diff --git a/CLAUDE.md b/CLAUDE.md index d37e3f8..0f29b4e 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -87,9 +87,14 @@ performance budgets docs/11 §6, CI workflows per docs/12 §2 (explicit `permissions:` blocks on every caller — and note the ops repo has **no** `browser-extension-*` family), CHANGELOG, README (badges, screenshots ⚠, install, privacy summary, donate), full manual QA (docs/11 §5), `wxt zip` -clean-install test. Also: point About's "Changelog" link at `CHANGELOG.md` -once it exists (`src/lib/links.ts` currently targets `/releases`), and split -`EditTab.svelte`, which is past the 500-line hard limit. +clean-install test. Also carried in from the Phase 4 review: +- point About's "Changelog" link at `CHANGELOG.md` once it exists + (`src/lib/links.ts` currently targets `/releases`); +- split `EditTab.svelte`, which is past the 500-line hard limit; +- `Button.svelte`'s `.danger`/`.success` variants hardcode `color: #fff`, which + measures **2.75:1** on dark-theme `--danger` (#ff6b7d) — under the 4.5:1 + docs/06 §5 requires. Phase 4 fixed the same class of bug for `--accent-fg`; + these need `--danger-fg`/`--success-fg` tokens and a visual decision. **Exit:** ready for the docs/13 §1 runbook (human performs store submission). ## Working style diff --git a/docs/07_I18N.md b/docs/07_I18N.md index 07f7043..d30232c 100644 --- a/docs/07_I18N.md +++ b/docs/07_I18N.md @@ -129,6 +129,20 @@ stay raw. a tiny helper selects by locale. VI/JA have one grammatical form, so they set `one` and `other` to the same string. They must still carry both keys, or the `satisfies Dict` shape check that §2 relies on for completeness fails. + + Implemented as `selectPluralForm(count, tag)` in `i18n/format.ts` (backed by + `Intl.PluralRules`, narrowed to the two forms — en yields `one`/`other`, vi + and ja only ever `other`, so two forms cover every locale we ship) and + `tPlural(path, count, params?)` in `index.svelte.ts`, which supplies `{n}` + already grouped for the locale so callers pass a raw count. Two tests in + `tests/unit/i18n/dictionaries.test.ts` enforce the shape: every plural key + carries both forms in all three locales, and carries nothing else. + + The current plural keys are `import.warnings.title` and **all seven** + `warnings.*` codes — including the two that interpolate no count, because + `ParseWarning` always carries one and a uniform shape lets `WarningList` call + `tPlural` unconditionally instead of maintaining a list of which codes count + things. Anything else stays a plain string. - Legal Gate: the four legal documents are **English-only** (canonical), with a localized one-line note: "The legally binding documents are in English." - Store listing: EN required; VI + JA listings via CWS dashboard localized diff --git a/docs/15_ROADMAP.md b/docs/15_ROADMAP.md index 4fadd09..3369901 100644 --- a/docs/15_ROADMAP.md +++ b/docs/15_ROADMAP.md @@ -29,8 +29,15 @@ review found real defects (7, 13 and 14 in phases 1–3) — several in the path that delete bookmarks. Budget that review into the remaining phases; a green test run has not once meant "done" on this project. -Phase 4 found three latent defects in code that 338 green tests had already -signed off, all of them unreachable until Phase 4 gave them a caller: +Phase 4's own review filed 33 findings across six lenses; 19 survived +adversarial verification. The most serious was a first-run dead-end: the gate +rendered only on the three routes it blocked, and those three tabs were the +only in-app navigation to them — so a user who reached `#settings` (via the +popup's new gear, which opens a fresh tab with no history) could never get +back to the accept UI. + +Phase 4 also found three latent defects in code that 338 green tests had +already signed off, all unreachable until Phase 4 gave them a caller: `writeSettings` handed a `$state` proxy to `chrome.storage.local.set` (would throw `DataCloneError` in a real browser — `fakeBrowser` stores the reference without cloning, so the suite could not see it); `updateSettings` replaced the @@ -116,3 +123,8 @@ attestation resolver, deadlocking it until reload. | 2026-07-26 | The gate's acceptance checkbox deliberately resets if the user navigates away and back — an un-submitted legal affirmation is not worth persisting | 14 §2 | | 2026-07-26 | Both entrypoints await settings (Manager also legal) **before** `mount()`, trading a blank frame for no locale/theme/gate flicker | 02 §2 | | 2026-07-26 | **Known limitation:** two open Manager tabs do not sync settings or acceptance to each other; last write wins. No `storage.onChanged` listener in v1 | 03 §4 | +| 2026-07-26 | Gate-blocked tabs stay **clickable**; what blocks a route is the gate replacing the tab body, not an inert button. Disabling them removed the only in-app path back to the accept UI | 06 §3 | +| 2026-07-26 | Failure UI shows the localized sentence and puts the raw browser error in a separate detail block — `detail ?? t(key)` made every translation dead code | 02 §7 | +| 2026-07-26 | `--accent-fg` is redefined in the dark blocks alongside `--accent`; white on the lightened violet measured 3.67:1 | 06 §5 | +| 2026-07-26 | Plural keys implemented per `07 §5` (`selectPluralForm` + `tPlural`); `import.warnings.title` and all seven `warnings.*` codes converted | 07 §5 | +| 2026-07-26 | The import session stores an i18n **key**, not a resolved string — `t()` is only reactive where it is called, so a stored sentence never follows a language switch | 03 §5 |