fix(test): repair Android replay fixtures against live device reality

[thymikee]

Three Android fixture defects from the #1482/#1484 full-tier suite, none of which ever executed in CI (both nightlies since failed on adb infra before the suite ran). All three verified live during the #1525 evidence arc on a fresh API 36 emulator with pixel_7 geometry and a Release fixture APK:

• 01-navigation-scroll.ad clicked label=Catalog, but the expo-router NativeTabs cart badge leaks '0 new notifications' into the tab's content description even while hidden, and unselected native tabs expose no child text node — exact match can never hit. Target the composed label the device actually exposes (deterministic at fixture start: cart is 0 after --relaunch). The badge does not leak on iOS, so the iOS twin keeps label="Catalog".
• checkout-form-android.ad opened by iOS display name 'Agent Device Tester'; Android open resolves packages (the APK label is 'Agentdevicelab'), so APP_NOT_INSTALLED was guaranteed. Use the package id, matching gesture-lab-android.ad.
• gesture-lab-android.ad aimed every gesture at y=700, above the gesture card (targets span y754–1329 on pixel_7 geometry; the home screen gained content above the card since authoring). Re-aim pans inside the exact-two-pointer zone, flings on the image clear of that zone, and pinch/rotate/transform at the card center. Verified: full suite passes 2/2 via the public test command (20 + 32 steps replayed).

Validation

Live evidence from the #1525 arc (2026-07-31, documented in its PR comment): consolidated suite green 2/2 on emulator-5556 (Pixel_7_CI, API 36, freezer disabled), checkout 20/20 + gesture-lab 32/32 steps replayed through the public test command, JUnit non-empty. These fixture fixes are also a prerequisite for the test-app:replay:android live-evidence requirement on #1536.

Refs #1478

[github-actions]

Size Report

Metric	Base	Current	Diff
JS raw	1.92 MB	1.92 MB	0 B
JS gzip	615.5 kB	615.5 kB	0 B
npm tarball	733.5 kB	733.5 kB	0 B
npm unpacked	2.57 MB	2.57 MB	0 B

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	26.5 ms	28.2 ms	+1.6 ms
CLI --help	57.2 ms	59.6 ms	+2.4 ms

Top changed chunks: no changes in the largest emitted chunks.

[thymikee]

Reviewed head 5db5831. The three fixture corrections match the cited live Android evidence, and all CI is green, but one portability gap remains:

• P2 — constrain or document the gesture fixture's validated geometry. examples/test-app/replays/gesture-lab-android.ad now hard-codes every target around y=1040, based on the API 36 Pixel 7 run. However, pnpm test-app:replay:android selects only --platform android, the script declares only kind=emulator, and examples/test-app/README.md presents that command as the general local Android suite. Any booted emulator can therefore receive these Pixel-7-specific coordinates; on another size/density the gesture card can move and the test will hit the wrong view. Constrain the route to the validated geometry, make target positioning geometry-independent, or at minimum document the strict Pixel 7/API 36 prerequisite and a reproducible device-selecting invocation.

The package-id fix and Android composed Catalog label are otherwise sound. No ready-for-human label yet.

[thymikee]

Addressed in 071326034 (rebased onto current main): took the document half of "constrain or document" — the gesture fixture's header now names the validated geometry (CI's pixel_7, 1080×2400 @ 420 dpi, with the measured gesture-card span) and states that a mismatched emulator fails honestly at the canary waits rather than silently passing; the README's test-app:replay:android section says the same and distinguishes the selector-driven checkout twin, which is size-independent. Parse check green for both files.

I stopped short of a hard constraint (e.g. asserting wm size at open) because the .ad grammar has no device-geometry predicate and inventing one for a fixture would be scope creep — but if you'd rather gate than document, a wait-able geometry canary rendered by the test app itself would be the clean mechanism and I can file that as a follow-up.

Generated by Claude Code

[thymikee]

Re-reviewed exact head 071326034. The prior portability finding is resolved: the Android gesture fixture and local-suite README now pin the validated pixel_7, 1080x2400 @ 420 dpi geometry, explain how mismatches surface through the positive gesture canaries, and distinguish the size-independent selector-driven checkout replay.

The corrected fixture bytes remain identical to the cited live Android run, and each regression edit is guarded by a positive state/count wait or deterministic launch/selector failure. All exact-head CI is green, including replay provenance, integration, coverage, Android smoke, and release builds. No blocking findings — ready for human merge review.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-08-01 11:50 UTC