ci: stop cancelling in-progress runs on new pushes#2191
Conversation
With cancel-in-progress: true, any new push to a PR (including autofix.ci formatting commits) immediately kills the running CI. This causes phantom failures where all tests were passing but the run shows as cancelled/failed. Setting cancel-in-progress: false lets existing runs complete. New pushes still get their own run — they just queue instead of killing the previous one. Intra-run fail-fast (matrix job cancellation on first failure) is unaffected.
Greptile SummaryThe PR disables Confidence Score: 5/5Safe to merge — the one-line change is well-reasoned and does not touch any job logic. The change is a single boolean flag flip in the concurrency block. The rest of the workflow — including the cross-job fail-fast job, the matrix fail-fast, and the per-PR concurrency group — is completely untouched and continues to behave as before. No files require special attention. Important Files Changed
Reviews (1): Last reviewed commit: "ci: stop cancelling in-progress runs on ..." | Re-trigger Greptile |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
|
we discussed this https://discord.com/channels/1341146487186391173/1500695132876439582/1506589906867327027 not super sold |
Problem
With
cancel-in-progress: true, any new push to a PR immediately kills the running CI workflow. This includes commits pushed byautofix.ci(formatting fixes), which means:Example: https://github.com/dimensionalOS/dimos/actions/runs/26152674707
Fix
Set
cancel-in-progress: false. Existing runs complete normally. New pushes still trigger their own CI run — they just queue instead of killing the previous one.What is NOT affected
strategy.fail-faston the matrix, completely independentTradeoff
Slightly more runner minutes on stale commits (old run finishes even if a new commit arrives). But no more phantom failures from autofix races.