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
🐛 Detect terminal-workflow CI reporting stalls without unsafe reruns #15938
Owner explicitly authorized implementation of workflow/check reconciliation. Focused detection/escalation follow-up; not a duplicate of #15905, preparation rail/native integration, settings or another session's #15849.
Observed #15841 exact head1731402a81418f7207f6337eafa68fe971b5b279: pending/queued required check runs persisted while their owning Actions workflow41108783466 REST run was completed/failure. This is reporting inconsistency, NOT proof of an infrastructure cause. Owner performed one justified same-head rerun; do not duplicate it. New active attempts must supersede old terminal observations.
Existing shared CI evaluation trusts check states. Its opaque-workflow guard ignores runs which produced jobs, so terminal workflows with unfinished reporting are not explained. Existing post-mutation shared-CI marker comments and /data/convergence/mutation filesystem flock cannot establish atomic pre-effect ownership across independently stored hives.
Implement read-only, bounded cached/event-aware exact-current-head reconciliation: Actions App/check/suite/run/latest-attempt/job provenance, independent propagation-grace observations and head/attempt consistency rereads. Ordinary active workflows are not stale. Unknown/unsupported/permission/API/head-changing states are explicit and never fabricated green. Preserve CI/review/hold/authorization/current-base gates and stage compatibility; surface exact run/attempt/check through structured observations and existing reason tooltip, no tracker layout/row/sort change.
Automatic rerun recovery is NOT included unless actual shared pre-effect authority is proven. Current delivery is detection plus owner inspection/action, independent of default-OFF integration feature. Future recovery must atomically claim exact repo/subject/head/run/attempt/action episode before App RPC, enforce designated executor/observer-only others, journal ambiguous intent without blind TTL steal/retry, and remain honest that Actions RPC provides no fence/idempotency token. No claim of automatic failover or exactly-once; no permission widening or live settings changes.
CI is sole verdict per AGENTS; owner authorizes guarded DCO admin source squash once all applicable exact-head CI completes green. No local suites, live mutation, additional workflow rerun, production rollout or configuration activation.
@clubanderson this issue was closed by @hivecommons-hive[bot] as completed.
If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.
If you are still seeing the problem, comment /reopen and it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.
Activity
hivecommons-hive commentedon Oct 31, 2026
Fixed by #15941 (merged 85e9ec7). Closing. Reply /reopen if the problem persists.
github-actions commentedon Oct 31, 2026
@clubanderson this issue was closed by @hivecommons-hive[bot] as completed.
If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.
If you are still seeing the problem, comment
/reopenand it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.