test(ios-smoke): answer deep-link prompt after readiness exhaustion - #3044
Conversation
Size Report
Startup median (7 runs, lower is better):
|
There was a problem hiding this comment.
All reported issues were addressed across 2 files
Reply with feedback, questions, or to request a fix.
Fix all with cubic | Re-trigger cubic
|
Reviewed at cb7f1a6. The harness now answers the deep-link confirmation when the first wait ends with readiness exhausted, instead of returning early and leaving the prompt open. The new regression covers that branch. Not blocking: The link to the #3041 and #3032 failures comes from the PR description; I did not read those CI artifacts. Both Smoke Tests jobs are still queued. The iOS smoke calls this helper at its deep-link step, so a failure there would be related to this change. A green run may not hit the readiness-exhausted case, so the regression test is the main proof. |
|
Reviewed at b93c2eb. The new commit limits the prompt retry to target-discovery exhaustion, and the regression test covers that condition. Not blocking: the raw The regression test uses a mocked device, so the live iOS smoke run on this head is still the real check. Both Smoke Tests jobs are still queued; if one fails, please check whether it fails at the deep-link confirmation step. Ready for human review. |
b93c2eb to
1fd5e01
Compare
|
Summary
Keep the iOS smoke deep-link confirmation probe active when the destination wait ends during typed target discovery readiness (
wait_readiness_exhausted). The helper previously returned beforealert get, leaving iOS's “Open in Agent Device Tester?” confirmation unanswered. A runner startup readiness failure still returns after one wait without prompt probes. Two test-harness files changed; production behavior is unchanged.Failure on #3006 and failure on #3032 both show a successful deep-link
open, a failed first 15-second destination wait, no alert probe, thenAPP_NOT_RUNNINGon the following wait. Final error details reportwait_capture_stalled,readinessPhase: target-discovery, and zero readable captures; screenshots show the unanswered system confirmation. The first wait's exact reason is absent from uploaded artifacts, so the missing branch is inferred from control flow and verified by regression.Validation
At
1fd5e01362f168d0ae545a44072337b8da4da6ea, rebased onto main3fe2e6929: frozen install, build, repository-widepnpm format, focused Node test (12 passed), and exact-headpnpm check:affected --runpassed, including Node integration (119 passed, 12 skipped), lint, typecheck, and fallow. Removing the target discovery branch makes the confirmation test fail with only the first wait recorded; the branch was restored. Live iOS smoke on the rebased head is pending.