Repository navigation
test(maestro): keep optional-scroll continuation assertion visible - #4538
Conversation
Size Report
Startup median (7 runs, lower is better):
|
|
Thanks for the fix. The code at 94bd208 looks right, but I can't call it ready yet because the iOS side has no evidence. The scenario also runs on the iOS differential lane, and nothing shows that iOS exposes a visible element labelled exactly 'Home' after the scroll. If it doesn't, the flow fails there with a false divergence. Please run optional-warned-scroll-and-wait on the iOS differential lane at this head and show both engines passing. If that lane does not run this scenario, please say where that is configured. The Android result is author-reported and has no attached artifact, and I did not rerun it. All 13 checks pass, but the diff only touches a fixture that unit CI does not run. The failed affected receipt (android-alert.test.ts:128 timeout) is in a provider path this diff does not touch. I know of no conflicts. Before merge, the scenario needs to pass on the iOS lane at this head, or be shown to be Android-only. Not blocking: a one-line comment in the flow saying why the final assertion uses the fixed tab and not the page heading. You can take it or leave it. |
Apex Sentinel · ✅ ApproveNo open findings. Reviewed with this repository's own Sentinel configuration ( 💬 Talking to SentinelIn a finding's thread, reply with:
On this conversation, comment |
Summary
Keep the optional-scroll conformance flow's continuation assertion visible after scrolling. The fixed Home tab replaces reliance on an off-screen page heading. Both optional operations, their timeouts and the initial title assertion remain unchanged.
One test fixture, +1/−1; part of #3377. No runtime changes.
Validation
Head
94bd20831bb796e0b3fe9dda7df573a036c98569; basee401de82d5a23bedbd88de68156862f8cc49ddca.On unchanged base, upstream Maestro and agent-device both fail the original final heading assertion. The committed head passes the same Android differential scenario on both engines through the T3-bound source daemon. Cleanup passed. Updated iOS differential verification also passed both engines, with warnings on both optional steps and successful final Home. Source-daemon routing and cleanup were verified; the cached fixture build SHA is unrecorded.
Independent source review found no regressions. Hosted checks: 7 success, 5 skipped, 1 neutral.
Pinned pnpm 11.17.0 frozen install/build and the single post-review affected run exited 0. Receipt
run-xGUbuq: 42 passed, 25 skipped,exactCommit=true,checkoutChanged=false; provider integration passed 288 tests. Defaultorigin/mainselection was preserved; actual PR diff was recorded separately.Earlier
run-bl6mNYremains a failed receipt due to an Android-alert timeout; its cause is unproven. This new pass neither erases nor explains it. Canonical Maestro performance and npm prerequisites remain separate C3 blockers. The fixture correction has independent source review, exact-head admission and Android/iOS evidence. It does not clear the separate plugin-extraction blockers.