You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Evaluate the final AWS readiness observation before timing out #4421
Parent: #4368
Audit entries: PROV-13
Source-check baseline: 3924da9474eded3d40c4abbaccc79fecae894366. Source inspection is not runtime reproduction.
Required behavior
Every completed readiness poll receives a status decision, including the final poll at the deadline boundary. After a completed poll, preserve cancellation priority; if the caller has not canceled, consume its RUNNING observation before synthesizing the polling timeout. Do not start another poll or widen the aggregate budget to do this. Do not stop a session whose accepted final observation is RUNNING. Preserve caller cancellation/deadline policy, typed ended-session errors, and exactly-once cleanup of failed billed sessions.
Prevention design and owner
Make polling observation and status evaluation one ordered loop at the current AWS session owner. Remove the unchecked last-observation exit; do not add another retry mechanism or increase budgets.
Exercise public prepare-session with fake client and virtual time: final RUNNING, final terminal status, pending timeout, cancellation, missing endpoint, cleanup counts. The old loop fails the final-observation regression. No paid cloud session or live device required for this deterministic owner change.
Record base/head or mutation detection, independent review, full repository formatting and exact-head affected validation. Keep failures, skipped CI/manual lanes and device evidence attributed to their tested SHAs. Preserve #4050 full-unit fallback; no blanket test deletion, gate relaxation, new compatibility fallback, merge, release or tags.
Dependencies and coordination
Recheck current main and open PR overlap before editing. No semantic prerequisite identified at selection.
The issue reports existing AWS device-farm readiness polling behavior as wrong: it may synthesize a polling timeout before evaluating the final completed RUNNING observation, and the old loop fails a final-observation regression. It identifies the owning source file and concrete required behavior/test coverage, making it actionable.
This child is assigned to an isolated Terra worker in the first wave of the second batch. The coordinator owns independent review, exact-head validation and publication. Check this claim before starting another implementation. No PR, merge, release or device evidence is claimed yet.
Current state: implementation is local at 6f6ddff30fa4a07751e4397d29bfbf8b27117ddd, based on 3924da9474eded3d40c4abbaccc79fecae894366. Two files change. Completed AWS polls receive their cancellation/status decision before a polling timeout is synthesized; no budgets or provider cleanup ownership change.
The focused public prepare-session suite passes 11 tests. In a disposable checkout at the base, the same final tests fail three cases (RUNNING at the final retry, terminal status at the boundary, and cancellation priority), with eight passing. Applying the corrected production file makes all 11 pass. Setup passed in that checkout; it is removed, and the reviewed SHA/status is preserved. Full repository formatting and diff checks pass.
Exact-head affected validation is running. Independent source review, CI and publication remain outstanding. No AWS allocation, live-device run, PR, merge or release is claimed. Other workers are independent and are not being polled.
Current state: 6f6ddff30fa4a07751e4397d29bfbf8b27117ddd passed all eight runnable exact-head affected checks. The receipt reports exactCommit=true, checkoutChanged=false, status=passed, exitCode=0; initial/final tracked status is clean. Base/merge-base: 3924da9474eded3d40c4abbaccc79fecae894366.
Passed: format, lint, typecheck, layering, Fallow, build, related tests and the full production-unit fallback. Related: 323 passed. Unit: 19,025 passed, three skipped. Seven selected lanes remain skipped with their owners: web-smoke (GitHub), replay-ios/replay-ios-device/replay-android (parked manual dispatch), replay-macos/replay-linux/linux-command-evidence (GitHub smoke owners). Those are not passes.
The same focused prepare-session command yields three failures/eight passes with base production and final tests, then 11 passes with corrected production in the disposable experiment. The experiment is removed; reviewed SHA/status was preserved. Frozen install/build/full formatting passed. Independent Luna review of this final SHA found no findings; it ran no checks.
This change is deterministic host/provider polling and cleanup; no paid AWS session or live-device run was used. Fresh PR CI/review remain obligations. No scope expansion, fallback change, test deletion, merge or release occurred.
Current state: PR #4429 is open and conflict-free at 6f6ddff30fa4a07751e4397d29bfbf8b27117ddd. CI has 16 successful checks and one skipped cache matrix entry; the skip is not counted as a pass. Android, Apple and Linux smoke, integration, coverage, repository compatibility, package checks and both review services are green. Cubic and Apex report no findings, there are no review threads, and the maintainer review reports no remaining pre-merge work. Exact local validation and base regression proof remain recorded above. No live AWS allocation is required by this deterministic polling contract. Monitoring remains active; no merge was performed.
Current state: maintainer-merged PR #4429 at head 6f6ddff30fa4a07751e4397d29bfbf8b27117ddd, merge 519565ccc029ff444bdb204d3408cfd13444b743, on 2026-11-11 11:53:21 UTC. The current rollup has17successful checks/one skipped cache entry, no unresolved review threads and no failures. Prior exact local validation/regression proof remains recorded above. This deterministic polling child is settled; no live AWS allocation was required. The coordinator did not merge it and is retiring its event subscription.
Activity
apex-sentinel-staging commentedon Nov 11, 2026
Apex Sentinel · Issue review
Suggested label:
bugThe issue reports existing AWS device-farm readiness polling behavior as wrong: it may synthesize a polling timeout before evaluating the final completed RUNNING observation, and the old loop fails a final-observation regression. It identifies the owning source file and concrete required behavior/test coverage, making it actionable.
Required sections
thymikee commentedon Nov 11, 2026
Decision: implementation ownership:
This child is assigned to an isolated Terra worker in the first wave of the second batch. The coordinator owns independent review, exact-head validation and publication. Check this claim before starting another implementation. No PR, merge, release or device evidence is claimed yet.
thymikee commentedon Nov 11, 2026
Current state: implementation is local at
6f6ddff30fa4a07751e4397d29bfbf8b27117ddd, based on3924da9474eded3d40c4abbaccc79fecae894366. Two files change. Completed AWS polls receive their cancellation/status decision before a polling timeout is synthesized; no budgets or provider cleanup ownership change.The focused public prepare-session suite passes 11 tests. In a disposable checkout at the base, the same final tests fail three cases (RUNNING at the final retry, terminal status at the boundary, and cancellation priority), with eight passing. Applying the corrected production file makes all 11 pass. Setup passed in that checkout; it is removed, and the reviewed SHA/status is preserved. Full repository formatting and diff checks pass.
Exact-head affected validation is running. Independent source review, CI and publication remain outstanding. No AWS allocation, live-device run, PR, merge or release is claimed. Other workers are independent and are not being polled.
thymikee commentedon Nov 11, 2026
Current state:
6f6ddff30fa4a07751e4397d29bfbf8b27117dddpassed all eight runnable exact-head affected checks. The receipt reports exactCommit=true, checkoutChanged=false, status=passed, exitCode=0; initial/final tracked status is clean. Base/merge-base:3924da9474eded3d40c4abbaccc79fecae894366.Passed: format, lint, typecheck, layering, Fallow, build, related tests and the full production-unit fallback. Related: 323 passed. Unit: 19,025 passed, three skipped. Seven selected lanes remain skipped with their owners: web-smoke (GitHub), replay-ios/replay-ios-device/replay-android (parked manual dispatch), replay-macos/replay-linux/linux-command-evidence (GitHub smoke owners). Those are not passes.
The same focused prepare-session command yields three failures/eight passes with base production and final tests, then 11 passes with corrected production in the disposable experiment. The experiment is removed; reviewed SHA/status was preserved. Frozen install/build/full formatting passed. Independent Luna review of this final SHA found no findings; it ran no checks.
This change is deterministic host/provider polling and cleanup; no paid AWS session or live-device run was used. Fresh PR CI/review remain obligations. No scope expansion, fallback change, test deletion, merge or release occurred.
thymikee commentedon Nov 11, 2026
Current state: PR #4429 is open and conflict-free at
6f6ddff30fa4a07751e4397d29bfbf8b27117ddd. CI has 16 successful checks and one skipped cache matrix entry; the skip is not counted as a pass. Android, Apple and Linux smoke, integration, coverage, repository compatibility, package checks and both review services are green. Cubic and Apex report no findings, there are no review threads, and the maintainer review reports no remaining pre-merge work. Exact local validation and base regression proof remain recorded above. No live AWS allocation is required by this deterministic polling contract. Monitoring remains active; no merge was performed.thymikee commentedon Nov 11, 2026
Current state: maintainer-merged PR #4429 at head
6f6ddff30fa4a07751e4397d29bfbf8b27117ddd, merge519565ccc029ff444bdb204d3408cfd13444b743, on 2026-11-11 11:53:21 UTC. The current rollup has17successful checks/one skipped cache entry, no unresolved review threads and no failures. Prior exact local validation/regression proof remains recorded above. This deterministic polling child is settled; no live AWS allocation was required. The coordinator did not merge it and is retiring its event subscription.