys`); nothing was measured. Since beta.21 its folder and its Steam id refuse it before any injection *(row added 2026-10-13 from the agent's log)* | 2026-10-12 | [spike-notes §19](spike-notes.md) |
| Wuthering Waves | Anti-Cheat Expert, kept beside the game's executable four folder levels in, where the check did not look before beta.21: a session with hooking on was injected and unhooked 37 seconds later, when the 30-second re-check met ACE's driver (`ACE-BASE.sys`); nothing was measured. Since beta.21 its folder and its Steam id refuse it before any injection *(row added 2026-10-13 from the agent's log)*; two sessions recorded without measurement on 2026-10-31 | 2026-10-12 → 10-31 | [spike-notes §19](spike-notes.md), database |
| Arknights: Endfield | Anti-Cheat Expert's folder (`AntiCheatExpert`) beside the game; one session recorded without measurement | 2026-10-31 | database |
| Stella Sora | Anti-Cheat Expert's folder (`AntiCheatExpert`); three sessions recorded without measurement | 2026-10-30 → 10-31 | database |
| Umamusume: Pretty Derby | CrackProof, on FrameLedger's title list since beta.11 (Steam 3224770) — measured before the list named it (above) | 2026-10-03 | [19_SAFETY](19_SAFETY_AND_ANTICHEAT.md) |
| Counter-Strike 2 | Valve Anti-Cheat — refused by FrameLedger's title list since beta.8. On 2026-08-04 the checks alone allowed it (VAC is invisible to them); nothing was injected | 2026-08-04, beta.8 | [spike-notes §13](spike-notes.md), [19_SAFETY](19_SAFETY_AND_ANTICHEAT.md) |
Also refused before any injection: **Deadly Heart Gambit** — a 32-bit game (2026-08-03).
@@ -74,4 +102,6 @@ Also refused before any injection: **Deadly Heart Gambit** — a 32-bit game (20
The Unreal Engine version (beta.10) was checked against ten installed Unreal games without running them: Hell Is Us
5.5.4, Black Myth: Wukong 5.0.0, Clair Obscur: Expedition 33 5.4.4, Lies of P 4.27.2, SILENT HILL 2 5.1, UMIGARI 5.5.4,
AbyssMemory 4.26.1, **Martha Is Dead 4.27.0**; Cronos (demo) and Rune Factory name no version in their executables and
read *not found*. ([spike-notes §15](spike-notes.md), [05_DETECTION §Engine version](05_DETECTION.md))
read *not found*. ([spike-notes §15](spike-notes.md), [05_DETECTION §Engine version](05_DETECTION.md)) Read since on the games
in the database: KEEP IT ON HER 5.8 and Wuthering Waves 4.26.2; Phantom Blade Zero, Hogwarts Legacy and Mithril name no
version in their executables, and an Xbox game's executable cannot be read outside its package (Valor Mortis).
+6 -1
View File
@@ -11,7 +11,7 @@ This folder is for people who build, change or review FrameLedger. **If you only
| Find your way around first | The [wiki](https://github.com/poli0981/frameledger/wiki)'s developer pages (source: [`../wiki/`](../wiki/), beta.17) — a short page per part, each linking here |
| Learn the system | [`../CLAUDE.md`](../CLAUDE.md) (the rules and the layout), then the reading order below |
| Build it | [`12_BUILD.md`](12_BUILD.md) — `./build.ps1 check` is the whole gate, the same one CI runs |
| Contribute or fork | [`../CONTRIBUTING.md`](../CONTRIBUTING.md), [`../FORKING.md`](../FORKING.md), [`../NOTICE`](../NOTICE) |
| Contribute or fork | [`../CONTRIBUTING.md`](../CONTRIBUTING.md), [`../FORKING.md`](../FORKING.md), [`../NOTICE`](../NOTICE), [`../CODE_OF_CONDUCT.md`](../CODE_OF_CONDUCT.md) *(it and `SECURITY.md`, below, added 2026-11-09: the index named neither)* |
**Status lives in four places, and nowhere else** — check a status claim against the code before you plan on it:
@@ -69,6 +69,7 @@ are safety items, and something listed there as unresolved should not be built t
| [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) | Telemetry layers (DXGI/PDH, LibreHardwareMonitor, NVAPI), the game process's memory, licences |
| [`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) | The anti-cheat guard, what a finding does, the user-mode exception, what will never be built |
| [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md) | Every unresolved question, defect and measurement, with its status marker |
| [`21_INPUT_DEVICES.md`](21_INPUT_DEVICES.md) | Controllers and keyboard-and-mouse activity: what is read from outside the game, from where, and what never is |
| [`HANDOFF.md`](HANDOFF.md) | Sequencing, the open owner-only items, the index of the owner's decisions |
| [`handoff/`](handoff/) | Since beta.24 (D66): the history of every train and phase, the passed owner-only items, the traps — moved out of `HANDOFF.md` verbatim |
| [`open-questions/`](open-questions/) | Since beta.24 (D66): the entries §S24 marks resolved, moved out of `20_OPEN_QUESTIONS.md` verbatim — a table there lists them |
@@ -87,5 +88,9 @@ are safety items, and something listed there as unresolved should not be built t
- [`../legal/`](../legal/) — the documents a user accepts, the accuracy statement, the third-party notices and licence
texts, the trademark note.
- [`../LIMITATIONS.md`](../LIMITATIONS.md) — what the whole software cannot do, in a player's words.
- [`../ACCESSIBILITY.md`](../ACCESSIBILITY.md) — the accessibility statement (beta.35, D88): the goal, what works with the
keyboard, a screen reader and contrast themes, the known limits and how a barrier is reported; shown in the App under
Help ▸ Accessibility. Its checks are [`08_UI.md`](08_UI.md) §Accessibility.
- [`../SECURITY.md`](../SECURITY.md) — the private route for a vulnerability, and why the guard has no override.
- [`../rules/detection-rules.json`](../rules/detection-rules.json) — engine, platform and capability signatures and the
anti-cheat blocklist, validated by `tools/rules-validate.ps1`.
+33
View File
@@ -0,0 +1,33 @@
<!-- The second alpha's notes (D73, D82), on main as every release's are; the alpha branch takes them with its sync. CHANGELOG.md lists every release. -->
## [0.1.0-alpha.2] - 2026-11-03
> ⚠ **EXPERIMENTAL BUILD — YOU USE IT AT YOUR OWN RISK.** This is an **alpha**: the source of `0.1.0-beta.33` built on
> **.NET 11 Release Candidate 3** (SDK `11.0.100-rc.3.26522.148`, runtime `11.0.0-rc.3.26522.148`), a version of .NET
> Microsoft has not finished, with the C# 15 compiler. It has not been tested the way a beta is, and it can crash, record
> wrongly or fail to update in ways no beta has. **If you want results you can rely on, use the beta.** Before it runs it
> shows the *Alpha notice* (`legal/ALPHA_NOTICE.md` 1.0), which you must accept, with its own box to tick.
**The second alpha: `0.1.0-beta.33` on .NET 11 RC3** — games from the Xbox app and the Microsoft Store (the beta's
notes, `docs/changelog/0.1.0-beta.33.md`, say what that brings). What an alpha is and how to go back: the user guide's
*Alpha builds* page (`guide/11-alpha-builds.md`), and the *Alpha builds* section of `LIMITATIONS.md`.
**Updating from `0.1.0-alpha.1`:** Help ▸ Check for updates, then *Restart to update* with no game running — an alpha
updates only to newer alphas. Before it first opens your database it copies it to
`%LOCALAPPDATA%\FrameLedger\backups\ledger-before-0.1.0-alpha.2.db` (the newest three copies are kept). The terms window
shows the **Privacy Policy 2.13** once and, as every time an alpha's terms window opens, waits for the Alpha notice's box
to be ticked again; the notice itself stays 1.0. The ledger stays at schema 21; the detection rules move to `2026.11.1`.
**Installing it over a beta:** download `FrameLedger.App-alpha-Setup.exe` below, verify its hash against
`SHA256SUMS.txt`, and let it replace the FrameLedger already installed; the Alpha notice is shown first. **Going back:**
install the newest beta's `FrameLedger.App-win-Setup.exe` over it; your database stays where it is.
**Windows and your PC:** as for the first alpha — Microsoft supports .NET 11 on Windows 11 24H2 and later, and on Windows
10 only in its LTSC and IoT editions; it needs a processor with SSE4.2 and POPCNT. .NET 11 RC3 is supported until .NET 11
ships (10 November 2026); the next alpha moves to the final release.
### What differs from 0.1.0-beta.33
The same as `0.1.0-alpha.1` differed from `0.1.0-beta.32` (`docs/changelog/0.1.0-alpha.1.md`): the runtime and the
compiler, the three libraries serviced with the runtime, what .NET 11 itself changes — and the part of FrameLedger loaded
into a game is the beta's.
+36
View File
@@ -0,0 +1,36 @@
<!-- The third alpha's notes (D73, D87), on main as every release's are; the alpha branch takes them with its sync. CHANGELOG.md lists every release. -->
## [0.1.0-alpha.3] - 2026-11-06
> ⚠ **EXPERIMENTAL BUILD — YOU USE IT AT YOUR OWN RISK.** This is an **alpha**: the source of `0.1.0-beta.34` built on
> **.NET 11 Release Candidate 3** (SDK `11.0.100-rc.3.26522.148`, runtime `11.0.0-rc.3.26522.148`), a version of .NET
> Microsoft has not finished, with the C# 15 compiler. It has not been tested the way a beta is, and it can crash, record
> wrongly or fail to update in ways no beta has. **If you want results you can rely on, use the beta.** Before it runs it
> shows the *Alpha notice* (`legal/ALPHA_NOTICE.md` 1.0), which you must accept, with its own box to tick.
**The third alpha: `0.1.0-beta.34` on .NET 11 RC3** — DLSS 5 and DLSS 4.5 named from what the game and the driver did,
filters on Games and Compare, and a Games page that opens at once (the beta's notes, `docs/changelog/0.1.0-beta.34.md`,
say what that brings). What an alpha is and how to go back: the user guide's *Alpha builds* page
(`guide/11-alpha-builds.md`), and the *Alpha builds* section of `LIMITATIONS.md`.
**Updating from `0.1.0-alpha.2`:** Help ▸ Check for updates, then *Restart to update* with no game running — an alpha
updates only to newer alphas. Before it first opens your database it copies it to
`%LOCALAPPDATA%\FrameLedger\backups\ledger-before-0.1.0-alpha.3.db` (the newest three copies are kept). The terms window
shows the **Disclaimer 2.13** and the **Privacy Policy 2.14** once and, as every time an alpha's terms window opens, waits
for the Alpha notice's box to be ticked again; the notice itself stays 1.0. The ledger moves to schema 22, which an
alpha.2 cannot open (the copy is the database as alpha.2 left it; going back is to the newest beta, which opens it); the
detection rules move to `2026.11.2`.
**Installing it over a beta:** download `FrameLedger.App-alpha-Setup.exe` below, verify its hash against
`SHA256SUMS.txt`, and let it replace the FrameLedger already installed; the Alpha notice is shown first. **Going back:**
install the newest beta's `FrameLedger.App-win-Setup.exe` over it; your database stays where it is.
**Windows and your PC:** as for the first alpha — Microsoft supports .NET 11 on Windows 11 24H2 and later, and on Windows
10 only in its LTSC and IoT editions; it needs a processor with SSE4.2 and POPCNT. .NET 11 RC3 is supported until .NET 11
ships (10 November 2026); the next alpha moves to the final release.
### What differs from 0.1.0-beta.34
The same as `0.1.0-alpha.1` differed from `0.1.0-beta.32` (`docs/changelog/0.1.0-alpha.1.md`): the runtime and the
compiler, the three libraries serviced with the runtime, what .NET 11 itself changes — and the part of FrameLedger loaded
into a game is the beta's.
+42
View File
@@ -0,0 +1,42 @@
<!-- The fourth alpha's notes (D73, D98), on main as every release's are; the alpha branch takes them with its sync. CHANGELOG.md lists every release. -->
## [0.1.0-alpha.4] - 2026-11-11
> ⚠ **EXPERIMENTAL BUILD — YOU USE IT AT YOUR OWN RISK — AND THE LAST ALPHA FOR NOW.** This is an **alpha**: the source
> of `0.1.0-beta.37` built on **.NET 11.0.0**, Microsoft's final release of .NET 11 (SDK `11.0.100`) — the same .NET the
> betas run on from `0.1.0-beta.37`. It exists to bring an installed alpha off the release candidate. **After it the
> alphas pause**, and a paused alpha gets no update — the list of anti-cheat software FrameLedger refuses included.
> **Install the latest beta over it.** Before it runs it shows the *Alpha notice* (`legal/ALPHA_NOTICE.md` 1.1), which
> you must accept, with its own box to tick.
**The fourth alpha: `0.1.0-beta.37` on .NET 11.0.0.** The alphas were made to try .NET 11 before the betas relied on it.
.NET 11 shipped on 10 November 2026 and the betas moved to it the next day (the beta's notes,
`docs/changelog/0.1.0-beta.37.md`), so an alpha and a beta now differ in nothing but what an alpha adds for itself. This
one is published so that a copy of `0.1.0-alpha.3`, which runs on a release candidate Microsoft no longer supports,
reaches the finished .NET 11 by an update — an alpha never updates to a beta.
**Then install the beta.** No alpha will follow until there is a newer .NET to try (.NET 12's first previews, expected
in 2027). Until then this build stands still: no fixes, and **its anti-cheat list is the one it shipped with**, because
that list reaches FrameLedger only inside an update. Its banner says so each time it starts. Download the newest beta's
`FrameLedger.App-win-Setup.exe`, verify its hash against that release's `SHA256SUMS.txt`, and let it replace the alpha;
your database stays where it is, and the beta opens it. What an alpha is and how to go back: the user guide's *Alpha
builds* page (`guide/11-alpha-builds.md`), and the *Alpha builds* section of `LIMITATIONS.md`.
**Updating from `0.1.0-alpha.3`:** Help ▸ Check for updates, then *Restart to update* with no game running. Before it
first opens your database it copies it to `%LOCALAPPDATA%\FrameLedger\backups\ledger-before-0.1.0-alpha.4.db` (the
newest three copies are kept). The terms window shows the **EULA 1.7**, the **Disclaimer 2.14** and the **Alpha notice
1.1**, and waits for the notice's box to be ticked. The ledger stays at schema 22 and the detection rules at
`2026.11.2`. The download is large, because the .NET runtime inside the package is replaced whole. **One limit:**
FrameLedger's updater looks at the newest ten releases, so once ten betas have been published after this alpha, an
older alpha no longer finds it — install the beta by hand then, as above.
**Windows and your PC:** as for the beta from `0.1.0-beta.37` — Microsoft supports .NET 11 on Windows 11, and on Windows
10 only in its LTSC and IoT editions; FrameLedger is untested and unsupported on Windows 10 22H2; a processor without
SSE4.2 and POPCNT cannot start it.
### What differs from 0.1.0-beta.37
Nothing in the runtime, the compiler or the libraries: both are built on .NET 11.0.0 from the same source, and the part
of FrameLedger loaded into a game is the beta's. What an alpha adds is its own: the Alpha notice in the terms window, the
ALPHA badge and the banner — which in this build says that the alphas are paused — the locked update channel, and the
copy of the database before it is first opened.
+81
View File
@@ -0,0 +1,81 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.34 was cut, 2026-11-06 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.34] - 2026-11-06
**The thirty-fourth pre-release: DLSS 5 and DLSS 4.5 named from what the game and the driver did, filters on Games
and Compare, and a Games page that opens at once.** The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**Updating from `0.1.0-beta.33`:** Help ▸ Check for updates, then *Restart to update* with no game running. The
Disclaimer (2.13) and the Privacy Policy (2.14) are shown once for you to accept; the EULA does not change. The ledger
moves to schema 22, which a beta.33 cannot open. The detection rules move to `2026.11.2`, so every game in the
library is read again once. DLSS 4.5 under the NVIDIA App's *Dynamic* frame generation reads on sessions you already
have; whether DLSS 5's neural rendering ran is kept for sessions recorded from now on.
### Added
- **Filters on Games and Compare** (owner request 2026-11-06, decision D85). Games: store, engine, state, installed and
played, beside the search and sort, with *Clear filters*; the title counts what is shown of all. Compare: game, tier
and frame generation over every session (it listed the newest 200), with a count; a ticked session stays listed.
### Changed
- **The list of games FrameLedger has been run with is current again** (`docs/LIST_GAME_TESTED.md`, decision D35's
method): refreshed from a copy of the owner's database taken 2026-11-06 and reviewed by the owner — fifteen more
games measured hooked, among them Phantom Blade Zero under the NVIDIA App's *Dynamic* mode, Hogwarts Legacy loading
DLSS 5's runtime and the first two Xbox app games — and the sessions recorded without measurement or refused by the
guard since 2026-09-27. The guide's Settings page says a language change stays on Settings; the closed §S30 notes the
fifth Streamline id decoded (`docs/open-questions/closed-s.md`).
- **The Compare page's layout** (the owner's screenshot): the selection bar beside the list instead of a window's width
away, with *Clear selection*; columns sized to what they hold — *Tier* and *FPS* touched, and wide gaps split the rest —;
the date sorts by when; the settings-changed column says so in words or shows a dash (it was blank); the result
table's session labels sit over their columns.
- **The selected page keeps its name in the left pane** (D86): WPF UI's pane animates the selected item's label away
(Compare's, in the screenshot — every page's while selected); FrameLedger uses a copy of its template without that.
- **Games opens at once from any page** (owner request 2026-11-06: switching to Games stalled 0.5–1.5 s). The window
keeps the library page it built and shows it again on every later visit, reading the library again behind it; the
library is read off the UI thread; the cards appear before the executables are checked — a USB drive that sleeps
took a second or more to answer, and no card was shown until every one had — and each card gets its *Not installed*
pill when its check answers; a read keeps every card that would look the same instead of rebuilding them all; the
list view builds only the rows on screen (`spike-notes` §27).
- **Changing the language keeps the page you are on** — Settings, after a change made there. The App remembered only
the pages it opened from code, so the language changed on Settings came back on the last game page opened
(`src/FrameLedger.App/Services/ShellHost.cs`).
- **DLSS 5 is named where the game ran it** (owner decision D84, amending D64). The owner's Hogwarts Legacy, Neural
Rendering on, loaded DLSS 5's runtime (`nvngx_dlssnr.dll` 310.9.16.0, from the game's folder) and the session could say
only *DLSS 5 runtime loaded* — nothing measured whether it ran. The Overlay's Streamline hook now decodes DLSS 5's
feature (`kFeatureDLSS_NR`, Streamline 2.14) into its own per-frame fact, and the Agent keeps the share of the frames
the game rendered through Streamline that carried it (`sessions.dlss_nr_pct`, **schema 0022** — a ledger this build
migrated no longer opens in an earlier one). Seen running, the DLSS line reads *DLSS 5 (neural rendering on 97.3% of
rendered frames)*, above every other generation; loaded and seen not running, *DLSS 5 runtime loaded, not seen
running*; where nothing could tell (an earlier Overlay, a title outside Streamline), *DLSS 5 runtime loaded*, as
before. Its library is listed with its version and source like the others, its Streamline plugin `sl.dlss_nr.dll`
is recorded when loaded, and a game shipping either file shows *Supports DLSS 5 Neural Rendering* (rules
`2026.11.2`, capability `dlss_nr`). The **Privacy Policy 2.14** names neural-rendering libraries among those whose
versions are read; the **Disclaimer 2.13**'s accuracy block says when DLSS 5 is named. One hook-path change: one more
compare in the `slEvaluateFeature` detour and one more test where a present is recorded (`spike-notes` §26).
- **Streamline's headers are `v2.14.16`'s** (developers): the vendored copy was `v2.12.0`, byte for byte; three of its
eleven files change — `kFeatureDLSS_NR`, the uplift buffer types 70–72, `kStructVersion7`, the version — and
nothing FrameLedger consumed otherwise (`src/native/third_party/streamline/README.md`).
- **A session under the NVIDIA App's *Dynamic* frame generation reads DLSS 4.5** (owner decision D83, amending D64).
Dynamic and its ×6 mode exist only in DLSS 4.5, and the DLSS line named a generation from the measured multiplier
alone: the owner's Phantom Blade Zero under *Dynamic, up to ×3* held ×2.0 and ×3.0 and read *DLSS 4 (a ×3.0
frame-generation state)*. It reads *DLSS 4.5 (Dynamic multi frame generation, set in the NVIDIA App)* where the game's
driver profile sets Dynamic, the driver reports its frame-generation override applied in the game's process, and DLSS
Frame Generation was identified (whether or not its factor was counted); a state measured at ×5 or more still ranks first, and a *Fixed* multiplier names nothing
by itself (The Witcher 3 under Fixed ×6 still presented ×2.0). The game page's chip and the CSV's `# dlss:` line
(`basis=dynamic_frame_generation`) follow; the **Disclaimer 2.13** says so in its accuracy block
(`src/FrameLedger.Domain/Detection/DlssGeneration.cs`, `src/FrameLedger.App/Services/DlssText.cs`; `spike-notes` §26).
- **The guard's source reads in order again** (developers only; no behaviour changes): the note that says what runs only
for a target the anti-cheat evaluation allowed sits above the function it describes again — beta.33's two isolation
checks had come between them — and names that isolation check among what runs
(`src/native/FrameLedger.Injector/src/fl_guard.cpp`).
### Fixed
- **An Xbox game's page no longer fails on a strange picture name** in the game's own `MicrosoftGame.config` (beta.33's):
a name that points at a stream of a file (`a:b.png`) made the search for its scaled variants throw — the page then
showed no picture and logged an error, and an action on the page that reloads it could end in the crash report. Such a
name is no picture a package ships, so it is passed over like a name outside the package
(`src/FrameLedger.Infrastructure/Art/XboxArtCache.cs`). Found by a self-review of #351 and #352 after their gates
passed; no install of the owner's eleven titles has such a name.
+91
View File
@@ -0,0 +1,91 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.35 was cut, 2026-11-09 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.35] - 2026-11-09
**The thirty-fifth pre-release: accessibility.** An accessibility statement (Help ▸ Accessibility); a screen reader
that reads every control and value in its words and is told what appears; headings and charts it can use; Windows
contrast themes followed; the keyboard following a page shortcut, and a message box opening over the window. And three
places where text read below WCAG's 4.5:1 in a Fluent theme, found on the way: the agent's pill, two kinds of short
message, a destructive button. The same unsigned installer: verify its hash against `SHA256SUMS.txt`, then *More info →
Run anyway*.
**Updating from `0.1.0-beta.34`:** Help ▸ Check for updates, then *Restart to update* with no game running. The legal
texts keep their versions, so nothing is shown for you to accept; the ledger stays at schema 22 and the detection rules
at `2026.11.2`. Nothing about capture or measurement changes.
### Added
- **An accessibility statement** (owner request 2026-11-09, decision D88). `ACCESSIBILITY.md` says what FrameLedger aims
for — WCAG 2.2 at level AA as W3C's WCAG2ICT applies it to desktop software, a goal and not a claim of conformance —
what works with the keyboard alone, a screen reader, a contrast theme or Magnifier, how it is checked (the project's
tests, and the owner's pass of 2026-11-09: keyboard only and Narrator on 0.1.0-beta.34 and 0.1.0-alpha.3, no fault),
what does not work yet and what to do instead, and how a barrier is reported and handled — by severity, with no
promised times. The app shows it under **Help ▸ Accessibility**, and after Limitations in the user guide's window; the
wiki gets an *Accessibility* page.
- **An *Accessibility problem* issue form** (`.github/ISSUE_TEMPLATE/accessibility.yml`) asking what was tried, what
happened and what the reporter uses — never anything about the reporter themself — and a user-interface section in the
pull-request template. The `accessibility` label exists, and so do `safety` and `priority-high`, which the *Safety gap*
form had named since its first day without either existing — so GitHub had never applied them.
- **A screen reader reads what the window shows** (decision D88's goal; measured in the UI Automation tree WPF builds on
every page, window and dialog, every tab — `PagesLoadTests.Automation`). Drop-down options were read as an internal
record (*Choice { Value = dark, Label = Dark }*), lists and table rows by their program names, table cells drawn by a
template as *Item: …, Column Display Index: n*, and the terms window's list by each document's whole text; each now
reads the words it shows, and an empty cell reads *empty*. The frame-rate readout says Native, Displayed and the
factor together (CLAUDE.md rule 6, in speech), a ray-tracing chip its value and where it came from, a change badge its
number. The safety notices' title and message reach a screen reader (they were hidden inside the notice's template),
and the notice's button says which notice it answers. A button repeated on each row names its row (*Open: …*, *Play:
…*, *Turn off: …*). Tab no longer stops on pictures, lists, Settings' rows, the FPS readout, the chips or the dialog
host, where it found nothing to say; an expander's button is named after its header.
- **A screen reader is told what appears** (owner decision D89): a safety notice as it arrives, the capture agent going
offline and coming back, an update's steps (not each percent), every short message at the bottom of the window, a
notice on a game's page and each result of Tools ▸ Database maintenance — in the words shown, and only while
FrameLedger's window is the active one, so a game that has the screen is never interrupted. The database line's live
region, which announced nothing, is gone.
- **Headings and charts a screen reader can use** (owner decision D90). Every page's title is a first-level heading and
each section title a second-level one, so a screen reader's heading keys move between sections. Every chart says what
it is and where its numbers are written as text; its keys — arrows to move, Ctrl with an arrow to zoom, A to see all of
it — now stay with the chart instead of also scrolling the page or moving the focus on, and its focus is drawn. No
chart tells two series apart by colour alone any more: Compare's five sessions, the GPU's temperature and its hot spot,
the two kinds of stutter, the two lows and the two limits each have their own line pattern, marker or hatch.
- **Windows contrast themes are followed** (owner decision D91). With a contrast theme on — Aquatic, Desert, Dusk, Night
sky, or one whose colours were edited, on Windows 10 or 11 — every colour in FrameLedger's windows is one of that
theme's, whatever Settings ▸ Theme says, and it changes the moment the theme is turned on, edited or turned off;
Settings ▸ Theme says so under its box. Before, *System* could apply WPF UI's own high-contrast files, most of whose
colours are a red placeholder, and the next window opened put the light or the dark theme back. The charts draw in the
theme's colours, their patterns telling the series apart; the frame-generation, ray-tracing and upscaler chips are the
theme's highlight and its text (WPF UI's files put the window text there, 1.26–1.78:1); a destructive button is the
theme's own button; a pressed menu item is no longer white on white; nothing is drawn translucent, and a game's picture
lends its page no colour, as before.
- **The keyboard follows a page shortcut.** **Ctrl+1** … **Ctrl+5** now move the focus into the page they open — its
first control — and a screen reader hears the page's title; before, the focus stayed where it was, often on the page
that had just gone, and Tab began again at the top of the window.
- **Pages no longer slide in while Windows' animation effects are off** (Settings ▸ Accessibility ▸ Visual effects),
followed as the setting changes.
### Fixed
- **A message box opens over FrameLedger's window** — Remove game, Delete sessions, the database's and the update's
questions. It had no owner, so it could fall behind the window it came from and, closing, give the focus to another
program; it is owned by the window that opened it now, and the focus comes back there.
- **A double-click on a chart shows all of it again**, as the user guide said; it toggled ScottPlot's render-time overlay.
- **The capture agent's state in the title bar reads in the light theme.** It was drawn on WPF UI's fixed green, blue,
orange and red under white — 2.78, 2.63, 2.16 and 3.68:1 — and is on the theme's own success, caution and critical
fills now, as the notices are; its words say the state.
- **The short messages at the bottom of the window read.** A *done* or a *warning* message was white on a fixed green or
orange, 2.78:1 and 2.16:1 in both themes; every message is in the theme's colours now, its icon saying which kind.
- **A destructive button's label reads in the dark theme** — Remove on a game's page, Delete in the confirmations: the
theme's dark text on the red, 5.3:1 and 5.7:1, where white read 3.68:1.
### Changed
- **Documents corrected**: charts do take focus and keys (`docs/08_UI.md` said neither — `ChartKeyboardTests` measures
it); the database-maintenance result's live region announced nothing, since WPF raises one only when the app does; the
"no Narrator pass" gap is the owner's pass now, beside what that pass could not show; the open question "Accessibility
/ DPI: no design anywhere" is closed; NFR-9 names screen readers and contrast themes (D88–D91). `LIMITATIONS.md` gains an
*Accessibility* section and loses its one struck sentence — a screen reader reads struck text as current, so no page
the app shows strikes text through. The guide's menu table names its menu on every row, read one row at a time.
+41
View File
@@ -0,0 +1,41 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.36 was cut, 2026-11-10 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.36] - 2026-11-10
**The thirty-sixth pre-release: the same software, its C# read by a second analyser and tidied.** Nothing you see,
record or measure changes. Qodana — JetBrains' code analysis — now reads the C# and the XAML of every change, and what
its first run reported is settled: about 400 small corrections in the code, none changing what the program does. The
same unsigned installer: verify its hash against `SHA256SUMS.txt`, then *More info → Run anyway*.
**Updating from `0.1.0-beta.35`:** Help ▸ Check for updates, then *Restart to update* with no game running. The legal
texts keep their versions, so nothing is shown for you to accept; the ledger stays at schema 22 and the detection rules
at `2026.11.2`. Nothing about capture or measurement changes.
> **Added the day of the release, after the tag (2026-11-10).** The package carries the **.NET runtime 10.0.15**, where
> beta.35's carried 10.0.14: the release was built with SDK 10.0.404, November's servicing release, which the build
> machine had that day (10.0.403 the day before), and `global.json` lets a build take the newest 10.0 SDK. Nothing in
> FrameLedger's own code asked for it, and the notes above did not name it when the tag was cut. It is why the update
> from beta.35 replaces 343 of the package's 456 files (a 5.8 MB download). The NuGet libraries stay at their 10.0.14
> versions.
### Changed
- **Qodana reads the C# and the XAML from CI** (developers only; owner decision D93). The owner made the project on
Qodana Cloud and set the `QODANA_TOKEN` secret; `.github/workflows/qodana.yml` is JetBrains' workflow for its Qodana
Scan action, changed where FrameLedger differs: the linter is Qodana Community for .NET — ReSharper's inspections in
a Linux container, which runs under every licence, where the Rider-based linter needs the Ultimate licence the
organisation holds only as a 30-day trial — over `FrameLedger.slnx`, with `EnableWindowsTargeting` passed to the
container so that the Windows-only projects build and load there; the action is pinned to a commit; the job's
permissions are the ones it uses. It reads no C++ (CodeQL and SonarCloud do). Its gate is zero problems
(`qodana.yaml`), and it is advisory: a red job beside a green `check`, never a blocked merge. An inspection that
argues with a choice this project made is switched off in `.editorconfig`, where Rider reads it too, with the reason
on the line above — sixty-three keys (`docs/13_CI_CD.md` §Qodana).
- **What Qodana's first run reported is settled** — 3,679 problems, 2,050 of them the missing-XML-comment warning the
project had already made a suggestion. About 400 were fixed in the code, in 181 files, none changing what the
software does: redundant qualifiers, casts, `using` directives, `!` suppressions and `unsafe` modifiers removed (290);
`var` for built-in types and one-line block bodies brought to `.editorconfig`'s own style (52); every XAML item
template names its `DataType`, and a binding that reached its page through `DataContext.ViewModel` names the page's
type (35), so the editor resolves them; two fields nothing read (`LogsViewModel`, `PipeServer.Client`), two
initialisers a constructor overwrote and an unused parameter removed; `SpokenTextAutomationPeer` no longer assumes
the base peer's name is a string. The members nothing in the solution uses are listed in `docs/13_CI_CD.md` §Qodana
and left as they are.
+77
View File
@@ -0,0 +1,77 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.37 was cut, 2026-11-11 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.37] - 2026-11-11
**The thirty-seventh pre-release: FrameLedger runs on .NET 11.** The same program, built on .NET 11.0.0 — released the
day before — where every beta until now was built on .NET 10 (owner decision D94). Nothing about how a game is measured
changes: the part of FrameLedger loaded into a game is native code, and it is 0.1.0-beta.36's. What changes is what
FrameLedger needs from your PC, and two of its legal texts. The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**Before you update from `0.1.0-beta.36`, three things:**
- **Your processor.** .NET 11 does not start on a processor without SSE4.2 and POPCNT. Every processor Windows 11
accepts has them; one from before about 2010 may not. On such a PC this version installs and then does not start, and
0.1.0-beta.36 cannot check that for you. Installing 0.1.0-beta.36 again brings FrameLedger back, with your data as it
was.
- **Windows 10.** Microsoft supports .NET 11 on Windows 11, and on Windows 10 only in its LTSC and IoT editions.
FrameLedger does not block Windows 10 22H2 — but it has never been tested there, on any version, and nobody supports
it there (D95).
- **The update is a large download** — about 70 MB, where a usual one is a few: the .NET runtime inside the package is
replaced whole.
Then: Help ▸ Check for updates, and *Restart to update* with no game running. **The terms window appears once**, with
the **EULA 1.7** and the **Disclaimer 2.14** to accept — what an alpha build is, now that .NET 11 has shipped, and a
paragraph on Windows and the processor. The ledger stays at schema 22 and the detection rules at `2026.11.2`, so going
back to 0.1.0-beta.36 is installing it over this version.
**For anyone on an alpha:** `0.1.0-alpha.4` follows this release on the same .NET 11.0.0, and after it the alphas pause
— install the beta (`guide/11-alpha-builds.md`).
### Changed
- **FrameLedger runs on .NET 11** (owner decisions D94 and D95, 2026-11-11). The betas are built on .NET 11.0.0 —
released 2026-11-10 — where they were built on .NET 10: SDK 11.0.100, `net11.0-windows10.0.22621.0`, and the three
libraries serviced with the runtime (`Microsoft.Extensions.Hosting`, `Microsoft.Data.Sqlite`,
`System.Diagnostics.EventLog`) at 11.0.0. The runtime is inside the package as before; nothing is installed beside
it. **What a user should know:** Microsoft supports .NET 11 on Windows 11, and on Windows 10 only in its LTSC and
IoT editions — FrameLedger does not block Windows 10 22H2, and it is untested and unsupported there; **on a
processor without SSE4.2 and POPCNT it no longer starts** (every processor Windows 11 accepts has them); the
ledger's schema is unchanged, so installing 0.1.0-beta.36 again is a way back. The part of FrameLedger loaded into
a game is native code and does not change. `docs/12_BUILD.md` §The main line's .NET says what decided the move
and what was set aside.
- **The capture agent ends with a crash report when a piece of its own background work fails**, where it used to
stop and exit as if nothing had happened: .NET 11 passes that failure on, and the Agent's existing crash path — a
Fatal log entry and a crash dump — takes it. Nothing was added for it (`docs/10_LOGGING_AND_BUG_REPORTS.md` §Crash
handling).
- **The NuGet licence texts follow the package set** (`legal/licenses/nuget/`, 69 packages where there were 78):
nine `Microsoft.Extensions.*` libraries are part of the .NET 11 runtime, and SQLite's engine is SourceGear's
`SQLite` package through SQLitePCLRaw 3.0.5.
- **For developers:** `release.yml` installs a .NET 10 runtime beside the SDK for `vpk` 1.2.161, which has no build
that runs on .NET 11 alone; Qodana's container installs the .NET 11 SDK before it builds (an interim, until the
Ultimate linter); two `string.StartsWith`/`EndsWith(char)` calls in the App's tests name the comparison .NET 11's
analyzer asks for; `HostBackgroundFailureTests` holds .NET 11's `IHost.RunAsync` behaviour.
- **What an alpha is, now that .NET 11 has shipped** (owner decisions D97 and D98). Until 0.1.0-alpha.3 an alpha was
"a beta's source built on a prerelease .NET", and every text said so. 0.1.0-alpha.4 runs on .NET 11's release — the
betas' own from this version — so an alpha is now described as *a beta's source built to try a newer version of
.NET, at times a prerelease one*: **EULA 1.7** (§2B), **Disclaimer 2.14** (§0, which also gains a paragraph on
Windows and the processor), **Alpha notice 1.1** (what an alpha is, the hardware, and a new part on the pause), the
accuracy statement's line, the guide and `LIMITATIONS.md`. The terms window shows the EULA and the Disclaimer once
more; an alpha asks for its notice's tick again.
- **The alphas pause after 0.1.0-alpha.4, and a paused alpha says so** (D98). With the betas on the same .NET there is
nothing for an alpha to try, so none is published until a newer .NET has a preview — and a paused alpha gets no
update, the anti-cheat list included, which reaches FrameLedger only inside one. An alpha cut while the line rests
shows, above every page, that the line is paused and to install the latest beta; the words "a prerelease .NET" now
appear only in an alpha that really runs on one. For developers: `FlAlphaLineRests` in `Directory.Build.props`,
stamped into every assembly, `Shared.AlphaLine`, `App/Services/AlphaTexts`; three new strings in en, vi and ja.
- **.NET 11's API where it reads better, and no C# 15 syntax yet** (developers only; owner decision D96). Sixteen
`char` searches name their comparison through the `string` overloads .NET 11 adds
(`StartsWith`/`EndsWith`/`LastIndexOf(char, StringComparison)`) where they went through a span to satisfy both .NET
10 and 11, and the App finds the Agent it started with `Process.TryGetProcessById` instead of a caught exception;
nothing changes in what the software does. **C# 15's constructs were asked for and are not used**: each was tried in
a file of its own, and ReSharper's engine — what Qodana runs — reads none of the four (it does not parse `closed`
or `union`, and calls collection expression arguments and labeled jumps preview features), so every file using one
would be an error on a gate of zero; a union also draws a warning from the SDK's own analyzers.
`docs/13_CI_CD.md` §C# 15 and the analyses holds the table and what unblocks each; `docs/spike-notes/29-…` where
each would go. `codeql.yml` takes `workflow_dispatch`, so a branch can have a full analysis — a pull request's C#
analysis is an overlay on `main`'s database.
+69
View File
@@ -0,0 +1,69 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.38 was cut, 2026-11-11 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.38] - 2026-11-11
**The thirty-eighth pre-release: the App opens web and mail links only.** One small hardening, and nothing else you
see, record or measure changes. The one place in the App that hands a link to Windows now refuses any address that is
not `http`, `https` or `mailto`, where it trusted each caller to have checked — Windows runs whatever it is handed, and
a `file:` address is a program. Every link the App shows already was one of those three, so no link you could click
before stops working. The rest of this release is for developers: Qodana — JetBrains' code analysis — now reads the C#
with its Rider-based linter and, as a second report, the C++. The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**Updating from `0.1.0-beta.37`:** Help ▸ Check for updates, then *Restart to update* with no game running. The legal
texts keep their versions, so nothing is shown for you to accept; the ledger stays at schema 22 and the detection rules
at `2026.11.2`. Nothing about capture or measurement changes: the code FrameLedger loads into a game is
0.1.0-beta.37's, line for line.
**No alpha follows this release**: the alpha line rests (`guide/11-alpha-builds.md`).
### Changed
- **Every place that starts a process was read, and the App now opens web and mail links only.** JetBrains' rule for
"a process started with something that is not a literal" reported 35 places — 11 in the product, 24 in tests. Each
was read: what it starts, and where the path and the arguments come from (the table in `docs/13_CI_CD.md` §Qodana).
None takes a value from outside the PC or from a file FrameLedger did not write, except the path and arguments of
the game the user added. One was hardened: `ShellUrlOpener` — what opens a link in the browser or the mail client —
now refuses any address that is not `http`, `https` or `mailto` itself, where it trusted each caller to have
checked; the shell runs whatever it is handed, and a `file:` address is a program. Every caller already passed one
of the project's own pages or a link held to those schemes, so no link you could click before stops working. The
rule stays on: the files its results were read in are listed in `qodana.yaml`, so a new file that starts a
process is reported — by path on the plugin's one inspection, since a rule's own id is honoured nowhere (three
runs measured it). And a test of the repository's own, `ProcessStartSitesTests`, in the gate, fails on a new file
that starts a process, or registers a scheduled task, until it has been read and listed.
- **Qodana for .NET, on the Ultimate licence, reads the managed code** (developers only; owner decisions D99 and D100).
The owner took an Ultimate licence for the Qodana Cloud organisation and asked for Qodana over every language. The
job — `managed`, *Qodana (managed)* in a pull request's checks — now runs Qodana for .NET (Rider's engine) in native
mode on a Windows runner, after a `dotnet build`, where the Community linter ran in a Linux container. Besides the
C#, the XAML and the `.resx` it reads the Markdown and XML around the solution, and JetBrains' security rules read
the C#. Its first run reported 3,953 problems, 3,607 of them **hints** — the level an editor offers without
underlining anything, which the Community linter never reported; hints are off as a severity. Of the 347 left: a vi
or ja string equal to the English one (271) is off in `.editorconfig` for `.resx` files — names, units and format
strings are not translated; two Markdown inspections (table padding, and anchors computed otherwise than GitHub
does) and the application manifest's XML are off in `qodana.yaml`; six results in `PrimaryDisplay.cs` are the
linter's own fault (an internal assertion, in its log) and are excluded for that file; one test expression was
simplified. The gate stays zero problems and the job stays advisory. An inspection is now switched off **where the
engine that raised it reads** — one of ReSharper's in `.editorconfig`, what has no key there in `qodana.yaml` — which
amends D93's "never in `qodana.yaml`" (`docs/13_CI_CD.md` §Qodana). The C++ goes to a second report, the next pull
request's.
- **Qodana reads the C++ too, as a second report** (developers only; owner decision D99). The job `native` has CMake
write a Visual Studio solution for `src/native` (the preset `x64-vs-analysis`; nothing ships from it) and runs
Qodana for .NET over it — ReSharper's C++ inspections and the clang-tidy checks they bundle — with settings of its
own (`qodana-native.yaml`), reporting to a Qodana Cloud project of its own. Its first run reported 4,695 problems,
4,040 of them hints; of the 655 left, 494 are distinct (a file compiled into three targets is reported three
times). Nearly all of them are not faults, and `.editorconfig` now says why in fifty-four lines, each with its
reason: clang's opinion of code that only MSVC builds (`__try`, a function pointer got by name, struct padding);
the engine misreading `std::atomic_ref`, zero-initialised API structures and tests of `constexpr` functions;
proposals to restate working code in a newer idiom; and what a probe or a test does on purpose. Eleven things
were fixed, none changing what a shipped binary does: `std::size` for four hand-written array lengths, two
multiplications done in a narrower type than their result, two unused member functions, an unused type alias, two
unused includes. **In code loaded into a game nothing was changed**: the report found no defect there, and what it
did say about those files is listed in `docs/13_CI_CD.md` §Qodana — among it one function in the Overlay that
nothing calls (`LogNoteW`), which is recorded for the owner and not removed (D100). The gate is zero problems and the job is advisory, like the other.
### Fixed
- **A test that failed on a busy machine** (developers only). `ShmDrainIntegrationTests`' drain under supervision
ended after one guard evaluation whenever the first scan was slow enough for fifty presents to be waiting, and
then failed its own "supervised more than once"; the loop now supervises at least twice whatever the first
drain holds.
+125
View File
@@ -0,0 +1,125 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.39 was cut, 2026-11-11 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.39] - 2026-11-11
**The thirty-ninth pre-release: Advanced metrics.** A new setting, **off unless you switch it on** — Settings ▸
Recording ▸ *Advanced metrics* — and a page of its own. With it on, the capture agent reads more from the graphics
driver and the hardware sensors during a session and keeps it with the session: the graphics card's memory, load,
clocks, fans, voltages and power in detail; what each display can do with HDR and variable refresh, and what the
game's display did; the driver's settings, global and for the game; and — when the agent runs as administrator with
PawnIO installed — the processor per core, the memory and the motherboard's sensors. The **Advanced** page (the last
item on the left, **Ctrl+6**) shows it under what it belongs to: **This PC**, read only while the page is open and
never stored; **Session**, what one session kept, in tables and a chart; **Sources**, every reading FrameLedger knows
of and, for each one it does not read, why. With the setting off, nothing more is read or stored and the page is not
there. The same unsigned installer: verify its hash against `SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** None of these readings has been checked against another tool. The processor's, the memory's
and the motherboard's sensors have never been read on real hardware — the first run as administrator with PawnIO is
still to come. The driver's own figures exist on NVIDIA cards only, and no AMD or Intel card was tested. A drive's
temperature is not read. Nothing is read inside the game by this release: the readings that need a hook come with
later ones. `LIMITATIONS.md` §Sensors lists every limit.
**Updating from `0.1.0-beta.38`:** Help ▸ Check for updates, then *Restart to update* with no game running. The terms
window shows the **Privacy Policy 2.15** and the **Disclaimer 2.15** once — they name what Advanced metrics reads and
stores, only while it is on, and that PawnIO and the administrator mode then read more than the CPU temperature. The
ledger moves to **schema 23** (two tables are added; nothing that was there changes), so a 0.1.0-beta.38 cannot open
it afterwards. The detection rules stay at `2026.11.2`. With the setting off, nothing about capture or measurement
changes: the code FrameLedger loads into a game — the overlay, the Vulkan layer, the shared-memory layout, the
injector — is 0.1.0-beta.38's, line for line.
*The entries below are in the order the seven pull requests landed. One that says "not yet" describes that pull
request; a later entry is what followed.*
### Added
- **"Advanced metrics" is planned, and its inventory is written — nothing reads it yet.** The owner asked for a
setting, *Thông số nâng cao*, with a navigation item of its own that shows as much as the NVAPI and Streamline
headers and PawnIO make available, each value under what it belongs to: one game session, or the PC. This pull
request is the first of that train and changes nothing you can see or record. It adds the list everything else in
the train reads — `AdvancedMetricCatalog`, 137 entries: what the Agent will read from NVAPI and
LibreHardwareMonitor (the card's firmware, architecture and cores; its memory with evictions and promotions; its
utilisation by domain; what a display can do with HDR and variable refresh; the driver's profile; and, as
administrator with PawnIO, the processor per core, the motherboard, the memory modules and the drives'
temperatures), what can only be read inside a game and waits for a later release, **and what is never read with the
reason** — a serial number, a display's EDID, the list of programs that are encoding, the camera's position, what
a vendor's licence closes. `docs/18_GPU_VENDOR_APIS.md` §Advanced metrics holds the design, the owner's decisions
D101–D108 and the same list as a table; a test fails when the table and the catalog differ in either direction.
- **The NVIDIA bridge can read what Advanced metrics will show — and nothing calls it yet.**
`FrameLedger.NvapiBridge.dll`, the library the Agent loads into its own process and never into a game, gains five
reads (its ABI 3): the card's facts — firmware version, architecture, ray-tracing and tensor core counts, base and
boost clocks, its memory sizes, the driver's package; what it says now — free and used video memory with evictions
and promotions, the frame-buffer, video and bus utilisation, the performance state, the encoder's load; what each
connected display can do with HDR and variable refresh; the state of the display a game presents on; and every
setting of a driver profile. On the development PC's RTX 5080 all twenty facts and eight of the ten live groups
answer — the card has no board or power-supply temperature sensor, so those two will read N/A.
**It reads no identifier**: no serial number, no unique id, no display EDID, and not which programs are encoding.
- **"The bridge never calls a setter" is now a build check.** A new gate, `tools/nvapi-readonly-check.ps1`, compares
every NVAPI call in the bridge's source with a committed list of forty-four reads, in both directions, and refuses a
name that sets, saves, restores, registers or controls something — or returns an identifier — even when someone
adds it to the list.
- **The reader behind Advanced metrics exists, was measured on a real PC, and is still not switched on.** One
reader for the whole Agent, on a thread of its own, that runs only while a session or the page needs it: NVIDIA's
figures every second, and every sensor LibreHardwareMonitor has for the graphics card — each fan and its duty, the
core voltage, PCI Express traffic, the load of every engine. When the Agent runs as administrator with PawnIO
installed it will also read the processor per core, the memory and the motherboard's sensor chip.
**A drive's temperature is not read**: the library cannot list drives without touching removable ones, so that
group is never opened and the value will read N/A with that reason.
- **`FrameLedger.CaptureHost probe-advanced`** (a developer tool, not shipped): runs that reader for a few seconds
on your own rights and prints everything it read and what it cost. On the development PC it read 35 facts and 58
values a second — and showed that one refresh of the sensor library takes 14 ms of processor time, so the reader
refreshes it every five seconds, where NVIDIA's own figures (0.4 ms) stay at one a second.
- **The ledger can hold Advanced metrics — and holds none yet.** Schema 23 adds two tables: one for the facts of
the PC, stored once per distinct set, and one for what a session recorded with the option on knows — its own
facts, which sources answered and why the others did not, and the statistics of its extra series. Nothing is
added to the session row itself, so what the JSON export and a bug bundle carry does not grow. The extra series
are stored beside the others under names of their own, with a real "no reading" for a second in which a sensor
was not read. One hour of them is about 0.2 MB. **A session recorded with the option off stores nothing of this**,
and the option does not exist yet. *A 0.1.0-beta.38 cannot open a ledger a later build has migrated.*
- **Settings ▸ Recording ▸ Advanced metrics — the switch exists, off by default.** On, the capture agent reads more
from the graphics driver and the hardware sensors during a session and stores it with the session: the graphics
card's memory, load, clocks, fans, voltages and power in detail, what each display can do with HDR and variable
refresh and what the game's display did, the driver's settings for the game — and, when the agent runs as
administrator with PawnIO installed, the processor per core, the memory and the motherboard's sensors. **Off,
nothing more is read or stored**, and the shared reader is not even started. It applies from the next session, and
a session recorded while it was off is not filled in later. NVIDIA's figures are read once a second, the sensors
every five. **No page shows what is stored yet** — that is the next pull requests' — and none of it is in an
export or a bug report.
- **The App can ask the agent what this PC reads right now — and no page asks yet.** A new request,
`GetAdvancedMachine`, answers what the agent's shared reader last read: the PC's facts, the newest value of every
series, and which sources answer or why they do not. It never touches a sensor itself, so it cannot hold up the
pipe. Each request also keeps the reader running for five seconds, which is how "with no game running, sensors are
read only while the page is open" works without anything telling the agent that a page closed. With Advanced
metrics off it answers that it is off, and nothing is read. On the App's side `AdvancedMachineFeed` asks once a
second while a page watches it and keeps each value with the moment it was read.
- **The Advanced page.** While Advanced metrics is on, **Advanced** is the last item in the list on the left, and
**Ctrl+6** opens it. **This PC** shows what the graphics driver and the sensors say right now — the card's
firmware, cores and memory, what each display can do, the driver's settings, and every reading with the time it
was read at and a chart of its last five minutes. It is read only while the page is open, and is not stored.
**Session** shows what a session recorded with the option on kept: the PC as it was, what belonged to that session,
the lowest, average and highest of each reading, and a chart over the session. **Sources** lists every reading
FrameLedger knows of and, for each one it does not read, why — the agent is not an administrator, PawnIO is not
installed, it comes with a later release, or it is never read. A session's summary has an **Advanced metrics…**
button that opens that session there. In `en`, `vi` and `ja`; the user guide has a page for it.
- **The terms window shows the Privacy Policy 2.15 and the Disclaimer 2.15 once.** The Privacy Policy names what
Advanced metrics reads and stores, only while it is on, and what it never reads: a serial number, a display's
identification data, a list of other programs. The Disclaimer's two sentences about PawnIO and about running the
agent as administrator said "the CPU temperature"; they now also name the processor's, the memory's and the
motherboard's sensors, read with Advanced metrics on. The same sentence changed in Settings, the first run, the
administrator dialog, the guide and `LIMITATIONS.md`.
### Changed
- **A crash file from a newer build no longer loses its tail in an older one.** The file a session writes as it goes
(so that a crash keeps what was recorded) is made of chunks; a reader that met a chunk of a kind it did not know
stopped there. It now passes over a whole chunk it does not know and reads on.
### Fixed
- **A sentence in the documents that kept a function from being called.** `docs/18_GPU_VENDOR_APIS.md` said the
vendored NVAPI headers do not declare `NvAPI_GPU_GetMemoryInfoEx`. They do — in `nvapi_lite_common.h`, which
`nvapi.h` includes; the search that found nothing had looked in `nvapi.h` alone. The row is corrected, and the
bridge will call it for the advanced memory rows. Nothing about the memory figures FrameLedger shows today changes.
+99
View File
@@ -0,0 +1,99 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.40 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.40] - 2026-11-12
**The fortieth pre-release: Advanced metrics, inside the game.** With **Advanced metrics** on (Settings ▸ Recording;
off unless you switch it on), a session of a game FrameLedger measures from inside now also keeps **what the game
itself tells its upscaler**: for DLSS the output size, the model preset for each quality mode, whether the game said
its colour buffers are HDR, auto-exposure, alpha upscaling and the exposure values; for FSR the sharpening, the
largest jitter offset, the exposure, the near and far planes, the **range** of the field of view, the flags, the
frame time the game reported and FSR's own video memory when the game asked for it. The **Advanced** page shows them
under the session, in the vendor's own notation, and the session says in a sentence what became of them. With the
setting off nothing more is read inside a game: the component in the game runs the same reads as before, and a test
that injects into a window sending all of it finds the new memory region empty to the last byte. The same unsigned
installer: verify its hash against `SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** **None of this has been read from a real game** — only from the project's own test window.
**Many games will show nothing**: most set their DLSS options through a call FrameLedger cannot intercept, and the
FSR rows need FSR 3.1 or later through AMD's own library (a game that upscales through the older FSR 3.0 interface
shows none). The session then says *the game made none of the calls these are read from* — which never means the
game uses no upscaler; its summary names the upscaler as before. These are the game's words to the upscaler, not a
check of what the upscaler did. **Never read:** where the camera is, which way it looks, its matrices — a build
check refuses those names anywhere in the code loaded into a game. No new hook is installed by this release; two
further steps of this train add hooks, each after this one has been run on real games. `LIMITATIONS.md` §Sensors
lists every limit.
**Updating from `0.1.0-beta.39`:** Help ▸ Check for updates, then *Restart to update* with no game running. The terms
window shows the **Privacy Policy 2.16** once — it names what is stored of the game's upscaler parameters, only with
Advanced metrics on and only for a game FrameLedger injects into, and what is never read. The Disclaimer stays 2.15.
**A game that is running when you update must be restarted before FrameLedger measures it again**: the shared memory
between the game and the agent has a new layout (version 5), and the component inside the game is new. The ledger
stays at schema 23 and the detection rules at `2026.11.2`.
*The entries below are in the order the four pull requests landed. One that says "not yet" describes that pull
request; a later entry is what followed.*
### Added
- **Advanced metrics inside the game is planned — three releases, and nothing reads it yet.** The readings that
exist only inside a game — the options a game gives DLSS and FSR, FSR's own memory, later the constants a game
gives Streamline and its frame-generation configuration — come in three steps (owner decisions D110–D116):
first more fields of calls FrameLedger already intercepts, with no new hook; then two new hooks; then two more.
Each step waits for the one before to be run on real games. With Advanced metrics off, nothing more is read
inside a game either. This pull request changes nothing you can see or record: it records the decisions, and
**which version of each vendor struct carries which field**, compared across every upstream release of Streamline
2.x and of the FidelityFX API — a field is read only where it is proven to be there.
- **`tools/agent-cost.ps1`** (a developer tool): what the shipped capture agent costs while it records, measured on
the real program against the project's own test window, with Advanced metrics off and on.
- **The shared memory between a game and the agent has room for what a game tells its upscaler — and nothing
writes it yet.** Its layout is version 5: one new region, sized once for the three releases of this train, and
one new word by which the agent says "this session records Advanced metrics". With that word at zero — every
session, until the next pull request, and afterwards every session with the option off — the region stays empty and
the component inside the game does exactly what it did before. **A game that is running when you update must be
restarted before FrameLedger measures it again**, as after every update.
- **The component inside a game can read what the game tells its upscaler — and only when asked.** For a game
FrameLedger injects into, and only while the agent asks for it (nothing asks yet: that is the next pull request,
and then only with Advanced metrics on), the calls FrameLedger already intercepts are read a little further: the
options a game hands DLSS with each evaluation (output size, presets, exposure, three switches), what it hands
FSR with each upscale (sharpening, the jitter's size, exposure, the depth range, the field of view as a range,
flags, the frame time it reports), and FSR's own memory when the game itself asks for it. **No new hook**, and
with the request off the same code as before runs: a test injects into a window that sends all of it and finds
the new memory region empty to the last byte. **Never the camera's position, its axes or its matrices** — a
build check now refuses those names anywhere in the code loaded into a game (`tools/overlay-fields-check.ps1`).
Many games set their DLSS options through a call that cannot be intercepted, and will show nothing here.
Measured cost while asked: 22–23 billionths of a second per frame for FSR, 10–12 for DLSS; not measurable when
not asked.
- **A second session of a game that kept running starts its summaries afresh.** The ranges and counts kept inside the
game belong to the session that asked for them: each session asks with a request of its own, and the agent reads
only what answers it.
- **Advanced metrics: what the game tells DLSS and FSR, kept with the session.** With Advanced metrics on, a session
of a game FrameLedger measures from inside now also keeps what the game itself handed to its upscaler: for DLSS the
output size, the model preset for each quality mode, whether the game said its colour buffers are HDR,
auto-exposure, alpha upscaling and the exposure values; for FSR the sharpening, the largest jitter offset, the
exposure, the near and far planes, the **range** of the field of view, the flags, the frame time the game reported
(as a chart) and FSR's own video memory at the moments the game asked. The Advanced page shows them under the
session, in the vendor's own notation. **Many games will show nothing**, and the session says why — most games
set their DLSS options through a call FrameLedger cannot see, and the FSR rows need FSR 3.1 or later through AMD's
library; it never means the game uses no upscaler. **Never read:** where the camera is, which way it looks, its
matrices. With the option off nothing more is read inside a game, exactly as before. Not yet read from a real game.
- **Privacy Policy 2.16** names the above: what is stored, only with the option on and only for a game FrameLedger
injects into, and what is never read. You are shown it once after updating.
### Changed
- **The agent's processor budget says what was measured.** The specification promised that the capture agent uses
at most 1 % of one processor core while recording; nobody had measured it. Measured on the development PC it is
1.93 % — 2.04 % with Advanced metrics on — nearly all of it the once-a-second read of the graphics card's sensors
that every session has always made, and its memory is 113 MiB (119 with the option on) against a limit of 150.
The owner chose to keep the readings and amend the limit: **NFR-2 is now ≤ 2.5 % of one core** (D117). Nothing the
agent does changes.
### Fixed
- **A crash dump is tried up to three times.** When FrameLedger or its agent stops on an error it never expected, a
small helper writes a dump of the program for a bug report - while the program's other threads are still
running. Windows can fail that once when the program changes under it, and one failure used to mean no dump at
all. It is now tried again, twice, a moment apart.
+41
View File
@@ -0,0 +1,41 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.41 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.41] - 2026-11-12
**The forty-first pre-release: one fix, for Advanced metrics with the agent as administrator.** If you run the
capture agent as administrator with PawnIO installed and have **Advanced metrics** on, 0.1.0-beta.40 and
0.1.0-beta.39, which has the same code and where it was not looked for, could lose sensors: after most sessions the Advanced page stopped showing the processor's readings
until it was reopened, and now and then a whole session was stored with the graphics driver's seventeen readings
and none of the two hundred from the sensor library. The agent was closing the sensor library in one place while
another place was still reading it. It now opens and closes it in one place only, and never while something is
reading. If you do not use that combination, nothing changes for you. The same unsigned installer: verify its hash
against `SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** The fix was made from the agent's log of ten real sessions; the fault needs an
administrator's agent with PawnIO to show, and has not been seen to be gone on such a PC yet. After a few sessions
the agent's log (Logs page) should no longer say *a device is left alone* or *the sensor library did not open*.
Everything 0.1.0-beta.40's notes say is not yet done still stands.
**Updating from `0.1.0-beta.40`:** Help ▸ Check for updates, then *Restart to update* with no game running. No
terms to read again, no change to the ledger (schema 23), the detection rules stay at `2026.11.2`. **The code
FrameLedger loads into a game — the overlay, the Vulkan layer, the shared-memory layout, the injector — is
0.1.0-beta.40's, line for line**; as after every update, a game that was running across it must be restarted
before it is measured again.
**A note on numbers.** The plan for Advanced metrics had named 0.1.0-beta.41 for its next step inside the game (two
new hooks). This fix took the number; that step is now 0.1.0-beta.42, and the ones after it move by one as well.
### Fixed
- **Advanced metrics: the processor's sensors no longer drop out after a session, and a session no longer loses its
sensors at the start.** With the agent running as administrator and PawnIO installed, the sensor library was being
closed by one part of the agent while another part was still reading it. After most sessions the Advanced page
lost the processor's readings until it was reopened, and one session in ten stored only the graphics driver's
seventeen series instead of two hundred and twelve. The agent now opens and closes the sensor library in one
place, and does not close it while anything is still reading. Found in the first sessions recorded that way; the
cause was read from the log and could not be reproduced without that setup, so the next sessions will show
whether it is gone — the agent's log should no longer say *a device is left alone* or *the sensor library did not
open*. Nothing changes in the code loaded into a game.
- **Two comparisons in the Advanced-metrics code say what they mean.** Whether a game gave FSR one field of view or
a range of them was decided by comparing two decimal numbers for equality; it is now decided by comparing the two
stored values bit for bit. Nothing you see changes. (Found by SonarCloud after 0.1.0-beta.40 was merged.)
+62
View File
@@ -0,0 +1,62 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.42 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.42] - 2026-11-12
**The forty-second pre-release: groundwork, and a log that can answer a question.** Four changes, none of which
adds a hook to a game. A session that saw two upscalers is now named after the one it ran — one real session had
been stored as "DLSS" with FSR's quality mode and FSR's render size. The part of FrameLedger inside the game can no
longer leave a hook in after it has stopped. Its log is now written during the session rather than only at the
start and the end, so it is whole after a game that simply exits, and it says which module each hook went into and
what kinds of call each AMD FidelityFX module received. And with **Advanced metrics** on, an FSR depth plane at
infinity is shown as "∞" where it used to be missing. The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** It does not fix a game whose upscaler reads "unknown" — Cyberpunk 2077 set to FSR is the
one known — it makes the next session of such a game say why, in `logs\overlay-<pid>-….log`. The two new readings
planned for Advanced metrics (the frame-generation settings a game gives FSR, and what a game tells Streamline
about each frame) are the next pre-release, after that log has been read. No real game has been run on this
build: the changes inside the game were tested with FrameLedger's own test program, and their cost was measured
on the code itself (about 4 billionths of a second per FidelityFX call for the log's new lines; nothing measurable
for the rest). Everything 0.1.0-beta.41's notes say is not yet done still stands.
**Updating from `0.1.0-beta.41`:** Help ▸ Check for updates, then *Restart to update* with no game running. No
terms to read again, no change to the ledger (schema 23), the detection rules stay at `2026.11.2`. **The code
loaded into a game changed in this release** — the first time since 0.1.0-beta.40 — so a game that was running
during the update keeps the old part until it is started again. Sessions already stored are not rewritten: one
that was named after the wrong upscaler keeps that name.
### Added
- **Advanced metrics: FSR's depth plane at infinity is shown as "∞".** Engines with an infinite depth range tell
FSR that one of the two camera planes is at infinity, and FrameLedger took that value for "not set" — so a
session's depth range showed one plane where the game had given two. The plane at infinity is now stored and
shown as such. Only for FSR, and only while Advanced metrics is on; nothing else about what is read changed.
### Changed
- **The Overlay's log is whole after a game exits, and says more.** The log FrameLedger's part inside the game
writes (`logs\overlay-<pid>-….log`) used to hold only the first half second of a session unless the session
ended while the game was still running: a game that exits takes the thread that writes the log with it. It is
now written during the session, whenever something new was noted. Each hook that went in is named by its module
and function (four AMD modules used to be four identical lines), a hook that could not go in says why, and the
log lists which kinds of call each AMD FidelityFX module received and which upscaling and frame-generation
modules the game loaded, and when. This is for finding out why a game's upscaler reads "unknown" — nothing in
it is a measurement, and nothing from it is stored in a session. Measured cost: about 4 billionths of a second
per FidelityFX call.
### Fixed
- **The part inside the game can no longer leave a hook in after it has stopped.** Two gaps in how the Overlay keeps
track of its own patches, found by reading the code and not by a failure: a patch made while the table of patches
was full would have been made anyway and never taken out, and a stop that arrived at the exact moment a hook was
going in could miss it. A patch now takes its place in the table before it is made — no place, no patch — and is
recorded before it can run, so the stop either removes it or the installer does. A patch that could not be made
is written to the Overlay's log with the reason, where it used to be silent. No hook's own code changed.
- **A session that saw two upscalers is named after the one it ran, with that one's quality and render size.** The
session's upscaler used to be the first one FrameLedger saw named, while the quality mode and the render size
came from whichever frames carried them. A game that evaluated DLSS briefly and then ran FSR was stored as "DLSS"
with FSR's quality and FSR's render size. The name is now the technology named on the most frames, the quality
and the size are read from that technology's frames only, and such a session is marked "settings changed during
session". What that technology did not say reads N/A rather than borrowing the other's value. Sessions already
stored are not rewritten.
+86
View File
@@ -0,0 +1,86 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.43 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.43] - 2026-11-12
**The forty-third pre-release: a game that was still loading is no longer lost, and two more readings inside the
game.** Two fixes first. A game that FrameLedger looked at while it was still loading its modules could be left
unrecorded — the session ended as a safety stop with nothing in it; FrameLedger now reads the list again before it
gives up, and what stops a session has not changed. And an FSR upscale is counted once per frame whichever way the
game's FidelityFX loader passes it on. Then, with **Advanced metrics** on, a session of a game measured from inside
also keeps what the game tells NVIDIA Streamline about each frame and how it configures FSR frame generation —
from two calls FrameLedger hooks only once Advanced metrics has been asked for in that game. The same unsigned
installer: verify its hash against `SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** The two new readings have been tested with FrameLedger's own test program and **have not
been run in a real game**: this release is the first time either hook is put into one. They are what the game
asked for, not a check of what happened, and Streamline's values are those of the first view a game draws.
Streamline 1.x is not read. Where the game's camera is, which way it looks and its matrices are never read — they
sit in the same data, and a check in the build refuses code that names them. The cost was measured on the code
itself: about 20 billionths of a second per frame for Streamline's call and 15 per call for FSR's while the option
is on, under 2 while it is off; nothing measurable for the two fixes. A game whose FSR is built in (Cyberpunk 2077's
FSR 2.1) still reads "unknown" as its upscaler. Everything 0.1.0-beta.42's notes say is not yet done still stands.
**Updating from `0.1.0-beta.42`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are shown once: Privacy Policy 2.17**, which names the two new readings. No change to the ledger (schema 23),
the detection rules stay at `2026.11.2`. **The code loaded into a game changed in this release**, so a game that
was running during the update keeps the old part until it is started again.
### Added
- **Advanced metrics: a session now shows what the game told Streamline each frame and how it configured FSR frame
generation.** For a game FrameLedger measures from inside, a session recorded with **Advanced metrics** on keeps
— beside what the game told DLSS and FSR — the largest jitter offset, the motion-vector scale and how motion
vectors are stored, the near and far planes and whether depth is inverted, the range of the field of view, the
aspect ratio, whether the projection is orthographic and how often the game reset the upscaler's history
(Streamline); and whether FSR frame generation was configured on, whether it may run asynchronously, its flags,
whether only generated frames are presented, and the part of the picture it generates in. A setting the game
changed during the session reads *on and off*. The line under the session now also says, for each of the two,
whether it was read, hooked and never called by the game, or not hooked because the game did not load the
library — so an empty page says which it is. These are what the game asked for, not a check of what happened;
neither has been confirmed on a real game yet. Never read: where the camera is, which way it looks, its
matrices.
- **The terms are shown once after this update: Privacy Policy 2.17.** It names the two new readings above and says
that the two calls they come from are intercepted in a game only once a session with Advanced metrics on has
started there. Nothing is sent anywhere, as before.
- **Advanced metrics, inside the game: how the game configures FSR frame generation** (the part inside the game;
what a session shows of it comes with the same release). With **Advanced metrics** on, FrameLedger now also
reads, from the call a game makes to configure AMD's frame generation, whether generation is on, whether it may
run asynchronously, its flags, whether only generated frames are presented, and the part of the picture it
generates in. Nothing else of that call is read. Like the Streamline hook, this one is put into a game only once
Advanced metrics has been asked for. Measured cost: about 15 billionths of a second per call while the option is
on, under 2 while it is off.
- **Advanced metrics, inside the game: what the game tells NVIDIA Streamline about each frame** (the part inside the
game; what a session shows of it comes with the same release). With **Advanced metrics** on, FrameLedger now also
reads, from a call a Streamline game makes every frame, the size of the jitter, the motion-vector scale, the depth
range, the range of the field of view, the aspect ratio and seven on/off settings (inverted depth, how motion
vectors are stored, whether the upscaler's history was reset). It never reads where the game's camera is, which
way it looks, or its matrices, which sit in the same data — a check in the build refuses any code that names
them. This is the first reading that needs a hook of its own, and that hook is put into a game only once
Advanced metrics has been asked for: with the option off, nothing in the game is touched for it. Measured cost:
about 20 billionths of a second per frame while the option is on, under one billionth while it is off.
### Fixed
- **FSR upscaling is counted once per frame whichever way the game's FidelityFX loader passes it on.** Newer
FidelityFX loaders hand a call on to the library behind them in a way that brought the same upscale to
FrameLedger twice; with frame generation off, the count of frames the game rendered would have doubled — Native
FPS twice too high. Seen in how Cyberpunk 2077's loader passes on its frame-generation calls; no game tested so
far upscales through such a loader, so this is a precaution. The upscale is now counted where the game calls it,
and nowhere else. (Owner decision D125. Measured cost: none that can be told from noise.)
- **A game is no longer left unrecorded because it was still loading when FrameLedger looked at it.** Before it
hooks a game, and again right after, FrameLedger lists the modules the game has loaded and checks them against
the anti-cheat list. A game that is starting loads and unloads modules for several seconds, and when the list
changed in the middle of being read FrameLedger treated it as unreadable and stopped: the session ended as a
safety stop with nothing recorded, and the game ran on unmeasured until it was restarted. It now reads the list
again — up to three more times, a quarter of a second apart — and only stops if the list never holds still. What
stops a session has not changed: an anti-cheat module found in any read stops it at once, a protected process
still stops it at once, and a list that cannot be read in the end is still a stop, never a pass. (Owner decision
D123; seen five times in one PC's logs in a week.)
- **A test of the Overlay's log reads its lines without `strlen`.** The helper that looks for a line in the log
took its text as C strings and measured one with `strlen`; it takes strings now. Test code only: nothing in the
installed software changes. (Found by SonarCloud after 0.1.0-beta.42 was merged.)
+83
View File
@@ -0,0 +1,83 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.44 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.44] - 2026-11-12
**The forty-fourth pre-release: what the first real games on 0.1.0-beta.43 showed, fixed — and two more readings
inside the game.** Nine sessions on five games were read the day 0.1.0-beta.43 came out. Its two new readings
worked on four of them; three things did not read right, and are fixed first. Cyberpunk 2077 tells NVIDIA
Streamline its field of view in degrees, and FrameLedger labelled the number as radians: a value too large to be
radians is now shown as the game gave it, and says so. A game whose hooking had once been switched off after two
early ends was switched off again by every single one after that: the count now starts again each time you switch
hooking back on. And an FSR frame-generation configuration with no flag set now says *None*. Then, with
**Advanced metrics** on, a session of a game measured from inside can also keep what Streamline answers when the
game asks **which version a feature is and what it requires** — from two more calls FrameLedger hooks only once
Advanced metrics has been asked for in that game. The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**What it is not, yet.** **Most sessions will show nothing of the two new readings.** A game asks Streamline about
its features while it starts, and FrameLedger is usually not in the game yet; the session then says the call was
hooked and the game did not ask again — which is the expected answer, not a fault. They have been tested with
FrameLedger's own test program and **have not been run in a real game**: this release is the first time either
hook is put into one. What they give is Streamline's answer as the game received it — not FrameLedger's own check
of the PC — and never the lists that answer points to (which buffers, which Vulkan extensions). A game that faults
while it is closing can still read *ended normally* in a session measured from inside; that is written down, not
changed. The two Cyberpunk 2077 sessions recorded on 0.1.0-beta.43 keep the line they were stored with. The cost
was measured on the code itself: 15 and 21 billionths of a second per call while the option is on, under 1 while
it is off, for calls a game makes a handful of times. Everything 0.1.0-beta.43's notes say is not yet done still
stands.
**Updating from `0.1.0-beta.43`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are shown once: Privacy Policy 2.18**, which names the two new readings. No change to the ledger (schema 23),
the detection rules stay at `2026.11.2`. **The code loaded into a game changed in this release**, so a game that
was running during the update keeps the old part until it is started again.
### Added
- **Two more calls are hooked inside a game, only where Advanced metrics was asked for** (Advanced metrics train B,
the third step): `slGetFeatureVersion` and `slGetFeatureRequirements` of `sl.interposer.dll` — what Streamline
answers when a game asks which version a feature is and what it requires. The original call runs first and its
answer is read only when it succeeded: the feature's Streamline and NGX versions; the graphics APIs it supports,
whether V-Sync must be off and hardware-accelerated GPU scheduling on, how many threads, viewports and buffers,
and the Windows and driver versions found beside the ones required. **The lists the answer points to — the
buffer tags, the names of Vulkan extensions and features — are never followed.** As with `slSetConstants`,
neither patch is made in a game no session has asked, and each stays for the life of the process once made. A
game asks these while it starts, so the two rows are the first the Overlay patches — and most sessions, attached
later, will still not see the calls. Nothing is shown of them yet: the Agent and the App read them in the next
pull request. No layout change (5), no schema change (23).
- **A session shows which version each Streamline feature is, and what it requires** (Advanced metrics). Two
catalog entries are now read inside the game: `sl.feature.versions` and `sl.feature.requirements` — Streamline's
answers to the two questions a game asks about a feature, in NVIDIA's own names (`kFeatureDLSS_G · versionNGX`,
`kFeatureDLSS · driverVersionRequired`). The line under the session says for each of the two calls whether it was
read, not hooked, or — the usual case — *hooked — the game asks this while it starts, and did not ask again after
FrameLedger attached*. **Most sessions will show nothing here**: a game asks before FrameLedger is in it. A game
that asks later is read then, and a second session of a game that kept running keeps what the first read. The
Agent's end-to-end tests now run a Streamline target as well as an FSR one.
### Changed
- **Privacy Policy 2.18**: it names what Streamline answers about a feature — the features asked about, their
versions, what each requires — and what of that answer is never read (the lists it points to: the buffers' names,
the names of Vulkan extensions and features); "two further calls" is four. The terms window shows once after the
update. The Disclaimer stays 2.15.
- **Switching hooking back on starts the count of early abnormal ends again** (owner decision D126). The count was
never set back, so a game that had once been switched off — after two ends within 60 seconds of injection — was
switched off again by every single such end after its user switched it on, under a reason that still said
"twice". Seen on 2026-11-12: Lies of P, two ends in October, one crash of its own that day. Two are needed again
each time. Nothing else about the rule moves: it only switches off, an anti-cheat block is never cleared, and the
user-mode exception's trial keeps its own record.
### Fixed
- **A field of view a game gives in degrees is no longer labelled radians** (Advanced metrics; owner decision
D128). Cyberpunk 2077 gives NVIDIA Streamline 51–90 where the library expects radians, and the session read
`51.0055 – 90 rad (2922.4° – 5156.6°)`. A value too large to be radians is now shown as the game gave it, with a
line that says it is not radians, and is not converted. Sessions recorded on 0.1.0-beta.43 keep the line they
were stored with.
- **An FSR frame-generation configuration with no flag set says "None"** (Advanced metrics). The row was missing,
where an upscale with no flag has always said so.
- **The accuracy statement and the limitations no longer say that what a game tells Streamline, and how it
configures FSR frame generation, have not been read from a real game**: they were, on 2026-11-12 — three games
and one. What is still unconfirmed is named: the DLSS options, and a configuration that reaches FSR through
AMD's 2.3 loader. A fault while a game is closing, which a session measured from inside the game does not see,
is written down as a limit (owner decision D127: not changed in this release).
+80
View File
@@ -0,0 +1,80 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.45 was cut, 2026-11-12 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.45] - 2026-11-12
**The forty-fifth pre-release: what a game does with NVIDIA Streamline's features.** With **Advanced metrics** on,
a session of a game measured from inside can now also keep, for each Streamline feature the game asks about or
controls: whether Streamline says it is **supported** — in NVIDIA's own word, a refusal included — whether it is
**loaded**, whether the game **switched it on or off** while it ran, how many times the game **allocated and
freed** what the feature needs, and how many times it **shut Streamline down**. They come from six more calls,
which FrameLedger hooks only once Advanced metrics has been asked for in that game. The pages that said most
sessions would show nothing of 0.1.0-beta.44's two readings are corrected: the first real sessions read both in
every Streamline game. The same unsigned installer: verify its hash against `SHA256SUMS.txt`, then *More info →
Run anyway*.
**What it is not, yet.** The six new calls have been tested with FrameLedger's own test program and **have not
been run in a real game**: this release is the first time any of them is patched in one, and it is six at once.
Two of them sit on a game's way out — freeing a feature's resources, shutting Streamline down — and two of the
games run so far already fault while closing without them; whether anything changes there is for the first real
sessions to say. What is shown are **calls the game made after FrameLedger was in it**, counted to fifteen: not
what the game did while it started, and not a statement that a feature ran or drew anything. **What a game
started Streamline with is not read and will not be**: the game makes that call once, before FrameLedger can be
inside it, and the Advanced page now says so instead of "a later release". Which graphics card a game asks about
is never read. The cost was measured on the code itself: at most 13 billionths of a second per call while the
option is on, under 2 while it is off, for calls a game makes a few times a second at most. A stop restores every
patch, each taking about a sixth of a second, and a Streamline game with Advanced metrics on now has six more of
them; the Overlay's log waits for them.
Everything 0.1.0-beta.44's notes say is not yet done still stands.
**Updating from `0.1.0-beta.44`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are shown once: Privacy Policy 2.19**, which names the new readings and two things that are never read. No
change to the ledger (schema 23), the detection rules stay at `2026.11.2`. **The code loaded into a game changed
in this release**, so a game that was running during the update keeps the old part until it is started again.
### Added
- **Three more of Streamline's calls are hooked inside a game, only where Advanced metrics was asked for** (owner
decisions D129–D131): `slIsFeatureSupported`, `slIsFeatureLoaded` and `slSetFeatureLoaded` — whether Streamline
says a feature is supported (its answer, a refusal included), whether it is loaded, and the game switching a
feature off and on while it runs. No vendor struct is read for these: the feature's id, a result code, a yes or
no. **The adapter a game names when it asks whether a feature is supported is never read**, and the build now
refuses code that names what identifies it. As with the four hooks before them, none of the three patches is made
in a game no session has asked. Nothing is shown of them yet: the Agent and the App read them in a later pull
request. No layout change (5), no schema change (23).
- **And three more, counted only**: `slAllocateResources`, `slFreeResources` and `slShutdown` — how many times a
game allocated and freed a feature's resources (up to fifteen each) and how many times it shut Streamline down.
The feature's id is the only argument looked at; the two that sit on a game's way out note the call and then
call the original, and run nothing after it. Six new calls in all in this release, each patched only where a
session asked.
- **`slInit` is not hooked, and will not be** (D129). A game starts Streamline once, before FrameLedger can be in
it — also when FrameLedger starts the game itself — so a hook there would be a patch that reads nothing. The
Advanced page's entry for what a game started Streamline with now gives that reason instead of "a later release".
- **A session shows whether each Streamline feature is supported and loaded, and what the game did with it**
(Advanced metrics). Two catalog entries are read from the six calls above: `sl.feature.support` — Streamline's
answer by NVIDIA's own name (`kFeatureDLSS_G · slIsFeatureSupported` = `eOk`; a refusal is named, and it is an
answer) and whether the feature is loaded — and `sl.feature.lifecycle` — the game switching a feature on, off or
both, how many times it allocated and freed the feature's resources (`≥ 15` past fifteen), and how many times it
shut Streamline down. The line under the session names the six calls in one sentence: read; hooked, the game did
not make the call; not hooked. They begin when FrameLedger is in the game — what a game did while it started is
not there — and they are calls the game made, not a statement that a feature ran. Not confirmed on a real game.
### Fixed
- **A stop that restores many patches no longer leaves its last lines out of the Overlay's log.** Restoring a patch
takes about 160 ms, and the wait for the stop's log lines was two seconds whatever the number of patches: with
fifteen the log ended "the stop was still writing". The patches were restored; the log did not say so. The wait
is now sized to the number of patches a game can hold.
### Changed
- **Privacy Policy 2.19**: it names, for each Streamline feature a game asks about or controls, whether Streamline
says it is supported and loaded, the game switching it on or off, and how many times its resources were allocated
and freed, and how many times the game shut Streamline down — and two more things that are never read: what
identifies the graphics card a game hands over with its question, and the content of a graphics resource. "Four
further calls" is ten. The terms window shows once after the update. The Disclaimer stays 2.15.
- **The limitations, the accuracy statement and the user guide no longer say that most sessions will show nothing
of what Streamline answers about a feature.** 0.1.0-beta.44's notes expected that, because a game asks while it
starts. The first real sessions (2026-11-12, five games) read both answers in every Streamline session: the
Unreal Engine games ask again later. The pages now say what was seen — and two things the readings showed: the
version Streamline gives for a feature can be newer than the file in the game's folder, and a game that does not
upscale through Streamline lists frame generation there and no upscaler. No code changes.
+42
View File
@@ -0,0 +1,42 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.46 was cut, 2026-11-13 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.46] - 2026-11-13
**The forty-sixth pre-release: hooks go into a game together, and come out together.** Putting a hook in or taking
it out stops every thread of the game for a moment. Until now that happened once for every hook: in the first real
sessions on 0.1.0-beta.45 the thirteen NVIDIA Streamline hooks took between 0.7 and 2.8 seconds of a game's start,
and stopping a session took about a sixth of a second for each hook. The hooks FrameLedger finds something for at
one moment are now put in under **one** stop of the game's threads, and a stop takes all of them out under one.
This release is that change and nothing else in the part that runs inside a game; the pages are also brought up to
date with what those first sessions showed. The same unsigned installer: verify its hash against `SHA256SUMS.txt`,
then *More info → Run anyway*.
**What it is not, yet.** It has **not been run in a real game**. On FrameLedger's own test program nine hooks go
in within 78–90 thousandths of a second where they took 0.6–0.7 seconds, and eighteen come out within 76–90 where
they took 1.4–1.6 seconds; how long one stop of the threads is in a game with a few hundred of them, and whether
any game minds its hooks arriving together, are for the first real sessions to say. Nothing new is read and
nothing a hook reads has changed, so the terms are the ones you accepted. Of 0.1.0-beta.45's six new readings five
have now been read in real games; no game run so far set up a feature's resources after FrameLedger was in it, so
that row has not been seen. Everything 0.1.0-beta.45's notes say is not yet done still stands.
**Updating from `0.1.0-beta.45`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are not shown again** (Privacy Policy 2.19 as before). No change to the ledger (schema 23), the detection
rules stay at `2026.11.2`. **The code loaded into a game changed in this release**, so a game that was running
during the update keeps the old part until it is started again.
### Changed
- **Hooks go into a game together, and come out together** (owner decision D134). Putting a hook in or taking it
out stops every thread of the game for a moment, and until now that happened once for every hook: the thirteen
Streamline hooks took 0.7 to 2.8 seconds of a game's start in real sessions (2026-11-13), and stopping a session
took about a sixth of a second for each hook — two to three seconds, sometimes on the thread that draws. The
hooks a moment finds are now put in under **one** stop of the threads, and a stop takes all of them out under
one. On FrameLedger's own test program: nine hooks in 78–90 ms where they took 0.6–0.7 s; eighteen out in
76–90 ms where they took 1.4–1.6 s. What a hook reads and when it is installed are unchanged, and no hook's own
code changed. The log of the part inside the game has a new line for each such moment (`PATCH_BATCH`, with the
time it took and the number of hooks). **Not run in a real game yet.**
- **The limitations, the accuracy statement and the user guide say what the first real sessions on 0.1.0-beta.45
showed** (2026-11-13, six games): five of its six new readings were read in real games — every Streamline game
asked whether its features are supported — and no game allocated a feature's resources after FrameLedger was in
it, so that row has not been seen. No code changes.
+92
View File
@@ -0,0 +1,92 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.47 was cut, 2026-11-14 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.47] - 2026-11-14
**The forty-seventh pre-release: with Advanced metrics on, FrameLedger reads NVIDIA's own report of where a frame's
time goes.** On an NVIDIA card, in a Direct3D 12 game that uses NVIDIA Reflex, a session recorded with *Advanced
metrics* on now also stores how long the graphics driver says each stage of a frame took — simulation, render
submit, present, driver, the operating system's queue, the graphics card — as seven lines under *Latency* on the
Advanced page, with a table of the driver's low-latency state. **They are the driver's numbers, not a measurement
by FrameLedger**, and they are not the delay from a click to the screen. The same unsigned installer: verify its
hash against `SHA256SUMS.txt`, then *More info → Run anyway*.
**This is the first release in which the part inside a game makes a call of its own.** Until now it only read what
a game itself passed to the graphics calls FrameLedger intercepts. To read this report it asks the graphics driver,
about once a second, on the game's own thread that presents its frames — two functions of the driver that only
read. It reaches them through NVIDIA's library only where the game had already loaded it, and never loads that
library itself; a copy sitting in the game's own folder is not the driver's and is not used. It is built to step
back rather than disturb a game: a request that takes two thousandths of a second or more, or that goes wrong, is
the last one until the game is closed, and the session says so. **With Advanced metrics off — the default — none
of this happens.** The Disclaimer and the Privacy Policy say all of it; the terms window shows once.
**What it is not, yet.** **No real game has been read.** The reading was written and tested against FrameLedger's
own test program with a stand-in for NVIDIA's library, and measured with a test tool against one driver on one
card (GeForce RTX 5080, driver 32.0.16.2114): there a request takes about three millionths of a second, and none
took a thousandth while the tool, like a game using Reflex, slept between frames. What a game's own threads, frame
generation, another driver or another card do with it is for the first real sessions to say; a game that enforces
Windows' Control Flow Guard on a driver that tool was not run against is not covered. On a busy PC the
two-thousandths limit will now and then switch the reading off for a pause that was Windows' and not the driver's;
the session then says *stopped* with the time, and the next start of the game reads again. The session summary's
latency fields still read N/A, and nothing here is read on Direct3D 11, Vulkan or OpenGL. No test runs the whole
path from the game to the ledger through the real capture agent; the Japanese texts are new and unreviewed.
**Also in this release.** The log of the part inside a game now says how a group of hooks went in or came out —
in one go, or one hook at a time — after two groups in the first real sessions on 0.1.0-beta.46 took far longer than
the rest and the log could not say why. And 0.1.0-beta.46's other half has now been seen in a real game: stopping
the part inside a running game took all twenty-nine hooks out under one stop of 83–86 thousandths of a second,
twice, and the game went on.
**Measured overhead this release.** A presented frame: one more read of one word; too small to measure on the
test program. The request to the driver, once a second while Advanced metrics is on: 2.5–2.6 millionths of a second
at the median after a frame that slept, 8–67 at worst, in three runs alone; 3.4 and 130 on a busy PC. Not measured
in a game.
**Updating from `0.1.0-beta.46`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are shown once** — Privacy Policy 2.20 and Disclaimer 2.16. No change to the ledger (schema 23) or to the
shared memory's layout (5); the detection rules stay at `2026.11.2`. **The code loaded into a game changed in this
release**, so a game that was running during the update keeps the old part until it is started again.
### Changed
- **The log of the part inside a game says how a group of hooks went in or came out** (owner decision D137). Since
0.1.0-beta.46 hooks are put into a game in groups, the game's threads stopped once for a group. In the owner's
first sessions two groups took much longer than the others (0.6 s and 1.3 s against 0.1-0.3 s), and the log
could not tell one long stop from a group that had to be finished one hook at a time. The `PATCH_BATCH` line now
reads `install` or `restore` for a group that went in one go, and `install singly` or `restore singly`, with a
count, for one that did not. Nothing else about hooking changed, and what a game is measured by is the same.
### Fixed
- **For developers: a native test that is cut short no longer counts as passed.** The quality gate's native tests
were judged by their exit code alone, and a test program that was stopped early with a clean exit code read as
*Passed* - seen once, when security software removed the test program while the gate ran. Every native test now
has to print the line it ends on, a test added without one stops the build's configure step, and four tests
exist only to prove the rule goes red when it should. Nothing a user installs changed.
### Added
- **The part inside a game can read NVIDIA's own latency report, with Advanced metrics on** (owner decisions D132, D136,
D138-D141). While a session records Advanced metrics, and only then, it asks the graphics driver about once a
second for its report of a frame's stages - simulation, render submit, present, driver, the operating system's
queue, the graphics card - and for its low-latency state. It is a call FrameLedger makes itself, on the thread
that presents the game's frames, through NVIDIA's `nvapi64.dll` only where the game had already loaded it;
Direct3D 12 only. A copy of that library sitting in the game's own folder is refused. A call that faults or takes
two milliseconds is the last one until the game is closed. **The Advanced page shows it** under *Latency*: seven
lines, one a stage, and a table of the driver's low-latency state; the line under a session says what became of
the read - read, the driver reported no frame, not read and why, or stopped and why. They are the driver's
numbers, not a measurement by FrameLedger, and no real game had been read when this was written. **Privacy
Policy 2.20 and Disclaimer 2.16** say what is stored and what the part inside a game now does; the terms window
will show once.
- **For developers: room in the shared memory for NVIDIA Reflex latency** (owner decisions D136, D138). Ten reserved
words of the block that carries Advanced metrics from a game are now named for the driver's latency report -
where the read stands, and a sum for each stage of a frame. Nothing writes them yet and nothing reads them; the
shared memory's layout version did not change.
- **For developers, nothing a user installs: a probe for NVIDIA Reflex latency** (owner decision D132). An
unshipped tool, `fl-probe-reflex`, and a stand-in `nvapi64.dll` among the test fixtures measure what it would take
to read the driver's latency report from inside a game - on FrameLedger's own test program and against the
installed driver, never in a game. What it found is `docs/spike-notes/31-reflex-route-d-harness.md`. A further
leg (owner decision D139) measured the same reads while the probe itself used Reflex's sleep, as a game does:
the reads do not wait behind it. The owner has said to go on to a real game (D136); that work is not in any
release yet. Nothing in the App, the Agent or the part that runs inside a game changed.
+90
View File
@@ -0,0 +1,90 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.48 was cut, 2026-11-14 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.48] - 2026-11-14
**The forty-eighth pre-release: NVIDIA's latency report has now been read in real games, and what they showed is
in it.** On the day 0.1.0-beta.47 came out the owner recorded eight sessions with *Advanced metrics* on. The
report was read in Phantom Blade Zero and in Cyberpunk 2077, with frame generation on and off: the driver answered
every second, the slowest request took 65 millionths of a second, nothing was switched off, and no game was
disturbed that the logs can show. Two games that are not Direct3D 12 and one that never loaded NVIDIA's library
were left alone, each with its one line saying why. The same unsigned installer: verify its hash against
`SHA256SUMS.txt`, then *More info → Run anyway*.
**What the first game showed, and what this release does about it.** In Phantom Blade Zero the line called *GPU
render* was longer than the whole frame — 16 thousandths of a second where the total was 5. That line runs from
the moment the graphics card starts on a frame to the moment it ends, **and includes the time the card waited in
between**. NVIDIA's driver also works out the figure without the waiting, and the time from one frame's end to the
next. Both are in the report FrameLedger already asks for, and a session now stores them as **a second chart under
*Latency*, with two lines**: *GPU active render* and *GPU frame time*. Nothing more is asked of the driver. The
old line stays, under a name that says what it includes. **The stages of a frame overlap; they were never parts
that add up to the total**, and the texts now say so.
**Three more things the sessions showed.**
- The table of the driver's low-latency state kept only the *last* frame-generation multiplier: a session that
generated frames at ×2 and ended in a menu showed 1. It now lists every value the session saw — *1, 2*.
- **Sensors were read every 1.3 seconds where the setting says every second** — on the owner's PC, in nearly every
session since the series began. The capture agent waited the whole interval *after* each reading, and a reading
there takes about a third of a second. It now keeps the period. No chart was wrong, because every sample is
stored with its own time; but a session now has about a third more samples, and readings of the latency report
no longer fall between two of them. What makes a reading take that long there is not established.
- A session could begin with the last reading of the session before it, in a game that kept running between the
two. It no longer does.
**What it is not.** Two games on one driver and one card (GeForce RTX 5080, driver 32.0.16.2114) are not every
game. The two new figures have not been seen from a real game yet: a driver may give neither, and the chart is
then simply not there. Reflex switched off in a game's menu, and a game started with Advanced metrics off, have
still not been tried on a real game. They remain the driver's numbers, not a measurement by FrameLedger, and not
the delay from a click to the screen; the session summary's latency fields still read N/A. Nothing here is read
on Direct3D 11, Vulkan or OpenGL. The Japanese texts are new and unreviewed.
**Measured overhead this release.** A presented frame: unchanged. The request to the driver, once a second while
Advanced metrics is on: summing the two further figures adds about 36 billionths of a second to work that follows
the request itself. In the real games of 2026-11-14 the slowest request took 20 to 65 millionths of a second.
**Updating from `0.1.0-beta.47`:** Help ▸ Check for updates, then *Restart to update* with no game running. **The
terms are shown once** — Privacy Policy 2.21 and Disclaimer 2.17. No change to the ledger (schema 23) or to the
shared memory's layout (5); the detection rules stay at `2026.11.2`. **The code loaded into a game changed in this
release**, so a game that was running during the update keeps the old part until it is started again.
### Added
- **Two more lines from NVIDIA's latency report: how long the graphics card was really at work** (owner decision
D142). In a real game the driver's *GPU render* stage turned out longer than the whole frame: it runs from the
card's start to its end and includes the time the card waited. With Advanced metrics on, a session now also
stores two figures the driver works out itself — **GPU active render**, the same stage without the waiting, and
**GPU frame time**, from the end of one frame's work to the end of the next — as a second chart under *Latency*.
They come from the report FrameLedger already asks the driver for; nothing more is asked. A driver may give
neither, and the chart is then not there. Sessions recorded before this release do not have them.
- **For developers: the part inside a game reads two more figures of NVIDIA's latency report, and keeps every
frame-generation multiplier a session saw** (owner decisions D142, D144). In a real game the driver's span between its two GPU marks was longer than the frame's whole
— it includes the card's waiting. The driver's own *active* render time and its GPU frame time are members of
the report FrameLedger already fetches; they are read from it, with **no further call into the driver**. Four
reserved words of the shared memory carry them and the multipliers seen (layout 5 as it was). `fl_reflex_summary`
and the injected `[latency]` cases hold it. Measured: +36 ns a read, once a second.
### Changed
- **The *GPU render* line says what it includes** (owner decision D143): *GPU render, start to end (idle time
included)*, in the three languages. The numbers are the same.
- **The driver's frame-generation multiplier is every value the session saw** (owner decision D144). The table of
the driver's low-latency state kept only the last one: a session that generated frames at ×2 and ended in a menu
showed 1. It now shows *1, 2*. Sessions already recorded keep what they stored.
- **The texts say what was read** (owner decision D145). The accuracy statement, the README and the guide said no
real game had been read; two have been, on one driver. They also say now that the stages overlap and do not add
up to the total. **Privacy Policy 2.21** names the two new averages; **Disclaimer 2.17** changes only with the
accuracy statement it carries. The terms window shows once.
### Fixed
- **Sensors were read less often than the setting says** (owner decision D146). With the sensor interval at
1000 ms, nearly every session on the owner's PC got a sample every 1.3 seconds: the Agent waited the whole
interval *after* each reading, and a reading there takes about a third of a second (what in it does is not
established; on a PC where the Agent is not run as administrator it takes a millisecond, and the samples were
already a second apart). It now keeps the period. No chart was wrong — every sample is stored with its own time — but a session
has about a third more samples now, and about a quarter of the readings of NVIDIA's latency report no longer
fall between two samples. For developers: `sensor_blobs.hz` was the constant 1 and is now the rate the
session's samples had.
- **A session could start with the last read of the session before it.** In a game that kept running between two
sessions, a read of the driver's latency report that ended after the first session's last tick was taken as the
second session's first point. A new session now drops it.
+79
View File
@@ -0,0 +1,79 @@
<!-- Moved from CHANGELOG.md's [Unreleased] when 0.1.0-beta.49 was cut, 2026-11-14 (D66). CHANGELOG.md lists every release. -->
## [0.1.0-beta.49] - 2026-11-14
**The forty-ninth pre-release: what the owner's sessions on 0.1.0-beta.48 showed, the same day.** Eighteen
sessions, sixteen of them with *Advanced metrics* on — Phantom Blade Zero, Cyberpunk 2077 and Clair Obscur:
Expedition 33, with frame generation on and off, with Reflex on and off. Sensors are now read every second; NVIDIA's driver gave both
of its own GPU times in every report of every game; the table lists every frame-generation multiplier; a game
started with the option off was left alone; the driver was read just the same with Reflex off; and the owner's
word on how the games felt: no hitch. The same unsigned installer: verify its hash against `SHA256SUMS.txt`, then
*More info → Run anyway*. **No legal text changed, so the terms window does not appear.**
**Two things those sessions showed, and what this release does about each.**
- **The *Latency* charts had holes.** After a session's first reading, between 0.4 and 16 % of its samples had
none. There were two reasons, one on each side. The part inside the game did its once-a-second round a second
*after* the last one finished — so it fell behind — and whenever the game loaded a library it was woken early,
published what it had, and asked the driver for nothing new. And the capture agent kept one reading until a
sample could carry it, and replaced it if another arrived first. Both are mended: the round inside the game is
now due at fixed moments, and the agent keeps the reading that waits.
- **In Cyberpunk 2077 without frame generation, NVIDIA's *GPU active render* time is longer than the time from the
card's start on a frame to its end** — which should contain it. Nothing stored can say whether that is the
driver's arithmetic or the way FrameLedger averages the two (one over every frame, one only over the frames that
gave a time). **This release changes no number you see.** It counts, frame by frame, how the two stand to each
other, and writes the counts to the game component's own log, together with the driver's low-latency state and
its frame-generation multiplier — the owner asked that Reflex and frame generation always be read together. The
next sessions say which it is; what to do about it is decided then.
**What it is not.** Not a promise that every sample now carries a reading: with both sides on one second a slip
can still leave one without, and the log says how often. The reason given for the agent's half was read from the
code and fits the pattern in the ledger; it has not been measured in a session. Nothing here is read on
Direct3D 11, Vulkan or OpenGL, and the figures remain the driver's, not a measurement by FrameLedger. The log of
the component inside a game grows by about seven lines a minute while *Advanced metrics* reads the driver — a few
tens of kilobytes an hour — and a bug report carries those logs, as it always carried them. A game that closes by
itself loses the last batch of counts, under a minute.
**Measured overhead this release.** A presented frame: unchanged — no line that runs with a present was touched.
The reading of the driver's report, once a second while a session with *Advanced metrics* runs in a Direct3D 12
game: 65 billionths of a second more (424–435 → 490–499 billionths, the same instrument as before), after the
request's own time has been taken. The background thread inside the game: ten readings in 9 996 ms, twenty in
20 002, sixty in 59 997 on the test harness — and the same with another thread waking it eighteen thousand times
in ten seconds, where the code before it published 35 720 times in eight seconds and read nine times.
### Added
- **The frames NVIDIA's latency report gives are counted, for the Overlay's own log** (owner decisions D148,
D151; nothing new is stored, shown, asked of the driver or read from its report). The owner's sessions on
0.1.0-beta.48 showed both of the driver's own GPU times in two games — and in Cyberpunk 2077 without frame
generation an *active render* time longer than the span from the card's start on a frame to its end, which the
two averages cannot explain: one is over every frame, one only over the frames that gave a time. While a session
with *Advanced metrics* reads the driver, the Overlay now places each frame by its own two numbers and writes
the counts to `overlay-<pid>-….log` as `LATENCY_FRAMES` lines, in batches that each belong to **one state of
the driver's low-latency status and its frame-generation multiplier** — the counts are read together with
Reflex and frame generation, never apart. About six lines a minute; a game that exits by itself loses the last
batch, under a minute. A bug report carries those logs, as it carried them before. Measured: 65 billionths of a
second more for each reading of the report, once a second; a presented frame costs what it did.
### Changed
- **The guide says whose switches the latency table shows**: the driver's, not the game's menu. `bLowLatencyMode`
is also on when *Low Latency Mode* is set for the game in the NVIDIA App, and `bUseGameSleep` says whether the
game itself paces its frames through Reflex — in some games only with frame generation on (the owner's word,
2026-11-14, and what Clair Obscur's sessions of that afternoon read).
### Fixed
- **NVIDIA's latency report is read on a grid of one second, and a busy loader no longer leaves holes in it**
(owner decision D150; the Overlay's half). After a session's first report, 0.4 to 16 % of its samples carried
none — the *Latency* charts had gaps. The Overlay's background thread waited a second *after* each round of its
work, so it fell behind the capture agent, which samples exactly every second; and whenever the game loaded a
library that thread was woken early, published what it had and asked for no new reading, so the round after it
had nothing to publish. Its once-a-second round is now due at fixed moments whatever the work took, and a round
woken early no longer publishes or takes a reading. Two readings are never less than half a second apart.
Nothing on a game's own threads changed.
- **A reading that waits for a sample is no longer overwritten by the next one** (D150; the capture agent's
half). The agent kept one reading from the game until a sensor sample could carry it, and replaced it when
another arrived first. When a reading landed just after a sample, that sample went without and a reading was
lost a little later — in one session of the owner's, one sample in five for four minutes. Two readings may now
wait and the older is stored first; a point can stand up to a second after its reading was taken. With the
sensor interval above one second, not every reading is stored, as before.
+85
View File
@@ -0,0 +1,85 @@
# HANDOFF history — 0.1.0-beta.33
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.34 was cut, 2026-11-06 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood.*
## **START HERE** — 2026-11-03, 0.1.0-beta.33: games from the Xbox app and the Microsoft Store
*Status is `docs/changelog/0.1.0-beta.33.md`; this is what no other file carries.* The owner: "Trong phiên này, ta sẽ tập
trung xử lý, đánh giá, phân tích cho các game thuộc MS Store/ Xbox App/Game Pass cho (các) phiên bản tiếp theo nhé. Quy
trình vẫn như mọi lần." — handle, evaluate and analyse the games of the Microsoft Store, the Xbox app and Game Pass for the
next versions, the usual train; the owner's games are in `C:\XboxGames` and `D:\xbox`. #351 (the library, the install
root, detection, the art, Privacy 2.13), #352 (Play, the guard's two refusals, one exe-path reader) and this release;
`0.1.0-alpha.2` follows (D82).
**Decisions (asked while planning, do not re-ask):**
- **D77 — measure first, on the owner's titles**: the launch chain was measured by activating six of the eleven packages
with no FrameLedger hook, read-only, each game closed by its window and never killed, with the decision table committed
before the run (`spike-notes` §25, its two commits in #351).
- **D78 — GDK only**: an app-container (UWP) process is refused with a reason of its own, `TargetAppContainer`, Tier 2
(`19_SAFETY` §The target must be able to take the Overlay); `LIMITATIONS.md` says so.
- **D79 — the platform is "Xbox"** (id `xbox`), full scope: import, art, Play, Tier-1 capture, detection, docs and legal.
- **D80 — Play on an Xbox game always activates its package**, in both D59 modes (`08_UI` §Games): started by its path a
GDK game is never the game in its package — the run corrected the plan's premise that `CreateProcess` fails with error 5
(it hands the launch executable back to the package, so the game is not the launcher's descendant; an Unreal shipping
executable runs outside it). The Agent records such a game in attach mode.
- **D81 — title lists stay keyed by Steam id**: an Xbox copy is matched by everything but that entry; the gap is
`20_OPEN_QUESTIONS` §G *Xbox app / Microsoft Store titles*.
- **D82 — alpha.2 right after beta.33** (the owner runs `0.1.0-alpha.1`, and the alpha's channel is locked): on RC3; if
the train reaches the alpha after .NET 11's release on 2026-11-10, the standing rule applies (SDK 11.0.100, packages
11.0.0) — no new decision.
The Privacy Policy 2.13's wording was approved with the plan (D36).
**Choices made here:**
- **A GDK package root is an install-root boundary**, for the guard's check 4 and for detection alike, and only where no
store boundary matched — so it can only widen what is walked (`19_SAFETY` §Pre-injection checks, amended 2026-11-03).
The guard gained `GamesInstalled\<X>` with it, and one case list read by both test suites
(`tests/fixtures/install-roots/cases.txt`) holds the two resolvers together.
- **`root_file`, a new detection signal**, answered from the install root's own listing, so the `xbox` platform rule —
first, because a GDK FINAL FANTASY ships Steam's DLL — is decidable even where a very large install's walk ran short.
- **The two refusals read the target once, before the primitive**, never at the 30 s re-scan; "could not tell" refuses as
`ProcessUnreadable`, and only enforcing flags count (`SetContextIpValidation` without strict mode is on every
CET-compatible process: `pwsh.exe` reads `0x105`).
- **`TargetResolver` reads the watcher's spelling** (`ProcessImagePath`) but keeps the guard's rights, so D21's
`TargetUnreadable` is unchanged.
**Measured on the way** (`spike-notes` §25): eleven GDK packages, full trust, the user's token, medium integrity, none of
the four mitigations; the image path and the loader's main module both in the `Content` spelling; every game `.exe`
refuses `FILE_READ_DATA`/`FILE_EXECUTE` outside its package, so no Unreal version and no *Runs as* for these titles;
`GetProcessMitigationPolicy` answers through `PROCESS_QUERY_LIMITED_INFORMATION`, and an unknown policy answers
`ERROR_INVALID_PARAMETER`.
**Traps met on the way:**
- **Malwarebytes quarantines `fl_guard_test.exe`** (`Malware.Heuristic.108256`, records in
`C:\ProgramData\Malwarebytes\MBAMService\RtpDetections`) on some builds and not others: ctest then reports `fl_guard` as
passed in 0.5 s with no output and `fl_prescan` with an empty command. Read `LastTest.log` for the assertion count before
believing a native pass; an exclusion for `E:\frameledger\build\` is the owner's to add (a security setting), and CI runs
both suites.
- **`build.ps1` finds the x64 clang-format; `Get-ChildItem … clang-format.exe` finds the ARM64 one first**, which will not
start on this PC.
- **SonarCloud's quality gate on #351 was 77.7 % new-code coverage** (≥ 80 % wanted) and one `cpp:S5813` (`wcslen`): both
closed in #352 — its own gate 95.4 %, main's after the merge 86.5 % on the code since beta.32 with no open issue. The
job is advisory, so a merge on green does not wait for it — read its verdict after each merge.
- **FluentAssertions' `OnlyContain` fails an empty collection**: a test of what this PC holds (`GamingRoots.OnThisPc()`)
passed here and failed on the hosted runner, which has no `.GamingRoot` (#352's first CI run) — assert that nothing
is wrong rather than that everything is right.
**Owner-only:** item 28 — the Xbox titles on real hardware.
### 0.1.0-alpha.2 — the second alpha, the same day (D82)
`alpha` took main at this pull request — `v0.1.0-beta.33` and these notes, no other change — by a merge commit, as
`13_CI_CD` §The alpha line says: no file the switch holds had moved on main, so nothing conflicted and the switch is the
first alpha's (SDK `11.0.100-rc.3.26522.148`, the three packages at `11.0.0-rc.3.26522.148`, the same NuGet licence
texts). Its gate on the 11 SDK must pass before the pull request into `alpha` is merged by hand (the branch has no
auto-merge); the release runs from the signed tag `v0.1.0-alpha.2` on `alpha`, and its delta from alpha.1 is the first
an alpha's release has to find — #350's lookup, by the name the pack lists. Its notes are
`docs/changelog/0.1.0-alpha.2.md`, which open with the experimental-build warning.
- **The next alpha moves to .NET 11's release**: from 2026-11-10 the switch is SDK `11.0.100` and the packages `11.0.0`,
and RC3 is out of support — move the alpha even if no beta comes.
**Owner-only:** item 29 — the second alpha on real hardware.
+90
View File
@@ -0,0 +1,90 @@
# HANDOFF history — 0.1.0-beta.34
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.35 was cut, 2026-11-09 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood.*
## **START HERE** — 2026-11-06, 0.1.0-beta.34: DLSS 5 and DLSS 4.5 from what the game and the driver did; filters; Games at once
*Status is `docs/changelog/0.1.0-beta.34.md`; this is what no other file carries.* The owner asked for seven things, the
usual train: (1) how DLSS 5 Neural Rendering is detected on Hogwarts Legacy (Neural Rendering on in the game), and its
version; (2) under the NVIDIA App's *Dynamic* multi frame generation the label read DLSS 4, though Dynamic and ×6 exist
only in DLSS 4.5 (Phantom Blade Zero); (3) filters on Games and Compare; (4) the Compare page's layout, from a
screenshot; (5) the documents and the tested-games list, from the database; (6) switching to Games stalled 0.5–1.5 s;
(7) a language change landed on the last content page instead of Settings. And while planning: "Bạn check Streamline các
bản gần đây có gì thay đổi và sửa luôn nếu thật sự cần nhé." — check what changed in recent Streamline versions and fix it
if really needed. #356 (two fixes from a self-review of #351 and #352), #357 (D83), #358 (D84, Streamline `v2.14.16`),
#359 (items 6 and 7), #360 (items 3 and 4, D85, D86), #361 (item 5) and this release; `0.1.0-alpha.3` follows (D87) —
the owner's screenshot is of an alpha, so the alpha is how these reach the owner's PC.
**Decisions (asked while planning, do not re-ask):**
- **D83 — Dynamic frame generation is DLSS 4.5's**: the owner's rule, "Dynamic and ×6 exist only in DLSS 4.5". The
evidence is the game's driver profile setting Dynamic (`NGX_DLSSG_MODE` = 4) **and** the driver reporting its
frame-generation override applied in the game's process (`FG_MODE` in its FG word), in a session where DLSS
Frame Generation was identified (`fg_mode = dlssg`; its factor need not have been counted); a state measured at ×5 or more still ranks first, and a *Fixed* multiplier names nothing by itself
(amends D64). The driver's `frameGenerationMode` — 3 under Dynamic and 1 under Fixed in every session read, and unnamed
by NVIDIA — is recorded in `spike-notes` §26, not keyed on.
- **D84 — DLSS 5 is the session's generation where its neural rendering was measured running**: the Overlay decodes
`kFeatureDLSS_NR` in its `slEvaluateFeature` detour and the Agent keeps its share of the Streamline batches
(`sessions.dlss_nr_pct`, schema 0022) — *DLSS 5 (neural rendering on N % of rendered frames)*; loaded and seen not
running, *runtime loaded, not seen running*; where nothing could tell, *runtime loaded*, as before (amends D64).
- **D85 — the filters**: Games — store, engine, state, installed, played; Compare — game, tier, frame generation, no
period; Compare lists every session (it listed the newest 200).
- **D86 — the left pane's selected item keeps its label**: our copy of WPF UI 4.3.0's LeftFluent item template, whose
`IsActive` trigger animated it away.
- **D87 — alpha.3 right after beta.34**, on RC3: the train reaches it before .NET 11's release on 2026-11-10.
Streamline was re-vendored at `v2.14.16` on the owner's instruction above, not as a decision of its own. The wording of
the Disclaimer 2.13 and the Privacy Policy 2.14, and the tested-games list (D35), were approved with the plan (D36).
**Choices made here:**
- **Neural rendering's share is over the Streamline batches**, not every present: a generated frame evaluates nothing,
and a title that never drains a batch reads *nothing could tell* (null), never 0 %.
- **A writer older than beta.34 is told apart by its family bit**, `FL_HOOK_SL_NEURAL_RENDERING` (bit 12), published with
the detour — an old Overlay's silence leaves `dlss_nr_pct` null instead of a measured 0 %; no layout change.
- **The Games page is the window's**, kept by our own page provider (`WindowPageCache`), not `NavigationCacheMode`: WPF UI
4.3.0 reads that only when neither a service provider nor a page provider is set, and `SetNavigationControl` sets one
(`16_WPFUI_SYNTAX` §Navigation). One per window, so a language change's window builds its own.
- **The cards come before the executable checks** — 39 of the owner's 41 Steam games are on a USB drive that sleeps — and
each card gets its *Not installed* pill when its check answers.
- **The pane's item template is a copy changed in one place**, reading WPF UI's own `NavigationViewConstants.xaml`: before
the WPF UI pin moves, the upstream template is read again.
**Measured on the way:** the owner's DLSS sessions of 2026-11-01 → 11-06, the driver's frame-generation numbering and
Streamline's `v2.12.0` → `v2.14.16` changes are `spike-notes` §26 (the ledger read read-only; the vendored copy was
`v2.12.0` byte for byte); the Games page before and after is §27 (first cards 822–830 ms → 10–15 ms with the first
executable check held 800 ms; a revisit 181–215 ms → 15–16 ms). The neural-rendering decode costs nothing measurable in
the detour and 0.14–0.19 ns on the present's feature byte (#358's body; NFR-1's budget is 1 µs per present).
**Traps met on the way:**
- **`notice-check` reads the tracked tree**: a file deleted from the working tree but not from the index fails it with
*Could not find file* — after every build and test had passed (#360's first local gate). Stage the change
(`git add -A`) before the gate.
- **A new hook family bit belongs in the native token test's exact mask** (`guard_test.cpp`), and a new Streamline fact
in the 0xF00D case's list of facts it must not light — #358's first gate (`0x100f` against `15`).
- **WPF UI 4.3.0**: `NavigationCacheMode` is ignored once a page provider is set; `INavigationService.SetNavigationControl`
hands the NavigationView the container's provider, so ours is set after it; LeftFluent's item template collapses the
selected item's label.
- **Microsoft.Data.Sqlite's async methods run synchronously**: a read awaited on the UI thread blocks it — `Task.Run` it.
**Owner-only:** item 30 — beta.34 on real hardware.
### 0.1.0-alpha.3 — the third alpha, the same day (D87)
`alpha` takes main at this pull request — `v0.1.0-beta.34` and these notes, no other change — by a merge commit, as
`13_CI_CD` §The alpha line says. One file the switch holds moved on main — `legal/THIRD_PARTY_NOTICES.md`, its
Streamline row (#358) — on a line the switch does not touch (the switch changes the *Transitive* row), so the merge
takes both without a conflict (`git merge-tree` before the pull request); main's package set did not move, and the
switch stays the first alpha's (SDK `11.0.100-rc.3.26522.148`, the three packages at `11.0.0-rc.3.26522.148`, the same
NuGet licence texts) — the train reached it before .NET 11's release. Its gate on the 11 SDK must pass before the pull
request into `alpha` is merged by
hand (the branch has no auto-merge); the release runs from the signed tag `v0.1.0-alpha.3` on `alpha`, with its delta
from alpha.2. Its notes are `docs/changelog/0.1.0-alpha.3.md`, which open with the experimental-build warning. It is
the first alpha whose ledger moves (schema 22): the copy it makes first is the database as alpha.2 left it, and going
back is to the newest beta, which opens schema 22.
- **The next alpha moves to .NET 11's release**: from 2026-11-10 the switch is SDK `11.0.100` and the packages `11.0.0`,
and RC3 is out of support — move the alpha even if no beta comes.
**Owner-only:** item 31 — the third alpha on real hardware.
+74
View File
@@ -0,0 +1,74 @@
# HANDOFF history — 0.1.0-beta.35
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.36 was cut, 2026-11-10 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then.*
## **START HERE** — 2026-11-09, 0.1.0-beta.35: accessibility — a statement, what a screen reader is given and told, headings and charts, contrast themes, focus
*Status is `docs/changelog/0.1.0-beta.35.md`; this is what no other file carries.* The owner: "Bạn lên plan và thêm
phiên bản mới (nếu thực sự cần) cho mục accessibility: 1. đọc docs, check, sync và sửa, cập nhật thêm file
ACCESSIBILITY.md vào đúng vị trí. 2. Check các Accessibility đã có, đề xuất thêm accessibility nếu cần. 3. Đã dùng: bàn
phím only (tắt chuột) và narrators, không lỗi." — plan accessibility, with a new version if it is really needed: read,
check, sync and fix the documents and add `ACCESSIBILITY.md` where it belongs; check what exists and propose more; the
owner had used the keyboard alone, mouse disconnected, and Narrator, with no fault. With opensource.guide's
accessibility practices and the open-source-accessibility toolkit's `ACCESSIBILITY.md` template (MIT: its structure,
not its text). #366 (the statement, D88), #367 (names and peers), #368 (announcements, D89), #369 (headings and charts,
D90), #370 (contrast themes, D91, and three contrast holes outside them), #371 (focus and motion) and this release; no alpha (D92).
**Decisions (asked while planning, do not re-ask):**
- **D88 — the accessibility statement**: `ACCESSIBILITY.md` at the repository's root, in the App under Help ▸
Accessibility and on the wiki. WCAG 2.2 at level AA, as W3C's WCAG2ICT applies it to desktop software, is a goal and
never a claim; a report is handled by severity, with no promised times; the owner's pass of 2026-11-09 — keyboard only
and Narrator, on 0.1.0-beta.34 and 0.1.0-alpha.3, no fault — is its first evidence by hand.
- **D89 — what appears is said to a screen reader**: UI Automation notifications for a safety notice, the agent's and
the update's banners, every snackbar, a game page's notice and Database maintenance's results, only while
FrameLedger's window is the active one.
- **D90 — headings and charts**: a page's title is a heading of level 1, a section's title level 2; every chart is
named with where its numbers are written; no chart tells two series apart by colour alone.
- **D91 — Windows contrast themes are followed**, whatever Settings ▸ Theme says.
- **D92 — beta.35 in this train and no alpha**; alpha.4 is the move to .NET 11's release (SDK `11.0.100`, packages
`11.0.0`), a task of its own after 2026-11-10.
**Choices made here:**
- **A contrast theme is the user's colours by role**, not WPF UI's high-contrast files (84 red placeholders among 97
colours, 17 brushes that throw, Windows 11's defaults) and not a hand list of keys: every key of the four files takes
the colour of the role it has in all of them, opaque colours only (`ContrastRoles`), over the light or the dark Fluent
theme as the contrast theme is; the text on the accent is the highlight's text, the App's choice.
- **What must change while it is shown binds `ContrastTheme.IsOn`**, a static property WPF hears through
`StaticPropertyChanged`: the implicit button style, the Settings line. The charts, the badges and the artwork band read
`ContrastTheme.Active` when they draw.
- **The snackbars lost their green and orange** (white on fixed colours, 2.16–2.78:1 in both themes): every message is
Secondary, its icon saying the kind. **The agent's pill is a `Pill` on the theme's fills**, no longer a badge.
- **A content dialog's focus is WPF UI's**: its host stores and restores the focused element; a test pins it. Message
boxes get the App's active window as owner (`MessageBoxes`), which WPF UI's leaves null.
- **A page shortcut focuses the page's first control and says its first level-1 heading** (`PageFocus`); a click in the
pane leaves the focus where WPF UI puts it.
- **Left as limitations, with what to do instead** (`ACCESSIBILITY.md` §Known limitations): a chart's points one by
one, headings inside the documents the App shows (a FlowDocument paragraph has no peer), the snackbars' four seconds,
Windows' text size, pausing only from the tray.
**Measured on the way:** the UI Automation tree WPF builds, on .NET 10.0.14 — items named by `ToString()`, peerless
focusable controls, a template's text hidden from the control view, template cells read as *Item: …* (`16_WPFUI_SYNTAX`
§Gotchas); WPF UI 4.3.0's high-contrast files, its Button, Badge and Snackbar appearances (fixed colours, the ratios
above), its message box's missing owner and its content dialog host's focus restore; ScottPlot 5.1.59's keys (focusable,
none marked handled) and its double-click benchmark.
**Traps met on the way:**
- **A `ContentDialog` test needs the dispatcher's context.** `ContentDialog.ShowAsync` takes the dialog out of its host
in a `finally` after an await; on a test thread with no `DispatcherSynchronizationContext` that never resumes on the UI
thread, the dialog stays in its host and the focus on its button — a first measurement concluded WPF UI kept the focus
there; it does not.
- **A test that listens to a static event hears every test's**: subscribe inside the one STA work item.
- **Inside the App, a brush that throws when WPF UI's file is read alone falls back to the application's resources**, the
Fluent theme's translucent fills: `MenuBarItemBackgroundPressed` read as the window text's white. Count only opaque
colours as evidence.
- **WPF calls a resource reference's callback when a merged dictionary changes, even for an equal value** (Aquatic's
window is the dark theme's `#202020`): no second signal was needed, measured and pinned.
- **A forward `StaticResource` in a style dictionary fails only when that resource is first used** — `AgentPill`
`BasedOn` `Pill`, placed above it: the dictionary loaded, the style threw when found.
- **Python's `"\b"` in a non-raw string is a backspace** — it went into a C# regex twice; write regexes raw.
- **`RenderTargetBitmap.Render(child)` draws the child at its offset**: sample a child's pixel through its parent.
**Owner-only:** item 32 — beta.35 on real hardware.
+79
View File
@@ -0,0 +1,79 @@
# HANDOFF history — 0.1.0-beta.36
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.37 was cut, 2026-11-11 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then.*
## **START HERE** — 2026-11-10, 0.1.0-beta.36: Qodana in CI, and what its first run reported settled
*Status is `docs/changelog/0.1.0-beta.36.md`; this is what no other file carries.* The owner: "Bạn có thể thêm vào,
trigger Qodana vào CI/CD và sửa lỗi, release (nếu có) với mẫu mà jetbrains cung cấp, có thể tuỳ biến (nếu cần)" — add
Qodana to CI from the workflow JetBrains gives (the Qodana Scan action on `ubuntu-latest`, `pr-mode: false`,
`QODANA_TOKEN`), changed where needed, fix what it reports, release if there is something to release; the usual train.
The owner had made the project on Qodana Cloud and set the `QODANA_TOKEN` secret the same day. #373 (the workflow, the
settings, the fixes and the documents, D93) and this release; no alpha (D92 stands: alpha.4 is the move to .NET 11's
release).
**Decisions (asked while planning, do not re-ask):**
- **D93 — Qodana Community for .NET, from a workflow of its own.** Asked which licence the Qodana Cloud organisation
holds — it decides the linter — the owner answered "Dùng thử 30 ngày", a 30-day trial. So not the Rider-based linter
(Ultimate; the job would turn red when the trial ends) but the Community one, which runs under every licence. A gate
of zero problems, advisory; `docs/13_CI_CD.md` §Qodana.
**Choices made here:**
- **An inspection is switched off in `.editorconfig`, never in `qodana.yaml`**: InspectCode reads its severities there
and so does Rider, so the job and the editor show one list, and a change to it is tested on a developer's PC without
CI. Sixty-three keys, each with its reason; the rule is CLAUDE.md §Coding conventions'.
- **What Qodana reports is fixed unless the inspection argues with a choice the project made.** The suggestions
(ReSharper's lowest reported level) that propose another style are off as a group: the style is `.editorconfig`,
`dotnet format` and the analyzers.
- **The members nothing in the solution uses were listed, not removed** (`13_CI_CD` §Qodana): whether each is
vocabulary, a piece not wired yet, or dead is the owner's to say (owner-only item 33).
- **One pull request, the gate shown red first**: the settings and the gate were pushed before the fixes, so the run
between them (42847369329) is the gate failing on the 409 problems the settings did not cover.
- **A release although nothing a user sees changes**: the fixes are in the shipped assemblies, so the version says
which code is installed; the changelog says in its first line that nothing visible moved.
- **The XAML bindings were re-pointed, not only annotated**: `{Binding DataContext.ViewModel.X, RelativeSource
AncestorType=Page}` became `{Binding ViewModel.X, RelativeSource AncestorType={x:Type pages:GamesPage}}` — the same
object, since each page and window sets `DataContext = this` (read in the five code-behinds) — because the inspection
resolves a typed ancestor and not `DataContext`.
**Measured on the way:** the Community linter's container builds and loads all thirteen `net10.0-windows` projects on
Linux once `EnableWindowsTargeting` is set, though JetBrains' page sends such projects to the Ultimate linter's native
mode; a clean worktree with no native DLLs builds with 15 warnings and no error; InspectCode 2026.2.3.1 on Windows and
in the container agree to within a handful of findings; `.editorconfig`'s `resharper_*` keys are read in the container.
**Traps met on the way:**
- **InspectCode on the owner's PC picks Visual Studio 18 Insiders' MSBuild and dies** ("No files to inspect were
found"): pass `--toolset-path` to the .NET SDK's `MSBuild.dll` and `--dotnetcore`. And **a `--caches-home` a failed
run wrote leaves a project unloaded** on the next run — 2,875 false compile errors in `Infrastructure.Tests`, and
hundreds of false "never used" — use a fresh folder.
- **A fix scripted from the report's positions needs the build as its judge**: a redundant tuple name's region stops
before its colon; two casts in one lambda (`SqliteGameMerge`) are each redundant alone and not together; a suggested
`m[^1]` sat in an expression tree (CS8790); `public bool On => false` on a bound row is CA1822.
- **`dotnet format style --diagnostics IDE0008 IDE0022 --severity info`** applies `.editorconfig`'s suggestion-level
style, which the gate's `--verify-no-changes` (warnings and up) never asks for.
- **`CrashDumpHelperTests.TheWriterHasTheAgentDumpThisProcessFromOutside` failed once on the hosted runner** (#373's
first run, two YAML files changed): "the dumper did not finish within 30 s". Green on the next runs; not looked into.
- **SonarCloud's gate failed on #373, on one condition: coverage on new code, 77 %** — 19 of the 87 changed lines it
counts were old lines no test ever ran (the update prompts, `App.xaml.cs`, chart code-behind), and a line a fix
touches is new code. No issue, no hotspot; the job is advisory and #373 merged on its required checks. `main`'s new
code is "since the previous version", so its gate reads the same until the first analysis after the tag — which
was the release commit's own (its `sonarcloud` job started after the tag was pushed, read version 0.1.0-beta.36 and
passed); had the tag come later, a `workflow_dispatch` of `ci.yml` on `main` would have been that analysis.
- **A release's delta says what the notes forgot**: read "Delta processed … patched" in the tag's log before
calling a release verified — 343 patched files was the runtime, which no line of the notes named.
- **A shell `cd` in a tool call stays** for the next one: use absolute paths.
**The release (the tag, 2026-11-10):** `v0.1.0-beta.36` on `deaa8f3`, a rehearsal of `release.yml` first (run
42853812423, on the release branch), the tag's run 42856948658 green; the seven digests of `SHA256SUMS.txt` are
GitHub's; the wiki published. **The package carries the .NET runtime 10.0.15, not beta.35's 10.0.14**: the hosted
runner installed SDK 10.0.404 that day, 10.0.403 the day before (`global.json` rolls forward within 10.0), so the delta patched 343
of 456 files (5,756,432 bytes) where a release with these changes alone would have patched a few dozen. The notes at
the tag did not say so; the changelog file and the release's text were corrected the same day, marked. Still open
from it: the 10.0.15 servicing packages (`Microsoft.Extensions.Hosting`, `Microsoft.Data.Sqlite`,
`System.Diagnostics.EventLog` are at 10.0.14) — Dependabot's cooldown holds them a week. `main` after the tag: `check`
green, Qodana 0 problems, SonarCloud's gate OK again (its first analysis with version 0.1.0-beta.36).
**Owner-only:** item 33 — Qodana's settings on the owner's side, the never-used list, and beta.36 on real hardware.
+205
View File
@@ -0,0 +1,205 @@
# HANDOFF history — 0.1.0-beta.37
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.38 was cut, 2026-11-11 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then.*
## **START HERE** — 2026-11-11, 0.1.0-beta.37: the main line on .NET 11, what an alpha is now, and why no C# 15 yet
*Status is `docs/changelog/0.1.0-beta.37.md`; this is what no other file carries.* The owner, with two screenshots — a
Qodana Ultimate licence in the JetBrains console, and the Qodana Cloud project's setup page: "Bạn có thể bắt đầu với phiên
cải tiến cho .net 11 và set up cho qodana cho tất cả ngôn ngữ (gói ultimate) nhé." The session D92 had left for after
.NET 11's release (2026-11-10), and Qodana for every language on the Ultimate licence. One plan, approved whole, in
three parts: **beta.37** (#376 the move, #379 what an alpha is, #380 .NET 11's API and the C# 15 finding, and this
release) → **alpha.4** → **Qodana Ultimate in two reports**. This section is the first part; the other two follow it on
`main` the same way.
**Decisions (asked while planning, do not re-ask):**
- **D94 — the main line moves to .NET 11 now.** Asked whether the betas move or only alpha.4 — with
`20_OPEN_QUESTIONS` §G's table quoted: two alphas on the released runtime and an owner's run of a hooked title first —
the owner chose "Chuyển main sang .NET 11 ngay". Condition (a) of that table was **set aside, not met**;
`12_BUILD` §The main line's .NET says what stood in for it, and owner-only item 34 is the run it asked for.
- **D95 — Windows 10 22H2 is not blocked, and is stated as untested and unsupported** ("Vẫn chạy, ghi rõ không được hỗ
trợ"). NFR-8 is amended; a processor needs SSE4.2 and POPCNT.
- **D96 — C# 15 in the same release** ("Áp dụng cả C# 15 ngay trong beta.37"). The plan narrowed it in writing, and was
approved with that: a construct only where the build, CodeQL and Qodana all read it.
- **D97 — all three legal texts reworded** ("Sửa cả ba văn bản"): EULA 1.7, Disclaimer 2.14, Alpha notice 1.1. Their
text was in the plan and approved with it.
- **D98 — alpha.4, then the alpha line rests** until a newer .NET has a preview ("alpha.4 rồi nghỉ tới .NET 12
Preview").
- **The order**: beta.37 → alpha.4 first, Qodana after; and Qodana in **two reports from the start** (the third part's
decision, D99).
**Choices made here:**
- **No C# 15 construct is used, although the owner asked for them.** The rule decided it, not a preference:
ReSharper's engine, which Qodana runs, reads none of the four (`13_CI_CD` §C# 15 and the analyses, `spike-notes`
§29). Unions would not have been the form in any case — the SDK's own CA1815 asks every union to hand-write its
equality, and the shapes that wanted one are closed record hierarchies. What the release takes is API.
- **.NET 11's `IHost.RunAsync` is accepted, not wrapped**: a failed background service now ends the Agent through its
crash report. A catch there would hide a fault, and could not reproduce .NET 10's exit exactly either.
- **A resting alpha says so in the build, not only in its notes**: the anti-cheat list reaches a user only inside a
release, so an alpha nobody updates carries an ageing one. `FlAlphaLineRests` is stamped on both lines; the `alpha`
branch sets it to false when it next tries a .NET.
- **"A prerelease .NET" is said only by a build that runs on one**, read from the runtime's own description — the next
preview alpha says it again with no change to the code.
- **`release.yml` names the .NET 10 runtime `vpk` needs**, rather than relying on the hosted image carrying one.
- **The Community Qodana job was kept alive across the move** by installing the SDK into its container — an interim;
the Ultimate linter replaces the job in the third part.
- **The 2026-10-31 forms stay where they read as well** (`json.AsSpan().TrimStart().StartsWith('[')` avoids a string):
.NET 11's `char` overloads replaced a span only where the span was there for .NET 10's sake.
**Measured on the way** (`spike-notes` §28 and §29): the gate on SDK 11.0.100 — two analyzer sites beyond RC3's
twenty-one, both in tests written after the last alpha; `IHost.RunAsync` rethrowing, observed for the first time; the
first delta across two runtime majors (372 of 443 files patched, 72.6 MB on the rehearsal); `main`'s first full CodeQL
C# analysis on `net11.0` (1,221 of 1,221 files, the same 23 unresolved `Windows.Win32` types as on .NET 10); the
SonarScanner 12.0.0 on the 11.0 runtime; InspectCode 2026.2.3.1 and 2026.3.0 against each C# 15 construct.
**Traps met on the way:**
- **A pull request's CodeQL C# run is an overlay analysis**: it extracts the changed files on top of `main`'s cached
database, and its "scanned N out of N" is `main`'s baseline. #376 wrote "CodeQL reads `net11.0` whole" from one and
was corrected, dated, two pull requests later with `main`'s own run. For anything the extractor has not met, dispatch
`codeql.yml` on the branch (it takes `workflow_dispatch` since #380).
- **`ShmDrainIntegrationTests.TheGuardInjectsTheOverlayAndTheReaderDrainsRealFrames` failed once on the dev box** —
"our own harness must not start matching the blocklist mid-run": the guard's re-scan refused the harness in one gate
run of five. The class passed three times in a row straight after and no guard code had changed; not looked into.
- **A wiki page names no version** (`wiki-check`): "since 0.1.0-alpha.4" in `wiki/Releasing.md` was red. The guide's
pages may.
- **The runner queue**: with three workflows per pull request and a probe's pull request open beside the train,
`check` waited about 25 minutes for a runner. Cancel a probe's runs once it has told what it can.
- **A Bash heredoc halves the backslashes** of an inline Python regex — write the script with the Write tool.
- **`x.Should().NotBeNull(); x!.Member`**: ReSharper reads the `!` as redundant (Qodana's one finding on #376) and the
compiler wants it. Keep the value the assertion needs, not the object.
- **CA2025** on a lambda that captures a `using` variable in a test: await inside a `try`, assert after.
- **A local gate reseeds `%LOCALAPPDATA%\FrameLedger\rules`** while the owner's installed copy runs beside it: the
files were hashed before and after every gate (identical each time — the repository's rules were the installed ones).
**The release (the tag, 2026-11-11).** Rehearsed first (`release.yml` dispatched on the release branch, run
42926539241), then the signed tag `v0.1.0-beta.37` on `1a7fbf3`: run 42930394186, published 2026-11-11T00:36:54Z.
Eight assets, the seven listed in `SHA256SUMS.txt` equal to the digests GitHub computed; `FL_BUILD_ID v0.1.0-beta.37`
in the shipped natives. The delta across two major runtimes exists — the stop condition the plan had named: "Delta
processed 0443 files. 0373 patched, 0070 unchanged, 0009 new, 0017 removed", 72,681,960 bytes beside a full package of
110,325,368. The wiki was published by its own run (42932792575).
**Owner-only:** item 34 — beta.37 on real hardware, a hooked session on .NET 11.0.0, and Windows 10 if a PC with it is
at hand.
### 0.1.0-alpha.4 — the fourth alpha, the same day, and the last for now (D98)
`alpha` takes main at this pull request — `v0.1.0-beta.37` and these notes, no other change — by a merge commit. **This
sync is not like the three before it**: main now holds what the switch held (`global.json`, the TFM, the three
packages, `legal/licenses/nuget/`, the *Transitive* row), so every switch file conflicts, and each is resolved by
taking main's — the branch's RC3 lines go, `RELEASE_LINE` stays `alpha`, and `git diff --name-only origin/main` on the
merged tree must print that one name. Its gate must pass on `alpha` before the pull request is merged by hand (the
branch has no auto-merge); the release runs from the signed tag `v0.1.0-alpha.4` on `alpha`, with its delta from
alpha.3. Its notes are `docs/changelog/0.1.0-alpha.4.md`: the warning, and that it is the last alpha for now.
- **After it the line rests** (`12_BUILD` §The alpha line): no sync follows a beta and no alpha is cut until a newer
.NET has a preview. `FlAlphaLineRests` in `Directory.Build.props` says so to the build; the next trial sets it to
`false` on `alpha` with the rest of the switch.
- **What a resting alpha costs its user** is in its notes, the guide and its banner: no update, the anti-cheat list
included; and once ten betas have followed it, a copy of alpha.3 or older no longer finds alpha.4 (Velopack reads
the newest ten releases).
**The sync and the release (2026-11-11).** #382 put the notes on main (`428b723`); #383 merged main into `alpha` by
hand (`134115b`): twenty-seven conflicts, every one resolved to main's side, and `git diff --name-only origin/main
origin/alpha` prints `RELEASE_LINE` alone. Rehearsed on `alpha` (run 42933603983: channel `alpha`, a delta from
alpha.3), then the signed tag `v0.1.0-alpha.4`: run 42935876535, published 2026-11-11T01:16:22Z as *FrameLedger
0.1.0-alpha.4 (experimental alpha)*, a pre-release. Eight assets, seven of seven digests equal; "Delta processed 0448
files. 0369 patched, 0079 unchanged, 0004 new, 0005 removed", 18,207,557 bytes; `FrameLedger.App-alpha-Setup.exe`,
`releases.alpha.json`; the wiki run skipped, as it does for an alpha.
- **One tag push started two release runs** (42935876535 and 42935878584, both `push`, created in the same second;
the tag was pushed once). `release.yml`'s concurrency group made the second wait for the first, and it then began
the whole build again. It was cancelled once the first had published: it could only have ended red, at
`gh release create` on a release that existed. If a tag ever shows two runs, let the first finish, verify the
release, then cancel the other.
**Owner-only:** item 35 — the fourth alpha on real hardware, and the beta installed over it.
### Qodana for .NET on the Ultimate licence — the managed report (D99, D100)
*Status is `docs/13_CI_CD.md` §Qodana and `CHANGELOG.md`; this is what no other file carries.* The third part of the
plan, in two pull requests: this one moves the existing job to Qodana for .NET (Rider's engine, native on a Windows
runner) and settles what it reports over the managed code; the next adds the C++ as a second report.
**Decisions:**
- **D99 — two reports from the start** ("Hai báo cáo ngay từ đầu"). The owner made a second Qodana Cloud project,
`frameledger-cpp`, and the secret `QODANA_TOKEN_NATIVE`; the managed report goes to `frameledger` with
`QODANA_TOKEN`. Both secrets are the owner's; no token was ever read or handled here.
- **D100 — where an inspection is switched off** (the plan's design, approved with it; it amends D93's "never in
`qodana.yaml`"): in the place the engine that raised it reads — one of ReSharper's in `.editorconfig`; what has no
key there in `qodana.yaml`. Hints are off as a severity. In code loaded into a game only defects are fixed.
**Choices made here:**
- **JetBrains' security rules stay on, scoped by path — which took four runs to find.** Every result was read first
(the table in `13_CI_CD` §Qodana), and one place hardened (`ShellUrlOpener` refuses anything but `http`, `https`
and `mailto`). Then the files read had to be told to the linter, and a rule's own id is honoured nowhere:
`ignore`, an `exclude` path and `enabled: false` each left all 36 results standing. The report's own
`log/effective.profile.xml` and `qodana_inspections_summary.csv` gave the answer — one inspection,
`OpenGrepInspection`, produces every rule's results, and it does take a path. Between the third run and the
fourth the rules were written as off, with a test in their place; the test stayed (`ProcessStartSitesTests`, in
the gate, and wider than the rule), the rules came back on.
- **`PrimaryDisplay.cs` was not changed to please the linter.** Its six results come from an internal assertion in
this linter's build (in the job's `backend.log`), not from the file; the first guess — that the engine misreads an
`unsafe` tuple — was wrong, and a rewrite made on it was reverted. Two inspections are excluded for that one file,
to be removed when a linter build reads it.
- **What the Ultimate licence offers and this job does not use**: quick-fixes (the job cannot push, on purpose), code
coverage in Qodana, FlexInspect. If the licence lapses the job is red and advisory; the way back is in `qodana.yaml`.
**Traps:**
- **A security rule's id is not an inspection.** The OpenGrep plugin's rules all report through
`OpenGrepInspection`; only that name works in `qodana.yaml`, and only by path. Read
`log/qodana_inspections_summary.csv` in the report artifact to see which inspection produced a result before
spending a run on a guess.
- **A report artifact is a zip inside the download, with backslash paths**: `unzip` extracts it and exits non-zero
on the warnings, so do not chain on its exit code.
- **A linter result can be the linter's fault**: read the job's log directory (the report artifact carries it) for an
`AssertionException` before changing the code it points at.
- **A Qodana run of the managed job takes fifteen to seventeen minutes**, most of it the scan; a pull request's own run
reads the whole solution (`pr-mode: false`), so there is no quick partial run to iterate on.
**Then:** the native report, in the pull request after it — the next section.
### Qodana — the native report: the C++ through a solution CMake writes (D99, D100)
*Status is `docs/13_CI_CD.md` §Qodana and `CHANGELOG.md`; this is what no other file carries.* The second pull request
of the third part: the job `native`, `qodana-native.yaml`, the preset `x64-vs-analysis`, fifty-four `.editorconfig`
lines for the C++ and eleven small fixes outside code loaded into a game.
**Choices made here:**
- **One linter for both reports.** Qodana for .NET is the only linter of JetBrains' that reads this C++ (MSVC-only;
`qodana-cpp` and `qodana-clang` are Linux images), and it wants a Visual Studio solution — so CMake writes one. The
generated solution is never built from, beyond the one header a build generates.
- **No `.clang-tidy` file.** The plan had expected one. Every clang-tidy check and clang diagnostic the engine runs
has an `.editorconfig` key of ReSharper's, so D100's "where the engine reads" is one file for all of the C++, and
the repository gains no clang-tidy configuration that a developer's editor would start obeying.
- **Most of what it reported is not a fault, and the settings say why, line by line**: clang's opinion of code only
MSVC builds; the engine misreading `atomic_ref`, zero-initialised API structures and constexpr tests; proposals of
a newer idiom. Eleven things were fixed. **In code loaded into a game nothing was changed**: no defect was found
there.
- **`LogNoteW` (`FrameLedger.Overlay/src/dllmain.cpp`) is dead code, and it is still there — the owner's call.**
Nothing calls it. D100 changes code loaded into a game for a defect only, and an unused function is not one, so
it was not removed on a linter's word; `.editorconfig` leaves that one file unread for an unused function, in a
section that says to remove it with the function. Removing it is eleven lines with no caller: a pull request of
its own, whose overhead statement is that the function was never reached.
- **"Possibly unused `#include`" is off, not obeyed.** Ten of the twelve reported outside the game are a file
including what it uses (`<cstring>` for `std::memset`) although `<windows.h>` already brought it; removing them
would have made the files depend on an accident.
**Traps:**
- **InspectCode cannot load the native solution on the owner's PC** (Visual Studio 18 Insiders' MSBuild crashes it,
2026.2.3.1 and 2026.3.0 alike), so C++ results are read from CI: dispatch `qodana.yml` on the branch, download
`qodana-report-native`, unzip the inner `qodana-report.zip`.
- **A file compiled into three targets is reported three times**: count distinct results (rule, file, line), not rows.
- **A shortened line is a clang-format change two lines away**: the gate's clang-format step failed on an alignment
that the edit above it had made possible. Run clang-format on every native file touched before the gate.
- **Do not switch branches while a gate runs** — a probe branch was made during a local native build, which changed
and restored two sources under it. Nothing came of it; the next gate was run on a still tree.
+104
View File
@@ -0,0 +1,104 @@
# HANDOFF history — 0.1.0-beta.38
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.39 was cut, 2026-11-11 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`.*
## **START HERE** — 2026-11-11, 0.1.0-beta.38: Qodana Ultimate in two reports — and the next work is the owner's "Thông số nâng cao"
*Status is `docs/changelog/0.1.0-beta.38.md`; this is what no other file carries.* This release closes the plan of
2026-11-11 — **beta.37** → **alpha.4** → **Qodana Ultimate in two reports**. The plan was one session's, and its START
HERE was moved whole: all three parts, the decisions D94–D100, the choices and the traps are
[`handoff/history/0.1.0-beta.37.md`](handoff/history/0.1.0-beta.37.md) — the two Qodana sections are its last two.
#384 moved the managed job to Qodana for .NET and settled what it reports; #385 added the C++ as a second report; this
release ships the one change to shipped code that came of it: the App's link opener refuses anything but a web or a
mail address.
### NEXT — the owner's request of 2026-11-11, for the session(s) after this one
Sent while this session was still running, and recorded here as it was written:
> "Xong phiên này, bạn cập nhật (những) phiên tiếp theo với các phiên bản về Setting "Thông số nâng cao" hoặc tương tự
> và hiển thị tại 1 navigation riêng, khai thác số liệu, dữ liệu, biểu đồ mức tối đa, toản bộ mà các header `.h` của
> Streamline và NVAPI hoặc pawnIO có thể sử dụng và hiển thị nhé."
That is: once this session is done, the following session or sessions — and their versions — are about a setting
called *"Thông số nâng cao"* (advanced metrics) or similar, shown in **a navigation item of its own**, that exploits
to the maximum the numbers, the data and the charts that the `.h` headers of **Streamline** and **NVAPI**, or
**PawnIO**, make available.
**Nothing of it was designed or begun here.** It is a new task, and it starts the way every train does — a plan, with
the owner's decisions asked while planning. What follows is this session's reading, not the owner's words:
- **An inventory comes first, from the headers in the tree**: `src/native/third_party/streamline`, `nvapi`,
`fidelityfx`; and what LibreHardwareMonitorLib reads through PawnIO (CPU and board sensors — optional, elevated
only). For each item: what it is, and **where it can be read from** — the Agent's own process (NVAPI through
`FrameLedger.NvapiBridge`, LibreHardwareMonitor) or only inside the game, through a hook (Streamline).
- **Anything read inside a game is a hook**, and the rules that govern hooks are not this feature's to bend: rules 4
and 5, `17_HOOK_ENGINE` §Hook inventory, the overhead stated, the open S-series of `20_OPEN_QUESTIONS`; the
anti-cheat guard and rule 2 untouched. Rule 6 governs any frame rate it shows beside frame generation, rule 7 any
Yes/No it shows, and rule 8 says it calls nothing on the network.
- **Questions that are the owner's**: on or off by default; what is stored in the ledger (a schema change, the size
of a session) and what is shown live only; what a PC without an NVIDIA card shows (N/A, never a guess); whether
the page asks for the elevation PawnIO needs; how many releases, and in what order.
- **What a new page owes**: strings in `en`, `vi`, `ja`; `ACCESSIBILITY.md` §For contributors for the page and each
chart; `LIMITATIONS.md` and the guide; `08_UI`.
#### Planned 2026-11-11, and approved by the owner — four trains, the first begun
The plan was made the same day, the owner's decisions asked while planning. **The design, the decisions D101–D108 and
the inventory are [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics**; what follows is only what that
section does not carry — the order, and what was asked.
- **The owner's four answers.** (1) Inside the game: **everything, in stages — and the refusal of 2026-08-09 is
re-examined**, not taken as final. (2) **Stored in full with the session.** (3) **Off by default, and off means
nothing more is read or stored**; the navigation item is hidden. (4) PawnIO: **the page says what is missing and
links to the admin mode's setting** (D34); it never asks for the rights itself.
- **Four trains, one release each, in this order**: **A** (0.1.0-beta.39) what the Agent can read — NVAPI,
LibreHardwareMonitor — with storage and the page, **and nothing changed in code loaded into a game**; **B**
(beta.40) inside the game, on calls already hooked and on functions Streamline exports, shared-memory layout 5;
**C** (beta.41) Reflex latency, after a spike; **D** (beta.42) the Streamline state behind `slGetFeatureFunction`,
after a spike. **Why this order**: A needs no hook, so it ships the page, the storage and the option first and
every later train only fills rows; B is hooks of a kind already built; C and D are routes nobody has proven safe,
and each ends in the owner's go or no.
- **The approval covers train A whole** — its eight pull requests through the release. **B, C and D each begin with a
plan of their own**, their decisions asked then. For D the owner will be asked one thing nothing else has needed:
handing the game a wrapper in place of the function it asked for is a write, and rule 4 as written forbids it.
- **Train A's pull requests**: (1) decisions, inventory, the catalog; (2) the NVAPI bridge's ABI 3 and the read-only
gate; (3) LibreHardwareMonitor's groups and the shared reader, not yet composed; (4) the `.partial` chunks and
schema 0023, produced by nobody yet; (5) the option goes live, with Privacy 2.15; (6) the pipe request and the
App's feed; (7) the page, the navigation item, Ctrl+6; (8) the release. A setting and its guide text, a navigation
item and its guide text, a schema script and its writer each land in ONE pull request — the guide tests, the
viewer-mode test and the registry test are red otherwise.
- **What the owner will be asked to run** (the release's owner-only item): the NVAPI rows on the RTX 5080 — CI has no
driver and never takes that branch; one run as administrator with PawnIO, which is also the first there has ever
been (`20_OPEN_QUESTIONS` §M10); the Agent's processor time and memory with the option on and off (NFR-2).
- **A trap already paid for**: LibreHardwareMonitor's storage group may touch every drive when it opens, and this
PC once had a failing USB drive. If the group cannot be opened without touching removable drives, storage stays
N/A in beta.39 — decided before the first run, not after it. **It cannot (read in the library's source,
2026-11-11), so it is not opened**: a drive's temperature reads N/A in beta.39, and the reason is on the page.
- **Measure before composing.** `FrameLedger.CaptureHost probe-advanced` ran the reader on this PC before any Agent
code used it, and found the sensor library's refresh at 14 ms of processor time — so the reader reads it every
fifth tick. The same number says the telemetry layer every session already uses costs more than NFR-2 allows
(`20_OPEN_QUESTIONS` §M16): **not this train's to fix, and the owner's to decide** — ask it when this train is out.
### Left open by this session
- **`LogNoteW`** (`src/native/FrameLedger.Overlay/src/dllmain.cpp`): a function nothing calls, found by the native
report and **not removed** — D100 changes code loaded into a game for a defect only, and dead code is not one. The
owner says whether it goes (owner-only item 36 (d)); `.editorconfig` has a section, naming it, that goes with it.
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, as beta.37's and alpha.4's were in #384.~~ **Recorded
2026-11-11, in the first pull request of the Advanced-metrics train: 0.1.0-beta.38 was released
2026-11-11T02:56:51Z** from the signed tag on `b7d6550`, by run 42947853293 — one run for the one tag push, where
alpha.4's push had started two. Eight assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed;
the shipped natives carry `FL_BUILD_ID v0.1.0-beta.38`. The delta from beta.37 is 699,201 bytes beside a full
package of 110,325,438 — "Delta processed 0452 files. 0022 patched, 0430 unchanged". The wiki was published by run
42949461191. The rehearsal on the release branch (run 42944533749) had made a delta of 727,036 bytes with the same
file counts.
- **The Qodana licence** is read from the job's log (*Qodana license plan: Ultimate*, and the day the subscription is
active until). When it lapses both jobs are red and advisory; `qodana.yaml`'s first comment is the way back.
- **C# 15's constructs** wait for a ReSharper engine that reads them (`20_OPEN_QUESTIONS` §G, the row added with
beta.37): the linter is now Qodana for .NET, whose engine is the same one — re-run the canary (`spike-notes` §29)
when the action's pin moves.
- **The alpha line rests** (D98). No sync and no alpha follow this release.
**Owner-only:** item 36 — beta.38 on real hardware, the links, the two Qodana Cloud projects, and `LogNoteW`.
+111
View File
@@ -0,0 +1,111 @@
# HANDOFF history — 0.1.0-beta.39
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.40 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`.*
## **START HERE** — 2026-11-11, 0.1.0-beta.39: Advanced metrics, train A — and train B begins with a plan of its own
*Status is `docs/changelog/0.1.0-beta.39.md`; this is what no other file carries.* The owner's request of 2026-11-11
— a setting, *Thông số nâng cao*, with a navigation item of its own — was planned the same day, approved, and its
**first train of four is this release**: what the Agent can read (NVAPI, LibreHardwareMonitor), stored with the
session, shown on a page, and **nothing changed in code loaded into a game**. Seven pull requests and this one:
#387 (decisions, inventory, the catalog), #388 (the NVAPI bridge's ABI 3, the read-only gate), #389 (the shared
reader, measured before composed), #390 (schema 0023, the `.partial` chunks), #391 (the switch, Privacy 2.15,
Disclaimer 2.15), #392 (the pipe request, the App's feed), #393 (the page, the navigation item, Ctrl+6). **The
design, the decisions D101–D108 and the inventory are [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced
metrics; the screen is `08_UI` §Advanced.** The request as the owner wrote it, the four answers and the order of the
four trains are [`handoff/history/0.1.0-beta.38.md`](handoff/history/0.1.0-beta.38.md) §NEXT.
### What this train decided that the plan had not
- **D109 — the Disclaimer goes to 2.15 with the Privacy Policy.** The plan approved the Privacy Policy's content only.
Writing #391 found two sentences of the Disclaimer (§5, §5A) that said PawnIO and the admin mode were for the CPU
temperature alone, which the option makes false; the owner was asked with both quoted and chose to amend them. The
admin mode's own disclosure followed (`agent-admin-dialog/2`).
- **Storage is not read at all** (D103's outcome, decided before the first run): LibreHardwareMonitor cannot list
drives without touching removable ones, so the group is never opened and a drive's temperature is N/A with that
reason. Power supplies and fan controllers are refused for the same kind of reason (a search of every USB device).
- **The sensors are read every fifth second**, NVIDIA's figures every second — set by a measurement
(`CaptureHost probe-advanced`: 14 ms of processor time per refresh of the sensor library), not by taste.
- **The page's hold on the reader is the request itself**: nothing tells the Agent that a page opened or closed
(`07_IPC` `GetAdvancedMachine`). A design with an "open" and a "close" message was not written, because a pipe
that drops would leave the reader running.
### NEXT
1. ~~**Ask the owner `20_OPEN_QUESTIONS` §M16 before anything else.**~~ **Asked 2026-11-11 — D110: measure it in a
real Agent first.** `tools/agent-cost.ps1` does, and §M16 holds the numbers: 1.93 % of one core with Advanced
metrics off, 2.04 % on. **D117: NFR-2 is amended to ≤ 2.5 %**; nothing in the code changes.
2. ~~**Train B (0.1.0-beta.40) starts with a plan of its own**~~ **Train B was planned and approved the same day**
(D111–D116): [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B is the design. What that section does not
carry:
- **The approval covers 0.1.0-beta.40 whole** — five pull requests: (1) the decisions, the struct-version table,
§M16 measured; (2) shared-memory layout 5, written by nobody yet; (3) the Overlay reads the extra fields of the
calls it already hooks, behind the control word, with `overlay-fields-check` and the overhead measured; (4) the
Agent's sampler, the catalog's new reach, the page's words, Privacy 2.16; (5) the release, owner-only item 38.
- **beta.41 and beta.42 are designed and not begun.** Each starts only when the owner says the release before it
passed its owner-only item on real games — that wait is what the split is for.
- **The design was reviewed against the code before approval, and the review changed it**: the split itself; the
struct-version table (the tree holds one version of each header); storing a vendor `Boolean` or preset as its
value plus one and refusing the vendors' sentinels (`INVALID_FLOAT` is FLT_MAX); flags as "ever seen" masks and
the frame time as a sum the watchdog drains, since a value sampled once a second aliases a per-frame one; the
new region written by the watchdog alone and only while requested; `AdvancedSources`' new member optional, so a
beta.39 row still parses; `kMaxPatches = 48` to be checked before a patch is added.
3. **Trains C and D** (Reflex latency; the Streamline state behind `slGetFeatureFunction`) each begin with a spike
and end in the owner's go or no. D needs an amendment of rule 4 if it goes ahead — handing the game a wrapper is
a write — and that question is put to the owner then, not assumed.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, as beta.38's were in #387.~~ **Recorded 2026-11-11, in the
first pull request of train B: 0.1.0-beta.39 was released 2026-11-11T13:52:07Z** from the signed tag on `e63cd49`
(#394), by run 43031168146 — green at the first attempt, one run for the one tag push. Eight assets; the seven
digests of `SHA256SUMS.txt` equal the ones GitHub computed; the shipped natives carry `FL_BUILD_ID v0.1.0-beta.39`.
The delta from beta.38 is 3,144,205 bytes beside a full package of 110,678,706 — "Delta processed 0452 files. 0026
patched, 0426 unchanged". The wiki was published by run 43033077805. The rehearsal on the release branch (run
43026873256) had made a delta of 3,159,090 bytes with the same file counts.
- **Nothing here was seen in the running App.** The page was rendered to images from a test over fixture data and
looked at; a chart does not render into such an image. No real Agent answered a real App, no session was recorded
with the option on, and no value was compared with another tool. All of that is owner-only item 37.
- **The first run as administrator with PawnIO is still to come** (`20_OPEN_QUESTIONS` §M10, §M13): the processor,
memory and motherboard groups have been opened on no PC. The reader leaves a device alone after two faults, and
runs on a thread of its own, so a sensor that hangs holds up neither a session's samples nor the pipe — that is
the design, and it has not been tried.
- **The `vi` strings were written by the assistant and the `ja` ones are marked `review`** — 245 keys each, every
catalog entry among them.
- **`LogNoteW`** is still there (owner-only item 36 (d)), and the alpha line still rests (D98): no alpha follows this
release.
### Traps this train paid for
- **A pull request whose workflows never start.** #390 was opened and GitHub started none of its three workflows —
`gh pr checks` said "no checks reported" ten minutes later, with nothing queued. Closing and reopening the pull
request starts them (`pull_request: reopened`); **closing drops auto-merge, so enable it again.** After every
`gh pr create`, ask the API for the runs of the head commit before walking away.
- **`ShmDrainIntegrationTests.TheGuardInjectsTheOverlayAndTheReaderDrainsRealFrames` was red once** in a local gate
(#392's first run): the guard's re-scan refused the harness mid-run, under the gate's load. It passed alone and in
the rerun; nothing in that pull request is on its path; it was not diagnosed. ~~If it shows again, read the verdict's
reason first — the assertion prints only that it was a refusal.~~ **It showed again the same day** (train B's first
local gate), on another line: the Overlay's status was still `Init` with fifty records already drained. That one
is a race in the test and is fixed — the Overlay publishes `Ready` after its last hook, later than its first
record, so the test now waits `Init` out, bounded. The refusal was not reproduced; the assertion now prints the
verdict, so the next one says why.
- **`PagesLoadTests.EveryTextFollowsAThemeSwitch` was red once on a hosted runner** (#393's first run, green three
times locally): *Settings: "Processor" shows in a fresh one only*. Every build of the Settings page in that fixture
made a `SystemInfoViewModel` of its own, whose lines arrive from the thread pool, so two pages built a moment apart
could differ by a row. The fixture now loads one and awaits it (`Surfaces.SystemInfo`) — fixed in the release's
pull request, test code only. A surface whose view model loads by itself is loaded once, awaited, then shared.
- **SonarCloud's coverage of new code** (advisory, 80 % asked) was 75.3 % on #391: the uncovered part was Win32
(`WindowGeometryProbe.MonitorDeviceOf`). A lookup that takes a named window, as `Observe` does, can be tested
against the harness's window; one that only searches for a visible window cannot, on a hosted runner.
- **The App's own accessibility tests will refuse a page** that sets `AutomationProperties.LiveSetting` in XAML, that
names a page-local resource key, whose item types have no `ToString`, or whose grid has an unnamed filler column —
and a navigation item or a guide page that the guide's contents do not name. All of it is one test run away:
run `FrameLedger.App.Tests` whole before the gate.
- **An image of a page from a test** (`RenderTargetBitmap` over the hosted frame) shows the layout and the text, and
costs a temporary test; ScottPlot draws nothing into it.
- **`sed -i` on a file the Write tool made** is harmless only because both are LF until the normalising step; on a
CRLF file it is the trap `handoff/traps.md` already names.
**Owner-only:** item 37 — beta.39 on real hardware: the update and its two terms, the switch and the page, a session
with a real game, the first run as administrator with PawnIO, and what the Agent costs with the option on and off.
+148
View File
@@ -0,0 +1,148 @@
# HANDOFF history — 0.1.0-beta.40
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.41 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its release numbers are those of the day it was written**: what it calls beta.41 and beta.42 are train B's second and third releases, now beta.42 and beta.43 (D118).*
## **START HERE** — 2026-11-12, 0.1.0-beta.40: Advanced metrics inside the game, the first of train B's three releases
*Status is `docs/changelog/0.1.0-beta.40.md`; this is what no other file carries.* Train B of the owner's
"Thông số nâng cao" was planned and approved on 2026-11-11 (D110–D117) as **three releases**, and this is the first:
what a hooked game tells its upscaler, read from **calls the Overlay already hooked — no new patch site** — only
while the session records Advanced metrics. Four pull requests and this one: #395 (the decisions, the struct-version
table, §M16 measured, NFR-2 amended), #396 (shared-memory layout 5, region 5), #397 (the Overlay's reads, the
`overlay-fields-check` gate, the overhead measured), #398 (the Agent's sampler, the catalog's reach `Game`, the
page's words, Privacy Policy 2.16). **The design is [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B; the
contract of the new region is [`07_IPC.md`](07_IPC.md) §Region 5; what each hooked call reads is
[`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory.** The plan as approved, the review that changed it and the
order of the trains are [`handoff/history/0.1.0-beta.39.md`](handoff/history/0.1.0-beta.39.md) §NEXT.
### What this train decided that the plan had not
None of these is an owner decision; each was found while building, and is written where it binds.
- **"On" is a request of the session's own, not a 1** (#397, `07_IPC` §Region 5). The plan had the control word as
0 or 1. A game outlives a session, and the Overlay's summaries with it — a second session of a game that kept
running would have read the first one's field-of-view range and counts. The Agent now writes a value no session of
that ring used, the watchdog starts its summaries afresh at a value it has not seen and names it in every
publication, and the reader keeps only a publication that names its own. Found while writing the injected test,
before anything shipped; layout 5 was changed in place (`reserved0` became `request`), which was possible only
because no release had carried it.
- **One module's calls are counted**, for the memory query as for the dispatch: a loader that passes a call on to a
leaf's export is seen at two hooks.
- **Five of the thirteen entries are tables**, not text (`sl.dlss.presets`, `sl.dlss.exposure`,
`ffx.upscale.sharpening`, `.jitter`, `.depth`): a row's name is the vendor's own member name, so nothing but a
switch's on/off and a flag's name needs translating. The catalog's unit and two summaries changed; no id did.
- **What became of the in-game entries is stored with the session** (`AdvancedSources.InGame`): read, nothing seen,
not hooked. Without it a session whose game set its DLSS options through `slDLSSSetOptions` would look like a
game with no upscaler.
- **The Disclaimer stays 2.15.** The accuracy statement gained an item, as in beta.19, beta.20 and beta.25, none of
which moved it; the Privacy Policy went to 2.16 with the content the plan approved.
- **A crash dump is tried three times** (#398). Not this train's subject: `CrashDumpHelperTests` failed under the
full gate for the third time in two days, and the cause — `MiniDumpWriteDump` of a running process that changes
under it — costs a real crash its only dump in exactly the same way.
### NEXT
1. **The owner runs item 38** — this release on real games. **Nothing of train B is begun before that**: the wait
is what the split into three releases is for (D113).
2. **0.1.0-beta.41, when the owner says item 38 passed**: two new hooks, `slSetConstants` and `ffxConfigure`. The
design is in `18` §Train B and the plan in [`handoff/history/0.1.0-beta.39.md`](handoff/history/0.1.0-beta.39.md)
§NEXT; region 5 already has their members (`sl*`, `ffxFg*`), so no layout changes. What to settle first, in a
plan of its own: `kMaxPatches = 48` against the patch count with the new rows; the window of `sl::Constants` that
is read, pinned by `static_assert` so that the camera members stay outside it; the overhead of a call made every
frame for every viewport, measured before the merge; `overlay-fields-check`'s list and its parameters for the two
new structs.
3. **0.1.0-beta.42, when the owner says the one before passed**: `slGetFeatureVersion` and
`slGetFeatureRequirements` — calls a game makes while starting, so most sessions will say the game did not make
them after FrameLedger attached. The release notes and the page say so.
4. **Trains C and D** (Reflex latency → beta.43; the Streamline state behind `slGetFeatureFunction` → beta.44) each
begin with a spike and end in the owner's go or no. D needs an amendment of rule 4 if it goes ahead.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, as beta.39's were in #395.~~ **Recorded 2026-11-12, in the
pull request after the release: 0.1.0-beta.40 was released 2026-11-11T19:30:13Z** (02:30 on the 12th in the
owner's time zone, which is the date the changelog carries; the legal documents' *Effective* date is the UTC one,
2026-11-11 — the fifth release to straddle midnight that way, after beta.9, .10, .12 and .14) from the signed tag
on `5268332` (#399), by run 43085044822 — green at the first attempt, one run for the one tag push. Eight assets;
the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the shipped natives carry `FL_BUILD_ID
v0.1.0-beta.40`. The delta from beta.39 is 2,413,166 bytes beside a full package of 110,707,319 — "Delta processed
0452 files. 0025 patched, 0427 unchanged". The wiki was published by run 43087844607. The rehearsal on the release
branch (run 43082208739) had made a delta of 2,426,184 bytes with the same file counts.
- **Nothing here was read from a real game, and nothing was seen in the running App.** The values come from
`hook-harness --advanced` and the FidelityFX stub; the page's new rows were loaded by its tests and looked at by
nobody. All of it is owner-only item 38.
- **Whether any shipping title chains `sl::DLSSOptions` into `slEvaluateFeature`** is not known. If none of the
owner's games does, the six `sl.dlss.*` entries read N/A everywhere until train D, and the page already says why.
- **The aliasable part of FSR's memory** is read and published and not shown: the catalog has one memory entry.
- **What the owner's first evening with it showed** (2026-11-12, read from a copy of the ledger and the Agent's log;
ten hooked sessions on eight games, no fault in any, the Overlay `Ready` at every end):
- **Read from a real game for the first time: Lies of P** (FSR 3 through `amd_fidelityfx_dx12.dll` 1.0.1, the
SDK 1.1.x monolith) — sharpening on at 0.5, jitter up to 0.48 / 0.49 of a pixel, pre-exposure 0.25–1, a vertical
field of view from 21.6° to 73.7°, no flag, and a reported frame time whose median is 16.67 ms beside a native
rate of 55.6 FPS. The series starts 83 seconds in and has no gap after. Whether these match the game's own
settings is still the owner's to say (item 38 (c)).
- **`cameraNear` is absent there and `cameraFar` is 10.** Most likely the game runs an inverted, infinite depth
and passes `FLT_MAX` for the plane at infinity, which the Overlay refuses as a value nobody set. "Infinite" would
be the true answer. It is a change in code loaded into a game: decide it in beta.41's plan, not before.
- **Every other game said *the game made none of the calls***, as expected: all ran DLSS. Dying Light: The Beast
evaluates DLSS through Streamline and chains no `DLSSOptions` — the first evidence on that open question, and
it is a no. Valor Mortis, Phantom Blade Zero, Clair Obscur and Kingdom Come: Deliverance II ship AMD's FSR 3.1+
libraries and did not upscale through them in these sessions, so **no FSR 3.1+ title has been read yet**.
- **The option off stored nothing** (Black Myth: Wukong, one session each way), and **a game under the user-mode
anti-cheat exception ran with it on** (Girls' Frontline 2, trial 1 of 2).
- **Valor Mortis ended with an access violation 51 seconds after injection, once**; the next launch ran fourteen
minutes on the same build. The Overlay counted no fault and Windows logged no application error. Not attributed
and not ruled out; its crash count stands at 1, and a second such end switches its hooking off.
- **The sensor library was closed under its other users** — a defect of train A that only an elevated Agent with
PawnIO can show. `20_OPEN_QUESTIONS` §M13 has what was seen; `LhmLibrary` is the fix; the owner's next sessions
say whether it holds.
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware".** They have now been read on one PC and compared with nothing.
The sentence is the owner's to reword.
- ~~**SonarCloud on #398** had not answered when this was written; #397's three findings are fixed in #398.~~
**It answered after the merge, with three findings on new code**, handled in the pull request after the release:
two `csharpsquid:S1244` in `InGameAdvancedSampler` — a lowest and a highest float compared with `==` to tell one
value from a range; they are now told apart by their bits, which is what was meant — and one
`csharpsquid:S3869`, a blocker, on `ParentDump`'s `DangerousGetHandle`. That one is **not new and not fixed**: the
owner accepted it in beta.24 (`13_CI_CD` §SonarCloud), #398 moved the same expression into a new method, and
SonarCloud tracks an issue by its place. It needs the owner's *Accept* again — item 38 (j).
- **The `vi` strings were written by the assistant and the `ja` ones are marked `review`** — fourteen more keys and
the reworded Settings text.
- **`LogNoteW`** is still there (owner-only item 36 (d)), item 37 is still unrun, and the alpha line still rests
(D98): no alpha follows this release.
### Traps this train paid for
- **A library that is "one object per user" may still be one per process underneath.** Three `Computer`s of
LibreHardwareMonitor shared a driver handle and mutexes; each was tested alone, unelevated, where the shared part
is never opened. Before composing a second instance of anything that talks to a driver, ask what its `Close`
closes. And `first ??= TryClose(x)` in a loop stops closing at the first throw — the test for "all are closed
even when one throws" caught it in the fix itself.
- **A heredoc in the tool's shell breaks on an odd number of apostrophes** — a script with `game's` in it fails with
"unexpected EOF while looking for matching `'`" before Python ever runs. Write the script to a file and run the
file.
- **A region that outlives its reader needs to say whose data it holds.** Region 4 never had the problem — it is
the swap chain's state, the same for any session. Region 5 holds summaries, and a summary has a beginning. Ask of
any new shared state: what does the second session of the same process read in its first second?
- **SonarCloud's engine does not see a Catch2 `REQUIRE` end a path**: `REQUIRE(OpenRingFor(pid, r))` followed by
`r.st->…` is a null dereference to it (`cpp:S2259`, a failed gate on new code). An explicit `return` after the
`REQUIRE` is what it follows. And a `LARGE_INTEGER` read through `QuadPart` is initialised through `QuadPart`
(`cpp:S6232`) — `13_CI_CD` §SonarCloud has had the rule since #336; new code still forgot it.
- **SonarCloud fails after the merge.** Its job is advisory and slower than the required checks, so auto-merge lands
the pull request first. Read its result on the merged pull request before building on it.
- **`MA0051` counts sixty lines, comments included**, and `CaptureSession.SuperviseAsync` was at fifty-nine: one
line more is a build error. What a session asks once belongs where the session's state is made, not in the loop.
- **`RCS1242` refuses `in` for a struct that is not `readonly`** — the mirror structs are not. Pass the object that
holds the struct and take `ref readonly` inside.
- **`/W4 /WX` refuses `alignas` on a struct it then has to pad** (C4324): align the variable, not the type.
- **The `.resx` files hold a literal apostrophe**, not `&apos;`: an edit script that searches for the entity finds
nothing and stops half-way through.
- **`ISessionRepository.FindAdvancedSensorsAsync` returns `t_ms` too**: "no advanced series" is "no series under
`adv.`", not an empty list.
- **PowerShell reads a backtick in a double-quoted string as an escape** (`` `r `` is a carriage return) and
`"$name: text"` as a scoped variable: in a gate's messages, quote names with apostrophes and write `${name}:`.
**Owner-only:** item 38 — beta.40 on real games: the update and the Privacy Policy 2.16, a game that was running
across the update, an FSR 3.1+ title and a DLSS title with the option on, the same with it off, a second session of
a game that kept running, a game under the user-mode anti-cheat exception, and what the Agent costs.
+131
View File
@@ -0,0 +1,131 @@
# HANDOFF history — 0.1.0-beta.41
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.42 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its release numbers are those of the hour it was written**: its item 3 calls train B's third step beta.43 and its item 4 calls trains C and D beta.44 and beta.45 — since D122, later the same day, they are beta.44, beta.45 and beta.46, and beta.43 is the two new hooks.*
## **START HERE** — 2026-11-12, 0.1.0-beta.41: one fix — the sensor library is no longer closed under its other users
*Status is `docs/changelog/0.1.0-beta.41.md`; this is what no other file carries.* 0.1.0-beta.40 was released in the
small hours of 2026-11-12 and the owner ran it that same night on eight games, as administrator with PawnIO — the
first elevated run there has been. A copy of the ledger and the Agent's log showed the in-game readings working
(Lies of P, the first real game read) and **a defect of train A that only that elevation can show**: three
LibreHardwareMonitor `Computer`s in one Agent, and the library's per-process state closed by one under the others.
This release is that fix and nothing else: #400 (0.1.0-beta.40's release facts; two comparisons SonarCloud flagged),
#401 (`LhmLibrary`), and this one. **Everything about train B — what was built, what the owner's sessions showed,
the traps — is [`handoff/history/0.1.0-beta.40.md`](handoff/history/0.1.0-beta.40.md)**; the fix is
[`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Advanced metrics and [`20_OPEN_QUESTIONS.md`](20_OPEN_QUESTIONS.md)
§M13.
### What this release decided
- **D118 — the fix is released at once as 0.1.0-beta.41, and train B's numbers move by one.** D113 had named
beta.41 for the train's second release. Told so, the owner chose to release the fix now. The train's second
release is **beta.42**, its third **beta.43**, train C **beta.44**, train D **beta.45**. Released changelogs and
moved HANDOFF sections keep the numbers of their day; so do the comments in `fl_shm.h` and `ShmLayout.cs` (the
members "0.1.0-beta.41" and "beta.42" fill), **deliberately**: touching them would have changed source that is
loaded into a game in a release that says the game-side code is beta.40's line for line. The release that fills
a member rewrites its comment.
- **The reads of the sensor library are not put under one lock**, only its opens and closes. A lock around every
read would have put a session's once-a-second poll behind the Advanced reader's, undoing the separate threads
that keep a hanging sensor from holding up the other reader.
### NEXT
1. ~~**The owner runs item 39** (this fix, on the PC that showed the fault) **and the rest of item 38** (an FSR 3.1+
title; two sessions of one running game; the cost; the values of Lies of P against the game's own settings).~~
**Both passed on 2026-11-12, on the owner's word** ("Bạn có thể đánh dấu là pass và tiếp tục") after the thirteen
sessions below; [`handoff/owner-only-passed.md`](handoff/owner-only-passed.md) says what those sessions covered and
what they did not.
2. **Planned and approved 2026-11-12 (D119–D122): the second step is two releases.** **0.1.0-beta.42 has no new
patch site** — five pull requests: (1) the session's upscaler label (D120); (2) the Overlay's patch registry
refuses a patch it cannot register, and registers before it enables; (3) the native log is flushed whenever
something new was logged, every inventory row's install line names its module and symbol, a failed patch says
why, and the first sight of each ffx descriptor type on each module is logged — so that one Cyberpunk session on
beta.42 says where its upscale goes (D121); (4) FSR's infinite depth plane (D119); (5) the release, owner-only
item 40. **0.1.0-beta.43 is the two new hooks** and begins only when that log was read; its design — reviewed
against the code, which changed it — is in the plan of that day and below. What the review found that the first
draft had wrong: there is no "dominant segment" (segments split on output size only), so the label is the modal
name over the dominant stream; the patch table drops a patch silently when full; the init thread's flush runs
after the watchdog starts; a second NaN pattern for "infinite" would have been dropped by the Agent — real
+infinity is stored instead; the new rows must be gated before `ResolveScoped`, which takes a module reference a
call; `slSetConstants` needs the struct's GUID checked on every call; `ffxConfigure` needs an owner slot of its
own; summaries across viewports must come from the first viewport seen.
~~**0.1.0-beta.42 — train B's second release — when the owner says item 38 passed begins now, with its plan**~~: `slSetConstants` and
`ffxConfigure`. A plan of its own first. What it must settle, beside what
[`handoff/history/0.1.0-beta.40.md`](handoff/history/0.1.0-beta.40.md) §NEXT lists (`kMaxPatches = 48`, the window
of `sl::Constants` pinned outside the camera members, the per-call overhead, the gate's list for the two new
structs):
- ~~**`FLT_MAX` in an FSR depth plane.** Lies of P gave `cameraFar = 10` and no `cameraNear`; most likely the
plane at infinity is `FLT_MAX`, which the Overlay refuses as "nobody set this". Decide whether it is stored as
"infinite" — an Overlay change, with its overhead stated.~~ *(Decided, D119, and built in beta.42:
`FL_ADV_FLOAT_INFINITE`, shown "∞". Whether Lies of P's missing plane IS `FLT_MAX` is still the hypothesis it
was — item 40 reads it.)*
- ~~**Valor Mortis.** One exit by access violation 51 seconds after injection under beta.40, none under an earlier
build, not attributed. Its crash count is 1; a second such end switches its hooking off. Before adding a hook
to that game's process, ask the owner whether it happened again.~~ **Answered by the owner, 2026-11-12: it was
the game's own — a shader-compilation error on entering the game, shown in the game's own crash dialog.** Not
FrameLedger's. It ran normally under beta.41 afterwards. Its crash count still stands at 1, because the crash
policy counts an abnormal end within 60 seconds of injection whoever caused it: a second one, from any cause,
switches its hooking off, and the owner then switches it on again from the game's page.
- **The release numbers in the comments** of `fl_shm.h`, `ShmLayout.cs`, `hook-harness` and `guard_test.cpp` for
the members that release fills (D118).
3. **0.1.0-beta.43 — the third — when the owner says the one before passed**: `slGetFeatureVersion` and
`slGetFeatureRequirements`.
4. **Trains C and D** (beta.44, beta.45) each begin with a spike and end in the owner's go or no.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request.~~ **Recorded 2026-11-12: 0.1.0-beta.41 was released
2026-11-11T22:38:39Z** (05:38 on the 12th in the owner's time zone) from the signed tag on `a6b95e8` (#402), by
run 43111028661 — green at the first attempt, one run for the one tag push. Eight assets; the seven digests of
`SHA256SUMS.txt` equal the ones GitHub computed; the shipped natives carry `FL_BUILD_ID v0.1.0-beta.41`. The delta
from beta.40 is 807,692 bytes beside a full package of 110,708,416 — "Delta processed 0452 files. 0022 patched,
0430 unchanged". The wiki was published by run 43113525941. The rehearsal on the release branch (run 43108189693)
had made a delta of 841,038 bytes with the same file counts.
- ~~**The fix has not been seen to work.** It was written from the log's timing; the fault needs an elevated Agent with
PawnIO. Item 39 is the check, and `20_OPEN_QUESTIONS` §M13 stays open until then.~~ **The owner's first seven
sessions on beta.41 show no fault** (2026-11-12, 06:38–07:01, read from the Agent's log and a copy of the ledger):
not one `advanced:` line, and every session with the option on stored its 212 or 213 series, the processor's 119
among them. On beta.40 the same morning the processor device faulted out after **each** of the six sessions
recorded between 04:34 and 05:07. The rarer fault — the reader failing to open — was one session in sixteen, so
seven clean ones do not yet say it is gone. ~~Whether item 39 has passed is the owner's to say.~~ The owner
marked it passed the same day.
- **What those sessions say of train B's first release** (the same log and ledger; the owner named which games are
first-person — Kingdom Come: Deliverance II, Cyberpunk 2077, Dying Light: The Beast, Valor Mortis — the rest
third-person, and Psycho-Sleuth 2D):
- **The first FSR 3.1 title read: Kingdom Come: Deliverance II** (FSR 3.1 chosen in the game;
`amd_fidelityfx_upscaler_dx12.dll` 4.0.2 behind the loader). Performance mode, 1280×720 to 2560×1440;
sharpening on at 0.5; `cameraNear` 0.05 and `cameraFar` 8000 — **both planes, the only title that gives both**;
one vertical field of view for the whole session, 63.1°; a reported frame time of 16.67 ms. A 41-second session
of the same game that never left its menu read nothing and says so.
- **Dying Light: The Beast** upscaling with FSR: sharpening off, 75.4°–78° vertical, 7.5 ms. **Clair Obscur**:
sharpening off, 5.7°–90.7° over four minutes. So far the two first-person titles hold a narrow range and the
two third-person ones a wide one (Lies of P 21.6°–73.7°) — what one would expect of cutscene and aiming
cameras, and no more than that: four titles.
- **Three engines give `cameraFar` alone, as a small number** — 10 (Lies of P), 1 (Clair Obscur), 0.1 (Dying
Light) — which is the near plane of an inverted, infinite depth with the other plane at `FLT_MAX`. That is the
case beta.42's plan has to decide.
- **Clair Obscur reports a frame time of 0.001 ms in some seconds.** It is what the game hands FSR, stored as it
was said; the lowest value of that series is not a frame time anyone saw.
- **Cyberpunk 2077 read nothing**, with FSR frame generation running and the render size known: its upscaling did
not come through an ffx-api upscale dispatch in those sessions. Which upscaler it was set to is the owner's to
say; `ffxConfigure` (beta.42) is where its frame generation would be read.
- **One session of Dying Light names DLSS as its upscaler and carries FSR's quality byte and render size** (81
seconds, both an FSR dispatch and a DLSS evaluation seen, "settings changed" not set). Not this train's code —
the session summary's choice between two upscalers seen in one session — and not looked into. **The owner,
2026-11-12: Dying Light and Expedition 33 were set, in the game, to FSR upscaling with DLSS frame generation.**
So "FSR beside DLSS frame generation" in their other sessions is the setting, not a misreading; in this one the
summary picked the wrong one of the two for its upscaler line, which is still to be looked at.
- **A session without consent says so**: two Cyberpunk sessions before its consent dialog was accepted are
Tier 2, and their Advanced row says *not measured inside the game*. **The option off stored nothing** again,
this time on a title that upscales with FSR 3.1 (Dying Light, beta.41).
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware".** They have been read on one PC and compared with nothing. The
owner's to reword.
- **SonarCloud's `csharpsquid:S3869` on `ParentDump.cs`** is open on `main` until the owner accepts it again (it was item
38 (j), the one part of that item not done; the owner said on 2026-11-12 that they will). Until then `main`'s `sonarcloud` job is red on that one line.
- **Items 36 (d) (`LogNoteW`) and 37 are still open**; the ledger shows part of 37 (e) happened — the elevated
groups opened and read — and nothing of it was compared with another tool. The alpha line still rests (D98).
**Owner-only:** ~~item 39~~ passed 2026-11-12, with item 38. Nothing new is open from this release; items 36 (d)
and 37 still are.
+204
View File
@@ -0,0 +1,204 @@
# HANDOFF history — 0.1.0-beta.42
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.43 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its §NEXT item 2 says 0.1.0-beta.43 is "under way"**: it was released the same day, and what it built and decided is HANDOFF.md's section for it.*
## **START HERE** — 2026-11-12, 0.1.0-beta.42: no new patch site — the label, the registry, a log that outlives the game, the plane at infinity
*Status is `docs/changelog/0.1.0-beta.42.md`; this is what no other file carries.* The owner marked items 38 and 39
passed on 2026-11-12 and train B's second step was planned the same morning. A review of the draft against the code
changed it, and the owner split it (D122): **this release adds no patch site**, and the two new hooks are
0.1.0-beta.43. Four pull requests and this one: #406 (the session's upscaler label, D120; the decisions D119–D122),
#407 (the patch registry), #408 (the native log and the descriptor-type census, D121), #409 (FSR's plane at
infinity, D119; Qodana's native report). **What the owner's sessions on beta.40 and beta.41 showed, and the plan as
approved, are [`handoff/history/0.1.0-beta.41.md`](handoff/history/0.1.0-beta.41.md)**; what each pull request built
is [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Unhooking and §Native logging, [`03_METRICS.md`](03_METRICS.md)
§Upscaling and [`07_IPC.md`](07_IPC.md) §Region 5.
### What this release decided, beyond D119–D122
- **The descriptor-type census is not behind the Advanced-metrics switch.** It is a diagnostic that reaches the log
file and nothing else — no record, no shared memory, no session — like `HOOK_INSTALLED`. Behind the switch it
could not answer D121 for a session recorded with the switch off, which is most of them. Its cost was measured
and stated instead (+3.8 ns a dispatch).
- **The log's text grew to 72 characters; nothing else about the ring did.** 256 events, static, 24 KiB.
- **A slot of the patch registry that was ever live is never reused**, and a failed create or enable leaves the
hook created in MinHook as it always did. Both are the smaller change: the first keeps the stop's walk lock-free
without a second word per slot, the second changes no failure path that has never been seen to run.
- **The live card still names the first upscaler a session saw** (`FgLadder.UpscalerIdentity`). The stored
session's name is the modal one (D120); for a session in progress the question is a different one.
- **Qodana's "never used" on the registry test's stand-in engines is switched off for that file**, not answered by
turning the template into function pointers: the header is loaded into a game, and D100 changes such code only
for a defect.
### What the owner's first four sessions on beta.42 showed (2026-11-12, 09:40–09:53)
Dying Light: The Beast, Kingdom Come: Deliverance II, Cyberpunk 2077 and Lies of P, each quit from its own menu;
read from the four `overlay-<pid>-….log` files, a copy of the ledger, and — for Cyberpunk — the game's own settings
file and executable on disk. All four hooked, no fault, no `PATCH_FAILED`, the Agent's log without an `advanced:`
line.
- **D121 is answered, and it is not a detection gap: in that session Cyberpunk 2077 upscaled with FSR 2.1, which is
compiled into the game.** Its `UserSettings.json`, written when the session ended, has `ResolutionScaling = FSR2`,
`FSR2 = Quality`, `FrameGeneration = FSR3`, `FSR3_FrameGeneration = true` — FSR **2** for the upscale, FSR **3**
for frame generation only. Quality is a ratio of 1.5, and the session's render size is 1707×960 at 2560×1440.
`Cyberpunk2077.exe` carries FSR 2's own function names (`ffxFsr2GetJitterPhaseCount`,
`ffxFsr2GetUpscaleRatioFromQualityMode`) and names one FidelityFX module, the loader; its folder has no FSR 2
DLL. **The log says the same from the other side**: the loader (2.3.0), the upscaler leaf (4.1.1) and the
frame-generation leaf (4.0.1) were all patched; the loader and the frame-generation leaf were handed `0x20004`
(PREPARE) and `0x20003` (FRAMEGENERATION); the loader and the upscaler leaf were asked `0x10002` (the ratio of a
quality mode) once at start; **no module was ever handed `0x10001`**, and there is no `UPSCALE_ROUTE` line. The
NVIDIA driver's words for the session say DLSS was not running. So the upscale goes through no export at all —
the case `LIMITATIONS.md` already names ("FSR 2 and upscalers built into the game are not identified"), and
"unknown" beside a measured render size is the true reading. ~~**The owner had said FSR 3 or 4; that the setting is
FSR 2.1 is read from the file, and is the owner's to confirm.**~~ **Confirmed by the owner the same morning with
a screenshot of the game's Graphics menu: Resolution Scaling reads "AMD FidelityFX Super Resolution 2.1"** (the
file's `FSR2`), with its own quality and sharpening rows under it. Nothing to fix for it in beta.43.
- **The loader of SDK 2.3 passes a dispatch on to the leaf's export — the older loader does not — and that is a
double count waiting for the first title that upscales through it.** In Cyberpunk each frame-generation dispatch
type is logged on the loader AND on the frame-generation leaf: the game calls the loader, the loader calls the
leaf's exported `ffxDispatch`. In Kingdom Come and Dying Light (loader 1.0.2) the UPSCALE is logged on the loader
alone, which is what `fl_hook_inventory.h` measured in September and what the fixture's loader stub copies.
Frame generation is unharmed — PREPARE is counted by its frame number, FRAMEGENERATION is a bit, and the
session's factor is ×2.0001. **An UPSCALE is counted once per hooked module that sees it**
(`g_ffxSeen`'s count): through a 2.3 loader that would be two a frame, and with frame generation off — no
PREPARE, so the count falls back to UPSCALE dispatches — the application-frame count would double. Not
observed: no UPSCALE was dispatched. **One session answers it: Cyberpunk with Resolution Scaling set to FSR 3
and frame generation off** — the log shows whether `0x10001` is on both modules, and the session's frame
counts show whether it was counted twice. If so, the fix is beta.43's: count at the first module seen, as
`fl::adv::Owns` already does for the Advanced reads.
- **The plane at infinity is measured, not a hypothesis any more (D119).** Lies of P: `cameraNear = ∞`,
`cameraFar = 10`. Dying Light: `cameraNear = ∞`, `cameraFar = 0.1`. Both gave one plane on beta.41. Kingdom
Come still gives both as numbers (0.05 and 8000). Clair Obscur was not run.
- **The logs are whole.** Each has its installs by module and symbol, the census as it grew, and lines long after
the first second — Dying Light's first UPSCALE at 71.6 s, Kingdom Come's at 37.5 s, when play began. One line
looks odd and was not looked into: Cyberpunk's frame-generation leaf's `ffxDispatch` is at `0x241cde7ba10`, far
from where the other modules sit (`0x7ff…`); the patch went in and was called.
- **Dying Light was named FSR, Quality, with DLSS frame generation ×3.99** — its settings as the owner gave them.
No session changed upscaler part-way, so D120's case was not exercised.
### And three more starts of Cyberpunk 2077, 10:05–10:15 the same morning
Asked for as "Resolution Scaling = FSR 3, frame generation off". Read the same way.
- **The game was still on FSR 2.** After the last of them `UserSettings.json` has `ResolutionScaling = FSR2`,
`FSR2 = Performance` (it was `Quality` at 09:50), `FSR3 = Auto` untouched, `FrameGeneration = Off`. So the quality
and the frame generation were changed and the upscaler was not: **the owner's screenshot of the menu, taken after
them, shows Resolution Scaling = "AMD FidelityFX Super Resolution 2.1", Performance, Frame Generation Off** —
the file's `FSR2` is the menu's FSR 2.1. Both
stored sessions (215 s and 158 s, attached to a game already running) agree with FSR 2.1 built in and frame
generation off: every hook in, **not one ffx dispatch or query of any type**, no Streamline evaluation, no render
size, upscaler "unknown", factor 1.0 from Streamline's frame tokens, ~290 FPS. **So the 2.3 loader's double count
is still unobserved either way**: what answers it is a session in which `UserSettings.json` says
`ResolutionScaling = FSR3` afterwards.
- **The first start was lost to the transient this file recorded an hour earlier — in a real session this time.**
10:05:20 the game appeared; the Overlay went in at 10:05:23; 0.47 s later it was told to stop, and the Agent's
log says why: `SafetyUnhook; Discarded (tier 1, exit=UnhookedSafety, frames=0); guard=ModuleScanFailed/a module
list came back incomplete`. The Overlay's log shows the four patches restored within 0.75 s. The game ran on
unrecorded for two minutes until the owner restarted it. **It is not new and not rare**: the Agent's logs kept on
this PC (2026-11-05 onwards) hold five such ends — Phantom Blade Zero three times on 2026-11-05 (two safety
unhooks and one refusal before injection), Crown & Coin on 2026-11-10, Cyberpunk today — beside the two test
runs. Each is a module list read while the target was still loading modules, at a launch. **The question for the
owner is no longer hypothetical**: a list that cannot be read is not a finding, and rule 2 treats it as a refusal
so that an unreadable game is never hooked; reading it again a bounded number of times before refusing would
keep that — a list that stays unreadable still refuses — and stop losing sessions to a moment of loading. It
changes what the guard does on a check, so it is the owner's decision, and it is not made here. *(Made the same
day: D123, §NEXT item 1b.)*
### NEXT
1. ~~**The owner runs item 40** — above all **one session of Cyberpunk 2077 with FSR upscaling** on beta.42, then
quits the game the ordinary way. Its `logs\overlay-<pid>-….log` is what beta.43 waits for.~~ **Run 2026-11-12;
the section above is what it showed.** What is left of item 40 is (d), a session that changes upscaler, and
the owner's word on the rest.
1a. ~~**Before beta.43's plan: one session of Cyberpunk 2077 with Resolution Scaling = FSR 3, frame generation
off** (the second bullet above), and a second with it on. It says whether FrameLedger names FSR 3 there —
the route through the loader that D121 was first taken to be about — and whether the 2.3 loader counts an
UPSCALE twice. *(Tried 2026-11-12 10:08 and 10:12: the game was still on FSR 2 — the section above. The check
that the right thing ran is the file: `ResolutionScaling = FSR3`.)*~~
**It cannot be run, the owner said the same day: Cyberpunk 2077 has FSR 2.1 as its only FSR upscaler and
FSR 3.1 as frame generation only — two separate settings, not a pair; there is no FSR 3 under Resolution
Scaling.** So D121 is closed for good — every Cyberpunk session upscaling with FSR reads "unknown", by the
limit `LIMITATIONS.md` states — and **no title run so far upscales through a 2.3 loader**. The double
count stays unobserved. What IS measured is that this loader hands a frame-generation dispatch to the leaf's
export; an UPSCALE through the same `ffxDispatch` would take the same path. beta.43's plan therefore counts
an UPSCALE at the first hooked module that sees it (as `fl::adv::Owns` does) as a precaution, with a fixture
whose loader forwards to the leaf's export — rather than wait for a session that cannot be had.
1b. ~~**The owner's decision on the guard's module-list read** (the section above): refuse at the first incomplete
list, as now, or read again a bounded number of times first. If the second, it is a pull request of its own
in the guard (`FrameLedger.Injector`), with `19_SAFETY` amended in it, before or with beta.43.~~ **Decided
2026-11-12 — read again (D123) — and built the same day**: `Collected::kUnsettled`, up to three re-reads,
`19_SAFETY` amended. Not released: it ships with the next pre-release, and until then a launch can still be
lost this way on beta.42. What the owner then checks is the Agent's log: no `guard=ModuleScanFailed` end at a
game's launch, and if one comes its signal now says which kind — *kept changing while it was read* (four reads
were not enough) or *came back incomplete* (more modules than the guard's buffer holds).
2. **0.1.0-beta.43 — planned and approved 2026-11-12 (D123–D125), and under way.** The guard's re-read (D123,
#413) ships with it and not on its own (D124). Five pull requests after it: (1) an UPSCALE is counted at the
first source that saw one (D125) — the fixture's loader bound through the leaves' exports read K = 1 as 2.0
before it; (2) `slSetConstants`; (3) `ffxConfigure` on the four ffx modules; (4) the Agent, the App and the
Privacy Policy 2.17, whose wording the plan carried and the owner approved with it; (5) the release, owner-only
item 41. The design is the one below, as the review of 2026-11-12's first plan left it; nothing in the logs
changed it except D125.
~~**Read that log, then plan 0.1.0-beta.43**~~ — `slSetConstants` and `ffxConfigure`, and the fix for D121 if the
log shows the route. What to look for: which `HOOK_INSTALLED` lines there are (which of the four ffx modules and
the FSR 3.0 host were patched); the `FFX_DISPATCH` lines — is `b=0x10001` (UPSCALE) on any module, or only
`0x20004` / `0x2000c` / `0x20003`; an `UPSCALE_ROUTE` line; `PATCH_FAILED`; the last `RUNTIME_CENSUS` word.
**No UPSCALE type anywhere and no route line means the upscale does not come through any export the Overlay
patches** — statically linked, or a module the inventory does not name — and then the fix is not a new arm in
a detour. The design for the two hooks, as the review left it, is the plan of 2026-11-12 (moved with the
section above) and these points: the new rows are gated at the top of `InstallByFamily`, **before**
`ResolveScoped`, and installed only once the process was ever asked; hook families 13 and 14 with latches of
their own; `slSetConstants` checks the struct's GUID and `structVersion ≥ 1` on every call and reads only
`jitterOffset`, `mvecScale`, `cameraNear` … `cameraAspectRatio` and the seven Booleans — never bytes 368..423,
pinned by `offsetof` of the allowed members; summaries from the first viewport seen; `ffxConfigure` has an
owner slot of its own and reads `frameGenerationEnabled`, `allowAsyncWorkloads`, `flags`,
`onlyPresentGenerated`, `generationRect` and nothing else; the detours also consult `MayObserve()`;
`overlay-fields-check` gains three parameter names, `structType` among the shared members and an `offsetof` per
listed `sl::Constants` member. The registry holds 48 and the worst case becomes 25
(`kPatchSitesOutsideTheInventory` stays 7; the inventory's row count goes from 13 to 18 and its
`static_assert` says so).
3. **0.1.0-beta.44** — `slGetFeatureVersion` and `slGetFeatureRequirements` — when the owner says beta.43 passed.
4. **Trains C and D** (beta.45, beta.46) each begin with a spike and end in the owner's go or no.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request.~~ **Recorded 2026-11-12: 0.1.0-beta.42 was released
2026-11-12T02:31:07Z** (09:31 in the owner's time zone) from the signed tag on `a708ae8` (#410), by run
43139775713 — green at the first attempt, one run for the one tag push. Eight assets; the seven digests of
`SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry `FL_BUILD_ID v0.1.0-beta.42`. The
delta from beta.41 is 710,943 bytes beside a full package of 110,716,228 — "Delta processed 0452 files. 0023
patched, 0429 unchanged". The wiki was published by run 43142009098. The rehearsal on the release branch (run
43137520796) had made a delta of 735,485 bytes with the same file counts.
- **No real game was run on this build.** The four changes were tested with the harness and the cost measured on
the headers. Item 40 is the first real session.
- **Whether the missing plane of Lies of P, Clair Obscur and Dying Light is `FLT_MAX`** is still the reading their
small `cameraFar` values point to, not a measurement. Item 40 (c).
- **A flake with a cause now, and a question in it for the owner.**
`ShmDrainIntegrationTests.TheGuardInjectsTheOverlayAndTheReaderDrainsRealFrames` was red once in this release's
local gate (and once on 2026-11-11, when the assertion did not yet print the verdict): the supervisor's first
re-scan, 0.8 s after the injection, came back `ModuleScanFailed` — *a module list came back incomplete* — while
the harness was still loading modules. The same gate was green on the same code before and after. In the test
it is a re-run. **In a real session the same answer is a safety unhook**: a re-scan that cannot read the module
list of a game that is still running stops the session (`19_SAFETY` §During a session; D48 covers only a game
that has exited). Whether the guard should read the list a second time before refusing is a change to rule 2's
checks and is the owner's to decide; nothing was changed. **A second one within the hour, in the gate of the pull
request that records the release's facts:**
`ToolhelpProcessSnapshotSourceTests.ThisProcessIsInTheSnapshotWithItsNormalisedImagePathAndACreationTime` got an
empty snapshot — `ToolhelpProcessSnapshotSource.Take` returns an empty list when the snapshot cannot be taken or
walked, and says nothing. Green on the run after. Both are Windows refusing, for a moment, to enumerate something
on a PC that was building, testing and running FrameLedger itself at once; what an empty snapshot does to the
watcher's tick in a real session was not looked into.
- ~~**SonarCloud has not answered for #406–#409**: `main`'s `sonarcloud` job was cancelled by the next merge each
time. It runs to its end on this release's commit; read it then.~~ **It answered on `a708ae8`: the quality gate
passed, and one issue was open on `main`** — `cpp:S5813`, a `strlen` in the helper the new log case of
`guard_test.cpp` reads lines with (#408). Fixed in the pull request that records these facts: the helper takes
strings. **`csharpsquid:S3869` on `ParentDump.cs` is no longer open** — SonarCloud shows it accepted at
2026-11-11T21:13Z (04:13 on the 12th, the owner's time), which is before the texts that still called it open
were written; they were not checked against SonarCloud then. **Qodana's two reports are green on the release's
commit**; the native one was red from #407 to #408 on seven problems in test code, answered in #409.
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. ~~**SonarCloud's `csharpsquid:S3869` on
`ParentDump.cs`** is the owner's to accept.~~ *(Done — the line above.)* **Items 36 (d) and 37 are still open.** The alpha line still rests
(D98).
**Owner-only:** item 40.
+154
View File
@@ -0,0 +1,154 @@
# HANDOFF history — 0.1.0-beta.43
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.44 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its §NEXT item 2 says 0.1.0-beta.44 is planned**: it was released the same day, and what it built and decided is HANDOFF.md's section for it.*
## **START HERE** — 2026-11-12, 0.1.0-beta.43: two hooks installed only where asked for, the guard reads a moving list again, an UPSCALE counted once
*Status is `docs/changelog/0.1.0-beta.43.md`; this is what no other file carries.* The owner ran beta.42 the morning
it was released; seven sessions, a copy of the ledger and the logs answered D121 (Cyberpunk 2077's FSR is 2.1, built
into the game), showed the guard losing launches to a module list that moved while it was read, and showed the
FidelityFX 2.3 loader handing a call on to the leaf's export. The owner decided D123–D125 and approved the plan the
same morning, Privacy Policy 2.17's wording with it. Five pull requests and this one: #413 (the guard reads again,
D123), #414 (an UPSCALE is counted at the first source that saw one, D125; D124), #415 (`slSetConstants`), #416
(`ffxConfigure` on the four ffx-api modules), #417 (the Agent, the App, Privacy Policy 2.17). **What the owner's
sessions on beta.42 showed, and the design as planned, are
[`handoff/history/0.1.0-beta.42.md`](handoff/history/0.1.0-beta.42.md)**; what each pull request built is
[`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) §Pre-injection checks,
[`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory (the two rows, with their measured cost) and
[`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B.
### What this release decided, beyond D123–D125
- **A hook that exists only for Advanced metrics is not put into a game that was never asked.** The two rows are
gated at the top of `InstallByFamily`, before the symbol is resolved (resolving holds the module). Once a
session with the option on has started in a process, the patch goes in on that watchdog tick and **stays for the
life of the process**: a later session with the option off finds the detour there, forwarding without looking.
Removing a patch between sessions would be a second patch event in a running game for nothing gained. So "off
means off" (D111) is, for these two: nothing read, nothing published — and, in a game no session ever asked,
nothing patched. Privacy Policy 2.17 §4 says it that way.
- **Privacy Policy 2.17, not 2.16 stretched.** beta.40's plan had 2.16 stand "once for the three releases"; its
text then listed DLSS's and FSR's parameters by name, and neither list holds Streamline's constants or the
frame-generation configuration. The plan of beta.43 carried the new sentences and the owner approved them with it.
**beta.44 reads feature versions and requirements — check 2.17's lists the same way before assuming it covers
them.**
- **For each of the two calls a session says which of three things happened** (`AdvancedInGameCall`): read; hooked
and the game never called; not hooked. "Not hooked" is three causes the session does not tell apart — the game
did not load the library, a Streamline 1.x interposer (refused), a patch the registry refused — and the Overlay's
log does: a `HOOK_INSTALLED` line naming the function, or a `PATCH_FAILED`.
- **A count is not part of what makes a session's facts new.** `sl.constants.resets` grows all session; the
recorder's context would be rewritten each second. `InGameAdvancedSampler.Conclude` brings it up to date once, at
the end. A session recovered after a crash holds the count as of the last time some other field moved — a lower
bound.
- **Streamline's summaries are the first viewport's**, and the configure count has an owner slot of its own (the
2.3 loader would otherwise bring one configure to two modules, as D125's dispatch).
- **The accuracy statement no longer says nothing of what a game tells its upscaler was read from a real game.**
FSR's upscaling values were (beta.41, beta.42: Lies of P, Dying Light: The Beast, Kingdom Come: Deliverance II).
The DLSS options, Streamline's constants and the frame-generation configuration are still unconfirmed, and the
three copies say so.
### Traps this train left
- **An unpatched export may already begin with a jump.** The loader stub's `ffxConfigure` starts with `E9` in the
file: "the first byte is not a jump" is not a test of "not patched". `ReadTargetExportBytes` compares the target's
bytes with the file's own (`inFile`).
- **A branch stacked on a pull request that is then squashed**: `git rebase --onto origin/main <the old tip>`, and
`git diff <old tip> <squash>` empty says the gate that ran still stands.
- **`mcp__ccd_pr__set_auto_merge` says "not bound" for a pull request opened from a script**: `gh pr merge <n>
--auto --squash`. A run's artifacts while it is still in progress: `gh api
repos/…/actions/artifacts/<id>/zip`.
- **clang-format re-wraps what an edit script wrote**, so a second script's anchors must be read from the formatted
file; a multi-file script that fails part-way has already written the earlier files.
- **One local gate printed 871 App tests where every other printed 872**, green. Not looked into.
### What the owner's nine sessions on beta.43 showed (2026-11-12)
Run 15:42–16:21 in the owner's time zone, read the same afternoon from a copy of the ledger (sessions 171–179), the
Agent's log, the nine overlay logs and Windows' application log. Advanced metrics was on for the first eight and off
for the last.
| # | Game | What it showed |
|---|---|---|
| 171 | Phantom Blade Zero, 649 s, DLSS frame generation ×3.0 | Streamline 2.14.1: **read** — jitter ≤ 0.47, motion-vector scale 0.0009 / 0.0016, near 0.01 and far 75,000 with depth inverted, field of view 0.6545–1.3612 rad, aspect 1.7767, 1 reset. FSR frame generation: hooked on the loader and both leaves, the game never called |
| 172 | Lies of P, 35 s, exit code 3 | 542 frames, nothing evaluated. **The owner: the game crashed after a shader failed to compile.** The crash policy counted it and turned the game's hooking off (below) |
| 173 | Lies of P, 130 s | Tier 2 — hooking was off; the owner switched it on again at 16:02 |
| 174 | Lies of P, 94 s, FSR 3 + FSR frame generation ×2.0 | **`ffxConfigure` read** on the one-module `amd_fidelityfx_dx12.dll`: frame generation on and off, not asynchronous, generated frames not presented alone, 2560 × 1440 at 0, 0; the log has `FFX_CONFIGURE` types `0x10007` and `0x20002`. Streamline: not hooked (the game has none) |
| 175 | Cyberpunk 2077, 72 s, no upscaler named, no frame generation | Streamline 2.14.1: **read** — jitter ≤ 0.5, motion-vector scale 1 / 1, near 0.02 and far 16,000 inverted, aspect 1.7778, **field of view 51.0055–90**, 2,575 resets |
| 176 | Cyberpunk 2077, 145 s, DLSS + DLSS frame generation ×2.0 | the same values; 1,821 resets. FSR frame generation: hooked on the loader and both leaves, never called |
| 177 | Kingdom Come: Deliverance II, 135 s, FSR | Streamline: not hooked. `ffxConfigure` hooked on the loader and the upscaler, never called — the game ships no frame generation |
| 178 | Hogwarts Legacy, 158 s, DLSS frame generation ×4.0 | Streamline 2.14.16: **read** — jitter ≤ 0.47, field of view 0.8872–2.0071 rad, aspect 1.7763, near 0.01 and far 75,000 inverted, 11 resets. `ffxConfigure` hooked on `amd_fidelityfx_dx12.dll`, never called |
| 179 | Lies of P, 129 s, **Advanced metrics off**, a new Agent, the game started afresh | no Advanced row; bits 13 and 14 clear in `hooks_installed_mask`; the log has `ffxDispatch` and `ffxQuery` and **no `ffxConfigure` line** |
- **The two hooks did what they were built to do on four real games**, three Streamline builds apart (2.14.1 twice,
2.14.16) and one FSR library. No `PATCH_FAILED`, no fault counted, in nine logs. `legal_acceptance` holds Privacy
Policy 2.17 accepted once, four seconds after the update; no session since the update ended with
`guard=ModuleScanFailed` (the day's one is beta.42's, at 10:05).
- **Cyberpunk 2077 gives Streamline its field of view in degrees.** The vendor's header says radians; the game
passes 51–90. Every other field of the same struct is sane (the aspect ratio beside it is 1.7778), so the layout
is not in question — the game's unit is. The page prints `51.0055 – 90 rad (2922.4° – 5156.6°)`: **a defect of
the label, not of the reading**, and beta.44's first pull request. The two stored sessions keep the string they
were stored with.
- **A resets count has no denominator.** 2,575 in a 72 s session that was mostly menus and 1,821 in 145 s on
Cyberpunk, beside 1 and 11 on the two Unreal games. It is what the game said (`reset == eTrue`, the first
viewport's calls); the page does not say out of how many calls.
- **A frame-generation configuration with no flag set has no flags row**, where an upscale with none says `none`.
- **Lies of P's 35 s end turned its hooking off, and one such end is now enough.** `hook_crash_count` is never set
back: the game's two ends of 2026-10-19/20 (`19_SAFETY` §Crash & stability safety, the hand-off to Steam) had
brought it to 2, the owner had switched hooking on again, and today's single early end made 3 — at or over
`CrashesToDisable`, so hooking went off at once, under a reason that says *twice*. Whether switching hooking back
on starts the count again is the owner's decision; it is asked with beta.44's plan.
- **Hogwarts Legacy faults in the driver's `_nvngx.dll` as it exits, hooked or not.** Windows' application log has
the same fault (0xc0000005 at +0x6aa85, `_nvngx.dll` 32.0.16.2114) six times since 2026-11-06, each at the end
of one of the game's seven recorded sessions — the first of them session 131, which nothing was injected into.
That one reads `crashed`; the five hooked ones, today's 178 among them, read `normal` with exit code 0, the
fault logged in the second the session was saved. **Why a hooked session does not see the crash event that an
unhooked one does was not looked into.**
- **Frame rates as before**, where the settings were: Lies of P 59 → 118 (beta.42: 58.9 → 118.0), Kingdom Come 59.3
(59.4), Hogwarts Legacy 58.8 → 235.8 at ×4 (alpha.3: 56.1 → 225.1). Phantom Blade Zero's factor moves between
sessions under the driver's dynamic mode (×2.9 to ×4.1 since 2026-11-05) and Cyberpunk was run with other settings.
- **Not exercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame generation) — the
configure count's owner slot has not met a real game; a game with neither library; Dying Light: The Beast.
Phantom Blade Zero and Hogwarts Legacy name no upscaler beside DLSS frame generation, as on every release before.
### NEXT
1. ~~**The owner runs item 41**~~ **Run and read 2026-11-12** (§What the owner's nine sessions on beta.43 showed).
The owner's word was to go on if it passed, and it is judged passed: both hooks read real games, nothing
failed to patch, off meant off. What was not exercised is listed there and stays open.
2. **0.1.0-beta.44** — first the field of view a game gives in degrees (and what else the nine sessions found),
then `slGetFeatureVersion` and `slGetFeatureRequirements`. Both are calls a game makes while starting, so most
sessions will have missed them: plan what "not seen" says. ~~**Planned with the owner before any code.**~~ **The
plan was approved 2026-11-12 with three decisions, D126–D128** (the count of early abnormal ends starts again
at a grant; the fault at exit is written down only; a field of view over π is shown as given), and with Privacy
Policy 2.18's wording. Four pull requests: the fixes; the two hooks in the Overlay; the Agent, the App and the
policy; the release.
3. **Trains C and D** (beta.45, beta.46) each begin with a spike and end in the owner's go or no.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said of #413–#417 (the
`sonarcloud` job on `main` is cancelled by the next merge, so it answers on the release's commit).~~ **Recorded
2026-11-12: 0.1.0-beta.43 was released 2026-11-12T06:57:57Z** (13:57 in the owner's time zone) from the signed
tag on `6c7ba74` (#418), by run 43169675460 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry
`FL_BUILD_ID v0.1.0-beta.43`. The delta from beta.42 is 1,934,264 bytes beside a full package of 110,736,398 —
"Delta processed 0452 files. 0025 patched, 0427 unchanged". The wiki was published by run 43171842051. The
rehearsal on the release branch (run 43167388933, the same tree before it was rebased onto #417's squash) had
made a delta of 1,944,086 bytes with the same file counts. **SonarCloud answered on `6c7ba74`: the quality gate
passed and no issue is open on `main`. Qodana's two reports are green on that commit**, and were on #417's.
- ~~**No real game was run on this build.** Two new patch sites, tested with the harness and the stubs; their cost
measured on the headers. Whether a real `sl.interposer.dll` 2.x lays `sl::Constants` out as the vendored header
does is checked on every call by the struct's GUID and version, not by a real game yet.~~ **Nine sessions on five
games, 2026-11-12** (§What the owner's nine sessions on beta.43 showed): interposers 2.14.1 and 2.14.16 lay the
struct out as the header does. **The accuracy statement's three copies and `LIMITATIONS.md` still say Streamline's
constants and the frame-generation configuration "have not been confirmed" on a real game** — reworded with
beta.44's first pull request. **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy and
the owner's review.
- **Item 40 (d)** — a session that changes upscaler — was not exercised, and the owner has not said item 40 passed.
**Items 36 (d) and 37 are still open.**
- **What an empty process snapshot does to the watcher's tick in a real session** was not looked into
(`ToolhelpProcessSnapshotSource.Take` returns an empty list and says nothing; seen once in a test on 2026-11-12).
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** what item 41 left unexercised — (d) through the FidelityFX 2.3 loader, (e) a game with neither library; item 40 (d).
+182
View File
@@ -0,0 +1,182 @@
# HANDOFF history — 0.1.0-beta.44
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.45 was cut, 2026-11-12 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its §NEXT says 0.1.0-beta.45 is planned**: it was released the same day, and what it built and decided is HANDOFF.md's section for it.*
## **START HERE** — 2026-11-12, 0.1.0-beta.44: what the first real games on beta.43 found, fixed; Streamline's answers about a feature, hooked only where asked
*Status is `docs/changelog/0.1.0-beta.44.md`; this is what no other file carries.* The owner ran nine sessions on
beta.43 the afternoon it was released and said: read them, write down the result, go on if it passes. They were read
from a copy of the ledger, the Agent's log, nine overlay logs and Windows' application log (#420); item 41 was judged
passed, and the owner decided D126–D128 and approved the plan the same afternoon, Privacy Policy 2.18's wording with
it. Four pull requests and this one: #421 (the fixes: a field of view in degrees, the count of early abnormal ends,
`flags` of `none`; D126–D128), #422 (`slGetFeatureVersion` and `slGetFeatureRequirements` in the Overlay), #423
(the Agent, the App, Privacy Policy 2.18). **What the nine sessions showed, session by session, is
[`handoff/history/0.1.0-beta.43.md`](handoff/history/0.1.0-beta.43.md)**; what each pull request built is
[`19_SAFETY_AND_ANTICHEAT.md`](19_SAFETY_AND_ANTICHEAT.md) §Crash & stability safety,
[`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory (the row of the two calls, with their measured cost) and
[`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B.
### What this release decided, beyond D126–D128
- **What is read of these two calls is the answer, and only a good one.** The detours call the original first and
read the struct the game passed when it returned `eOk`; a refusal leaves the struct as the game made it, and the
fixture's stub leaves 99s in it to prove nothing of a refused answer is kept. The pointers the answer carries —
the list of required buffer tags, the names of Vulkan extensions and features — are never followed; the stub's
are wild, and no fault is counted.
- **The feature slots are the one group of region 5 a new request does not empty.** Every other group is a
session's — ranges, counts, switches seen. Which version a feature is and what it requires are facts about the
process, and a game asks once: a second session of a game that kept running would have nothing where the first
had the answer. `Reset` keeps the slots and their bit of the mask. They were read only while a session asked.
- **The two rows are the first of the table, because rows are patched in the table's order.** Nothing in the
Overlay paces the installs: the 0.1–0.25 s between `HOOK_INSTALLED` lines in the owner's logs is MinHook stopping
every thread of the game for each patch. In a process that was asked, the two calls a game makes while starting
are the ones that cannot wait behind five others; in a process never asked, the two rows are passed over before
anything is resolved.
- **"Hooked, and the game did not call" has words of its own for a call made while starting**
(`AdvInGameCall_NotAskedAgain`). For a call made every frame it is a finding about the game; for these two it is
what every late attach will say. One bit of the mask covers both questions, so which was answered is told by the
slots (`InGameAdvancedSampler.FeatureCall`).
- **Privacy Policy 2.18, not 2.17 stretched**: 2.17 says "two further calls" and lists what they give. 2.18 also
says what of the answer is never read.
- **The rows are the vendor's words**: `kFeatureDLSS_G · versionNGX`, the requirement flags by their enumerators,
an id this build has no name for as `feature <number>`. A version of all zeros has no row, except `versionSL`.
- **The Agent's end-to-end tests run a Streamline target now** (until this release only an FSR one): the constants
and the two answers, through the shipped Agent and the real Overlay.
- **A field of view over π is two rows** (D128): the number, and `cameraFOV unit` with the token `not_radians` —
a token replaces a row's value in the App, so a note beside a number is a row of its own.
- **The count of early abnormal ends is set back by the consent grant and by nothing else** (D126) — the one
statement that turns hooking on. `hook_autodisabled_reason` and `_at` stay as the row's history; the App shows
them only while hooking is off.
### Traps this train left
- **`vswhere -latest` names SQL Server Management Studio on this PC.** Ask for the C++ tools as `build.ps1` does
(`-requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64`). The `x64-release` preset writes no
`compile_commands.json`: `ninja -C build/native/x64-release -t commands FrameLedger.Overlay` gives the Overlay's
flags for a bench.
- **Run clang-format on the native files an edit script wrote before the gate** — the format check comes after the
native build and tests, and a miss costs the whole run.
- **`tools/resx-gen.ps1` needs `-Resx src/FrameLedger.App/Strings.resx`**; started without it from a shell that
has no console it waits for the missing parameter until it is stopped.
- **A Python script passed to a shell as a here-document loses its backslashes** before Python sees them: write
the script to a file.
- **`gh pr checks --watch` returns at once when no check has started yet**; wait a moment before starting it.
- **Qodana's native report is advisory, and auto-merge does not wait for it.** On #422 it reported three
`bugprone-implicit-widening-of-multiplication-result` in the test and the harness (an index into the feature
slots computed as a product of two 32-bit values: compute it in `size_t`) while the required checks were still
running; auto-merge was switched off until the fix was pushed. A finished job's log while its run is still going:
`gh api repos/…/actions/jobs/<id>/logs --allow-escape-sequences`; the report's SARIF is a zip inside the
artifact's zip.
- **GPG's pinentry timed out once** (the cache had expired); the second attempt went through when the owner was
at the keyboard.
### What the owner's eight sessions on beta.44 showed (2026-11-12)
Run 19:52–20:46 in the owner's time zone, read the same evening from a copy of the ledger (sessions 181–188), the
Agent's log, eight overlay logs and Windows' application log. Advanced metrics was on throughout.
| # | Game | What it showed |
|---|---|---|
| 181 | Phantom Blade Zero, 796 s, DLSS frame generation ×4.0 | Streamline 2.14.1. **Both questions read**: versions of `kFeatureReflex`, `kFeatureDLSS_G`, `kFeaturePCL`; requirements of those and `kFeatureDeepDVC`. The constants as on beta.43 |
| 182 | Clair Obscur: Expedition 33, 107 s, FSR + DLSS frame generation ×3.0 | Streamline 2.7.30. **Both read** (`kFeatureDLSS_G`, `kFeatureReflex`, `kFeaturePCL`); the constants and FSR's upscale values read beside them |
| 183 | Valor Mortis (Microsoft Store), 51 s, **crashed** — exit code `0xC0000005` | Streamline 2.14.1. Both read before it ended. The game's second early abnormal end, so hooking went off (below) |
| 184 | Valor Mortis, 935 s, ×6.0 | hooking switched on again at 20:11; both read; ended normally |
| 185 | Black Myth: Wukong, 167 s, ×4.0 | Streamline 2.7.4. Both read |
| 186 | Cyberpunk 2077, 257 s, DLSS + DLSS frame generation ×2.0 | **The field of view reads `51.0055 – 90` with a row `cameraFOV unit` = *not radians*** (D128). Requirements **read** for `kFeatureDLSS`, `kFeatureDLSS_RR`, `kFeatureDLSS_G`, `kFeatureReflex`, `kFeatureNIS`; the version question *hooked, not asked again* |
| 187, 188 | Kingdom Come: Deliverance II, 39 s and 289 s | no Streamline: the constants and both questions *not hooked*; `ffxConfigure` hooked and never called. Neither session reached an FSR dispatch (the upscaler's first query came 11 s before the second one ended) |
- **The two questions were read in six of six Streamline sessions, on five games** — the plan, the release notes
and four documents said most sessions would show nothing, because a game asks while it starts. Unreal's
Streamline plugin (four of the five games) asks again after the attach, and Cyberpunk 2077 asks for requirements
again; only Cyberpunk's version question was *not asked again*. The three answers of the line were each seen.
The documents are corrected in the same pull request as this text.
- **What was read agrees with the PC.** `driverVersionDetected` 621.14.0 is the snapshot's driver
(`32.0.16.2114`), `osVersionDetected` 10.0.29648 its Windows build, `versionNGX` of `kFeatureDLSS_G` the file
version of the game's `nvngx_dlssg.dll` (310.9.1, 310.6.0), `versionSL` of Reflex and PCL the interposer's.
- **`kFeatureDLSS_G`'s `versionSL` is 2.14.16 on three games that ship 2.14.1 or 2.7.30** (Phantom Blade Zero,
Valor Mortis, Clair Obscur): Streamline answers for a newer frame-generation plugin than the game's own folder
holds. Which file that is was not looked into — an updated plugin delivered by NVIDIA's app is the likely one.
It is the first value of this kind that a file's version could not have given.
- **No Unreal game asked about `kFeatureDLSS`.** Their DLSS upscaling does not go through Streamline (Cyberpunk's
does: it is asked about, and the session names DLSS). That is the likely reason Phantom Blade Zero, Hogwarts
Legacy, Valor Mortis and Black Myth: Wukong name no upscaler beside DLSS frame generation — the question left
open since beta.43. Naming it would take a hook on NGX, which this project does not have.
- **Valor Mortis crashed 51 s into a hooked session, as it did on beta.40.** Session 141 (2026-11-12 02:44,
beta.40: `0xC0000005` after 51 s, 498 frames) and session 183 (20:09, beta.44: the same code after 51 s, 8,077
frames); the session after each ran normally under the same build. So it is not this release's hooks. Whether
FrameLedger has any part in it is not known: Windows logged no event for either (a Store package), and there is
no unhooked start to compare. **The crash policy did what D126 says**: today's was the game's second early end, so
hooking went off at 20:09:52; the owner switched it on at 20:11:03 and the count is 0.
- **Black Myth: Wukong faults in Steam's `crashhandler64.dll` as it exits** (0xc0000409 at +0x39b49) — at each of
its three recorded exits, all hooked (beta.40 twice, beta.44 once); each session reads `normal`, the shape D127
wrote down for Hogwarts Legacy. **There is no unhooked exit to compare**: one session of it with hooking off
would say whether FrameLedger has a part.
- **No `PATCH_FAILED`, no fault counted, in eight logs**; Privacy Policy 2.18 was accepted once, eight seconds
after the update. The two new rows are the first Streamline `HOOK_INSTALLED` lines of every log. Frame rates as
before where the settings were: Clair Obscur 106 → 323 (beta.41: 106 → 318), Valor Mortis 39.6 → 237 (37.7 →
228). Wukong's native rate was 90 where beta.40's two sessions had 124; nothing says whether the scene and
settings were the same.
- **Not exercised: off means off** — every session had Advanced metrics on. And, from item 41: a configure through
the FidelityFX 2.3 loader.
### NEXT
1. ~~**The owner runs item 42**~~ **Run and read 2026-11-12** (§What the owner's eight sessions on beta.44
showed): both hooks read real games — far more often than planned for — the degrees fix reads right on
Cyberpunk 2077, a game with no Streamline says *not hooked*. *Off means off* was not exercised.
2. **The next version, by the owner's word of 2026-11-12: `slInit`, `slIsFeatureSupported` "and some other parts"**
— said the evening beta.44 came out, to be done if beta.44 passed, and then "update and go on" once these
sessions were in. It reopens D112 (`slInit` is not hooked in train B; if ever: never `applicationId`,
`projectId`, the paths), and `slIsFeatureSupported` is handed an `sl::AdapterInfo` whose `deviceLUID` and
`vkPhysicalDevice` are not to be read (D104). ~~**Planned with the owner before any code**: which other parts,
what of each is read, the Privacy Policy's wording, and where train C goes.~~ **Planned and approved 2026-11-12
(D129–D131)**: `slInit` stays out — shown why it cannot be seen, the owner chose so; six calls go in, in one
release, **0.1.0-beta.45** — `slIsFeatureSupported`, `slIsFeatureLoaded`, `slSetFeatureLoaded`,
`slAllocateResources`, `slFreeResources`, `slShutdown`; Privacy Policy 2.19's wording approved with the plan.
Pull requests: three hooks, three hooks, the Agent and the App with the policy, the release.
3. **Train C (Reflex latency per frame)**: its spike is written beside beta.45, a document only; code waits for the
owner's go after reading it. **Train D** after that, the same way.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-12: 0.1.0-beta.44 was released 2026-11-12T12:24:26Z** (19:24 in the owner's time zone) from the signed
tag on `f9e9fdb` (#424), by run 43214789700 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry
`FL_BUILD_ID v0.1.0-beta.44`. The delta from beta.43 is 1,866,183 bytes beside a full package of 110,750,960 —
"Delta processed 0452 files. 0025 patched, 0427 unchanged". The wiki was published by run 43217647799. The
rehearsal on the release branch (run 43204362797, at `56b1157` — the released tree but for two namespace
qualifiers #423 dropped afterwards) had made a delta of 1,878,774 bytes with the same file counts. **SonarCloud
answered on `f9e9fdb`: the quality gate passed and no issue is open on `main`. Qodana's two reports are green on
that commit.** Qodana found two things on the way, each fixed before its pull request merged: three index
products in the injected test and the harness (#422, native), and two namespace qualifiers a new `using` made
redundant (#423, managed) — #423's auto-merge was held until that report was in. **`check` failed once on
#424**, a pull request that changes no code: the CaptureHost's launch-mode end-to-end test
(`LaunchModeStartsTheHarnessItselfAndReportsWhatRanUnhooked`) read 19.8 presented frames a second from the
harness on a slow runner where it asks for more than 30; the job was run again and passed. A rate asserted
against a wall clock on a shared runner — the one flake of this train, not looked into further.
- ~~**No real game was run on this build.** Two new patch sites, tested with the harness and the stub; their cost
measured on the headers.~~ **Eight sessions on six games, 2026-11-12** (§What the owner's eight sessions on
beta.44 showed). **The release notes on GitHub still say "most sessions will show nothing of the two new
readings"** — a prediction the sessions did not bear out; `docs/changelog/0.1.0-beta.44.md` is a released file
and is not edited, and the published notes were left as they are.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed). Looking
again later would make every session of Hogwarts Legacy `crashed`, and under a user-mode exception would end the
exception for a fault no hook caused; the decision is the owner's, and is not taken.
- ~~**Phantom Blade Zero and Hogwarts Legacy name no upscaler beside DLSS frame generation**, as on every release
before; not looked into.~~ **Likely answered 2026-11-12**: no Unreal game asks Streamline about `kFeatureDLSS`, so
their DLSS upscaling does not pass through the calls this project hooks.
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **What an empty process snapshot does to the watcher's tick in a real session** was not looked into (seen once in
a test on 2026-11-12).
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** item 42 (f), and one session each of Valor Mortis and Black Myth: Wukong with hooking off; what
item 41 left unexercised; item 40 (d).
+175
View File
@@ -0,0 +1,175 @@
# HANDOFF history — 0.1.0-beta.45
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.46 was cut, 2026-11-13 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its §NEXT says 0.1.0-beta.46 is to come**: it was released the next day, and what it built and decided is HANDOFF.md's section for it.*
## **START HERE** — 2026-11-12, 0.1.0-beta.45: six more of Streamline's calls, hooked only where asked; `slInit` stays out; the Reflex spike as a document
*Status is `docs/changelog/0.1.0-beta.45.md`; this is what no other file carries.* The owner ran eight sessions on
beta.44 the evening it was released and said: add the three new ones, update, go on. They were read from a copy of
the ledger and the logs (#426); asked to reopen D112, the owner was shown why `slInit` cannot be seen and chose to
leave it out (D129), chose six other calls for one release (D130), and approved the plan with Privacy Policy 2.19's
wording. Four pull requests and this one: #427 (`slIsFeatureSupported`, `slIsFeatureLoaded`, `slSetFeatureLoaded`
in the Overlay; D129–D131), #428 (`slAllocateResources`, `slFreeResources`, `slShutdown`), #429 (the Agent, the
App, Privacy Policy 2.19), #430 (train C's spike, a document: `spike-notes` §30). **What the eight sessions on
beta.44 showed, session by session, is [`handoff/history/0.1.0-beta.44.md`](handoff/history/0.1.0-beta.44.md)**;
what each pull request built is [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Hook inventory (the two rows of the six
calls, with their measured cost) and [`18_GPU_VENDOR_APIS.md`](18_GPU_VENDOR_APIS.md) §Train B (the fourth step).
### What this release decided, beyond D129–D131
- **Everything the six calls give is word 1 of a feature's slot, and one reserved word.** Bits 2–4 the two answers
(supported, with the `sl::Result` in bits 8..15; loaded and its value), bits 5–6 the game switching the feature
off and on, bits 16..19 and 20..23 two counts that stop at fifteen and never carry; `slShutdowns` took the first
of three reserved words (@308). Layout 5 and schema 23 as they were. **The answers are the process's and a new
request keeps them; the switching, the counts and the shutdowns are a session's and a new request clears them.**
- **A refusal is the answer.** `slIsFeatureSupported`'s detour calls the original first and keeps whatever it
returned; the page names it by NVIDIA's enumerator (forty codes, the same order at sixteen upstream tags from
v2.0.0 to v2.14.16), `result <n>` past the list, 255 for a code a byte could not carry.
- **The two calls on a game's way out note the call, then call the original, and run nothing after it** —
`slFreeResources` and `slShutdown`. The fixture shuts Streamline down and goes on using the stub through every
detour.
- **The adapter is never read** (D131), and the build refuses the names of `sl::AdapterInfo`'s members in the code
loaded into a game (`overlay-fields-check`); the fixture hands a wild LUID pointer.
- **A stop restores its patches one after another, about 160 ms each.** MinHook stops every thread of the process
for each `MH_DisableHook`. With six more patches the stop's last log lines were cut at the two seconds the
Overlay waited for them; the wait is now sized to the patch table (`250 ms × kMaxPatches`). The patches were
always restored; `UNHOOKED` is written after the last. **Not changed, and proposed**: queueing the restores (and
the installs) so the threads are stopped once — a change to the hook engine's order of operations, which wants
its own pull request, its own measurement and the owner's word. Until then a stop in a Streamline game with
Advanced metrics on takes two to three seconds on the thread that stops.
- **Six calls are one sentence on the page**, each by its function's name under which of three things became of
it; a row stored before this release has none of the six and no such sentence.
- **`AdvancedReach.BeforeAttach` is a reason of its own** for `sl.init.preferences`: not "a later release". No
catalog entry carries `NewHook` any more; the value stays, for rows a later train may add.
- **Where the rows sit in the table** (rows are patched in its order): the two questions after beta.44's two, at
the head; `slSetFeatureLoaded` after `slSetConstants`; the three counted calls last of the Streamline block.
- **Train C's spike is a reading** (`spike-notes` §30): hooking NVAPI's or Streamline's latency functions is
refused or is train D's wrapper; the Overlay calling `slReflexGetState` breaks the vendor's thread contract;
what remains is the Overlay calling two driver getters with the game's device, and a no-injection ETW route
that gives half the number. It ends in three questions that are the owner's.
### Traps this train left
- **The catalog table of `18_GPU_VENDOR_APIS` is compared with the catalog row for row by a test**: a catalog
entry and its row go in the same commit, or the Application tests are red.
- **Qodana's native report again found an index into the feature slots computed as a 32-bit product** (#427, the
harness's probe): write `size_t{2} * FL_ADV_FEATURE_WORDS`. It is advisory and auto-merge does not wait for it;
both reports were read before each merge.
- **A second `region` inside the probe's block is C4456**, and warnings are errors: name the inner one otherwise.
- **A test method over sixty lines is MA0051** in the test projects too; split the assertions by what they cover.
- **The injected test waits for the stop**: with eighteen patches a stop takes about three seconds, so its wait is
200 iterations.
- **The shell tool refuses `sleep N; command`**: wait in a background `until` loop.
- **After a lower pull request is squashed**, `git diff <old tip> origin/main` empty proves the squash is the
branch; then `git rebase --onto origin/main <old tip> <next branch>` and a forced push with lease.
### What the owner's seven sessions on beta.45 showed (2026-11-13)
The owner ran seven sessions on six games the day after the release and said: check the logs and go on. Read from
a copy of the ledger (sessions 190–196), the seven overlay logs, the Agent's log and Windows' application log —
no FrameLedger binary was run against the owner's data. **Every session is hooked, `normal`, with no fault
counted, no dropped record and no `PATCH_FAILED`; the count of early ends is 0 on all six games.** The update
showed the terms once: Privacy Policy 2.19 was accepted at 07:23:36.
| Game (session, length) | Interposer | `slIsFeatureSupported` | `slIsFeatureLoaded` | `slSetFeatureLoaded` | allocate | free | shutdown |
|---|---|---|---|---|---|---|---|
| Phantom Blade Zero (190, 2 h 48 min; 191, 1 h 32 min) | 2.14.1 | **read** — Reflex, DLSS frame generation, PCL `eOk`; ImGUI and DeepDVC `eErrorFeatureMissing` | **read** — Reflex, frame generation *on* | not called | not called | **read** — frame generation, 1 | not called |
| Cyberpunk 2077 (193, 7 min) | 2.14.1 | **read** — frame generation, DLSS, ray reconstruction, Reflex `eOk`; NIS `eErrorFeatureMissing` | **read** — the four *on* | **read** — the four, *on and off* | not called | not called | not called |
| Dying Light: The Beast (194, 4 min) | 2.8.0 | **read** — DLSS, frame generation, PCL, Reflex, NIS `eOk` | not called | **read** — all five, *on and off* | not called | not called | **read** — 1 |
| Hogwarts Legacy (195, 4 min) | 2.14.16 | **read** — frame generation, Reflex, DeepDVC, DLSS 5's `kFeatureDLSS_NR`, PCL `eOk`; ImGUI `eErrorFeatureMissing` | **read** — the first four *on* | not called | not called | not called | not called |
| Black Myth: Wukong (196, 3 min) | 2.7.4 | **read** — frame generation, Reflex, PCL `eOk` | not called | not called | not called | not called | not called |
| Kingdom Come: Deliverance II (192, 4 min) | none | not hooked | not hooked | not hooked | not hooked | not hooked | not hooked |
- **`slIsFeatureSupported` was read in every Streamline session, and `slAllocateResources` in none.** Item 43
wrote the meaning of that before the run: a call no game makes after the attach gives nothing in attach mode, and
whether it keeps its patch is the owner's to say. `slFreeResources` and `slShutdown` were each read once, in one
game each; `slIsFeatureLoaded` in three games, `slSetFeatureLoaded` in two.
- **A refusal reads as one.** Phantom Blade Zero and Hogwarts Legacy ask about `kFeatureImGUI`, Streamline's
debug overlay, and are told it is missing; Cyberpunk 2077 is told the same of NIS, which Dying Light is told is
supported. Phantom Blade Zero is told DeepDVC is missing although Streamline answered its requirements.
- **The count says how often, not when.** Dying Light's one shutdown and Phantom Blade Zero's one freeing of frame
generation's resources may each be the game closing; nothing in a session says. Dying Light switched all five
features off and on.
- **The way out is as it was.** Hogwarts Legacy faulted at its exit in `_nvngx.dll`, `0xc0000005` at offset
`0x6aa85` — the offset of its six before, since 2026-11-06, hooked or not — and Black Myth: Wukong in
Steam's `crashhandler64.dll`, `0xc0000409` at `0x39b49`, the offset of its three before (Steam's file is now
11.23.65.88). Windows' log has no fault for the other four games; the two hooks on a game's way out were in all
five Streamline games.
- **Rates are where they were**, for what a comparison without the same scene is worth: Phantom Blade Zero ×4 at
64 → 255 (beta.44: 63 → 251), Wukong 85 → 341 (90 → 361), Cyberpunk 2077 164 → 327 (137 → 275), Dying Light ×4
at 48 → 194 (beta.42: 54 → 216), Kingdom Come 58.8 presented (59.3).
- **An install stops the game's threads once a patch, as a restore does.** The thirteen interposer patches took
0.74 s on Phantom Blade Zero, 1.2 s on Cyberpunk 2077, 1.7 s on Dying Light and Wukong (twelve there) and
**2.8 s on Hogwarts Legacy** — 57 to 215 ms a patch, while the game starts. That is the cost the proposal to
queue patches (§What this release decided) would cut, measured on real games.
- **No stop was exercised.** Every session ended with the game closing, so no log has an `UNHOOK_RESTORED` line.
Item 43 (g) said a log "ends with an `UNHOOK_RESTORED` line for each patch": that is true only of a session
stopped while its game runs, and the item was worded wrongly.
- **Not exercised: off means off** — Advanced metrics was on in every session. **Not visible in the data**: (c), whether *loaded* agrees with the game's own menu, and (h), the
catalog's wording on the page — both are what the owner sees.
- **For train C's spike**: Streamline says Reflex is supported in all five games and loaded in the three that
ask; Cyberpunk 2077 and Dying Light switch it off and on (`spike-notes` §30 has the note).
### NEXT
1. ~~**The owner runs item 43**~~ **Run and read 2026-11-13** (§What the owner's seven sessions on beta.45
showed): five of the six calls read in real games, nothing different on the way out, no fault. *Off means off*
was not exercised, and `slAllocateResources` was called by no game.
2. ~~**The owner reads `spike-notes` §30 and answers its three questions.**~~ **Answered 2026-11-13 (D132,
D133)**: route D goes to a spike on `hook-harness` and a stub — the owner ruled that a driver's read-only getter
called with the device of the hooked swap chain is inside rule 4 — and no real game is touched before the owner
has seen its result (`spike-notes` §31, to be written after 0.1.0-beta.46 is tagged); route C is refused; the
ETW route stays closed. No shipped code changes for the spike.
3. ~~**The stop's cost** (above): whether to batch the restores is the owner's to say.~~ **Decided 2026-11-13
(D134): batch them, installs too** — a pull request of its own, measured on the harness, released alone as
**0.1.0-beta.46**. The plan was reviewed against MinHook's source before it was approved: `MH_ApplyQueued`
applies every entry whose queued state differs, not only the caller's, so the queue-to-apply sequence is
serialised, only targets that are still ours are ever queued for a restore, and nothing is queued "for all".
4. **`slAllocateResources` keeps its patch** (D135); looked at again with more sessions.
5. **Train D** after train C, the same way: a spike first.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-12: 0.1.0-beta.45 was released 2026-11-12T16:40:42Z** (23:40 in the owner's time zone) from the signed
tag on `550d830` (#431), by run 43257803723 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry
`FL_BUILD_ID v0.1.0-beta.45`. The delta from beta.44 is 1,762,465 bytes beside a full package of 110,767,499 —
"Delta processed 0452 files. 0024 patched, 0428 unchanged". The wiki was published by run 43261335359. The
rehearsal on the release branch (run 43253780609, at `f4e3fe2`, the released tree) had made a delta of 1,779,418
bytes with the same file counts. **Qodana's two reports are green on `550d830`.** On the way Qodana's native
report found two index products in the harness's probe (#427), fixed before that pull request merged; every
later report was clean, and each was read before its merge. No required check failed on this train.
**SonarCloud's quality gate failed on `550d830`, on one condition: duplicated lines on new code, 3.7 % where the
gate allows 3 %.** No issue is open and coverage on new code is 98.6 %. The 39 lines are the six new members of
`FlHookFamily` (`ShmLayout.cs`) and of its Domain mirror `HookFamilies` — two enums that are the same on purpose
(Domain references nothing, and `MetricEnumMirrorTests` fails when they differ), which six members at once took
over a threshold that two at a time had stayed under. The job is advisory and the release was already cut. The
pull request that records this leaves the Domain's mirror out of duplication detection, with the reason in
`SonarQube.Analysis.xml` and in `13_CI_CD` §SonarCloud; **whether the gate is green again is for the analysis of
that commit on `main` to say**, and it is not in this text. *(It said so the same night: green on `d3e8dd3`,
duplicated lines on new code 0.0 %, no issue open.)* One more thing the train showed: main's branch
protection wants a pull request up to date, so of two that were green at once the second (#430) had to be
rebased after the first merged, and ran its checks again.
- ~~**No real game was run on this build.** Six new patch sites, tested with the harness and the stub; their cost
measured on the headers.~~ **Seven sessions on six games, 2026-11-13** (§What the owner's seven sessions on
beta.45 showed). The published release notes say the six calls "have not been run in a real game"; a released
file is not edited.
- **`slAllocateResources` has a patch no game has used**: not called after the attach in any of six sessions on
five games. ~~Whether it stays is the owner's.~~ It stays for now (D135, 2026-11-13).
- **Off means off (item 42 (f)) has never been exercised on a real game**; it is item 43 (f) now, for all ten.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed).
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **What an empty process snapshot does to the watcher's tick in a real session** was not looked into (seen once in
a test on 2026-11-12).
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** item 43; the three questions of `spike-notes` §30; one session each of Valor Mortis and Black
Myth: Wukong with hooking off; what item 41 left unexercised; item 40 (d).
+247
View File
@@ -0,0 +1,247 @@
# HANDOFF history — 0.1.0-beta.46
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.47 was cut, 2026-11-14 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its §The train after 0.1.0-beta.46 says the release is held**: item 44 (d) was run on 2026-11-14 and 0.1.0-beta.47 was released that day; what it built and decided is HANDOFF.md's section for it.*
## **START HERE** — 2026-11-13, 0.1.0-beta.46: patches go in and come out under one stop of the game's threads; Reflex route D goes to a harness spike
*Status is `docs/changelog/0.1.0-beta.46.md`; this is what no other file carries.* The owner ran seven sessions on
beta.45 the day after it was released and said: check the logs and go on. They were read from a copy of the ledger
and the logs (#433), and the same day the owner took four decisions on what they showed (D132–D135) and approved
the plan. Two pull requests and this one: #433 (what the sessions showed; the decisions; rule 4's note in
`CLAUDE.md`), #434 (the batches, D134). **What the seven sessions on beta.45 showed, session by session, is
[`handoff/history/0.1.0-beta.45.md`](handoff/history/0.1.0-beta.45.md)**; what the batches are, what the queue
makes dangerous and what was measured is [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Unhooking, *Batches*.
### What this release decided, beyond D132–D135
- **The first design was wrong, and reading MinHook's source before writing a line is what found it.** It marked
every hook for disabling and then un-marked the ones it must not restore. `MH_ApplyQueued` applies every hook
whose mark differs from its state; un-marking a hook that is already out marks it for *enabling*, and the stop's
own apply would have put a patch back in after the stop — logged as "declined", with the status `UNHOOKED`. The
rules that replaced it are in `fl_patch_registry.h`: no mark left standing, only a patch that is still ours
marked for a restore, nothing marked "all", and mark–apply–commit inside a gate.
- **The gate is the Overlay's one lock, and it is on no hook body's path.** A slim lock, exclusive, held for one
apply, never across a symbol lookup (the loader's lock). The one way from a hook to it is the stop, which always
waited for MinHook's own lock and a stop of the threads. Its owner re-enters it without waiting: a slim lock taken
twice by one thread never returns.
- **The verdict on an install is the engine's, not the bytes'.** Reading the target back after the apply would take
a foreign overlay that patched over us in that instant for a failure, and mark a restore nobody had compared.
- **An apply stops at its first failure and says only that one failed.** The hooks after it are enabled — in the
stop, restored — one at a time, so one bad patch still costs one patch, and a stop still takes everything out.
- **One stop of the threads costs what it cost for one patch**: 77–91 ms on the fixture for four, nine or eighteen.
What it costs in a game with a few hundred threads is not known; the `PATCH_BATCH` line's `a=` is there to say.
- **Init's three DXGI patches, the loader detour and OpenGL are as they were** — three already went in together,
the other two are one each. Ray tracing's two are a batch of their own, apart from the upscaler rows', so that a
batch's size in a test does not depend on which tick a D3D12 device appeared in.
- **`fl_guard`'s timing assertion is on the restore only** (under a second for eighteen patches): the time from an
Advanced request to its families includes the wait for the watchdog's tick, and batching changes the tick's
phase, so that number went *up* in the test while the install itself went from 0.6 s to 0.08 s. The install is
asserted by count — every lazily installed patch under a `PATCH_BATCH` line — not by time.
- ~~**The spike of route D is still to be written**~~ **The spike of route D was run the same day** (`spike-notes`
§31; `tools/fl-probe-reflex`, ctest `fl_reflex_probe_stub`): a decision table written before the runs, a stand-in
`nvapi64.dll`, then the installed driver in the probe's own process. None of §30's stop conditions fired. Two
things it decided that a plan must keep: **the call belongs on the presenting thread, once a second** — a call
the driver holds stops the thread that made it, so not the watchdog, and at a microsecond or so it needs no thread
of its own — and **the newest frames of a report are not final** (one run read a span of half an hour in the
newest frame). Rule 4's note (D132) is in `CLAUDE.md`; nothing shipped calls a driver.
### Traps this train left
- **GPG's agent had restarted and the pinentry timed out three times in a row**; nothing could be committed until
the owner unlocked the key (`echo test | gpg --clearsign` in the owner's own terminal). Say so and go on with
what needs no commit — planning did not.
- **A here-document still eats backslashes** (a `\n` inside a C++ string in an edit script): write the script to a
file. It cost one failed edit again.
- **A Catch2 `WARN` longer than the console's width wraps**, and a filter on its first words loses the number at
its end: keep a measured line short.
- **To measure the old build beside the new one**: `git stash push -- <the changed sources>`, build, run the one
test case by name (its `INFO` log prints when a new assertion fails, which it does on the old build), `git stash
pop`, build again.
- **SonarCloud's gate was red on #434 before it merged, on three things it was right about**: a `LARGE_INTEGER`
initialised with `{}` and read through `QuadPart` (cpp:S6232 - write `{.QuadPart = 0}`, as the rest of the file
does since #336), a `strlen` in a test, and a test model indexed with what a lookup returned unchecked. Its job is
advisory and auto-merge does not wait for it: the pull request was held until the fix was analysed.
- **`docs/14_TESTING` and the changelog are where a non-per-present cost goes**; "measured overhead" in a pull
request body means what the change can move — here the install and the stop, not a present.
### What the owner's five sessions on beta.46 showed (item 44, read 2026-11-13)
The owner updated at 20:29 (+07) the day of the release, ran five Streamline games with Advanced metrics on — one
session each, four to nine minutes, sessions 198–202 — and said: check and go on. Read from a copy of the ledger
and the logs, and from Windows' application log. None of the five is under the user-mode exception.
| Session | Game | Install batches in the log's order: `a=` in microseconds · patches | The interposer's patches on beta.45, first line to last |
|---|---|---|---|
| 198 | Hogwarts Legacy | **1 260 835 · 13** (Streamline) — 261 920 · 3 — 227 061 · 2 | 2.80 s, 0.19–0.31 s a patch |
| 199 | Phantom Blade Zero | 81 417 · 3 — 92 512 · 6 — 88 455 · 13 (Streamline) — 99 253 · 2 | 0.74 s and 0.85 s |
| 200 | Black Myth: Wukong | 187 931 · 12 (Streamline) — 157 010 · 3 — 109 025 · 2 | 1.65 s |
| 201 | Cyberpunk 2077 | 93 168 · 16 (Streamline's thirteen and the FSR loader's three) — **608 700 · 3** — 177 970 · 3 — 120 878 · 2 | 1.18 s |
| 202 | Clair Obscur: Expedition 33 | 168 527 · 3 — 165 405 · 16 (Streamline's thirteen and FSR frame generation's three) — 218 172 · 2 | no session on beta.45 |
- **The batches went in whole.** In every log the `HOOK_INSTALLED` lines above a `PATCH_BATCH install` carry one
time to the thousandth of a second; there is no `PATCH_FAILED` and no `UNHOOK_DECLINED`; and the hooks are the
ones beta.45 installed for the same game, compared name for name (four games; Clair Obscur had no log on beta.45).
- **Sixteen of the eighteen batches cost 81–262 ms, whatever they held** — two patches or sixteen. That is what one
patch cost in the same game on beta.45, as expected.
- **Two did not, and neither is explained.** Hogwarts Legacy's thirteen took **1.26 s** — under half the 2.8 s they
took one by one, and about five times what one stop cost in that game the same afternoon. Three of Cyberpunk
2077's took **0.61 s**, which is what three patches cost there one by one. **The log cannot tell one long stop
from an apply that failed part-way and was finished one patch at a time**: a batch writes a line only for a patch
that fails for good (item 44's text said a fallback leaves a `PATCH_FAILED`; that was wrong). What speaks for one
long stop: the queued apply fails only where a page's protection cannot be changed, and thirteen single stops in
Hogwarts Legacy would have cost 2.5 s or more. *A reading, not a measurement*: a stop waits until every thread of
the game has taken its suspension, and how long that takes follows what the game is doing at that instant — both
happened while the game was loading — and not how many patches the stop covers.
- **What that means for a ceiling on a batch** (item 44's own words for a second or more, written before the run):
the eighteen batches say the cost does not follow the count — sixteen patches in 93 ms, three in 609 ms — so
splitting a batch would add stops, not shorten one. What is missing is the line that says how a batch ended.
- **Nothing else moved.** All five hooked, `normal`, no fault counted, no record dropped. Rates and factor beside
each game's earlier sessions: Cyberpunk 2077 149 → 298 (×2.0) beside 164 → 327 and 137 → 275; Hogwarts Legacy
61 → 245 (×4.0) beside 59 → 236; Black Myth: Wukong 78 → 315 (×4.0) beside 85 → 341; Phantom Blade Zero 87 → 261
(×3.0) beside 81 → 244 at ×3.0 on beta.43; Clair Obscur 112 → 339 (×3.0) beside 106 → 323 on beta.44.
- **The way out is as it was.** Hogwarts Legacy faults in `_nvngx.dll` at `0x6aa85` and Black Myth: Wukong in
Steam's `crashhandler64.dll` at `0x39b49` — the module and the offset of every build read so far. The other
three games leave nothing in Windows' application log.
- **The Advanced rows are as on beta.45**, and Clair Obscur gave its first reading of the six calls (interposer
2.7.30): `slIsFeatureSupported` and `slIsFeatureLoaded` read, among them a feature no session had named before,
`kFeatureLatewarp` (Streamline's answer: `eErrorFeatureMissing`). `slAllocateResources` and `slShutdown` were
called in none of the five (D135 stands). That game's `hook_crash_count` of 3 is old: it dates from 2026-10-20
and is the same in the copy read for item 43.
- **Not exercised: the stop.** Every session ended with its game exiting; no log has a `STOP` or a `PATCH_BATCH
restore` line. **`RestoreAll` has still run in no real game** — item 44 (d). Nor (e), a game with no Streamline,
nor (g), off means off: `telemetry.advanced` was on throughout. (c), how a game's start feels, is the owner's to
say. (a) held: the terms window did not show.
### The train after 0.1.0-beta.46: Reflex route D in a game, and D137 (plan approved 2026-11-13)
*Nothing of this is in a release. The release pull request is not opened before the owner has run item 44 (d).*
**Where it stands (2026-11-14):** all five are merged — #440 (PR-0), #441 (D137), #442 (the words), #443 (the
Overlay), #444 (the Agent, the App, the legal texts, owner-only item 45); `main` is `cd4f69a`. What is left is the
release, which waits for the owner: **item 44 (d) was tried on 2026-11-14 and not exercised** (the switch was on
before the games started — item 44's last paragraph). **Not verified by any test: the whole path through the real Agent** —
the injected tests stop at the shared memory, the sampler's tests start from it; no end-to-end case loads the
stand-in module under an Agent session. The Japanese strings are new and unreviewed, as every `ja` string is.
The owner approved the plan the evening item 44 was read. Its order: **PR-0** the decisions and the probe's leg
with a sleep (done: `spike-notes` §31 — no stop condition) → **PR-1** D137 → **PR-2** the words in region 5, written
by nobody → **PR-3** the Overlay, the gates, the fixtures → **PR-4** the Agent, the App, strings, documents, legal
texts and owner-only item 45 → **the release, held**. What the plan's review against the source decided — a draft
of mine was wrong in six places — and that no other file carries yet:
- **Resolve on the watchdog, once for the process; call on the presenting thread.** The watchdog asks Windows for
an `nvapi64.dll` that is *already loaded* (never loads it), only while a session asks for Advanced metrics and a
Direct3D 12 device is held; it keeps one reference and never releases it; it refuses a module in the executable's
own folder (D140); it fetches `nvapi_QueryInterface` and the two ids under a guard and checks with `VirtualQuery`
that both addresses lie inside ~~that module~~ an image the process has loaded. "No getter" and "refused" are
final for the process; "no module" is asked again each second. **Found when the header was run against the
installed driver: `nvapi64.dll` is a shim and the two getters are in `nvapi64_impl.dll`** (`spike-notes` §31,
*What moving the sequence into the Overlay found*) — so the check accepts another loaded image, a reference is
kept on that image as well, and the code that is called *is* a Control Flow Guard image (the plan's review had
reasoned from the shim, which is not).
- **One atomic gate with four states** — idle, armed, reading, off — not a busy flag beside accumulators (two
presenting threads could read back to back, and the watchdog could drain half a read). The watchdog drains only
when it sees *idle* and arms at most once a second by its own clock (the loader detour wakes it more often); a
present does one relaxed load, takes *armed* to *reading*, reads into a static buffer (the report is 15.4 KB: never
on a game thread's stack), and stores *idle* or *off*. A request that goes off takes *armed* back.
- **After the record is published, and only for a swap chain whose device is the one held** (`g_d3d12Device` is the
first device seen, not "this chain's": a flag in the chain's slot, set in `ResolveApi`).
- **The latches, each until the game exits**: a fault in a getter (its own guard with `FlFilter`; counted once
through `NoteFault`, so the session shows it and it cannot reach the three that self-disable); one call of 2 ms or
more (D141); the watchdog finding *reading* on two ticks running — logged and flushed at once, published as
stalled, and **nothing can stop a call that does not return**. An error status or an empty report is recorded and
reading goes on.
- **What D141 costs, measured once on a busy PC** (`spike-notes` §31, *A run on a busy PC*): the 2 ms latch will
sometimes fire on a pause that was Windows' and not the driver's — a few sessions in a hundred at worst, by a
reading of one run. The owner chose one call over two knowing it errs towards the game; item 45's outcomes must
say how the two are told apart (the microseconds published with *slow*). **Asked again with this run in hand,
the owner kept it: one call (2026-11-13). Do not re-ask.**
- **The summary**: only frames newer than the last one summed; the two newest of a report skipped (§31: not final);
a zero end point is "span not reported" (published as all ones), not a refused frame; a frame is refused for a
span that runs backwards, one over a second, or no total; frame ids that go backwards restart the sequence.
- **Layout 5 stays**: the ten reserved words of region 5's first line take it (`slShutdowns` is the precedent; the
handshake compares the build id anyway, and a trial must be removable without a layout 7). The sleep interval is
not stored. No new request word (D138). `FL_HOOK_REFLEX`, `FL_MEASURED_LATENCY` and `reflexLatencyUs` stay unset:
there is no hook, and the numbers are a series beside the frames; `sessions.reflex_active` stays NULL.
- **Managed side**: a new reach, `GameDriver`, not a relaxed test of `Game` (whose meaning is "the arguments of a
call the Overlay hooks"); `nvapi.reflex.frame_report` as seven series, `nvapi.reflex.sleep_status` as a table of
switches *seen on / seen off*; `nvapi.reflex.mode`, `sl.reflex.state` and `sl.pcl.markers` become *refused* with
the reason; `InGameAdvancedSampler.Take` must read the latency state when the mask is zero and keep the new bit
out of `SawAnything`.
- **The gates**: `overlay-fields-check` learns the two NVAPI structs (`frameID` joins the shared members; `version`
is written outside the header); `nvapi-readonly-check` gets a pass over the Overlay — no `NvAPI_*(` token, every
id that equals one of `nvapi_interface.h`'s on a list of exactly two, `Sleep` a writer verb; `hookinventory-check`
refuses an `nvapi` import. The injected tests prove nothing about Control Flow Guard (the harness does not enforce
it): the probe includes the header that ships, so `fl_reflex_probe_stub` tests the shipped sequence under it.
- **The legal texts change in PR-4** (Disclaimer 2.16, Privacy Policy 2.20, the accuracy statement): the part
inside the game now *makes a call of its own*, once a second, on the game's presenting thread.
### NEXT
1. ~~**The owner runs item 44**~~ **Read 2026-11-13** (§What the owner's five sessions on beta.46 showed): the
install half held in five games. **Still to run, the owner's: (d) a stop while a game runs** — *Disable all
hooking* with a game being recorded — which is the half of beta.46 no real game has exercised; (e) a game with
no Streamline; (g) off means off. Not first on a game under the user-mode exception's trial (D38).
~~**Proposed, not done: `PATCH_BATCH` says how the batch ended**~~ **Decided the same evening (D137):
`PATCH_BATCH` says how the batch ended** — under one apply, or finished one patch at a time, and how many — so
that a long `a=` can be read. It goes with the next release, not in one of its own. **Written 2026-11-13**:
`install` / `restore` for one apply, `install singly` / `restore singly` with the count in the high half of `b`
(`17_HOOK_ENGINE` §Unhooking, *Batches*). It explains nothing about the two long batches already seen; it makes
the next one readable.
2. ~~**The spike of Reflex route D on `hook-harness`**~~ **Done 2026-11-13** (`spike-notes` §31): on the harness
and against the installed driver the two getters are harmless, cost about a microsecond and return a readable
report, and the Direct3D 12 device the Overlay holds is what the driver takes. **The owner's next word is go or no
for one real title** — a go is a plan of its own before any code (where the numbers go, the Disclaimer's and the
Privacy Policy's wording, `nvapi-readonly-check` over the Overlay). Routes C and E are closed (D133).
**The owner said go on 2026-11-13 (D136): a plan first, for Cyberpunk 2077.** No code before the plan is
approved, and no release that carries it before item 44 (d) — the stop while a game runs — has been run.
**The plan was approved the same evening, with D138–D141** (§The train after 0.1.0-beta.46).
3. **`slAllocateResources`** keeps its patch (D135); looked at again when there are more sessions.
4. **Train D** after train C, the same way: a spike first — and §30 found its hook would miss a game that fetched
its functions before the attach.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-13: 0.1.0-beta.46 was released 2026-11-13T11:10:28Z** (18:10 in the owner's time zone) from the signed
tag on `47f010a` (#435), by run 43412434074 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry
`FL_BUILD_ID v0.1.0-beta.46`. The delta from beta.45 is 634,023 bytes beside a full package of 110,767,120 —
"Delta processed 0452 files. 0021 patched, 0431 unchanged". The wiki was published by run 43414965433. The
rehearsal on the release branch (run 43409330961, at `7b4ecd7`, the released tree) had made a delta of 668,012
bytes with the same file counts. **SonarCloud answered on `47f010a`: the quality gate passed and no issue is open
on `main`. Qodana's two reports are green on that commit.** On the way SonarCloud was red on #434 (six issues
in the new code, fixed before it merged — §Traps), and no required check failed.
- **A local gate once counted `fl_guard` as passed after 0.66 s** (2026-11-13 17:30, the spike's branch). The
security software on the development PC was switched on for three minutes while the gate ran; the freshly linked
`fl_guard_test.exe` was removed from the build tree, **ctest reported the run that was cut short as `Passed`**
(the suite takes 181 s), and the gate went red only because the next test, `fl_prescan`, runs the same executable
and could not find it. Run again with that software off, the gate was green. Nothing was changed for it: a test
whose process ends early with a zero exit code is a pass to ctest, and the suite has no line it must print to
count. ~~Worth a guard of its own (a `PASS_REGULAR_EXPRESSION` on Catch2's last line, or a minimum number of test
cases); proposed, not done.~~ **Done the same evening, on the owner's word**: every native test names the line it
ends on and ctest requires it; a test registered without one fails the configure; four canaries prove the rule
red and green (`14_TESTING` §Native unit tests; `src/native/CMakeLists.txt`). Two things it turned up: four tests
passed on a line they print in the middle of their run, and `fl_nvapi_probe` passed whatever it printed after its
branch line — a pass expression makes ctest ignore the exit code.
- ~~**No real game was run on this build.**~~ Five were, on 2026-11-13 (above): the order in which patches reach a
game changed for every game, and the five took it. **The restore under one apply has run in no real game.**
- **Two batches took 1.26 s and 0.61 s and nothing explains them** (Hogwarts Legacy, Cyberpunk 2077); the log has
no line that tells a long stop from a batch finished one patch at a time.
- **Off means off has never been exercised on a real game** (item 43 (f), now item 44 (g)); nor (c) and (h) of
item 43, which only the owner sees.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed).
- **For the inline patches the comparison has never been re-made under the write window** (`17_HOOK_ENGINE`
§Compare-and-restore says so now): MinHook owns that window.
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **What an empty process snapshot does to the watcher's tick in a real session** was not looked into (seen once in
a test on 2026-11-12).
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** item 44 (d), (e), (g) and what (c) felt like — **(d) is what the next release waits for**; item 45 once that release exists; the go or no on `spike-notes` §31 (one real title); one session each of Valor Mortis and
Black Myth: Wukong with hooking off; what item 41 left unexercised; item 40 (d).
+146
View File
@@ -0,0 +1,146 @@
# HANDOFF history — 0.1.0-beta.47
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.48 was cut, 2026-11-14 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its NEXT and its "Left open" were written through the morning of 2026-11-14 as the owner's sessions were read**: what they led to is HANDOFF.md's section for 0.1.0-beta.48.*
## **START HERE** — 2026-11-14, 0.1.0-beta.47: the part inside a game asks NVIDIA's driver for its latency report; the batch log says how a batch ended
*Status is `docs/changelog/0.1.0-beta.47.md`; this is what no other file carries.* The owner ran five sessions on
beta.46 the evening it was released and said: check and go on. They were read (#439), the owner took six decisions
(D136–D141) and approved the plan of the train the same evening, and the five pull requests of it were merged
overnight: #440 (the decisions; the probe's leg with a sleep mode), #441 (D137), #442 (the latency words of region
5), #443 (the Overlay makes the read; the gates; the fixtures), #444 (the Agent, the App, the legal texts,
owner-only item 45). The release was held for item 44 (d), which the owner ran on 2026-11-14 (#445, and below).
**What the five sessions showed, the design of the train as the plan's review left it, and the traps of beta.46
are [`handoff/history/0.1.0-beta.46.md`](handoff/history/0.1.0-beta.46.md)**; what the read is and what bounds it
is [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Calls the Overlay makes into a driver; what was measured is
[`spike-notes/31-reflex-route-d-harness.md`](spike-notes/31-reflex-route-d-harness.md).
### What this release decided, beyond D136–D141
- **It is the first thing the Overlay does in a game that is not reading a hooked call's arguments**, and it is
listed as that — its own table in `17_HOOK_ENGINE`, its own reach in the catalog (`GameDriver`, not a stretched
`Game`), its own paragraph in the Disclaimer. A second call of this kind needs a ruling of the same kind (D132).
- **Resolve on the watchdog, once for the process; call on the presenting thread, after the record is published,
behind one gate of four states** — idle, armed, reading, off. The plan's first draft had a busy flag beside
accumulators: two presenting threads could have read back to back, and the watchdog could have published half a
read. The whole design as reviewed is in the history file, §The train after 0.1.0-beta.46.
- **`nvapi64.dll` is a shim.** The check that the two addresses lie inside the module it asked refused the real
driver: the getters are in `nvapi64_impl.dll`, in the driver's store. An address is accepted when it is executable
code of a loaded image, and a reference is kept on that image too. Only the run against the installed driver
could find it — the stand-in answers with addresses of its own.
- **The code that is called is a Control Flow Guard image**; the plan's review had reasoned from the shim, which is
not. The probe enforces CFG and makes the calls from the header that ships, so this driver's getters are valid
targets. A CFG-enforcing game on another driver is not covered, and a refusal there is a fast-fail.
- **The layout stayed 5**: ten reserved words of region 5 took the read. `FL_HOOK_REFLEX`, `FL_MEASURED_LATENCY`
and `reflexLatencyUs` stay unset — there is no hook, and the numbers are series beside the frames.
- **D141 was asked twice.** A run on a busy PC showed the 2 ms latch can fire on a pause that was Windows'; the
owner kept one call. Do not re-ask without a real game's numbers.
- **What the game asked of Reflex, and Reflex through Streamline, are `RefusedRoute`** — a patch at a run-time
address, and a call the vendor forbids from another thread (D133). `sl.pcl.markers` waits for train D's spike.
### What item 44's last part showed (2026-11-14)
- **A first try did not exercise it** (sessions 203–205, #445): *Disable all hooking* was on before the games were
started, so two sessions were `RefusedKillSwitch` with nothing injected, and the hooked one ran to its exit.
- **The second try did, twice, in Phantom Blade Zero** (sessions 206 and 207): the game started with the switch
off, was recorded for a minute and a half, and the switch was turned on while it ran. Each log has `STOP` at the
next safety scan, **an `UNHOOK_RESTORED` line for every one of the twenty-nine patches, all at one time**, one
`PATCH_BATCH restore` — **83 309 µs and 85 578 µs for twenty-nine** — and `STOP_FLUSHED … a=1`. Nothing was
declined. Both sessions ended `KillSwitchEngaged`, tier 1, no fault, no record dropped, writer status
*unhooked*; the game ran on for three and four minutes and closed with nothing in Windows' application log.
Restoring twenty-nine one by one would have cost about 2.4 s at this game's 84 ms a stop.
- **Not said by the logs**: whether the owner felt a hitch at the stop. Not exercised: (e) a game with no
Streamline and (g) off means off — both are parts of item 45 now.
### Traps this train left
- **A bash here-document ate backslashes again** and wrote a NUL byte into a `.cpp`. Edit scripts are files.
- **A file made by the Write tool is LF**; `dotnet format --verify-no-changes` fails a new `.cs` for it. Convert.
- **`NvAPI_Status(__cdecl*)(…)` is a call by name to `nvapi-readonly-check`**: the status type has an alias in
`fl_reflex.h`. And a member of the Overlay's own named like a vendor's (`fgMultiplier`) fails
`overlay-fields-check` outside the header.
- **A copy of the harness in `%TEMP%` leaves an overlay log the run's sweep does not own** (it matches this
build's harness by path): the test that makes the copy sweeps it.
- **Stacked pull requests, with a dirty tree**: once the lower one is squash-merged GitHub retargets the next to
`main`; `gh pr update-branch` works while the upper branch only adds, and conflicts once it edited lines the
lower one added — then `git rebase -S --onto origin/main <lower tip> <branch>` and a forced push with lease.
- **Advisory checks found real things after the local gate was green**: SonarCloud a `wcslen`, Qodana's native
report a `switch` that did not name two enumerators, its managed report two redundant qualifiers and a
redundant `!`. Hold a merge until both have reported.
- **The probe's numbers depend on what else the PC does**: a run beside a build read 3.4 µs where three alone read
2.6, and saw reads of 2 ms. Say which kind a number is.
### NEXT
1. ~~**The owner runs item 45**~~ **Run on 2026-11-14 in Phantom Blade Zero, three sessions: the read works in a
real game** (item 45's last paragraphs) — the driver was found, read every second, and no call took more than
45 µs. **Cyberpunk 2077 followed the same morning, twice, and three games the read must leave alone** (sessions
211–215): read, 65 µs at the slowest, and *not read* where it should be. **Still to run of it**: Reflex off in
a game's menu, off means off; and the owner's word on whether the games felt different.
2. **The owner answered on the *GPU render* line (2026-11-14): read the driver's active time.** Two more members
of the report FrameLedger already fetches — no further call into the driver. ~~**Plan first**~~ **Planned and
approved the same day as the train of 0.1.0-beta.48 (D142–D147)**: the native half first, then the Agent, the
App and the texts, then the sampling period, then the release and owner-only item 46 — one batch of tests for
the owner, with what is left of item 45 in it. What the plan gathered: the multiplier row that keeps only the last value, the pages that still say no real game
was read, the Agent's 1.32 s between samples beside `hz` 1.0. The owner tests what is left once, on that
release. Still the owner's to say: more titles; the session summary's latency fields; D141 (the slowest call
seen in a game is 65 µs, thirty times under it).
3. **`slAllocateResources`** keeps its patch (D135); looked at again when there are more sessions.
4. **Train D** after train C, the same way: a spike first — `sl.pcl.markers` waits for it, and §30 found its hook
would miss a game that fetched its functions before the attach.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-14: 0.1.0-beta.47 was released 2026-11-14T01:11:19Z** (08:11 in the owner's time zone) from the signed
tag on `3d7c273` (#446), by run 43537436561 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed; the four shipped natives carry
`FL_BUILD_ID v0.1.0-beta.47`. The delta from beta.46 is 1 820 144 bytes beside a full package of 110 795 507 —
"Delta processed 0452 files. 0024 patched, 0428 unchanged". The wiki was published by run 43539059710. The
rehearsal on the release branch (run 43534606602, at `43426fe`, the released tree) had made a delta of 1 838 295
bytes with the same file counts. **SonarCloud answered on `3d7c273`: the quality gate passed and no issue is open
on `main`. Qodana's two reports, CodeQL and `check` are green on that commit.** On the way the advisory checks
found three things after local gates were green (§Traps), and no required check failed.
- **The release branch's local gate was red once, for a reason outside the code** (2026-11-14 07:20): the freshly
linked `fl_guard_test.exe` was gone from the build tree seconds after the gate began — Windows Defender's log has
three configuration changes at 07:20:53–56 and no detection; which security program removed the file is not
known, and nothing of its settings was looked at further or touched. `fl_guard` ended after 0.72 s **and was
counted as failed, not passed**: the rule of #438 (a native test names the line it ends on) did what it was
written for. Run again, the gate was green.
- ~~**No real game has run the read.**~~ Two have (Phantom Blade Zero and Cyberpunk 2077, 2026-11-14). **The accuracy
statement, `LIMITATIONS.md` and the guide still say no real game had been read**: true when written, to be
brought up to date with the next release, as beta.46's pages were.
- **In Phantom Blade Zero the driver's *GPU render* span is longer than its total** (item 45): shown as reported,
not explained. In Cyberpunk 2077 it is not. The owner's answer: read the driver's active time (NEXT 2).
- ~~**The Agent's samples are 1.32 s apart while `sensor_blobs.hz` says 1.0**~~ (seen 2026-11-14 in the `t_ms`
series). **Looked into the same day (D146)**: the poller waited its whole interval after a poll's work; only
the App's series loader reads `hz`, and only where a row has no `t_ms`, so no chart was wrong. The poller now
keeps a fixed period and `hz` is measured from the samples (`06_DATA_MODEL` §sensor_blobs) — in no release yet.
- **Where the third of a second goes is not established.** Unelevated, a poll takes about a millisecond (the
layers composed as the Agent composes them, each read under a stopwatch). In the owner's ledger 180 of 186
sessions are 1.32 s apart and six are 1.0 s: the two with no `cpu_temp` series (an Agent not elevated) are among
the six — which points at the sensor library's processor update, run on the poll's thread as administrator
with PawnIO — **but four sessions that have the series kept one second** (22, 72, 73, 136). **Not measured**:
that call itself (it needs elevation), and whether it is processor time or waiting. Not changed; the period no
longer depends on it. The owner's to say whether it is worth timing on their PC and moving off the poll's thread.
- **No test runs the whole path through the real Agent**: the injected tests stop at the shared memory and the
sampler's start from it; no end-to-end case loads the stand-in module under an Agent session.
- **Two install batches of beta.46 took 0.61 s and 1.26 s and nothing explains them** (Hogwarts Legacy, Cyberpunk
2077). Since this release the log says whether a batch went singly; the next long one can be read.
- **The Japanese strings are new and unreviewed**, as every `ja` string is.
- **Off means off has never been exercised on a real game** (item 43 (f), item 44 (g), now item 45 (g)); nor (c)
and (h) of item 43, which only the owner sees.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed).
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** what is left of item 45 (Reflex off in a game's menu; off means off; how it felt); one session
each of Valor Mortis and Black Myth: Wukong with hooking off; what item 41 left unexercised;
item 40 (d).
+115
View File
@@ -0,0 +1,115 @@
# HANDOFF history — 0.1.0-beta.48
*Moved verbatim from `docs/HANDOFF.md` when 0.1.0-beta.49 was cut, 2026-11-14 (owner decision D66); `docs/HANDOFF.md` §How this file is split lists every part. Nothing below was edited, so its "below" and "above" mean HANDOFF.md as it stood then, and a path it writes is relative to `docs/`. **Its NEXT and its "Left open" were written through the afternoon of 2026-11-14 as item 46 was read**: what they led to is HANDOFF.md's section for 0.1.0-beta.49.*
## **START HERE** — 2026-11-14, 0.1.0-beta.48: two real games read; the driver's own GPU times beside a span that includes idle; every multiplier seen; samples a second apart
*Status is `docs/changelog/0.1.0-beta.48.md`; this is what no other file carries.* The owner ran eight sessions on
beta.47 the morning it was released and said, twice: check. They were read (#448, #449): Phantom Blade Zero three
times, Cyberpunk 2077 twice, two Direct3D 11 games and a Direct3D 12 game with no NVIDIA library — **the read
works in real games**, and four things came out of it that the owner ruled on the same morning (D142–D147) and
approved as one train: #450 (the Overlay reads two more members of the report and keeps the multipliers seen),
#451 (the Agent, the App, the texts — Privacy Policy 2.21, Disclaimer 2.17), #452 (the poller's period), and this
release. **What the eight sessions showed is item 45's last paragraphs, below; beta.47's own section, its traps
and its release facts are [`handoff/history/0.1.0-beta.47.md`](handoff/history/0.1.0-beta.47.md)**; what the
read is and what bounds it is [`17_HOOK_ENGINE.md`](17_HOOK_ENGINE.md) §Calls the Overlay makes into a driver.
### What this release decided, beyond D142–D147
- **Two more members of a report already fetched are not a new call.** Rule 4's note stands as written: the same
two getters, the Overlay's own buffer, the two names on `tools/overlay-allowed-fields.txt`. The owner ruled on
them by name (D142); a third member is asked the same way.
- **The driver's own times have a count each, and zero means not given.** An older driver leaves both zero, and
so did the probe's frames. They are added after the frame is accepted, so which frames the seven spans are made
of did not move — a session's old series mean what they meant.
- **An entry of their own in the catalog** (`nvapi.reflex.gpu_times`), not an eighth and ninth line of the frame
report: `AdvancedPresenter.MaxLines` is 8, and nine series of one entry are drawn as a band.
- **The layout stayed 5 again**: line 0 of region 5 was full, so the four words came out of the seven reserved
words that ended the `sl::DLSSOptions` line. Three remain there, two on the `sl::Constants` line, two on the
FidelityFX line.
- **The fact's key no longer holds the last multiplier.** It held it since beta.47, so every menu rewrote the
session's context.
- **`hz` is measured from the samples, not taken from the setting**: the finalizer cannot see the setting, and a
session recovered from a `.partial` file never knew it.
- **The plan named `tools/agent-cost` for the poller's change and it was not run**: it runs the Agent unelevated,
where a poll takes a millisecond and the period was already a second. Said in #452 rather than run for show.
### Traps this train left
- **A cause I wrote down before checking it.** "The third of a second is the processor's temperature" went into
five files on two facts — unelevated a poll takes a millisecond; the sessions without a `cpu_temp` series are a
second apart — and then four sessions *with* the series turned out to be a second apart too. The texts were
taken back to what is established before the commit. **Count the exceptions before naming a cause.**
- **A bash here-document ate a backslash again** (a Python one-liner with `'\\'`). Edit scripts are files — the
rule already existed.
- **`FETCH_HEAD` belongs to the worktree that fetched.** A merge in a second worktree needs the commit's id.
- **A squash-merge of the lower pull request conflicts the upper one in `CHANGELOG.md` every time** both added to
`[Unreleased]`. With Auto-fix on, the app asks for a merge of `main`, not a rebase: do it in a worktree of its
own when a gate is running in the main one, keep the upper branch's file (it is the superset), and check that
the merged tree equals the branch's tip — then the gate already run still stands.
- **The gate script's last lines do not reach a background task's output**: read the log's last line and hash the
`rules` folder by hand.
- **An analyzer wants a `<summary>` once any `<param>` is written** (RCS1139).
### NEXT
1. ~~**The owner runs item 46**~~ **Run and read 2026-11-14** (the item, below): the samples are a second apart,
both games gave both figures, the multipliers are listed, off means off. **The owner's word, the same day: no
hitch.** What is left of it is whether session 221 was the one with Reflex off.
2. **Then the owner's to say**: more titles; D141 on a larger sample (the slowest call seen in a game is 65 µs);
whether the processor update is worth timing on their PC and moving off the poll's thread (below).
3. **`slAllocateResources`** keeps its patch (D135); looked at again when there are more sessions.
4. **Train D** after it, the same way: a spike first — `sl.pcl.markers` waits for it, and §30 found its hook
would miss a game that fetched its functions before the attach.
### Left open by this session
- ~~**This release's own facts** — the tag's run, the digests, the delta — are not in any file yet: the release is cut
after this text. Record them in the next pull request, with what SonarCloud and Qodana said.~~ **Recorded
2026-11-14: 0.1.0-beta.48 was released 2026-11-14T04:49:19Z** (11:49 in the owner's time zone) from the signed
tag on `aec1736` (#453), by run 43556901499 — green at the first attempt, one run for the one tag push. Eight
assets; the seven digests of `SHA256SUMS.txt` equal the ones GitHub computed, and the portable package hashed
here equals its line; the five native libraries in it carry `FL_BUILD_ID v0.1.0-beta.48`, none of them
`-dirty`. The delta from beta.47 is 1 944 037 bytes beside a full package of 110 800 248 — "Delta processed
0452 files. 0024 patched, 0428 unchanged". The wiki was published by run 43558546949. The rehearsal on the
release branch (run 43555020289, at `0c210d0`, the released tree) had made a delta of 1 966 174 bytes with the
same file counts. **SonarCloud answered on `aec1736`: the quality gate passed and no issue is open on `main`. Qodana's two reports, CodeQL and `check` are green on that
commit**, as they were on each of #450–#453 before its merge; no advisory check found anything in this train.
- ~~**No real game has given the two new figures yet.**~~ **Two have (item 46, 2026-11-14)**, in every frame
report: Phantom Blade Zero's active render is the short number the reading expected (9.6 ms inside a span of
16.4 ms at 60 FPS). **In Cyberpunk 2077 without frame generation active render is longer than the span between
the two GPU marks** (3.17 against 2.79 ms) **and why is not established** — the two averages are over different
frames, or the driver counts work outside the marks. The label says *idle time included*, which that game's
numbers do not show; the texts say the figures are the driver's own. Not changed.
- **After a session's first report, 0.4–7.3 % of its samples carry none** (item 46): the Overlay's reads are a
second apart or two — its watchdog's tick — and the Agent's samples exactly one. Read in the code, not measured;
not changed.
- **Where the third of a second of a poll goes is not established.** Unelevated, a poll takes about a millisecond
(the layers composed as the Agent composes them, each read under a stopwatch). In the owner's ledger 180 of 186
sessions are 1.32 s apart and six are 1.0 s: the two with no `cpu_temp` series (an Agent not elevated) are among
the six — which points at the sensor library's processor update, run on the poll's thread as administrator
with PawnIO — **but four sessions that have the series kept one second** (22, 72, 73, 136). **Not measured**:
that call itself (it needs elevation), and whether it is processor time or waiting. The period no longer
depends on it; a poll that takes a third of every second is still a poll that takes a third of every second.
- **With the sensor interval above one second, some readings of the driver's report are not stored**: the newest
stands until a sample carries it (`SessionRecorder`). Said in the guide; not changed.
- **No test runs the whole path through the real Agent**: the injected tests stop at the shared memory and the
sampler's start from it; no end-to-end case loads the stand-in module under an Agent session.
- **Two install batches of beta.46 took 0.61 s and 1.26 s and nothing explains them.** Cyberpunk 2077's eight
batches on beta.47 took 85–221 ms, each in one apply; Hogwarts Legacy has not been run since.
- **The Japanese strings are new and unreviewed**, as every `ja` string is.
- ~~**Off means off has never been exercised on a real game**~~ **It was, 2026-11-14** (item 46 (e), session
226: no `LATENCY` line, none of the hooks Advanced metrics adds, no row). (c) and (h) of item 43, which only the
owner sees, are still not.
- **Valor Mortis's crash at 51 s** (twice, beta.40 and beta.44) **and Black Myth: Wukong's fault in Steam's crash
handler at every exit** are not explained. One session of each with hooking off is the comparison that is missing.
- **What item 41 left unexercised**: a configure through the FidelityFX 2.3 loader (Cyberpunk 2077 with FSR frame
generation), a game with neither library.
- **A fault while a game is closing is not seen by a hooked session** (D127: written down, not changed).
- **`docs/LIST_GAME_TESTED.md` is as of 2026-11-06**: refreshing it is a new copy of the ledger and the owner's review.
- **Item 40 (d)** was not exercised and the owner has not said item 40 passed. **Items 36 (d) and 37 are still open.**
- **`LIMITATIONS.md` and the accuracy statement still say the processor's, the memory's and the motherboard's
sensors "have never been read on real hardware"** — the owner's to reword. The alpha line still rests (D98).
**Owner-only:** one session each of Valor Mortis and Black Myth: Wukong with hooking off; what item 41 left
unexercised; item 40 (d).
+38 -2
View File
@@ -1,6 +1,6 @@
# Owner-only items that passed
*Moved verbatim from `docs/HANDOFF.md` §Owner-only on 2026-10-19 (beta.24, owner decision D66): the items the owner reported passed or took, numbers kept — a number is never reused. The open ones stay in `docs/HANDOFF.md`.*
*Moved verbatim from `docs/HANDOFF.md` §Owner-only on 2026-10-19 (beta.24, owner decision D66): the items the owner reported passed or took, numbers kept — a number is never reused. The open ones stay in `docs/HANDOFF.md`. Items that passed later are added below as they pass (38 and 39 on 2026-11-12).*
3. ~~**Which titles** go in `blockedExecutables` — a product decision with false-refusal
consequences. The list ships empty until it is taken.~~ **Taken 2026-09-25 (D30):** per-title
@@ -66,4 +66,40 @@
page and choose its longest session, switch the page's tabs; visit the Dashboard a few times, then end a session —
one "Session saved". (e) The Privacy Policy 2.11 is shown once after the update. (f) The `ja` strings marked
`review`.
38. ~~**beta.40, Advanced metrics inside a real game**~~ **Passed 2026-11-12** — the owner ran it on ten games over one night (beta.40, then beta.41), was shown what a copy of the ledger and the Agent's log said of those sessions, and answered: "Bạn có thể đánh dấu là pass và tiếp tục." What the sessions covered: (b) the option off storing nothing, on a DLSS title and on an FSR 3.1 one; (c) three titles upscaling through FSR 3.1 — Kingdom Come: Deliverance II (chosen in the game), Dying Light: The Beast and Clair Obscur: Expedition 33, the last two with **FSR upscaling beside DLSS frame generation, which is how the owner had set them in the game** — and Lies of P through FSR 3; (d) every DLSS title saying *the game made none of the calls*, none chaining its options into the evaluation; (e) the sentence under the session in each of its three forms; (g) Girls' Frontline 2 under the user-mode exception. **Not exercised, and passed on the owner's word**: (f) two sessions of one game that kept running; (h) the cost; (i) the words. **(j) is not done**: the SonarCloud acceptance is still the owner's to click, and stands in HANDOFF §Left open. Valor Mortis's one crash was the game's own (a shader-compilation error, its own crash dialog — the owner). *(The item as written:)* (D111–D116; train B's first release — beta.41 is not begun
until this passes). (a) **The update from beta.39**: the terms window shows the **Privacy Policy 2.16** once, and
not the Disclaimer; a game that was **running across the update** is refused until it is restarted, with the
reason, and measured normally after. (b) **The option off, a hooked game that uses DLSS or FSR**: the game runs
and is measured as in beta.39; the session is not on the Advanced page's list; nothing in the Agent's or the
Overlay's log mentions the upscaler's options. (c) **The option on, a title that upscales through FSR 3.1 or
later** (AMD's `amd_fidelityfx_*.dll` in its folder; a title on the older FSR 3.0 interface, such as one that
ships `ffx_fsr3_x64.dll` alone, is expected to show none): on *Session*, under Upscaling, sharpening on or off
and its strength against the game's own setting; the field of view as a range, against the game's slider (it is
the **vertical** angle, in radians with degrees beside it) — and wider after zooming or aiming; the near and far
planes; the flags; the chart of the frame time the game reported beside the session's own frame time; FSR's
memory only if the game asks for it. (d) **The option on, a DLSS title**: most are expected to say *the game made
none of the calls these are read from*. If one shows rows, this is the first sight of them anywhere: the output
size against the display resolution, the preset letters against the NVIDIA app's override or the DLSS
indicator, HDR input against the game's HDR setting. Write down which title did. (e) **The sentence under the
session** says which of the three it was, and the session's summary still names the upscaler and its quality as
before. (f) **Two sessions of one game that kept running**, both with the option on: the second one's field-of-
view range and counts are its own, not the first one's carried over. (g) **A game under the user-mode anti-cheat
exception**, the option on: it runs, the session is hooked, and its trial is not ended by anything new. (h) **What
it costs**: the game's own frame rate with the option on and off, and the Agent's processor time in a hooked
session (NFR-2: 2.5 % of one core). (i) **The words**: the fourteen new `vi` strings and the Settings text; the
`ja` ones are marked `review`. (j) **SonarCloud**: accept `csharpsquid:S3869` on
`src/FrameLedger.Infrastructure/Diagnostics/ParentDump.cs` again — the same call the acceptance of beta.24 covers,
raised anew because #398 moved it (`13_CI_CD` §SonarCloud says why it is safe); until then `main`'s reliability
rating on new code shows E for that one line.
39. ~~**beta.41, the sensor library on the PC that showed the fault**~~ **Passed 2026-11-12** — the same answer. Seven sessions on beta.41 as administrator with PawnIO, no `advanced:` line in the Agent's log, 212 or 213 series in each session with the option on; on beta.40 the same morning the processor device faulted out after each of six sessions. **Not exercised**: (c) the page after a session was not looked at by anyone but the owner; (d) the unelevated run; (e) no value beside HWiNFO — that stays item 37 (e). The rarer fault, the reader failing to open, was one session in sixteen; `20_OPEN_QUESTIONS` §M13 keeps the count. *(The item as written:)* (D118; `20_OPEN_QUESTIONS` §M13). (a) **The
update from beta.40**: no terms window, the ledger opens as before; a game running across the update is refused
until restarted, as after every update. (b) **As administrator with PawnIO, Advanced metrics on**: five or more
sessions of different games, some with the Advanced page left open on *This PC* throughout, some with it closed,
one started within a few seconds of another ending. **Afterwards the Agent's log (Logs page, or
`agent-<date>.log`) holds no `advanced: a device is left alone` and no `advanced: the sensor library did not
open`**, and every one of those sessions has its processor, memory and motherboard readings on *Session* (about
two hundred series; a session with seventeen is the fault). (c) **The page after a session ends**: the
processor's rows on *This PC* are still there and still moving, without reopening the page. (d) **Not as
administrator**: nothing differs from beta.40 — the card's sensors and the driver's figures, and *Sources* says
why the rest is not read. (e) While there, the part of item 37 (e) nobody has done: a few of the processor's and
the motherboard's values beside HWiNFO.
+134
View File
@@ -63,6 +63,140 @@ are kept because they record what each version changed and why (owner decision D
>
> Version 2.5 (2026-09-27, 0.1.0-beta.10) adds how a tracked game's window is shown — read from Windows — and the engine build name read from an Unreal Engine game's executable. It is also the first version to say two things the software already did: it reads the **content** of a library game's executable (markers in its first 8 MB and its import table, since the detection of engines and features; and, when a session starts, the graphics-library markers in the whole file), and it reads the processor load and the memory in use (since 0.1.0-beta.3). Nothing here leaves your device, as before.
### What version 2.22 of the Privacy Policy and 2.18 of the Disclaimer changed (0.1.0-beta.50)
**Both documents said FrameLedger never reads input, and it now reads that a controller, or the keyboard or mouse,
is in use** (owner decisions D155–D163; the sentences were shown to the owner with the plan of 2026-11-14 and
approved with it). Until this version the Privacy Policy's §4 said it "does not read game saves, chat, input
content", the Disclaimer's §1 that it never reads "save files, input, chat", the README that it "never touches saves
or input", and the guide that it never reads "your input". The **Privacy Policy** (2.22) adds to §1 what a session
stores of controllers and input and what it never does — which button, stick, trigger or key, text, the mouse's
position, a serial number, a Bluetooth address; to §3A that these numbers, unlike Advanced metrics', **are** in an
export and in a bug report (D160); to §4 what is observed — the controllers' reports, looked at and not kept,
XInput's state and battery, the time of the last keyboard or mouse input, the window in front — and that nothing is
opened or sent; and narrows the sentence after it to what is still true. It also says, in §1 and §4, that nothing
of this is read until that version has been accepted on the PC (D156): the option is on unless switched off, and an
Agent that started before the App showed the document records without it. The **Disclaimer** (2.18) gains a
paragraph in §1 for what is read from outside the game about controllers, the keyboard and the mouse, and its "what
it never does" names what is kept instead of the word "input", and adds that nothing is sent to a controller
(D159). The **accuracy statement** — and so the README and the Disclaimer — gains a paragraph that ends: none of
this had been read from a real controller when it was written. The trademarks notice names Sony's and Nintendo's
products, which the App now shows as a controller's model.
### What version 2.21 of the Privacy Policy and 2.17 of the Disclaimer changed (0.1.0-beta.48)
**Two real games had been read, and the texts still said none** (owner decisions D142–D145; the wording's main
points approved with the plan of 2026-11-14). The accuracy statement — and so the README and the Disclaimer, which
carry it — said the driver's report had been "read with a test program against one driver and one card; no real
game had been read when this was written". On 2026-11-14 the owner read it in Phantom Blade Zero and Cyberpunk
2077, and the first of them showed something the statement did not say: the stages overlap, and the graphics
card's stage — the driver's start mark to its end mark — includes the card's waiting and was longer than the whole
frame. So the statement now says both, names the two figures of the driver's own that are shown beside that stage
since this release, and says a driver may give neither. The **Privacy Policy** (2.21) adds those two averages to
what a session stores, and "its frame-generation multiplier" becomes the multipliers the driver gave during the
session. The **Disclaimer** (2.17) changes nowhere else: the component makes no new call, so §2's paragraph on the
one call it makes stands as written.
### What version 2.20 of the Privacy Policy and 2.16 of the Disclaimer changed (0.1.0-beta.47, train C)
**The part inside a game makes a call of its own** (owner decisions D132, D136, D138–D141; the wording's main
points approved with the plan of 2026-11-13). Every earlier version could say that the component reads only what a
game passes to the graphics calls it intercepts. With Advanced metrics on, in a Direct3D 12 game that had loaded
NVIDIA's driver library, it now also asks the driver, about once a second, for the driver's report of the last
frames' stages and its low-latency state. So: the **Privacy Policy** (2.20) names what a session stores of it —
durations and five switches, never a time of day, never the moment of an input — in §1's item on Advanced metrics
and in §4's list of what the component observes. The **Disclaimer** (2.16) changes three places: "What it observes
inside the game" names the report; the sentence "API arguments and nothing else" is replaced by what is true; and
§2, after the paragraph on the background thread, gains the paragraph on this call — that it is made on the game's
presenting thread, through a library the game loaded and FrameLedger never loads, that two files are kept loaded
for it, that it calls nothing that sets or delays, what ends it, and **that a call which never returned could not be
stopped**. The accuracy statement loses "PC latency (Reflex)" from "not measured at all" — end-to-end latency
still is not measured — and gains the item saying these are the driver's numbers, read from no real game when it
was written.
### What version 2.19 of the Privacy Policy changed (0.1.0-beta.45)
**Six more calls read inside the game, only while Advanced metrics is on** (train B's fourth step, owner decisions
D129–D131; the wording approved with the plan of 0.1.0-beta.45). Version 2.18 named Streamline's answers to two
questions about a feature — its version, what it requires — and "four further calls". This release reads two more
answers and four things the game does, none of which is in that list, so the version moves. §1's item on Advanced
metrics gains, **for each Streamline feature the game asks about or controls: whether Streamline answers that it is
supported, whether it is loaded, whether the game switched it on or off while it ran, and how many times its
resources were allocated and freed; and whether the game shut Streamline down, and how many times**. It also says
two more things that are never read: **what identifies the graphics card that the game hands over with its
question** — `slIsFeatureSupported` takes an `sl::AdapterInfo`, the adapter's LUID or a Vulkan physical device,
which D131 forbids and a check in the build refuses by name — and **the content of a graphics resource**: the two
resource calls take a command buffer and a viewport, forwarded and never looked at. §4's "four further calls" is
ten, its "two questions" are four, and its list of what is stored names the new facts and repeats that what
identifies the card is not among them. What the game started Streamline with (`slInit`) is not in the policy
because it is not read, and will not be (D129). The accuracy block's item says what the new rows are — calls the
game made, counted to fifteen, not a statement that a feature ran. The Disclaimer stays 2.15. §2 is unchanged:
nothing is sent.
### What version 2.18 of the Privacy Policy changed (0.1.0-beta.44)
**Two more calls read inside the game, only while Advanced metrics is on** (train B's third step, owner decisions
D126–D128 beside it; the wording approved with the plan of 0.1.0-beta.44). Version 2.17 named Streamline's per-frame
constants and the frame-generation configuration, and "two further calls"; this release reads what Streamline
**answers** when the game asks about a feature, which is in neither list, so the version moves. §1's item on
Advanced metrics gains **which features the game asked about, the version of each and of its NGX model, and what
each requires** — the graphics APIs it supports, whether V-Sync must be off and hardware-accelerated GPU scheduling
on, how many processor threads, viewports and buffers, and the versions of Windows and of the graphics driver that
Streamline found beside the ones the feature needs. It also says what of that answer is never read: **the lists it
points to** — the buffers' names, the names of Vulkan extensions and features — which a check in the build holds
(the count of required buffers is read; its list is not). §4's "two further calls" is four, and its list of what is
stored names the versions and requirements. The versions of Windows and of the driver are the ones Streamline
reports to the game; FrameLedger already stores the driver's version with a session's hardware. The accuracy
block's item says what these are — Streamline's answers, not a check of the PC — and that a game asks while it
starts, so most sessions have none. The Disclaimer stays 2.15. §2 is unchanged: nothing is sent.
### What version 2.17 of the Privacy Policy changed (0.1.0-beta.43)
**Two more calls read inside the game, only while Advanced metrics is on** (train B's second step, owner decisions
D122–D125; the wording approved with the plan of 0.1.0-beta.43). Version 2.16 named what a session stores of DLSS and
of FSR's upscaling item by item, and neither list holds what this release reads, so the version moves rather than
the words being stretched. §1's item on Advanced metrics gains **what the game told NVIDIA Streamline about each
frame** — the size of the jitter, the motion-vector scale, the depth range and whether depth is inverted, the lowest
and highest field of view, the aspect ratio, whether the projection is orthographic, how its motion vectors are
stored and how often it reset the upscaler's history — and **how it configured FSR frame generation** — on or off,
whether it may run asynchronously, its flags, whether only generated frames are presented, and the rectangle it
generates in. The sentence on what is never read (where the camera is, which way it looks, its matrices) stands as
it was: those sit in the same data as Streamline's constants, and a check in the build refuses code that names them.
§4 says what is new about *how*: until now every parameter read was of a call FrameLedger intercepted anyway; these
two calls are intercepted in a game **only once a session with the option on has started there**, and with the
option off nothing in the game is touched for them. The accuracy block's item on what a game tells its upscaler
names the two, and no longer says none of it was read from a real game — FSR's upscaling values were, on
0.1.0-beta.41 and beta.42. The Disclaimer stays 2.15, as before. §2 is unchanged: nothing is sent.
### What version 2.16 of the Privacy Policy changed (0.1.0-beta.40)
**Advanced metrics inside the game** (train B, owner decisions D111–D116; the content approved with the plan, once
for the train's three releases). §1's item on Advanced metrics gains what a session stores **for a game FrameLedger
injects into, only while the option is on**: what the game passed to its upscaler in the calls FrameLedger
intercepts — the output size, presets and exposure values it gave DLSS, whether it said its colour buffers are HDR,
auto-exposure and alpha upscaling; the sharpening, the jitter's size, the exposure, the depth range, the lowest and
highest field of view and the flags it gave FSR, the frame time it reported, FSR's own memory when the game asked —
and what is never read: where the camera is, which way it looks, its matrices. It also says that switching the option
off stops the reading inside a game as well. §4 had always named "the parameters your games pass to the graphics APIs
it intercepts"; it now says that with the option on more of the upscaling calls' parameters are read **and stored**,
which is what is new. The accuracy block gains an item: these are the game's words to the upscaler, not a check of
what the upscaler did, and many games show none. The Disclaimer stays 2.15, as the accuracy statement's additions
have not moved it before (beta.19, beta.20, beta.25). §2 is unchanged: nothing is sent.
### What version 2.15 of the Privacy Policy and of the Disclaimer changed (0.1.0-beta.39)
**Advanced metrics** (owner decisions D101–D108; the Privacy Policy's content approved with the plan, the Disclaimer's
two sentences by the owner's answer of 2026-11-11). Privacy Policy §1 gains one item — what a session stores **only
while the option is on**: more sensor series of the graphics card, and of the processor, the memory and the motherboard
when the Agent runs as administrator with PawnIO; a description of the hardware; what each display can do and what the
game's display did; the graphics driver's settings, global and for the game — and what is never read: a serial
number, a display's EDID, a list of other programs. §3A says an export and a bug report do not carry it. §4 names the
same as things the software can observe, only while the option is on. The Disclaimer's §5 and §5A said PawnIO and the
admin mode were for the CPU temperature alone; both now name the processor's, the memory's and the motherboard's
sensors with the option on. The accuracy block gains an item that says none of these readings has been checked
against another tool. §2 is unchanged: nothing is sent, as before. (Versions 2.12 to 2.14 of the Privacy Policy and
2.12 to 2.14 of the Disclaimer are recorded in the changelog of the release that carried each.)
### What version 2.11 of the Privacy Policy changed (0.1.0-beta.23)
A game's page shows its artwork (beta.23 PR-3, owner decisions D60/D61; the wording approved with the plan). §1 adds the
+9
View File
@@ -121,6 +121,15 @@ rather than of an instance.
> `spike-notes` §8 and §H5 — so item 3's counter has nothing to count on this route. That is
> a separate finding with its own consequences for HANDOFF item 3, and it does not belong to
> this entry.
>
> **2026-11-06 (beta.34, D84): a fifth id is decoded, and not by guess.** `kFeatureDLSS_NR` (1004),
> DLSS 5's neural rendering, comes from Streamline `v2.14.16`'s own `sl_core_types.h`, re-vendored for it
> (`src/native/third_party/streamline/README.md`) — row O4's warning holds: no id is mapped from a number seen
> arriving. Until then an evaluation of it fell to `FL_FEAT_SL_UNDECODED`, which the Agent does not store. It
> is its own fact (`FL_FEAT_SL_NEURAL_RENDERING`, from the drain word, never from the decoded `upscaler`
> byte — the rule above), never an upscaler; the census counts it in a bucket of its own
> (`SlCensus.WithNeuralRendering`), so the buckets still add up. Whether a real game evaluates it through
> `slEvaluateFeature` is owner-only item 30(a), with its table in `spike-notes` §26.
<details>
<summary>The entry as it stood while it was open, including the pre-committed table with its two holes</summary>
+6
View File
@@ -49,3 +49,9 @@ Each section below lives in its own file under `docs/spike-notes/`, moved verbat
| §23 | The owner's runs of beta.26 on real hardware (2026-10-20) | [`spike-notes/23-owner-runs-2026-10-20.md`](spike-notes/23-owner-runs-2026-10-20.md) |
| §24 | The whole gate on .NET 11 RC3 — what the alpha line's SDK asks of the source (2026-10-31, beta.32, D73–D76) | [`spike-notes/24-dotnet11-probe.md`](spike-notes/24-dotnet11-probe.md) |
| §25 | Xbox app / Microsoft Store (GDK) titles — the layout, the launch chain, the image paths (2026-11-03, beta.33, D77–D82) | [`spike-notes/25-xbox-gdk.md`](spike-notes/25-xbox-gdk.md) |
| §26 | DLSS 5 on Hogwarts Legacy, Dynamic frame generation on Phantom Blade Zero, and Streamline 2.12 → 2.14.16 (2026-11-06, beta.34, D83, D84) | [`spike-notes/26-dlss5-dynamic-streamline.md`](spike-notes/26-dlss5-dynamic-streamline.md) |
| §27 | Switching to Games, and a language change that lands elsewhere (2026-11-06, beta.34) | [`spike-notes/27-games-page-2026-11-06.md`](spike-notes/27-games-page-2026-11-06.md) |
| §28 | The main line on .NET 11's release — the whole gate, and what the analyses read (2026-11-11, beta.37, D94–D96) | [`spike-notes/28-dotnet11-main.md`](spike-notes/28-dotnet11-main.md) |
| §29 | C# 15 against the checks — the canary, and where its constructs would go (2026-11-11, beta.37, D96) | [`spike-notes/29-csharp15-canary.md`](spike-notes/29-csharp15-canary.md) |
| §30 | Reflex latency per frame — the routes there are, and what each would cost; a reading, nothing run (2026-11-12, beside beta.45, D130; §M14) | [`spike-notes/30-reflex-latency.md`](spike-notes/30-reflex-latency.md) |
| §31 | Reflex route D on the harness — the driver's two getters called from an unshipped probe, against a stand-in module and the installed driver; no game touched (2026-11-13, after beta.46, D132; §M14) | [`spike-notes/31-reflex-route-d-harness.md`](spike-notes/31-reflex-route-d-harness.md) |
+1 -1
View File
@@ -99,7 +99,7 @@ The last thing P1 owes (§R4 moved it here because it needs P2's drain and recor
| 2 | OFF | | | |
| 3 | OFF | | | |
Gates: FPS delta ≤ 0.5 %, Agent CPU ≤ 1 % of a core, Agent RSS ≤ 150 MB. **Overlay resident ≤ 8 MB is
Gates: FPS delta ≤ 0.5 %, Agent CPU ≤ 1 % of a core *(≤ 2.5 % since 2026-11-11, D117)*, Agent RSS ≤ 150 MB. **Overlay resident ≤ 8 MB is
not measured by the runbook** and is left blank rather than guessed: it is a module inside the game
process, and reading the target's memory is what CLAUDE.md rule 4 forbids — an external tool that
inspects module sizes is the way to fill it.
@@ -0,0 +1,85 @@
## 26 · DLSS 5 on Hogwarts Legacy, Dynamic frame generation on Phantom Blade Zero, and Streamline 2.12 → 2.14.16 *(2026-11-06, beta.34, owner decisions D83, D84)*
**Why this was measured.** The owner asked, on 2026-11-06, how FrameLedger detects DLSS 5's Neural Rendering on
Hogwarts Legacy — turned on in the game — and why Phantom Blade Zero under the NVIDIA App's *Dynamic* multi frame
generation, up to ×3, read "DLSS 4" when Dynamic and ×6 exist only in DLSS 4.5; then, while the plan was written,
what had changed in Streamline since the copy the Overlay is built against.
**Where, and how.** The owner's ledger was read **read-only** — Python's `sqlite3` with `mode=ro`, no FrameLedger
binary, the Agent of `0.1.0-alpha.2` running beside it — on the dev box of §Environment (Windows 11 Pro build 29648,
i7-14700KF, RTX 5080, 2560×1440 at 240 Hz; NVIDIA driver 32.0.16.2101, then 32.0.16.2114 from 2026-11-05 18:29).
Streamline's headers were compared with the upstream repository's tags through GitHub's API, byte for byte.
### The NVIDIA App's frame-generation settings against the driver's report
`driver_profile` is the game's profile as the Agent read it at the session's start; the FG word and `frameGenerationMode`
are `NvAPI_NGX_GetNGXOverrideState`'s, from `ngx_driver_words`. Bits: `0x1F` initialised … selected, `0x20` preset,
`0x200`/`0x400` created/evaluated, `0x2000` `FG_MODE`, `0x8000` `FG_MULTI_FRAME`.
| Sessions | The App's setting (profile) | FG word | `frameGenerationMode` | Frames counted |
|---|---|---|---|---|
| Phantom Blade Zero 111, 112, 114 (11-01 → 11-02) | `NGX_DLSSG_MODE` 4 *dynamic*, ceiling 5, target `0x01000000` | `0x263F` | 3 | states ×3.0 … ×4.0, ×5.2, ×6.0 |
| Phantom Blade Zero 116 → 129 (11-03 → 11-05), the eight hooked sessions the driver answered (130 was recovered without its report) | `NGX_DLSSG_MODE` 4 *dynamic*, **ceiling 2**, target `0x01000000` | `0x263F` (`0x203F` once) | 3 | ×2.0 and ×3.0 only; ×3.0 held 75–94 % of each session longer than an hour |
| SILENT HILL 2 60, 71 (10-19, 10-20) | `NGX_DLSSG_MODE` 4 *dynamic*, ceiling 5 | `0xA03F` | 3 | ×2.0, ×3.0, ×6.1 |
| Clair Obscur: Expedition 33 64, The Witcher 3 68, SILENT HILL 2 36 | `NGX_DLSSG_MODE` 2 *on*, multi-frame count 5 (*Fixed ×6*) | `0xA03F` | 1 | ×6.0; Witcher 3 ×2.0 |
| Cyberpunk 2077 69 (multiplier chosen in the game) | none set | `0x5` | 0 | ×1.9, ×3.0, ×4.0 |
So the driver reports `FG_MODE` applied under both of the App's modes and its own mode number tells them apart — 3 under
Dynamic, 1 under Fixed, on all sixteen sessions — but NVIDIA's header names neither number. **What D83 reads is the
profile's value, which the header does name (4, `NGX_DLSSG_MODE_DYNAMIC`), with the report's `FG_MODE` bit as proof the
driver applied it**, and DLSS Frame Generation identified in the session (`fg_mode = dlssg`, whether or not its factor
was counted). The ceiling's decode (frames generated per
rendered frame, `03_METRICS` §The mix) held: a ceiling of 5 reached ×6.0, a ceiling of 2 never passed ×3.0.
**Decided (D83):** a session the driver reports running the App's Dynamic mode is **DLSS 4.5**, below a measured ×5 and
above every DLSS 4 row; *Fixed* stays the frames' to name (`05_DETECTION` §DLSS generation).
### DLSS 5 on Hogwarts Legacy
Session 132 (2026-11-06 13:56, hooked, attached, 141 s, Neural Rendering on in the game): `runtime_modules` holds
**`nvngx_dlssnr.dll` 310.9.16.0, loaded from the game's folder** — the name `CensusNames` carried from community sources
(owner-only item 20(d)) seen on a real game for the first time — beside `nvngx_dlss.dll`, `nvngx_dlssd.dll`,
`nvngx_dlssg.dll` 310.9.16.0 and Streamline 2.14.16 (`sl.interposer.dll`, `sl.common.dll`, `sl.dlss_g.dll`,
`sl.reflex.dll`). The install ships the matching Streamline plugin, `Engine\Plugins\Runtime\Nvidia\StreamlineCore\
Binaries\ThirdParty\Win64\sl.dlss_nr.dll` (402,048 bytes; `nvngx_dlssnr.dll` is 266,424,944), which no list asked about.
The driver's super-resolution word read `0x80605` (created and evaluated, no override); the session identified DLSS Frame
Generation but refused a factor (non-uniform, ×1.97 overall); 140.2 FPS presented.
What the App made of it: *DLSS · Super Resolution 310.9.16, from the game · … · DLSS 5 runtime loaded* — loaded, never
whether it ran. Nothing measured that: an evaluation of a Streamline feature the Overlay does not decode lights
`FL_FEAT_SL_UNDECODED`, which the Agent does not store, and the session's `sl_tag_census` (`0x20E00`) shows only that
the game tags a buffer type the census does not name, per frame (bit 17) — which Streamline 2.14 names, below.
### Streamline: the vendored copy against the releases
The vendored closure (`src/native/third_party/streamline`, "main, 2026-08-09") is **byte-identical to tag `v2.12.0`**,
all eleven files. Releases since: `v2.14.1` (2026-09-08: V-Sync and frame-rate limiters in Dynamic multi frame
generation, `VK_NV_low_latency2`) and `v2.14.16` (2026-10-16: fixes) — the version Hogwarts Legacy ships and the NVIDIA
App's override loads (`1B0_E658700.dll` 2.14.16.0). From `v2.12.0` to `v2.14.16` three of the eleven change:
- `sl_core_types.h`: **`kFeatureDLSS_NR = 1004`**; buffer types 70–72, reserved until then, become
`kBufferTypeUpliftInputColor`, `kBufferTypeUpliftOutputColor` and `kBufferTypeUpliftControlMask` — "neural-net
uplift passes", DLSS 5's inputs; `Resource::reserved` (32 bits) becomes `internalFlags` + `reserved` (16 each, the
struct's size unchanged).
- `sl_struct.h`: `kStructVersion7`.
- `sl_version.h`: 2.14.16.
`sl_core_api.h` (`PFun_slEvaluateFeature`), `sl_dlss.h`, `sl.h` and `license.txt` are unchanged, and nothing of
FrameLedger reads `Resource::reserved`. **Decided:** re-vendor at `v2.14.16` (the owner: "check what changed … and fix it
if really needed") so the Overlay decodes DLSS 5's evaluations by the vendor's own constant (D84).
### What beta.34 builds on it, and owner-only item 30(a)'s decision table — written before the run
Built (D84): the Overlay decodes `kFeatureDLSS_NR` into `FL_FEAT_SL_NEURAL_RENDERING` and publishes
`FL_HOOK_SL_NEURAL_RENDERING` with the detour; the Agent stores `dlss_nr_pct` — the share of the Streamline batches that
carried it — only from a writer that published the family. The harness drives the id through an injected Overlay
(`guard_test`, *DLSS 5's neural rendering is its own fact*). What no fixture can say is whether Hogwarts Legacy evaluates
neural rendering through `slEvaluateFeature` at all; one short hooked session each:
| Neural Rendering in the game | `dlss_nr_pct` | What it means, and what follows |
|---|---|---|
| On | above 0 | the game evaluates it through Streamline: *DLSS 5 (neural rendering on N % of rendered frames)* — D84 holds as built |
| On | 0 | Streamline batches drained and none carried 1004: neural rendering runs another way (NGX directly, or another id). The line reads *runtime loaded, not seen running*; record the session's `sl_tag_census`, and the decode is revisited before anything is named |
| On | NULL | no Streamline batch drained, or the family is missing from `hooks_installed_mask` (bit 12): the detour never installed — record which |
| Off | 0, the runtime loaded | the runtime loads with the feature off: "loaded" never meant "on", which is why D84 waits for the evaluation |
| Off | NULL, the runtime not loaded | the game loads it only when on — recorded, nothing changes |
@@ -0,0 +1,41 @@
## 27 · Switching to Games, and a language change that lands elsewhere *(2026-11-06, beta.34)*
**Why this was measured.** The owner, 2026-11-06: switching from any page to Games stalls about 0.5–1.5 s before the
page appears; and changing the language in Settings brings back the last content page (a game's detail) instead of
Settings. §21 (beta.23) had measured the library at 275 ms for 48 games and left it, the grid non-virtualizing by the
owner's decision (2026-10-15); the library has 65 games now.
**What the code showed** (read, then measured): every page is transient and rebuilt per visit; `GamesViewModel` read the
settings, the games and the per-game session summary on the UI thread (Microsoft.Data.Sqlite completes its asynchronous
calls synchronously), then asked for every executable's file **and showed no card until all had answered** — 39 of the
owner's 41 Steam games are on the USB drive D:, which sleeps — and cleared and re-added every card on each read; the list
view's `DataGrid` had no height bound and built every row. WPF UI's `NavigationCacheMode` would not have helped: 4.3.0
reads it only when the NavigationView has neither a service provider nor a page provider, and this App sets one. The
language: `ShellHost` remembered only pages opened from code (`PageNavigator`), never a pane click.
**The decision table, written before the runs:** fix what keeps the first card from appearing, what rebuilds a page on
every visit, and what blocks the UI thread ≥ 50 ms; leave the grid non-virtualizing (the owner's decision); record the
rest.
**How.** A scratch xUnit test over the App's own `GamesViewModel` and `GamesPage` (not committed — it times, and a timing
assertion in CI is a flake), Release, on the dev PC of §21: a scratch ledger shaped like the owner's — 65 games, a third
with a session — with the first executable check held **800 ms** (a USB drive waking) and every later one answering at
once. Seven iterations a run, the median; three runs a side, the code before this change (PR-3's tree) and after it.
*First cards*: from the view model's construction to all 65 cards listed. *Build*: a new page laid out at 1200×900 with
its 65 cards. *Revisit*: before, a new page again (every visit built one); after, the kept page laid out again.
| | before (three runs) | after (three runs) |
|---|---|---|
| First cards | 822 – 830 ms | 10 – 15 ms |
| Build (first visit) | 165 – 197 ms | 164 – 223 ms — the same work, run-to-run noise |
| Revisit | 181 – 215 ms | 15 – 16 ms |
**Built** (beta.34): the window keeps the library page it built (`Services.WindowPageCache`, the window's own page
provider, set after `INavigationService.SetNavigationControl`) and the page reads the library again each time it is shown
(`GamesViewModel.Attach`); the library is read on the thread pool; the cards are listed first and the executable checks
follow, each card getting its *Not installed* pill when its check answers; a read keeps every card that would look the
same (`GameCardViewModel.LooksLike`, `GamesViewModel.Sync` — no `Clear`); the list view is bounded (`MaxHeight` 640).
`ShellHost` notes every page the live window shows, so a rebuild opens on Settings after a change made there.
**Left as it was:** the first visit's build (≈ 2.5–3.4 ms a card at 65) — the non-virtualizing grid's cost, unchanged by
decision; a page kept per window holds its 65 cards' visuals for the window's life.
+46
View File
@@ -0,0 +1,46 @@
## 28 · The main line on .NET 11's release — the whole gate, and what the analyses read *(2026-11-11, beta.37, owner decisions D94–D96)*
**Why this was measured.** The owner moved the main line to .NET 11 the day after it shipped (D94), setting aside the
condition `20_OPEN_QUESTIONS` §G had written for it — two alphas on the released runtime. §24 measured Release Candidate
3 on a throwaway branch; nothing had measured the release, and three things had never met .NET 11 at all: CodeQL's C#
analysis (a required check, main-only), SonarCloud's scanner, and a Velopack delta across two runtime majors. This note
is what stood in for the condition that was set aside.
**Where.** The dev box of §Environment (Windows 11 Pro build 29648, i7-14700KF, RTX 5080). SDK `11.0.100`, runtime and
Windows Desktop `11.0.0` (released 2026-11-10), installed beside the 10.0 SDKs. The hosted runner: image
`windows-2025-vs2026`, Visual Studio 18 Enterprise, MSVC 14.51.
**The switch, on main this time** (branch `beta37/pr1-net11`): `global.json` → `11.0.100`, `allowPrerelease` false;
`Directory.Build.props` → `net11.0-windows10.0.22621.0`; `Microsoft.Extensions.Hosting`, `Microsoft.Data.Sqlite`,
`System.Diagnostics.EventLog` → `11.0.0`; `legal/licenses/nuget/` regenerated (78 packages → 69, the alpha's set at
released versions). Nothing else in the build.
**What the release asked of the source that RC3 had not.** Two sites, both in tests written after the last alpha was
cut (beta.35's `ContrastThemeTests.Drawn.cs`): CA1307 on `string.StartsWith(char)` and `string.EndsWith(char)`, which
.NET 11 gives a `StringComparison` overload. They take that overload now — the main line compiles against .NET 11, so
§24's rule (a form .NET 10 also compiles) no longer binds. The 21 sites §24 fixed stayed quiet.
**`IHost.RunAsync`, observed for the first time.** §24 read Microsoft's note that a failed `BackgroundService` now fails
`RunAsync` and recorded that no test pinned it. `HostBackgroundFailureTests` does: a host with one service that throws
fails `RunAsync` with that exception on 11.0.0. The Agent's `--serve` therefore leaves through its crash report
(`10_LOGGING` §Crash handling) where it exited 0 on .NET 10; it was accepted, not wrapped.
**The gate, on the dev box (`./build.ps1 check`, 338 s, all gates passed).** Native: 32 of 32 suites. Managed: the build
with warnings as errors, 0 warnings; `dotnet format` clean; 2,561 tests passed and 2 skipped in the six projects (Domain
429, Infrastructure 578 and the 2 skipped, Application 568, CaptureHost 137 with its end-to-end cases, App 816, Agent
33), every one on `net11.0`; the coverage floors held (12 reports); `license-check` green on the regenerated texts. No
test changed its answer between .NET 10 and .NET 11.
**The pull request's own runs (#376, 2026-11-11), which were the first of each on .NET 11:**
| Run | What it showed |
|---|---|
| `check` (42916490997) | green in 12 min 22 s on the hosted runner, SDK 11.0.100 installed by `setup-dotnet` from `global.json` |
| `analyze (csharp, none)` (42916491104, CodeQL 2.27.4) | green in 4 min 5 s. Build mode `none` used the runner's 11.0.100 SDK for references (on .NET 10 it downloaded 10.0.100 itself): 1,221 source files found, 363 references resolved and none unresolved, 9 types unresolved (main's last run on .NET 10: 23 in three `Windows.Win32` namespaces), and "CodeQL scanned 1220 out of 1220 C# files" — the same line as main's last run. CodeQL's documentation lists C# up to 14 and .NET up to 10; *(Corrected the same day: this was an **overlay** analysis — the changed files, on top of `main`'s cached database — and its "1220 out of 1220" was `main`'s baseline. The whole source was read by `main`'s run after the merge, 42921492229: "Overlay mode not enabled", 1,221 files found and "CodeQL scanned 1221 out of 1221 C# files", 363 references resolved, none unresolved, 23 types unresolved as on .NET 10, 4 min 45 s.)* `net11.0` with source that uses no C# 15 construct is read whole |
| `analyze (cpp, manual)` | green, 7 min 7 s — the native layer did not change |
| `sonarcloud` | green, the gate passed: the SonarScanner for .NET 12.0.0 (a `netcoreapp3.1` tool that rolls forward to the newest major) ran on the 11.0 runtime |
| Qodana (42916491047), the Community linter with the interim | the `bootstrap` line installed SDK 11.0.100 into the container, InspectCode 2026.2.3.1 loaded all thirteen `net11.0-windows` projects, and it reported **1 problem**: a redundant `!` in the new test, fixed in the same pull request |
| `release.yml` rehearsal on the branch (42916488083, `0.1.0-beta.37`) | green. `setup-dotnet` installed SDK 11.0.100 and, for `vpk` 1.2.161, the .NET 10 line (SDK 10.0.404, runtime 10.0.15). **The first delta across two runtime majors**: "Delta processed 0443 files. 0372 patched, 0071 unchanged, 0009 new, 0017 removed" — 72,604,898 bytes, against a full package of 110,314,263 (beta.36's: 107,902,117) and a `Setup.exe` of 117,828,375 |
**Not measured here.** An installed build on real hardware and a hooked title on .NET 11.0.0 (owner-only item 34);
Windows 10 22H2, which no version of FrameLedger has been run on; a processor without x86-64-v2.
+61
View File
@@ -0,0 +1,61 @@
## 29 · C# 15 against the checks — the canary, and where its constructs would go *(2026-11-11, beta.37, owner decision D96)*
**Why this was measured.** The owner asked for C# 15 on the main line in the same release that moved it to .NET 11 (D96).
The plan's rule, approved with it: a construct is used only where (1) the build, its analyzers and `dotnet format` take
it without an analyzer being switched off, (2) CodeQL extracts its file without an error, and (3) Qodana does not report
errors that are only its engine not knowing the syntax. So each construct was written once, in a file of its own, and
shown to each tool before any product code used it.
**The canary** (branch `probe/csharp15-canary`, draft pull request #377, closed unmerged): four files under
`tests/FrameLedger.Domain.Tests/Canary/`, one construct each, each with a test that runs —
```csharp
HashSet<string> set = [with(StringComparer.OrdinalIgnoreCase), "a", "A"]; // collection expression arguments
private closed record State; private sealed record Idle : State; // a closed hierarchy, switched over
private union Pet(Cat, Dog); // a union, switched over
outer: foreach (int[] row in rows) { foreach (int cell in row) { if (cell < 0) { continue outer; } } } // a labeled jump
```
**What each tool made of them.**
| Construct | SDK 11.0.100's compiler, its analyzers (`latest-all`), Roslynator 5.0.1, Meziantou 3.0.319, VS Threading 18.12.1 | `dotnet format` | ReSharper's engine: InspectCode 2026.2.3.1 (what Qodana 2026.2 runs) | InspectCode 2026.3.0 (the newest released) |
|---|---|---|---|---|
| Collection expression arguments, `[with(…), …]` | builds, 0 warnings | clean | **error** — "Feature 'collection expression arguments' is not available. Please use language version 'preview'." | the same error |
| Closed hierarchy, `closed record` | builds, 0 warnings; the switch needs no default arm | clean | **not parsed** — "Cannot resolve symbol 'closed'", "Unexpected token", then CS8509 on the switch it reads as open | the same |
| Union, `union Pet(Cat, Dog)` | builds **only with CA1815 silenced**: the SDK's own rule asks every union — a struct — to override `Equals` and the equality operators | clean | **not parsed** — "Cannot resolve symbol 'union'", "Identifier expected" | the same |
| Labeled `continue` / `break` | builds, 0 warnings | **moves the label one indent to the left** (`csharp_indent_labels`), the statement staying on its line | **not parsed** — "; expected", "Cannot resolve symbol 'outer'", and the loop's tail read as unreachable | **error** — "Feature 'labeled break and continue' is not available. Please use language version 'preview'." |
The four tests passed. No analyzer threw on a syntax node it did not know (an `AD0001` would have been an error here).
InspectCode was run on the dev box with `--toolset-path` at SDK 11.0.100's `MSBuild.dll`, a fresh caches folder each
time; 2026.2.3.1 reported 29 results in the four files, 2026.3.0 24.
**CodeQL (2.27.4): not established, and why a pull request cannot establish it.** `analyze (csharp, none)` was green on
the canary's pull request in 3 min 10 s, and its log says "Reading overlay changes": on a pull request the C# analysis is
an **overlay** — it extracts the files the pull request changed on top of the database cached from `main` — and its log
prints no per-file extraction result. "CodeQL scanned 1220 out of 1220 C# files" is `main`'s baseline figure, printed
unchanged on a pull request that had added eight files. The SARIF the code-scanning API returns carries no extraction
notification. CodeQL's documentation lists C# up to 14. So this run neither shows that CodeQL reads these constructs nor
that it does not; a full analysis needs a run that is not a pull request's (`codeql.yml` takes `workflow_dispatch`
since this pull request, for that).
**The conclusion, by the rule.** No C# 15 construct is used on the main line in 0.1.0-beta.37: every one of the four
fails condition (3) — Qodana's engine would report errors in each file that used one, on a gate of zero — and the union
fails (1) as well. What the main line takes from .NET 11 in this release is **API, not syntax**: the
`string.StartsWith` / `EndsWith` / `LastIndexOf(char, StringComparison)` overloads (sixteen call sites that §24's rule had
written as spans, or that §28 fixed), and `Process.TryGetProcessById` in place of a caught `ArgumentException`.
**What unblocks each, and where it would go** — the survey of 2026-11-11 over the thirteen projects, kept so the next
attempt starts from it:
| Construct | Waits for | Where it would go |
|---|---|---|
| `[with(comparer), …]` | a ReSharper engine that reads C# 15 as released, in the Qodana linter | the seven sites §24 wrote with explicit `StringComparer.Ordinal` constructors for MA0002, and about fifteen more local `HashSet`/`SortedSet` constructions (`GameFileProbe`, `CaptureOrchestrator`, `LibraryImporter`, `GameTitleGuess`, …). Not the `IReadOnlySet`-typed statics (not a collection-expression target) nor `ConcurrentDictionary` |
| `closed` | the same; and CodeQL's reading, measured by a dispatched run | `Application/Watch/WatchEvent` (the one abstract record in `src/`); and, as closed record hierarchies rather than unions, the hand-written "one of" shapes: `Domain/Detection/EngineMatch` (with `RuleEvaluator.MatchPlatform`), `Application/Recording/FgVerdict` with the identity and the withheld reason it travels with, `Application/Telemetry/GameProcess`, the Agent's private `ElevationDecision` |
| `union` | the same, **and** CA1815 settled — a union that must hand-write its equality is not shorter than what it replaces | only a value that is one of several *existing* types: `IRingAttacher.TryAttach`'s `(ICaptureSink?, ShmAttachRefusal)` |
| Labeled `continue` | the same | one site: `Infrastructure/Detection/InstallRoot.cs`, the `hit` flag |
| Extension indexers | — | nowhere: the six extension methods in `src/` are not index-like |
**Left as they are, whatever the tools read:** the enums whose names are stored or cross a wire
(`SessionEndReason`, `AntiCheatRefusalReason`, `ShmAttachRefusal`, `Tri`), and `AntiCheatVerdict` (it mirrors a native
result, and "zeroed means not evaluated" is a rule of the guard). And the two preview features — runtime-async and the
updated memory-safety rules — which D76's "no preview feature" still excludes.
+232
View File
@@ -0,0 +1,232 @@
## 30 · Reflex latency per frame — the routes there are, and what each would cost *(2026-11-12, written beside 0.1.0-beta.45, owner decision D130; `20_OPEN_QUESTIONS` §M14)*
**Why this was written.** Train C of Advanced metrics is "Reflex latency per frame" (`18_GPU_VENDOR_APIS` §The
trains). §M14 asks whether any route to it stays inside rules 4 and 5, and pre-commits the answer for the case
that none does. The owner decided on 2026-11-12 (D130) that this spike is written beside 0.1.0-beta.45 **as a
document**, and that no code of train C is written until the owner has read it and said go or no.
**What this is, and what it is not.** A reading — of the vendored NVAPI headers
(`src/native/third_party/nvapi/nvapi.h`, `nvapi_interface.h`), of Streamline's public headers and plugin source
at tag `v2.14.16` (`include/sl_reflex.h`, `include/sl_pcl.h`, `source/plugins/sl.reflex/reflexEntry.cpp`,
`source/plugins/sl.pcl/pclstats.h`, `pclImpl.h`; fetched through GitHub's API, not vendored), of this
repository's own hook engine, and of a copy of the owner's ledger; one file on the dev box was looked at, the
export table of the installed `nvapi64.dll`. **Nothing was run in a game and no shipped code changed.** Every
"unknown" below is unknown, not assumed either way.
### What Reflex's numbers are, and who holds each
| What | Where it lives | The catalog entry that wants it |
|---|---|---|
| **The per-frame report**: for each of the last 64 frames, the times of input sample, simulation start/end, render submit start/end, present start/end, driver start/end, OS render queue start/end, GPU render start/end, and two durations (`gpuActiveRenderTimeUs`, `gpuFrameTimeUs`) | The **driver**. A program gets it by calling `NvAPI_D3D_GetLatency(pDev, NV_LATENCY_RESULT_PARAMS*)` with the game's own device; Streamline's `slReflexGetState` makes that call and hands the same 64 reports back as `sl::ReflexState::frameReport` | `nvapi.reflex.frame_report`, `sl.reflex.state` |
| **The driver's low-latency state**: low-latency mode on, the last sleep interval, whether the game calls the sleep, the frame-generation multiplier last used | The driver: `NvAPI_D3D_GetSleepStatus(pDev, …)` | `nvapi.reflex.sleep_status` |
| **What the game asked for**: low latency or low latency with boost, the frame-rate limit, whether markers may steer the driver | Only the **arguments** of `NvAPI_D3D_SetSleepMode` or of Streamline's `slReflexSetOptions` — no getter returns boost or the limit | `nvapi.reflex.mode` |
| **The markers the game sets** each frame (simulation start … present end, the latency ping, the flash trigger) | The arguments of `NvAPI_D3D_SetLatencyMarker` / `slPCLSetMarker`; the PCL plugin also writes each as an ETW event | `sl.pcl.markers` |
`FlFrameRecord.reflexLatencyUs` (@48, "0 = unavailable") has waited for this since the record was designed, and
`03_METRICS` lists *PC latency* as "mean/p95 of `reflexLatencyUs` when Reflex reports it". **A report cannot go
into its own frame's record**: the driver completes a frame's report after that frame's GPU work ends, several
presents after the record was written, and a ring record is never rewritten. Whatever route is taken, latency
arrives as a series beside the frames, not as a column of them.
### What the owner's sessions already say
From 0.1.0-beta.44's `sl.feature.versions` / `sl.feature.requirements` (sessions 181–186 of the ledger's copy;
`docs/HANDOFF.md`, item 42):
| Game (session) | Interposer | Asked Streamline about `kFeatureReflex` | about `kFeaturePCL` |
|---|---|---|---|
| Phantom Blade Zero (181) | 2.14.1 | version and requirements | version and requirements |
| Clair Obscur: Expedition 33 (182) | 2.7.30 | version and requirements | version and requirements |
| Valor Mortis (183, 184) | 2.14.1 | version and requirements | version and requirements |
| Black Myth: Wukong (185) | 2.7.4 | version and requirements | version and requirements |
| Cyberpunk 2077 (186) | 2.14.1 | requirements only | not asked |
So **every Streamline game run asks about Reflex**, and the four Unreal Engine games about PCL as well. That a
game asks about a feature is not that it uses it; 0.1.0-beta.45's `slIsFeatureLoaded` and `slSetFeatureLoaded`
rows will say whether Reflex and PCL are loaded and switched, on the owner's next sessions. No session says
anything about a game that uses Reflex without Streamline: none of the hooks sees NVAPI.
**Added 2026-11-13, from the first sessions on 0.1.0-beta.45** (190–196): Streamline answers `eOk` for
`kFeatureReflex` in all five Streamline games run, and `slIsFeatureLoaded` says it is loaded in the three that ask
(Phantom Blade Zero, Cyberpunk 2077, Hogwarts Legacy); Cyberpunk 2077 and Dying Light: The Beast switch it off and
on through `slSetFeatureLoaded`. `kFeaturePCL` is supported in the four games that ask about it. Loaded is still
not in use: whether a game sets markers is what no hook sees.
### The routes
**A — hook NVAPI's own functions** (`SetLatencyMarker`, `SetSleepMode`; what `17_HOOK_ENGINE` §Memory / latency
still names). The `nvapi64.dll` on the dev box (driver 32.0.16.2114, its export table dumped for this note)
exports two functions, `nvapi_QueryInterface` and `nvapi_Direct_GetMethod` — §M14 said one — and no latency
function: each of those is an address `nvapi_QueryInterface` returns for an id (`nvapi_interface.h`:
`NvAPI_D3D_GetLatency` is `0x1a587f9c`). A hook on one is a patch at an address
learned at run time — the class refused on 2026-08-09 — or a wrapper returned from a hooked
`nvapi_QueryInterface`, which is a write to what the game receives (rule 4). **Refused, as §M14 pre-committed.**
**B — hook Streamline's Reflex and PCL functions** (`slReflexSetOptions`, `slPCLSetMarker`, `slReflexGetState`).
None is an export of `sl.interposer.dll`. The game gets each through the exported `slGetFeatureFunction(feature,
name, &fn)`, and Streamline's own header does that **once**: `SL_FEATURE_FUN_IMPORT_STATIC` keeps the pointer in
a function-local static after the first call. Two consequences. A hook on `slGetFeatureFunction` could only see
the game's calls by handing back a wrapper — train D's question (§M15), which needs rule 4 amended. And a hook
that exists only for Advanced metrics is installed after the attach (D111), by which time a game that has
presented a frame has fetched its functions: **the same reason `slInit` is not hooked (D129) applies to the
fetch.** Not a route for train C; it stays train D's, and this finding weakens train D as well.
**C — the Overlay calls `slReflexGetState` itself**, having asked the interposer's exported, documented-thread-safe
`slGetFeatureFunction` for it. No patch, no wrapper, and Streamline uses the device it already holds. But the
header says of `slReflexGetState`: *"This method is NOT thread safe"*, and the source shows why — the function
**writes** two fields of the plugin's context (`lowLatencyAvailable`, `latencyReportAvailable`) from the result
of the two driver calls it makes, with no lock, and then calls into the PCL plugin. The game calls the same
function from a thread FrameLedger cannot know (the call is not hooked), so the Overlay cannot keep its call
from overlapping the game's. It would be FrameLedger breaking the vendor's stated contract inside the game, and
writing plugin state the game reads back. **Recommended refused**, on the header's own words.
**D — the Overlay calls the driver's two getters itself**: `NvAPI_D3D_GetLatency` and `NvAPI_D3D_GetSleepStatus`,
resolved through `nvapi_QueryInterface` of an `nvapi64.dll` **the game already loaded** (never loaded by
FrameLedger), with the device of the swap chain the Overlay already hooks. This is §M14's "route to try first".
What it would be:
- **No patch and no new hook.** Two calls, both getters; `tools/nvapi-readonly-check.ps1`'s list (D108) would
gain them, and the check would have to cover the Overlay as well as the bridge. The Overlay would not link
`nvapi64.lib`: it needs the two ids and the two structs from the vendored MIT headers.
- **It works without Streamline** — any Direct3D game that uses Reflex — and not on Vulkan (the Vulkan layer has
no NVAPI device; Reflex there is a Vulkan extension, a different study).
- **It gives** the report (`nvapi.reflex.frame_report`) and the sleep status (`nvapi.reflex.sleep_status`), and
through them most of `sl.reflex.state`. **It does not give** what the game asked for (boost, the limit:
`nvapi.reflex.mode` stays N/A, with the reason) or a count of markers.
- **First of its kind in two ways**, and both are the owner's to rule on. The Overlay would call a *vendor's
driver interface on its own initiative* inside the game — today it only forwards the game's calls and reads
their arguments. And it would use *the game's device object* for a call the game did not make: rule 4 allows
"COM/handle objects we legitimately own", and the device is the game's. The nearest precedent is the read of
`IDXGISwapChain::GetLastPresentCount` on the hooked chain (rule 6, 2026-09-06), a getter on the game's swap
chain inside the present detour.
- **Unknown, and what the spike would have to measure**: which object NVAPI wants for a Direct3D 12 game (the
device behind the queue the swap chain was made on — not verified); whether the getters are safe on a thread
that is not the game's — the header is silent; Streamline's own plugin calls them on whatever thread the game
asks for its state while the render thread sets markers, which suggests the driver tolerates it, and is an
inference; what a call costs, on the presenting thread or on the Overlay's own; what it returns in a game that
never set a marker (the header: valid only after markers with rising frame ids); whether a call that blocks
in the driver can be bounded — an SEH guard does not bound a wait.
- **How often**: one call returns 64 frames. At 240 presents a second that is four calls a second to lose none;
once a second keeps the newest 64 and says how many it skipped.
- **What else moves**: a series through region 5 (a layout change, or the reserved words), the Agent's sampler,
the page; the Disclaimer's sentence about what the Overlay's own thread does (a version of its own, the
owner's wording); the Privacy Policy (what is stored: times, no content).
**E — from outside the game: the PCL statistics ETW provider.** Streamline's PCL plugin (and the Reflex SDK's
`pclstats.h` in a game without Streamline) registers `PCLStatsTraceLoggingProvider`
(`{0d216f06-82a6-4d49-bc4f-8f38ae56efab}`) and writes an event for every marker (`PCLStatsEvent`: `Marker`,
`FrameID`), which is what NVIDIA's own overlay listens to. No injection, so it would also be the only route for
a game the guard refuses. What it costs:
- **It gives the game-side markers only** — simulation, render submit, present, the ping. The driver, OS queue
and GPU times are not in these events; NVIDIA's overlay has them from the driver.
- **A trace session needs administrator rights** (or membership of *Performance Log Users*); the Agent is
unelevated by default (D34).
- **Listening changes what the game does.** Enabling the provider sets the game's `g_PCLStatsEnable`, and its own
ping thread then posts a registered window message — or, where the game chose a key, `F13`–`F15` key messages
— to the game's window every 100–300 ms. It is the game's code and the vendor's design, and it is what any
listener causes; FrameLedger would still be the cause, in a game it is not even injected into.
- **It reopens a closed decision**: ETW left the product on 2026-08-27/28 (Tier 2 is not a measurement;
`20_OPEN_QUESTIONS` §G keeps the question of a no-injection measurement).
**F — NVAPI from the Agent.** `NvAPI_D3D_GetLatency` takes the game's device. Another process has none. Not a
route.
### What each catalog entry would become
| Entry | Route A, B | C | D | E |
|---|---|---|---|---|
| `nvapi.reflex.frame_report` | refused | (refused) | **read**, Direct3D | game-side times only |
| `nvapi.reflex.sleep_status` | refused | part | **read**, Direct3D | — |
| `nvapi.reflex.mode` | refused | — | N/A: no getter returns it | — |
| `sl.reflex.state` | refused | (refused) | the same numbers, under the NVAPI entries | — |
| `sl.pcl.markers` | refused | — | — | counted |
### What this reading settles, and what it leaves to the owner
- **Settled by reading**: A and B are the refused classes; C breaks the vendor's own contract; F does not exist.
§M14's pre-committed "no route without a patch at a run-time address" does **not** fire — D has no patch and E
has no injection — so the entries do not go to N/A on this document alone.
- **The owner's**, because each crosses a line the rules draw today:
1. **Route D — go to a spike, or no?** A go means: `hook-harness` and a stub `nvapi64.dll` first (the call
made, the report read, the overhead measured against NFR-1, a blocked call and a missing module shown to be
harmless), then one real title of the owner's choice, never first in a game inside the anti-cheat
exception's trial (D38). Pre-committed for that spike: a fault or a hang attributable to the call, an
overhead above NFR-1, or a Direct3D 12 device that cannot be obtained without reading what rule 4 forbids
→ Reflex latency stays N/A and the page says why. A go also needs the owner's ruling that calling a
driver getter with the game's device is inside rule 4, asked with the rule quoted, as §M15 asks for its
own amendment.
2. **Route C — recorded as refused?** (Recommended: yes.)
3. **Route E — leave §G closed, or reopen it for latency markers?** (Recommended: leave it; the route gives
the smaller half of the number and makes a game post messages to itself.)
- **Not looked into**: Reflex on Vulkan; AMD Anti-Lag 2 and Intel XeLL, which have their own interfaces; what
`nvapi_Direct_GetMethod` is (the vendored headers do not declare it; the vendored import library names it; nothing here
would use it); what a
real game's interposer and its Reflex and PCL plugins export (the above is the headers' and the source's word,
not a dump of a game's files).
### The owner's answers (2026-11-13, D132 and D133)
1. **Route D: go — to the spike on `hook-harness` and a stub first.** With it the ruling a go needed: calling a
driver's read-only getter with the device of the swap chain the Overlay hooks is inside rule 4 (`CLAUDE.md`
carries the note). No real game until the owner has seen the spike's result; no shipped code changes for it.
2. **Route C: refused.**
3. **Route E: left closed**; §G is not reopened.
The spike is [`spike-notes` §31](31-reflex-route-d-harness.md), run on 2026-11-13: none of the stop conditions
above fired on the harness, and it ends in the question of one real title.
### Added 2026-11-14: Streamline's two programming guides, read against this note and the code
`docs/ProgrammingGuideReflex.md` and `docs/ProgrammingGuideDLSS_RR.md` of NVIDIA's Streamline repository, both
"Version 2.14.16", fetched as text and read whole — at the owner's request. What they add, and what they change.
**Reflex**
- **Low latency and frame generation: the dependency runs one way.** The guide lists the plugin's parts as
"distinct to each other without any cross-dependencies", wants Reflex's default to be *On*, and its checklist
tests *Reflex On* with frame generation **off**. Frame generation is what needs Reflex. The owner's "without
frame generation the game cannot switch Reflex on" is true of some of the owner's games' menus, not of Reflex.
The sessions of 2026-11-14 agree (HANDOFF, item 47): low-latency mode on at a multiplier of 1 in four games;
Phantom Blade Zero switches it on with frame generation; Clair Obscur without frame generation has it off.
- **The sleep and the markers are sent whatever the mode.** "`slReflexSleep` and `slPCLSetMarker` must always be
called even when Reflex Low Latency mode is Off"; the verification overlay's *App Called Sleep* stays 1 with
Reflex off. So `bUseGameSleep` says the game calls the sleep, **not** that Reflex is on, and a report exists
with Reflex off — which is what the sessions on 0.1.0-beta.48 showed (HANDOFF, item 46). Phantom Blade Zero
without frame generation is exactly this: the game sleeps, low-latency mode off, 1 062 reads.
- **`slReflexSetOptions` is called once, and again only when an option changes.** A hook on it would see nothing
in most sessions even if it could be placed: `nvapi.reflex.mode` stays refused for a second reason.
- **The markers are the report.** `slPCLSetMarker` replaces `NvAPI_D3D_SetLatencyMarker`, and the times the
markers carry (simulation, render submit, present) are the ones the driver hands back in the report this
product already reads (route D). `sl.pcl.markers` would add a count of calls, the ping and the flash trigger —
nothing a player reads. **Recommended to the owner: leave Reflex and PCL out of train D** — **decided the
same evening (D153)**: whether a game has Reflex is all that is wanted.
- **PC latency is not in the report, and NVIDIA takes it from outside the game**: its PCL statistics, read by
tools started from an administrator's prompt (route E here). D147 stands on that; two design tables that still
promised "PC latency as reported by Reflex" were corrected today (`03_METRICS`, `05_DETECTION`).
- **The stages, in NVIDIA's words** (the verification overlay's table): *driver* is from the first buffer
submission to the last present submission; *OS queue* from the first significant buffer submission to the end
of the GPU's work; *GPU* from where its rendering starts to where it ends; *total render time* is "time spent
by GPU rendering". With frame generation the test's own columns split the GPU's part at "DLSS FG start". **Not
reconciled with what is stored**: an OS queue that ends with the GPU's work should not be shorter than the
GPU's span, and in Phantom Blade Zero it is (12.9 ms beside 16.4 ms). The guide's table describes an overlay,
not the fields of `NV_LATENCY_RESULT_PARAMS`; nothing is concluded from it, and no label was changed on it.
- **Where Reflex helps**, by the guide: GPU-bound, or V-Sync with variable refresh; not CPU-bound. FrameLedger
shows the driver's switches and draws no conclusion about benefit.
**DLSS Ray Reconstruction**
- **The code already has it right, and two comments had the reason wrong.** The guide: "When DLSS-RR is switched
on, it completely overrides DLSS (Super Resolution)", in the quality mode set for DLSS, through
`slEvaluateFeature(sl::kFeatureDLSS_RR, …)`. That is the present detour's RR arm (upscaler DLSS, the RR fact
beside it), measured on Cyberpunk 2077 on 2026-08-15. `fl_shm.h` and `dllmain.cpp` still said super resolution
and RR "run together" / "in the SAME frame"; they do not, and the comments now say what is true.
- **Off is a mode, not a release**: "set `sl::DLSSDOptions.mode` to `sl::DLSSDMode::eOff`, note that this does
NOT release any resources". So *loaded* and *allocated* (the Advanced tables of 0.1.0-beta.45) are not *in
use*; the RR verdict is taken from evaluations and from nothing else, which is why it is right.
- **If `sl.dlssd.options` is ever read** (it is a spike entry, set through a call no hook sees): the structure
carries `worldToCameraView` and `cameraViewToWorld` — camera matrices, never read (rule 4) — and RR ignores the
DLSS options `sharpness` and `useAutoExposure`, so a session's DLSS rows do not describe RR on those two.
What the entry would add is RR's preset for each mode and how normals and roughness are packed. Little.
- **No dynamic resolution under RR**: a change of input size re-initialises the denoiser. Nothing here depends
on it.
@@ -0,0 +1,221 @@
## 31 · Reflex route D on the harness — the driver's two getters, called from a probe *(2026-11-13, after 0.1.0-beta.46, owner decisions D132, D133; `20_OPEN_QUESTIONS` §M14)*
**Why this was measured.** §30 read the routes to Reflex latency per frame and left one that needs no patch: the
Overlay calling the driver's two getters, `NvAPI_D3D_GetLatency` and `NvAPI_D3D_GetSleepStatus`, with the device of
the swap chain it hooks. The owner said go on 2026-11-13 (D132) — **to a spike on the harness first** — and ruled
that such a call is inside rule 4 (`CLAUDE.md`, the rule's note). This is that spike. **No shipped code changed and
no game was touched**: the call sequence lives in an unshipped probe, and nothing loaded into a game includes it.
**The instrument.** `src/native/tools/fl-probe-reflex` (built with the Overlay's flags; ctest
`fl_reflex_probe_stub` for the half a build machine can run):
- `fl_reflex_candidate.h` — what would move into the Overlay: find `nvapi64.dll` **if the process already holds
it** (never load it), ask its export `nvapi_QueryInterface` for the two getters by the ids of the vendored
`nvapi_interface.h`, call them SEH-guarded, summarise the sixty-four frames. It does not link `nvapi64.lib`:
the static library loads the DLL itself, from a trusted location and with a signature check.
- `--stub`: against `tools/vendor-stubs/stub_nvapi64.cpp`, a stand-in module with knobs for what a driver cannot be
made to do on demand.
- `--real`: against the installed driver on this PC's NVIDIA adapter, **with the probe playing the game's part** —
it loads `nvapi64.dll`, calls `NvAPI_Initialize`, and **sets latency markers on its own device** for a few
hundred frames. That is a write, and it is said here plainly: a getter has nothing to return without it, the
process is the probe's own, and no shipped binary contains the call (D108's read-only rule is about the Agent's
bridge; rule 4's note is about a game). The owner was told and approved it with the plan.
### The decision table, written before any run
| # | Measured | Outcome → consequence |
|---|---|---|
| M1 | Stub: the module absent; a getter absent; an error; a report of nothing; sixty-four reports; both device kinds | Anything called or loaded when the module is absent, or a wrong classification → the candidate is wrong and is fixed before any driver run. Otherwise: these cases are harmless as written. |
| M2 | Stub: a call that does not return for 1.5 s, made on a thread of its own and made by a thread that has other work | Made by the busy thread, that thread stands still → **the call may not be made on the Overlay's watchdog** (it supervises, installs and flushes): a thread of its own, or the presenting thread if M6 allows. A thread of its own is a second background thread, and the Disclaimer's sentence about "one" changes. |
| M3 | Real: the getters **before** `NvAPI_Initialize` | A fault → a game that holds the module but has not initialised it would fault under us, and the Overlay cannot initialise NVAPI itself (not a getter): **route D stops** unless initialisation can be known without a call. An error status → harmless: read as "not available". |
| M4 | Real: initialised, 120 frames, **no marker** | Zero frames or an error → as the header says: a game that does not use Reflex reads N/A. Frames with content → record what the driver fills in by itself. |
| M5 | Real: 300 frames **with markers** | Frames with sensible spans → the report is readable by a third party in the process. Zero frames → markers alone are not enough (a sleep mode is needed too): route D reads only games that set one, and the note says so. |
| M6 | Real: the cost of one call (2000 calls, on the presenting thread) | NFR-1 allows 1 µs a present. Once a second on the presenting thread costs `call / presents-a-second` a present: at 60 a second a call may cost 60 µs. Median within that → the presenting thread, once a second, is allowed. Above it, or a worst case over 1 ms → **never on the presenting thread**; a thread of its own (M2). |
| M7 | Real: the getters on another thread four times a second for 60 s, beside a thread that presents and sets markers | **Any fault → route D stops** (pre-committed, §30). Error statuses or empty reports → recorded; they decide whether a read is trusted. Frames a second beside the poller more than 0.5 % below alone → the rate is too high or the route too costly: lower it and measure again, or stop. |
| M8 | Real, Direct3D 12: read with the device, read with the queue, markers on each | The device works → the Overlay already holds it (`g_d3d12Device`). **Only the queue works → Direct3D 12 is out of reach**: the Overlay does not hold the queue (`ResolveApi`: `GetDevice` does not return it) and may not go looking for it. Route D would then be Direct3D 11 only. |
| M9 | Real: the unit of the report's times | The gap between two frames' `simStartTime` against the probe's own frame time. Whatever it is, is recorded; a unit that cannot be told → the spans are not shown as times. |
Not in this spike, whatever it shows: anything in a game; where the numbers would go (shared memory, the Agent, the
page); the wording of the Disclaimer and the Privacy Policy. Those are a plan of their own, after the owner's go.
### Results
**Where.** The dev box of §Environment: Windows 11 Pro build 29648, RTX 5080, NVIDIA driver 32.0.16.2114
(`nvapi64.dll` of the same version), 2026-11-13. `--stub` on software devices; `--real` three times (and a fourth,
shorter, after a fix to the probe — row M6), on the NVIDIA
adapter, each with a 256×256 composition swap chain presented without a wait — some 14,000 to 17,000 frames a
second, which is what makes the cost of anything beside it visible and is nothing like a game.
| # | What was seen | Against the table |
|---|---|---|
| M1 | Stub, all 20 checks pass: with the module absent `Resolve` says so, calls nothing and loads nothing; a module without the getters is refused whole; an error comes back as the error; a report of nothing is zero frames; sixty-four reports are read as the header lays them out; the driver is handed the device and only the device, Direct3D 11 and Direct3D 12 alike; no device, no call. | Harmless as written. |
| M2 | A call held for 1.5 s on a thread of its own: another thread ticked 16 times in its first second. The same call made by that ticking thread: **0 ticks** in the second. | **Not on the watchdog.** A thread of its own, or the presenting thread (M6). |
| M3 | The getters **before** `NvAPI_Initialize`, with the module loaded: both resolve, `GetLatency` returns `NVAPI_OK` and sixty-four empty frames. No fault, three runs. | Harmless: an uninitialised NVAPI reads as "nothing reported". |
| M4 | Initialised, 120 frames, no marker: `NVAPI_OK`, zero frames with anything; the sleep status is `NVAPI_OK` with every field zero. | As the header says. A game that does not use Reflex reads N/A, and the read costs it nothing else. |
| M5 | 300 frames with markers, **no sleep mode set**: sixty-four frames, the newest one or two ids behind the last frame drawn. Per frame the driver fills simulation end, render submit start/end, present start/end, driver end, OS queue start/end and GPU start/end; `inputSampleTime` and `driverStartTime` stay zero (the probe sets no input marker). | **Readable by a third party in the process, from markers alone.** |
| M6 | 2000 calls back to back on the presenting thread: `GetLatency` median **0.50 µs** in the three runs (mean 0.53–0.54, worst 1.2 / 1.5 / 3.0 µs) and **1.50 µs in a fourth**, made that evening after a fix to the probe, with the PC doing other things (mean 1.54, worst 70 µs); `GetSleepStatus` median 0.10 µs in all four, worst 0.7 / 0.8 / 62 / 12 µs. | **Far within NFR-1 once a second**: at 60 presents a second a call may cost 60 µs, and the medians are 0.5–1.5. **Not within it at every present**: the fourth run's median alone is over the microsecond NFR-1 allows a present, and single calls of 62 and 70 µs were seen. Once a second is what is proposed. |
| M7 | 60 s, four reads a second from another thread beside the presenting, marker-setting thread, three runs: **228 reads each, none not OK, none faulted, none empty**; worst read 103 / 112 / 248 µs on the polling thread; no marker error. Frames a second beside the poller against alone: −1.3 % (one 60 s leg against one 5 s leg), then in six alternating 10 s legs −1.4 % and −0.6 % — **with single legs between +5.9 % and −9.9 %**. | **No fault: the route does not stop.** The half-percent line cannot be read off this: the loop's own drift from leg to leg is ten times it, in both directions. So a dose was measured instead — next row. |
| M7b | The same poller **never pausing**: 38.3 million reads in 30 s (**1.28 million a second**), none not OK, none faulted, worst read 2.7 ms; frames a second −3.3 % (legs −1.1 %, −5.3 %, −3.7 %). | Three hundred thousand times the proposed rate costs about three percent of an uncapped loop. Four reads a second is that divided by 320,000: not measurable, and not half a percent. **The driver also took 38 million reads beside some three million marker calls without an error** — the concurrency §30 could only infer. |
| M8 | Direct3D 12: markers set with the **device** — no error; read with the device and read with the queue both return the same sixty-four frames; the sleep status is `NVAPI_OK` with either; markers set with the queue work as well and are read back with the device. | **The device works**, and it is the object the Overlay already holds. |
| M9 | `simStartTime` of two neighbouring frames 91 and 43 apart in two runs where the probe measured 81 and 64 µs a frame, and the absolute values are the time of day (1 794 560 148 756 529 on 2026-11-13). | **Microseconds, on the clock of the Unix epoch.** Spans are differences of two of them. |
**One thing seen that is not understood.** In the first of the three runs the newest frame of the Direct3D 11
report had a `presentEndTime` 1 844 674 519 µs — half an hour — after its own `simStartTime`; the other two runs
show 36–72 µs there, and the Direct3D 12 reads 46–51 µs every time. The frame was the newest in the report, read
within microseconds of being presented. Whatever it was, **the newest frames of a report are not final**: a reader
must not take the last few as complete, and must refuse a span that no frame can have.
### What this settles, and what it does not
- **Settled, on this driver and this adapter**: the two getters can be found without loading anything and called
with a Direct3D 11 device or the Direct3D 12 device the Overlay holds; they do not fault before initialisation,
without markers, or beside a thread that sets markers at 16,000 frames a second; a call costs half a microsecond
to a microsecond and a half;
the report is readable from markers alone and its unit is known. None of §30's stop conditions fired: no fault or
hang attributable to the call, no overhead near NFR-1, and the Direct3D 12 device is the one already held.
- **Decided by M2 and M6 together**: the call belongs on the **presenting thread, once a second** — a microsecond
or so a second, with a worst case seen of 70 µs — not on the watchdog and not on a new thread. So the Disclaimer's "one background thread"
would not change; what would change is its list of what the part inside the game does at a present.
- **Not settled, because only a game can say**: a game's own device — many threads, deferred contexts, a device a
game is about to release; what a real title's report holds (input sample, the GPU times of a real frame, frame
generation's `aiFrameTimeUs`); whether a title that uses Reflex through Streamline alone leaves `nvapi64.dll`
findable by name (not looked at); a driver other than 32.0.16.2114, and any card but this
one; a call that blocks — the stub shows what a block does to its thread, and nothing here made the driver block.
- **Not looked into**: Direct3D 11 in the Overlay, which keeps no D3D11 device today (`ResolveApi` releases it):
reading there means holding one, a reference the Overlay does not take now; and the incomplete newest frame above.
### The question for the owner
**Go to one real title, or stop here?** A go means a plan of its own before any code: the two calls in the present
detour once a second, **only while a session asks for Advanced metrics** (D111), only where the game already
loaded `nvapi64.dll`, Direct3D 12 first (the device is already held); a place for the numbers — a series beside the
frames, since a report arrives after its frame's record is written (§30); `nvapi-readonly-check` extended to the
Overlay and taught that `Sleep` is a writer; the Disclaimer's and the Privacy Policy's wording; the first run on a
title of the owner's choosing that is not inside the anti-cheat exception's trial (D38), with what each outcome
means written before it. A no leaves `nvapi.reflex.*` and `sl.reflex.*` at N/A with the reason, and this note as
the record of why it was within reach.
**Answered 2026-11-13 (owner decision D136): go, to a plan.** The title is Cyberpunk 2077. No code before the plan
is approved, and no release that carries the calls before a stop while a game runs has been exercised on
0.1.0-beta.46 (owner-only item 44 (d)).
### The leg with a sleep (owner decision D139) — the table, written before the run
The plan's review found the largest thing the runs above never touched: **no run set a sleep mode or called
`NvAPI_D3D_Sleep`**, and a game that uses Reflex does both, every frame. If the driver makes a getter wait behind
the game's sleep, a read on the presenting thread holds a present for as long as the sleep lasts. The owner said:
measure it in the probe first (D139). The probe plays the game's part once more — `NvAPI_D3D_SetSleepMode` (low
latency on, a minimum interval of 4 166 µs, markers used) and `NvAPI_D3D_Sleep` at the start of every frame, on its
own Direct3D 12 device, in its own process. **Both are writes**, as the markers were; no shipped binary contains
either, and the sleep mode is switched back off before the device is released.
| # | What is measured | What it decides |
|---|---|---|
| S1 | The sleep mode takes: `SetSleepMode` returns OK, the sleep status then reads low-latency on, and a frame takes about the interval | Not OK, or the status does not change → the leg did not exercise a sleep; said so, nothing concluded from S2–S4. |
| S2 | The two getters on the thread that sleeps and presents: one `GetLatency` after each of 2000 frames, and 2000 calls back to back | **Median above 60 µs → stop** (NFR-1 at 60 presents a second). **Any call of 2 ms or more → stop**: on a presenting thread that is the latch of D141, in every game that uses Reflex. |
| S3 | The getters on **another** thread four times a second for 20 s, and without a pause for 10 s, while the first thread sleeps and presents — against the same two runs on the same device **before** the sleep mode was set | **A fault → stop** (§30). Reads of 2 ms or more that the control does not have → the getter waits behind the sleep: **stop**, and the note says the route reads only games that do not sleep. Reads of 1 ms or more at the same rate as the control → the scheduler, not the sleep. |
| S4 | What the report holds once the game sleeps: frames with anything, and the sleep status's fields | Recorded: it is what a real title's report would look like more closely than M5's. An empty report → a sleep mode changes what markers alone gave; recorded, the reader's "nothing reported" covers it. |
A stop here means no Overlay code is written and the owner is asked again.
### The leg with a sleep — results
Three runs of `fl-probe-reflex --real 10` on 2026-11-13 (RTX 5080, driver 32.0.16.2114), the leg after R5.
**No stop condition fired.**
| # | What was seen | Reading |
|---|---|---|
| S1 | `SetSleepMode` returned OK in all three runs. A frame then took 4 159 µs (the interval asked for is 4 166), 4 037–4 041 of them inside `NvAPI_D3D_Sleep`; no sleep error. The sleep status afterwards: interval 4 166 µs, `bUseGameSleep` 1 — and **`bLowLatencyMode` 0**, although the mode was set with it on. | The sleep was exercised: the thread stood inside it 97 % of the time. **The status's low-latency field is not the flag the game set**, on this driver; why is not known. It is shown as the driver's word, never as "the game switched Reflex on". |
| S2 | One `GetLatency` after each of 2000 sleeping frames, on the thread that sleeps and presents: median **2.5 / 2.6 / 2.6 µs**, mean 2.7–2.8, worst 66.9 / 16.8 / 8.0 µs, none of 2 ms. Back to back: median 0.7 µs (worst 10.2 / 81.6 / 4.1); `GetSleepStatus` median 0.2 µs. | Far inside both limits. A read after a frame that slept costs four or five times a read made back to back — a cold cache, by the look of it — and that is the Overlay's case: **about 2.6 µs, once a second**. |
| S3 | From another thread while the first slept and presented at 240 frames a second. Four reads a second for 20 s: 80 reads a run, none not OK, none faulted, worst 64 / 82 / 12 µs (the control before the sleep mode: 92 / 209 / 119 µs). Without a pause for 10 s: **10.26 / 10.26 / 10.43 million reads**, none not OK, none faulted, worst 442 / 635 / 286 µs, **none of a millisecond** (the control: 9.97 / 9.94 / 9.89 million, worst 1 605 / 1 390 / 203 µs, one read of a millisecond in two of the three runs). | **The getter does not wait behind the sleep.** Thirty-one million reads were made while another thread was inside the sleep nearly all the time, and none took a millisecond; the slow reads there are belong to the control. |
| S4 | The report: sixty-four frames with content. `driverStartTime` is filled in all sixty-four — it was zero in M5, without a sleep mode; `inputSampleTime` in none (the probe sets no input marker). Simulation start to present end 102–151 µs, to GPU end 150–191 µs; `gpuActiveRenderTimeUs` 0. | With a sleep mode the driver span is reported. A reader that treats a zero end point as "span not reported" covers both shapes. |
**What it does not say**: a game's own sleep, on a game's own threads, with a real frame's work between the sleep
and the present; frame generation's pacer; another driver. Those are what the first real title is for (D136).
### What moving the sequence into the Overlay found (2026-11-13, the same evening)
The call sequence became `FrameLedger.Overlay/include/fl_reflex.h`, with one check the candidate did not have: that
the two addresses the module answers with are code of the module. **Run against the installed driver, that check
refused the driver.** `C:\Windows\System32\nvapi64.dll` is a shim: it answers both ids with addresses inside
`nvapi64_impl.dll`, in the driver's own store (`…\DriverStore\FileRepository\nv_dispi.inf_amd64_…`), an image of
its own. Three consequences, all in the header now:
- An address is accepted when it is **executable code of an image the process has loaded** — which still refuses
the heap, a stack and nowhere — not only when it is inside `nvapi64.dll`.
- **A reference is kept on that image too.** The addresses are code only while `nvapi64_impl.dll` stays loaded, and
what the shim does with it when a game unloads NVAPI was not looked at.
- **The note above was wrong about Control Flow Guard.** It reasoned from `nvapi64.dll`, which is not a CFG image;
the code that is called is in `nvapi64_impl.dll`, which is one (`dumpbin`: instrumented, a function table of
5 332 entries). The probe enforces CFG and has made the calls from the shipped header, so on this driver the two
getters are valid targets. A game that enforces CFG, on a driver the probe was not run against, is not covered:
a target that were not valid is a fast-fail no guard catches.
The stand-in module answers with addresses of its own, which is why the stub leg could not find this; only the
real run could. The probe's `--real` now prints the image each address lies in.
### A run on a busy PC, and what it says about the 2 ms latch (D141)
The run that confirmed the corrected header (`Resolve` OK, the reads as before) was made while a native build and an
injected test ran on the same PC — the probe's own loop made 9 000 frames a second where it had made 16 000. It is
the only run so far that was not alone, and it is the nearer one to a game:
| | Alone (three runs) | Busy (one run) |
|---|---|---|
| `GetLatency` after a sleeping frame, on the thread that sleeps and presents (2000 reads) | median 2.5–2.6 µs, worst 8–67 µs | median 3.4 µs, worst 130 µs, none of 2 ms |
| From another thread without a pause, while the first sleeps: reads of 1 ms or more / of 2 ms or more | 0 / 0 in 10.3 million | **536 / 25 in 2.9 million**, worst 3.5 ms |
| The same before the sleep mode was set | 1 / 0 in 9.9 million | 6 / 2 in 3.2 million, worst 2.2 ms |
| Four reads a second for 20 s | worst 12–82 µs | worst 110 µs, none of 1 ms |
**A reading, not a measurement of a game**: a thread that does nothing but call is taken off its processor now and
then on a busy PC, and a read that was under way then is as long as the pause. Twenty-five such pauses of 2 ms or
more in ten seconds of calling is 2.5 a second; a read of 3 µs made once a second is inside one of them about once
in a hundred thousand reads. **So the latch of D141 will, on a busy PC, sometimes switch the read off for a pause
that was Windows' and not the driver's** — of the order of a few sessions in a hundred, for sessions of two hours,
if a game's presenting thread is paused as often as this probe's poller was; a presenting thread usually is not.
The session then says *stopped: slow* with the microseconds, and the next start of the game reads again. That is the
trade the owner chose with D141 (one call, not two): it errs towards the game. The number published with the state
is what tells the two apart afterwards — a driver that holds a call does it again and for longer; a pause is a few
milliseconds, once.
**The owner was asked again with this table in hand and kept D141 as it is: one call** (2026-11-13).
### The first real game (2026-11-14, 0.1.0-beta.47, the owner's sessions 208–210)
Phantom Blade Zero, Direct3D 12, three sessions — frame generation ×3, then off twice. What only a game could say,
and now has for one title on this driver:
- **The game's own device is the one the driver takes**, and `nvapi64.dll` is findable by name in a game that uses
Reflex through Streamline: resolved four to five seconds after the attach, the shim in System32 and the getters
in `nvapi64_impl.dll`.
- **No latch fired in 1 515 seconds of reading**: the slowest call was 24, 20 and 45 µs, under a game's real load
and with frame generation's pacer running. The busy-PC reading above put a false latch at a few sessions in a
hundred at worst; three sessions cannot confirm or deny that.
- **The report is real**: with frame generation ×3 the total is 23 ms; without it, 5.4 ms at 149 presents a second
and 15.9 ms at 67. `inputSampleTime` is not read; `driverStartTime` is reported (the driver series has points).
- **The sleep status' low-latency switch reads on in a game** (on at times, off at others) — it never did for the
probe. And the driver's `fgMultiplier` read 3 in the session FrameLedger measured at ×3.0.
- **`gpuRenderStartTime` lies before `simStartTime` in this game**: start to end is 16.4 ms where simulation start
to GPU end is 5.4 ms. The span includes whatever the GPU did before this frame's simulation began — it reads
like the interval between two frames' GPU ends. The header's `gpuActiveRenderTimeUs` is the active time and is
not read by FrameLedger. The probe could not show this: its frames have no GPU work to speak of.
Not said by these sessions: ~~Cyberpunk 2077 (REDengine's own NVAPI use)~~, a game that enforces Control Flow
Guard, another driver, and whether the player feels the read.
**Cyberpunk 2077, the same morning (sessions 211 and 212).** Frame generation ×2 for 105 s of the first, off in
the second at 238 FPS. Read in both; the slowest call 34 and 65 µs; a report in 97 % and 98 % of the samples.
- **The spans are sane there**: total 11.7 ms at 238 FPS, GPU render 2.7 ms, never longer than the total (0 of
429 samples). With frames generated the total is 13.5 ms; in the menus 4.1 ms. So `gpuRenderStartTime` before
`simStartTime` is Phantom Blade Zero's, where *GPU render* sits at 16.4 ms whether the game presents 149 or 67
times a second.
- **`fgMultiplier` read 1 in the session that generated ×2**: the word carries the last value, and the session
ended in a menu. The table needs the values seen, not the last one.
- **Samples are 1.32 s apart** (the Agent's `t_ms`), reads one second apart: a read is not always the one sampled.
- **Left alone as designed**: two Direct3D 11 games (state 1, one line) and a Direct3D 12 game that never loaded
`nvapi64.dll` (state 2, said once, asked again each second without a line).
**The owner's ruling on the span (2026-11-14): read the driver's active time** — `gpuActiveRenderTimeUs`, which
the header defines as the same interval "excluding the idles in between", and with it `gpuFrameTimeUs`. Both are
members of the report already fetched.
+2 -2
View File
@@ -1,7 +1,7 @@
{
"$comment": "Pins the SDK band. Without this, dotnet picks the newest installed SDK — on a machine with a .NET 11 preview present that silently changes analyzer behaviour under TreatWarningsAsErrors and makes local and CI disagree for reasons invisible in a diff. See docs/12_BUILD.md.",
"$comment": "Pins the SDK band: .NET 11 since 0.1.0-beta.37 (owner decision D94, 2026-11-11). Without this, dotnet picks the newest installed SDK — on a machine with a .NET 12 preview present that silently changes analyzer behaviour under TreatWarningsAsErrors and makes local and CI disagree for reasons invisible in a diff. The alpha branch carries this same file while its line rests. See docs/12_BUILD.md.",
"sdk": {
"version": "10.0.100",
"version": "11.0.100",
"rollForward": "latestFeature",
"allowPrerelease": false
}
+4 -1
View File
@@ -2,7 +2,10 @@
## What you need
- **Windows 10 (22H2) or Windows 11, 64-bit.**
- **Windows 11, 64-bit.** FrameLedger may also run on Windows 10 (22H2), but it has not been tested there and nobody
supports it: it is built on .NET 11, which Microsoft supports on Windows 11.
- **A processor Windows 11 accepts.** On an older one, without the SSE4.2 and POPCNT instructions, FrameLedger does
not start.
- Games FrameLedger can measure: **64-bit DirectX 11, DirectX 12 and OpenGL games.** Other games — 32-bit games,
DirectX 9 games, and Vulkan games you start yourself — are still recorded (how long you played, your PC's sensors,
the game's memory use), but not measured. [Limitations](../LIMITATIONS.md) has the full list.
+7 -4
View File
@@ -19,9 +19,11 @@ A value FrameLedger could not measure reads **N/A** — never 0, and never an es
settled. For FSR it is the mode the game asked AMD's library about and then rendered at; at a dynamic resolution
none is shown.
- **The DLSS line** names the DLSS the game ran: *DLSS 4.5* or *DLSS 4* only when the session showed it (the model
preset the NVIDIA driver reports, or a frame-generation multiplier of ×3 and above), each DLSS library's version and
whether it came from the game or from the graphics driver (the NVIDIA App can swap in the driver's copy), and *DLSS 5
runtime loaded* when it was. With nothing to add to the header's *DLSS*, it is not shown.
preset the NVIDIA driver reports, a frame-generation multiplier of ×3 and above, or the NVIDIA App's *Dynamic* frame
generation, which is DLSS 4.5's), each DLSS library's version and
whether it came from the game or from the graphics driver (the NVIDIA App can swap in the driver's copy). *DLSS 5*
heads the line when the game was seen running its neural rendering — with the share of the frames it rendered —; its
library loaded and nothing seen reads *DLSS 5 runtime loaded* (*not seen running* when FrameLedger looked). With nothing to add to the header's *DLSS*, it is not shown.
- **Frame generation in this session** (sessions recorded with 0.1.0-beta.24 or later) lists every frame-generation
state the session held — `×2.0`, `×4.0`… — each with its own frame rates and its share of the time, then the time
without frame generation, and *Not classified* for time in no state that lasted at least 10 seconds: switching,
@@ -70,7 +72,8 @@ correct it by hand — your correction is labelled as yours.
shows the same. *Sessions at a glance* numbers the same days the same way.
- **Compare** (in the navigation on the left) — pick sessions, even of different games, and see them side by side.
To look closer, hold **Ctrl** and turn the mouse wheel over a chart to zoom, drag to pan, and double-click to reset. The
To look closer, hold **Ctrl** and turn the mouse wheel over a chart to zoom, drag to pan, and double-click to see all of
it again — or Tab to the chart and use the arrow keys, **Ctrl** with an arrow to zoom, and **A**. The
wheel alone scrolls the page.
## Export
+4 -2
View File
@@ -16,8 +16,10 @@ your decision and your risk.
refuses — or stops measuring at the next check — and turns hooking off for that game. **There is no switch that
overrides this**, apart from the one narrow exception below.
- **It never hides.** It keeps its real name and does nothing to avoid being seen by security software.
- **It never reads or changes the game's memory**, its saves or your input. It only observes the graphics calls the
game makes.
- **It never reads or changes the game's memory** or its saves, and never what you type or which buttons you press.
Inside the game it only observes the graphics calls the game makes. From outside it — unless you switch
*Controllers and input* off in Settings — it notes whether a controller, or the keyboard or mouse, is in use, and
sends nothing to a controller.
## What it cannot promise
+38 -18
View File
@@ -20,27 +20,28 @@ What each part of FrameLedger shows and what each button does. What the numbers
update** (it waits while a game is being recorded); and the safety notices — a game's hooking refused, measurement
stopped, hooking turned off — each with one button that only says you have read it.
**The pages** are on the left: **Dashboard**, **Games**, **Compare**, **Logs**, and **Settings** at the bottom.
**The pages** are on the left: **Dashboard**, **Games**, **Compare**, **Logs**, and **Settings** at the bottom. **Advanced** joins them, after Logs, while Advanced metrics is switched on.
## The menus
| Menu | Item | What it does |
|---|---|---|
| **File** | **Add game…** | Pick a game's `.exe` file. It is added with hooking off, and its page opens. |
| | **Import library…** | A checklist of the games installed through Steam, GOG Galaxy, Epic, itch.io and the Xbox app (or the Microsoft Store). |
| | **Export** ▸ **Session CSV** · **Session JSON** · **Chart PNG** | Saves the session you last selected or opened. Read [Your data](05-your-data.md) before you share a file. |
| | **Exit** | Quits FrameLedger, even when *Minimize to tray on close* is on. |
| **File** | **Import library…** | A checklist of the games installed through Steam, GOG Galaxy, Epic, itch.io and the Xbox app (or the Microsoft Store). |
| **File** | **Export** ▸ **Session CSV** · **Session JSON** · **Chart PNG** | Saves the session you last selected or opened. Read [Your data](05-your-data.md) before you share a file. |
| **File** | **Exit** | Quits FrameLedger, even when *Minimize to tray on close* is on. |
| **Tools** | **Agent status…** | Opens the Dashboard, where the agent's card is. |
| | **Update detection rules** | The agent reads its detection rules again and checks your games again. |
| | **Vulkan layer registration…** | Opens Settings, where the Vulkan layer is. |
| | **Open data folder** | Opens `%LOCALAPPDATA%\FrameLedger` in Explorer. |
| | **Database maintenance…** | **Check**, **Back up…**, **Sweep…** and **Compact** your database ([When something goes wrong](09-troubleshooting.md#the-database)). |
| **Tools** | **Update detection rules** | The agent reads its detection rules again and checks your games again. |
| **Tools** | **Vulkan layer registration…** | Opens Settings, where the Vulkan layer is. |
| **Tools** | **Open data folder** | Opens `%LOCALAPPDATA%\FrameLedger` in Explorer. |
| **Tools** | **Database maintenance…** | **Check**, **Back up…**, **Sweep…** and **Compact** your database ([When something goes wrong](09-troubleshooting.md#the-database)). |
| **Help** | **User guide** | This guide. |
| | **Check for updates** | Asks GitHub now. |
| | **Report a bug…** | Builds a report on your PC ([When something goes wrong](09-troubleshooting.md#report-a-problem)). |
| | **Online documentation** | The project's page on GitHub. |
| | **Limitations** | What FrameLedger cannot do. |
| | **About FrameLedger** | The version, the licences and the notices. |
| **Help** | **Check for updates** | Asks GitHub now. |
| **Help** | **Report a bug…** | Builds a report on your PC ([When something goes wrong](09-troubleshooting.md#report-a-problem)). |
| **Help** | **Online documentation** | The project's page on GitHub. |
| **Help** | **Limitations** | What FrameLedger cannot do. |
| **Help** | **Accessibility** | What works with the keyboard alone, a screen reader or a contrast theme, what does not yet, and how to report a problem. |
| **Help** | **About FrameLedger** | The version, the licences and the notices. |
## Dashboard
@@ -59,6 +60,9 @@ stopped, hooking turned off — each with one button that only says you have rea
## Games
- **Search games**; sort by **Name**, **Last played** or **Playtime**; **Show as a list** or **Show as a grid**.
- **Filters** (since 0.1.0-beta.34): by **store**, **engine**, **state** (Hooking on, Hooking off, Anti-cheat, Exception,
Not recorded), **installed** and **played**; the title then says how many of your games are shown. **Clear filters**
shows them all again.
- **Refresh** reads your library again: it offers games newly installed through a store, and marks a game whose `.exe`
is gone as *Not installed*. The agent removes such a game by itself after ten minutes when it has no sessions and its
drive is connected — never one where anti-cheat was found (**Settings ▸ Data ▸ Remove uninstalled games** turns this
@@ -103,18 +107,27 @@ It opens from a session row, from the Dashboard, and from the notice when a sess
- **Frame generation in this session** — a bar and a table of every frame-generation state the session held, when it
held more than one, with how frame generation was set.
- **Tags and notes** with **Save**, and **Export CSV**, **Export JSON**, **Export PNG**.
- **Advanced metrics…**, for a session recorded with Advanced metrics on: it opens that session on the Advanced page.
## Compare
Tick **2 to 5 sessions**, from any of your games, and press **Compare**: one frame-rate curve per measured session, and
a table with the best value of each row in bold. Comparing a measured session with an unmeasured one asks first,
because they measured different things.
because they measured different things. The list holds every session; narrow it by **game**, **tier** (measured or
not) and **frame generation** — a session you ticked stays in the list. **Clear selection** unticks them all.
## Logs
What FrameLedger and its agent wrote down. Choose **App** or **Agent**, a level, and search; **Pause** stops the view from
following new lines. **Open logs folder**, and **Export bug bundle**, which does the same as Help ▸ Report a bug….
## Advanced
There only while Settings ▸ Recording ▸ **Advanced metrics** is on. Three tabs: **This PC** — what the graphics
driver and the sensors say right now, read only while the page is open; **Session** — what a session recorded with
the option on kept, in a table and a chart; **Sources** — every reading FrameLedger knows of, and for each one it
does not read, why. [Advanced metrics](12-advanced-metrics.md) describes each.
## Settings
Every switch and box is described in [Every setting](08-settings.md).
@@ -128,8 +141,15 @@ is ready.
## Keyboard and mouse
- **Ctrl+1** … **Ctrl+5** open Dashboard, Games, Compare, Logs and Settings.
- **Alt** or **F10** moves to the menus.
- **Ctrl+1** … **Ctrl+5** open Dashboard, Games, Compare, Logs and Settings, and put the keyboard in that page;
**Ctrl+6** opens Advanced while Advanced metrics is on.
- **Alt** or **F10** moves to the menus; **Alt** with a menu's underlined letter opens it (**Alt+F**, **Alt+T** and
**Alt+H** for File, Tools and Help in English).
- **Esc** closes a dialog the way its Close or Cancel button does.
- Over a chart, **Ctrl** and the mouse wheel zoom, dragging pans, and a double-click resets. The wheel alone scrolls the
page.
- Over a chart, **Ctrl** and the mouse wheel zoom, dragging pans, and a double-click shows all of it again. The wheel
alone scrolls the page.
- **Tab** reaches a chart; then the **arrow keys** move its view, **Ctrl** with an arrow zooms, and **A** shows all of it
again.
What works with the keyboard alone, a screen reader or a contrast theme — and what does not yet — is on its own page:
[Accessibility](../ACCESSIBILITY.md) (Help ▸ Accessibility).
+6 -4
View File
@@ -20,8 +20,10 @@ The link under them, *Read what the hook does and does not do*, opens the safety
| Setting | What it does | Default |
|---|---|---|
| **Minimum session length (s)** | A shorter session is not kept. 5 to 600 seconds. *Agent.* | 30 |
| **Sensor interval (ms)** | How often the PC's sensors are read during a session. 500 to 2000 ms. *Agent.* | 1000 |
| **Sensor interval (ms)** | How often the PC's sensors are read during a session. 500 to 2000 ms. With a value above 1000, not every reading of NVIDIA's latency report — made about once a second inside the game — becomes a point: each point is still one reading ([Advanced metrics](12-advanced-metrics.md)). *Agent.* | 1000 |
| **Raw series kept per game** | How many of each game's newest sessions keep their per-frame data; older ones keep their statistics. 0 keeps everything. *Agent.* | 20 |
| **Controllers and input** | Records what a session was played with: which game controllers were connected and in use, their battery level, and whether the keyboard or mouse was in use — as times, read from Windows and never from the game. No button, stick, key or text is kept, and nothing is sent to a controller. Nothing is read until you have accepted the Privacy Policy that describes it, so after an update the first sessions have no such figures until FrameLedger has been opened once. Off: nothing of this is read or stored. It applies from the next session; a session recorded while it was off is not filled in later. *Agent.* | On |
| **Advanced metrics** | Reads more from the graphics driver and the hardware sensors during a session — the graphics card's memory and load in detail, each fan and voltage, what your display can do, the driver's settings for the game — and stores it with the session. In a game FrameLedger measures from inside it also reads what the game tells DLSS, Streamline and FSR, and — on an NVIDIA card, in a Direct3D 12 game that uses NVIDIA's library — asks the graphics driver about once a second how long each stage of a frame took ([Advanced metrics](12-advanced-metrics.md)). Off: nothing more is read or stored. It applies from the next session; a session recorded while it was off is not filled in later. While it is on, the **Advanced** page is in the list on the left ([Advanced metrics](12-advanced-metrics.md)). *Agent.* | Off |
## Data
@@ -34,15 +36,15 @@ The link under them, *Read what the hook does and does not do*, opens the safety
| Setting | What it does | Default |
|---|---|---|
| **Run the agent as administrator** | Lets FrameLedger measure games that run as administrator, and read CPU temperature with PawnIO installed. Windows asks every time the agent starts, at logon too. A warning explains the risk first; read [Limitations](../LIMITATIONS.md#running-the-agent-as-administrator). The agent restarts to apply it, or at its next start if a game is being recorded. | Off |
| **Run the agent as administrator** | Lets FrameLedger measure games that run as administrator, and — with PawnIO installed — read CPU temperature and, when *Advanced metrics* is on, the processor's, the memory's and the motherboard's sensors. Windows asks every time the agent starts, at logon too. A warning explains the risk first; read [Limitations](../LIMITATIONS.md#running-the-agent-as-administrator). The agent restarts to apply it, or at its next start if a game is being recorded. | Off |
| **Start the agent at logon** | **Register**, **Repair** or **Remove** a task that starts the agent when you log on, for your user only and without administrator rights. Without it, FrameLedger starts the agent when you open it. | Not installed |
## Appearance
| Setting | What it does | Default |
|---|---|---|
| **Theme** | Follow system, Light or Dark. Applies at once. | Follow system |
| **Language** | English, Tiếng Việt or 日本語. The window is rebuilt in the new language at once. | English |
| **Theme** | Follow system, Light or Dark. Applies at once. While a Windows contrast theme is on, FrameLedger uses that theme's colours whatever you choose here, and says so under the box. | Follow system |
| **Language** | English, Tiếng Việt or 日本語. The window is rebuilt in the new language at once, and stays on Settings (since 0.1.0-beta.34; before, it could open on the last game you looked at). | English |
## Games
+1 -1
View File
@@ -32,7 +32,7 @@ While the game runs, a notice may say *hooking refused*.
| The game runs in an app container, or Windows protects it in a way the component can neither load into nor hook. | Nothing: an older Microsoft Store app (UWP), or a game under code protection, is recorded, not measured ([Limitations](../LIMITATIONS.md#xbox-app-and-microsoft-store-games)). |
| **Disable all hooking** is on. | Switch it off in Settings ▸ Capture. |
| The game was already running when FrameLedger was updated. | Close the game and start it again. |
| Hooking was switched off after the game crashed twice soon after it started. | Switch it on again in the game's page if the cause may be gone — after a driver update, for example. |
| Hooking was switched off after the game crashed twice soon after it started. | Switch it on again in the game's page if the cause may be gone — after a driver update, for example. The count starts again when you do (since 0.1.0-beta.44). |
| Hooking was switched off after **Play** started the game directly twice, and each time the game handed itself to Steam or closed at once. | Switch it on again, and set Settings ▸ Games ▸ **Play button** to *Through the store*: some Steam games will not start any other way. |
| The game's file is gone. | If only its drive letter changed, FrameLedger follows it within about 15 seconds; otherwise **Change executable…** on the game's page. |
| The component could not be loaded into the game. | [Report a problem](#report-a-problem). |
+3 -1
View File
@@ -16,7 +16,9 @@ What the words on FrameLedger's screens mean, in alphabetical order. How each fi
says. Sessions are still recorded.
- **Displayed FPS** — the frames you saw, generated frames included.
- **DLSS 4, DLSS 4.5** — NVIDIA's DLSS generations. FrameLedger names one only when the session shows it: the model
preset the NVIDIA driver reports, or a frame-generation multiplier of ×3 and above.
preset the NVIDIA driver reports, a frame-generation multiplier of ×3 and above, or the NVIDIA App's *Dynamic*
frame generation (DLSS 4.5 only).
- **DLSS 5** — NVIDIA's neural rendering. Named only when the game was seen running it (since 0.1.0-beta.34).
- **FG factor** — Displayed FPS divided by Native FPS, shown as `×1.9 FG`.
- **FPS** — frames per second over the time measured.
- **Frame generation** — the technology (DLSS Frame Generation, FSR Frame Generation) that inserts frames the game did
+23 -11
View File
@@ -2,14 +2,25 @@
## What an alpha is
Besides the betas, FrameLedger publishes **alphas**: the same program as a beta, built on a version of .NET that
Microsoft has not finished yet — .NET 11's release candidate first. They exist so that a new .NET can be tried before the
betas move to it. An alpha's version ends in `-alpha.N` (`0.1.0-alpha.1`), its page on GitHub says *experimental alpha*,
and its installer is `FrameLedger.App-alpha-Setup.exe`.
Besides the betas, FrameLedger publishes **alphas**: the same program as a beta, built to try a newer version of .NET —
sometimes one Microsoft has not finished yet. They exist so that a new .NET can be tried before the betas rely on it. An
alpha's version ends in `-alpha.N` (`0.1.0-alpha.1`), its page on GitHub says *experimental alpha*, and its installer is
`FrameLedger.App-alpha-Setup.exe`.
**If you want results you can rely on, use the beta.** An alpha can crash, record wrongly or fail to update in ways no
beta has. The part of FrameLedger that goes into a game is the beta's own and does not use .NET; everything that starts
it, reads it and stores what it measured runs on the new .NET.
it, reads it and stores what it measured runs on the alpha's .NET.
## Right now: the alphas are paused
The alphas tried .NET 11. Since 0.1.0-beta.37 the betas run on .NET 11 too, so **0.1.0-alpha.4 is the last alpha for
now**: it brings an installed alpha onto the finished .NET 11, and after it no alpha is published until there is a newer
.NET to try.
A paused alpha gets **no updates** — no fixes, and not the list of anti-cheat software FrameLedger refuses to touch, which
only arrives inside an update. Its banner tells you so. **Install the latest beta over it** (below); your data stays.
One more thing: once ten betas have been published after 0.1.0-alpha.4, an older alpha can no longer find it as an
update, and installing the beta by hand is the only way forward.
## Before it runs
@@ -18,7 +29,8 @@ tick: *I understand that this is an experimental build and I use it at my own ri
Until you accept, the alpha's capture agent does not run at all. The notice is shown again whenever its text changes.
While an alpha runs, the title bar carries an **ALPHA** badge (hover it to see the .NET version), and a yellow banner
above every page says what it is each time it starts; you can close the banner for that run. **Help ▸ About** names the
above every page says what it is each time it starts — while the alphas are paused, that they are, and to install the
beta; you can close the banner for that run. **Help ▸ About** names the
.NET it runs on.
## Your data
@@ -45,9 +57,9 @@ alpha.
## Windows and your PC
Microsoft supports .NET 11 on Windows 11 24H2 and later; on Windows 10 only in its LTSC and IoT editions. An alpha may
work on Windows 10 22H2 Home or Pro, but nobody supports it there. .NET 11 also needs a processor with SSE4.2 and POPCNT,
which every processor Windows 11 accepts has; on an older one the alpha does not start at all.
An alpha needs what a beta needs since 0.1.0-beta.37 ([Install](01-install.md)): Microsoft supports .NET 11 on
Windows 11; on Windows 10 (22H2) FrameLedger may run, but it is not tested there and nobody supports it; and on a
processor without SSE4.2 and POPCNT it does not start at all.
A release candidate stops being supported by Microsoft when the final version ships — for .NET 11's, 10 November 2026 —
and the next alpha moves to the final version.
When an alpha runs on a .NET Microsoft has not finished, Microsoft stops supporting that version when the finished one
ships, and the next alpha moves to the finished one.
+178
View File
@@ -0,0 +1,178 @@
# Advanced metrics
A session's summary shows the figures most people want. **Advanced metrics** is for looking closer: everything the
graphics driver and the hardware sensors will say about your PC, kept with each session and shown on a page of its
own.
It is **off unless you switch it on**, and while it is off nothing more is read or stored.
## Switch it on
Settings ▸ Recording ▸ **Advanced metrics**. A new item, **Advanced**, appears at the end of the list on the left
(**Ctrl+6**). Sessions you record from then on keep the extra readings; a session recorded while it was off is not
filled in later.
## This PC
What your PC says right now.
- **About this PC** — the graphics card's firmware, architecture and cores, its memory sizes, the driver's package,
what each display can do with HDR and variable refresh, and the driver's global settings. A row with a number in
brackets unfolds into a list.
- **Right now** — every reading with the time it was read at. NVIDIA's own figures are read every second; the
sensors (fans, voltages, loads) every five seconds, so a sensor's value can be a few seconds old, and its time
says so.
- **Over time** — choose a reading and see its last five minutes. More than eight lines at once cannot be told
apart, so a reading with many parts (every processor core, say) is drawn as its lowest, average and highest.
These readings are taken **only while this page is open or a game is being recorded**, and what the page shows is
not stored.
## Session
Choose a session recorded with Advanced metrics on. You see the PC as it was then, what belonged to that session
alone — whether variable refresh was in use on the display the game ran on, the colour space the game presented in,
the driver's settings for that game — the lowest, average and highest of every reading, and a chart over the
session's length.
### What the game told its upscaler
For a game FrameLedger measures **from inside** (hooking is on for it), a session recorded with Advanced metrics on
also keeps what the game itself handed to DLSS and FSR:
- **DLSS** — the output size, the model preset for each quality mode, whether the game said its colour buffers are
HDR, auto-exposure, alpha upscaling, and the exposure values.
- **FSR** — sharpening (on or off, and how strong), the largest jitter offset, the exposure, the near and far planes,
the **range** of the field of view over the session, the flags of the call, the frame time the game reported, and
FSR's own video memory at the moments the game asked for it.
- **Streamline** (NVIDIA's library that games use for DLSS; since 0.1.0-beta.43) — what the game tells it about
each frame: the largest jitter offset, the motion-vector scale and how the motion vectors are stored, the near and
far planes and whether depth is inverted, the **range** of the field of view, the aspect ratio, whether the
projection is orthographic, and how many times the game reset the upscaler's history (a camera cut, a loading
screen). Where a game draws several views, these are the first view's.
- **FSR frame generation** (since 0.1.0-beta.43) — how the game configured it: on or off, whether it may run
asynchronously, its flags, whether only generated frames are presented, and the part of the picture it generates
in. A setting the game changed during the session reads *on and off*. This is what the game **asked for**; how
many frames were generated is in the session's summary, as before.
- **Streamline's features** (since 0.1.0-beta.44) — what Streamline answered when the game asked about a feature:
which version it is (Streamline's own, and the DLSS model's), and what it requires — the graphics APIs it works
with, whether V-Sync must be off, whether hardware-accelerated GPU scheduling must be on, how many threads,
viewports and buffers, and the Windows and driver versions Streamline found beside the ones the feature needs.
The features carry NVIDIA's own names: `kFeatureDLSS` is DLSS, `kFeatureDLSS_G` DLSS frame generation,
`kFeatureDLSS_RR` ray reconstruction, `kFeatureReflex` Reflex. A game asks these while it starts; many ask
again later, and then they are read. Where a game asks only at its start, the line under the session says
*hooked — the game asks this while it starts, and did not ask again*. The version shown is the one Streamline
reports, which can be newer than the file in the game's folder.
- **Whether a feature is supported and loaded, and what the game did with it** (since 0.1.0-beta.45) — two more
tables, one row for each feature and call. `kFeatureDLSS_G · slIsFeatureSupported` is Streamline's answer in
NVIDIA's own word: `eOk` is yes, and a refusal is named (`eErrorFeatureNotSupported`, `eErrorFeatureMissing`).
`· slIsFeatureLoaded` is *on* or *off*. `· slSetFeatureLoaded` says whether the game switched the feature *on*,
*off* or *on and off* during the session; `· slAllocateResources` and `· slFreeResources` count how many times it
set up and released what the feature needs (up to 15, then `≥ 15`); `slShutdown` counts how many times it shut
Streamline down. These begin when FrameLedger is in the game: what the game did while it started is not there.
In the games run so far every one asks whether its features are supported; the other rows appear in some games
and not in others, and no game has shown an allocation.
These are the game's own words to the upscaler, shown in the vendor's notation (`preExposure`, preset `K`).
FrameLedger does not check what the upscaler did with them.
The last four are read from calls FrameLedger hooks **only for Advanced metrics** — ten calls in all: in a game
where no session has been recorded with the option on since it started, those hooks are not there. The line under
the session says, for each call, whether it was *read*, *hooked — the game did not make the call*, or *not hooked —
the game did not load the library that has the call*; the six calls of 0.1.0-beta.45 are named in one sentence,
under which of the three happened to each.
### What NVIDIA's driver says about each frame's stages
On an NVIDIA card, in a **Direct3D 12** game that has **NVIDIA Reflex** built in, a session with Advanced metrics
on also stores the driver's own latency report — whether Reflex is on or off in the game's menu, because a game
with Reflex marks its frames either way: how long the driver says each stage of a frame took — *simulation*,
*render submit*, *present*, *driver*, *operating system's render queue*, *GPU render*, and the *total* from the start
of simulation to the end of the graphics card's work. They are seven lines under **Latency**, one point about every
second, each the average of the frames of one reading, in microseconds. A table beside them shows the driver's
low-latency state as five switches, by NVIDIA's own names, each *On*, *Off* or *On at times, off at others*, and
`fgMultiplier`: every frame-generation multiplier the driver gave during the session (*1, 2* is a session that
generated frames at ×2 for part of it). **They are the driver's switches, not the game's menu.** `bLowLatencyMode`
is on when the game switched Reflex on *and* when **Low Latency Mode** is set for that game in the NVIDIA App —
an override there shows here, and can read *On at times, off at others* in a game whose own setting never
changed. `bUseGameSleep` says whether the game calls Reflex's sleep each frame. NVIDIA asks games to call it even
while Reflex is off, so *On* there does not say Reflex is on — `bLowLatencyMode` does; some games call it only
while frame generation is on. Frame generation needs Reflex and not the other way round: the mode can be on
with `fgMultiplier` at 1.
**The stages overlap; they are not parts that add up to the total.** And *GPU render* is the time between two
marks the driver sets, at the start and at the end of the card's work on a frame: where the card waits between
pieces of work, the waiting is in it — in one game it is longer than the whole frame. So a second chart, **the
graphics card's time per frame, as the driver works it out**, has two lines of the driver's own figures: *GPU
active render*, how long the driver says the card worked on the frame, and *GPU frame time*, from the end of one
frame's work to the end of the next. **The active time is usually the shorter of the two, and not always**: in
Cyberpunk 2077 without frame generation it is about a fifth longer than the two marks are apart, frame after
frame — the driver counts work its marks do not enclose. Both are shown as the driver gives them. A driver may
give neither; the chart is then not there, and nothing is wrong.
Three things to know about them:
- **They are the driver's numbers, not FrameLedger's.** FrameLedger does not measure these stages; it stores what
the driver reports and has not checked it against another tool. They are not the delay from a click to the screen.
- **FrameLedger asks for them from inside the game**, about once a second — the one thing it does in a game that is
not reading what the game itself passes along. It asks only through NVIDIA's library where the game had already
loaded it, and never loads that library itself. A copy of the library sitting in the game's own folder (some
mods put one there) is not the driver's, and FrameLedger does not use it.
- **It switches itself off rather than disturb the game.** If one request takes two thousandths of a second or
more, or goes wrong, FrameLedger makes no further request until the game is closed. The line under the session
then says *stopped*, and why; the next time the game starts it reads again. On a busy PC this can happen for a
pause that was Windows' and not the driver's.
The line under the session says what happened: *read*; *asked, and the driver reported no frame* (the game does not
use Reflex); *not read* with the reason — the game does not present through Direct3D 12, it has not loaded NVIDIA's
library, or that library is in the game's own folder; or *stopped*.
**Many games will show nothing here**, and the line under the session says why: *the game made none of the calls
these are read from* — most games set their DLSS options through a call FrameLedger cannot see, and only games that
use FSR 3.1 or later through AMD's own library show the FSR rows — or *the session was not measured inside the game*.
It never means the game uses no upscaler; the session's summary still names the upscaler as before.
**Where the camera is, which way it looks and its matrices are never read.** The field of view is kept as a lowest
and a highest value, never as a line over time.
**A game may use a unit of its own.** NVIDIA's and AMD's libraries expect the field of view in radians, and
FrameLedger shows it that way with the degrees beside it. Some games give degrees instead (Cyberpunk 2077 does, to
Streamline): a value too large to be radians is shown as the game gave it, with a line that says *Not radians*, and
is not converted — FrameLedger shows what the game said, not what it may have meant. The number of **resets** is
the number of calls in which the game told the upscaler to drop its history; menus and loading screens raise it. A
frame-generation configuration with no flag set says *None*.
A session's summary has an **Advanced metrics…** button that opens that session here.
## Sources
Every reading FrameLedger knows of, with what it belongs to (this PC, a session, or a session over time), where it
comes from, and whether it is read. Where it is not, the reason is written out:
- **the capture agent is not running as administrator**, or **PawnIO is not installed** — the processor's, the
memory's and the motherboard's sensors need both ([Every setting](08-settings.md), *Run the agent as
administrator*);
- **not present on this PC** — the driver's own figures exist on NVIDIA cards only;
- **read inside a game FrameLedger measures, when the game makes the call** — what the game tells DLSS and FSR
(above);
- **read inside a game FrameLedger measures, from the graphics driver** — NVIDIA's latency report (above);
- **not read: it would take a patch at an address found while the game runs, a function of FrameLedger's handed
to the game in place of the vendor's, or a call the vendor forbids from outside the game's own threads** — what
the game asked of Reflex, Reflex's state through Streamline, and what a game sets and asks through Streamline's
own functions for a feature: DLSS frame generation's options and state, the video memory DLSS estimates for
itself, and the options of ray reconstruction, Image Scaling and Dynamic Vibrance. No later release will read
these;
- **the game makes this call once, before FrameLedger is inside it** — what the game started Streamline with. It is
not read, and no later release will read it;
- **never read** — a serial number, a display's identification data, which other programs are encoding, the camera's
position in the game. FrameLedger does not read these at all.
## Good to know
- A reading that could not be taken is **N/A**, never a guess and never a zero.
- **A drive's temperature is not read**: the sensor library cannot list drives without touching removable ones.
- None of these readings has been checked against another tool yet.
- They are not part of an exported session or a bug report.
More limits are on the [Limitations](../LIMITATIONS.md) page, under *Sensors*.
+5 -3
View File
@@ -4,7 +4,7 @@ FrameLedger records how your games run on this PC — frame rate, frame times, t
with, temperatures, and how much memory the game uses — and keeps every session so you can compare them later.
Everything stays on your PC.
Each page below takes a minute or two to read. The first six are the guide; the next five are there to look things up.
Each page below takes a minute or two to read. The first six are the guide; the next six are there to look things up.
1. [Install and first start](01-install.md)
2. [Record a game](02-record-a-game.md)
@@ -17,10 +17,12 @@ Each page below takes a minute or two to read. The first six are the guide; the
9. [When something goes wrong](09-troubleshooting.md)
10. [Words FrameLedger uses](10-glossary.md)
11. [Alpha builds](11-alpha-builds.md) — the experimental builds on a new .NET, and how to go back
12. [The terms, in plain words](terms-in-plain-words.md)
12. [Advanced metrics](12-advanced-metrics.md) — everything the driver and the sensors say, if you switch it on
13. [The terms, in plain words](terms-in-plain-words.md)
What FrameLedger cannot do is listed on its own page: [Limitations](../LIMITATIONS.md) (in the app: Help ▸
Limitations).
Limitations). How it works with the keyboard alone, a screen reader or a contrast theme, and how to report a problem:
[Accessibility](../ACCESSIBILITY.md) (Help ▸ Accessibility).
**FrameLedger is beta software.** It certainly still has bugs. If something goes wrong, Help ▸ Report a bug… builds a
report on your PC that you can look at before you decide to share it.
Loaded 100 of 755 files, more files were not shown because too many files have changed in this diff. Show more