release: 1.5.3 - #203

No actionable comments were generated in the recent review. 🎉

Merge Risk: 🔵 Low · up to 3ef43

The WAnchor focus fix and 1.5.3 metadata updates are ready, but the changelog retains an unused Unreleased link reference that may fail markdown lint and should be cleaned up or accepted.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Version-bump-only release PR; every surface is consistent and the gate numbers reproduce locally, with one cosmetic duplicate heading in the promoted changelog section.

I verified the promotion itself rather than trusting it: grep for 1.5.2 across all .md / .yaml / .yml / .txt / .dart / .json / .lock outside CHANGELOG.md returns nothing, so no surface was left behind. skills/wind-ui/ correctly needs no move on a patch - the frontmatter version: 2.15.0 and the <!-- fluttersdk_wind 1.5.x | Skill v2.15.0 (2026-09-08) --> marker agree, and the 1.5.x half still reads right. The [Unreleased]: ...compare/1.5.3...HEAD link definition surviving with no ## [Unreleased] heading matches what release: 1.5.2 (589e6ad) did, so it is the repo's shape, not a dangling reference. The 2026-09-10 date matches the head commit's own timestamp (Thu Sep 10 01:13:51 2026 +0300).

Minor

CHANGELOG.md:9 — the promoted [1.5.3] section carries two ### Added subsections, at lines 3 and 12 of the section, split by ### Fixed. CLAUDE.md asks for entries under one of Added / Changed / Fixed / Removed / Quality / Security, and no other version section in the file repeats a heading. The WindAnchorState.hasPrimaryFocus bullet belongs in the first Added block. Carried in from the Unreleased section rather than introduced here, but a release promotion is the last moment to tidy it, and it will ship to pub.dev as-is (maintainability).

Tests

No test change, correctly - the diff is version strings and changelog prose. The behaviour being released is covered by the tests that landed with 92f20f5.

Checks I ran

• flutter pub get then dart analyze lib: No issues found! (a bare dart analyze before resolving deps reports ~28k uri_does_not_exist-family errors; that is the unresolved checkout, not this diff).
• dart format --set-exit-if-changed --output=none .: Formatted 391 files (0 changed), exit 0.
• flutter test: +1804 ~1: All tests passed! - matches the PR's claim of 1804 passing with 1 pre-existing skip.
• python3 tool/check-docs.py: checked 72 doc pages + 3 linking files against 171 demo routes: 0 issue(s).
• ./tool/coverage.sh 90: not run; a version-string diff cannot move coverage, and the suite above already passed clean.
• grep -rn "1\.5\.2" across docs/config/source/lockfiles excluding CHANGELOG.md: no matches.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@kodizm

Full re-review (no since_sha was passed): the two new commits fix a real state-staleness bug with a matching test and clear the duplicate-heading nit from my last pass, and every version surface is still consistent - no findings.

Since my earlier review: 1b789a9 clears _isFocused / _hasPrimaryFocus alongside _isHovering on the enabled → disabled transition in lib/src/widgets/w_anchor.dart:172, with a state-transition test in test/widgets/w_anchor/dpad_activation_test.dart:527; 3ef4306 moves the [1.5.3] date to 2026-09-11 and merges the two ### Added headings into one Added then Fixed section, which settles the only finding I raised.

The fix is in the right place. Focus.canRequestFocus flips to false on that transition, FocusNode.canRequestFocus's setter unfocuses in response, and _onFocusChange (w_anchor.dart:192) returns early for a disabled widget, so the listener structurally cannot observe the transition it is asked to clean up - didUpdateWidget is the only hook that sees both the old and new isDisabled. Clearing without setState is correct here because build follows didUpdateWidget unconditionally. I did not observe the test red before the fix (that needs an edit, which I do not make); the assertion path is sound from reading - the wrapper republishes the ancestor's hasPrimaryFocus at w_anchor.dart:273, so a stale true would surface on the inner context exactly as the test asserts.

Version surfaces: grep for 1.5.2 across .md / .yaml / .yml / .txt / .dart / .json / .lock outside CHANGELOG.md returns nothing, so nothing was left behind. The 2026-09-11 date matches the head commit's own timestamp (Fri Sep 11 00:13:55 2026 +0300), the same convention [1.5.2] - 2026-09-08 follows (589e6ad, Sep 8 2026). The surviving [Unreleased]: ...compare/1.5.3...HEAD link with no ## [Unreleased] heading matches release: 1.5.2, so it is the repo's shape.

The "CHANGELOG only" post-change sync claim holds: doc/widgets/w-anchor.md and skills/wind-ui/SKILL.md §11 describe the activation contract and the wrapper inheritance, neither of which moves here, and nothing on either surface asserted the stale-focus behaviour.

Tests

test/widgets/w_anchor/dpad_activation_test.dart:527 covers the transition (focus the enabled anchor, rebuild disabled, assert isDisabled true and both focus flags false); the existing static-disabled test covers the other half. No test needed for the version-string files.

Checks I ran

• flutter pub get then dart analyze lib test: No issues found!
• dart format --set-exit-if-changed --output=none .: Formatted 391 files (0 changed), exit 0
• flutter test test/widgets/w_anchor test/widgets/w_div: +223 ~1: All tests passed!
• flutter test: +1805 ~1: All tests passed! - matches the claimed 1805 with the one pre-existing skip
• python3 tool/check-docs.py: checked 72 doc pages + 3 linking files against 171 demo routes: 0 issue(s)
• ./tool/coverage.sh 90: not run - the diff adds 2 executable lines, both exercised by the new test, and the full suite is green
• grep -rn "1\.5\.2" across docs/config/source/lockfiles excluding CHANGELOG.md: no matches