Repository navigation
release: finish a release that a cancelled run tagged - #119
Conversation
|
Warning Review limit reachedYour organization has reached its usage spending cap. Adjust your spending cap in the billing tab. Next included review available in 8 minutes. View limit details
|
b450249 to
11309c2
Compare
[Medium impact] The changes affect the manually dispatched release workflow and its publishing path. The queued release must check out the latest branch tip before this is ready to merge. Findings
SummaryThe release workflow now queues overlapping runs and reuses an existing version tag when it points to the current commit.
Reviews (1) · Last reviewed commit: "release: reuse the tag that a cancelled ..." · Reviewed by Greptile |
Two dispatches of the release workflow raced to push the version bump, and the one that lost failed. A second dispatch now waits for the running one and bumps the branch's tip when it starts, not the commit it was dispatched at.
A run cancelled after pushing the version bump and its tag left a release that no later run could finish, since tagging again failed. The tag is now reused when HEAD descends from it and its crate has the version, and any other tag of that name still fails the job.
11309c2 to
275afd8
Compare
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #119 +/- ##
==========================================
- Coverage 97.26% 95.72% -1.55%
==========================================
Files 22 22
Lines 14040 13530 -510
==========================================
- Hits 13656 12951 -705
- Misses 384 579 +195 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
A release run cancelled right after it pushed the version bump and the
5.0.0tag left a release that no later run could finish, becausegit tagfailed on the existing tag. A second run dispatched a second later lost the race to pushmainand failed.releaseconcurrency group, so a second dispatch waits for the running one instead of racing it, and the "Bump version" job checks out the branch's tip when it starts rather than the commit it was dispatched at, so a queued run builds on the first one's bump.Cargo.tomlhas the version, since the test, release and crate jobs check out the tag. Any other tag of that name fails the job. Once this is merged, dispatching5.0.0again releases ddf5291.