fix(ci): ignore pull_request statuses in the tag sibling wait

nalum commented

2026-08-16 12:59:13 +00:00

Owner

The tag job on main failed after the #90 merge with no red push in sight. The sibling wait reads every status on the merged commit. Fast-forward merges make the PR head and the main head the same commit, so that commit also carries the PR runs's statuses. Forgejo cancels the PR's in-flight run when the merge closes it, and those cancelled jobs land as failure statuses (check / report (pull_request), check / web (pull_request), android / build (pull_request), android / report (pull_request)). The wait counted them and refused to tag, while every (push) context was green.

The wait now filters to (push) contexts. Those are the statuses that speak for the push under release. The pull_request contexts speak for the closed PR. Verified the new jq filter against the live status API on 2c966a94: it returns only the six push contexts, none failed.

The tag job on main failed after the #90 merge with no red push in sight. The sibling wait reads every status on the merged commit. Fast-forward merges make the PR head and the main head the same commit, so that commit also carries the PR runs's statuses. Forgejo cancels the PR's in-flight run when the merge closes it, and those cancelled jobs land as `failure` statuses (`check / report (pull_request)`, `check / web (pull_request)`, `android / build (pull_request)`, `android / report (pull_request)`). The wait counted them and refused to tag, while every `(push)` context was green. The wait now filters to `(push)` contexts. Those are the statuses that speak for the push under release. The `pull_request` contexts speak for the closed PR. Verified the new jq filter against the live status API on `2c966a94`: it returns only the six push contexts, none failed.

nalum added 1 commit

2026-08-16 12:59:13 +00:00

fix(ci): ignore pull_request statuses in the tag sibling wait

check / commits (pull_request) Successful in 5s

Details

check / web (pull_request) Successful in 4m6s

Details

check / go (pull_request) Successful in 4m54s

Details

check / report (pull_request) Successful in 4s

Details

check / commits (push) Successful in 8s

Details

check / web (push) Successful in 2m7s

Details

check / go (push) Successful in 3m9s

Details

check / report (push) Has been skipped

Details

tag / tag (push) Successful in 3m46s

Details

release / binaries (push) Successful in 1m54s

Details

release / docs (push) Successful in 2m54s

Details

release / image (push) Successful in 4m45s

Details

release / android (push) Successful in 6m41s

Details

release / manifests (push) Successful in 1m24s

Details

release / module (push) Successful in 1m29s

Details

release / release (push) Successful in 13s

Details

db1e0a7ba7

Fast-forward merges make the PR head and the main head the same
commit, so the merged commit carries the PR runs' statuses too.
Forgejo cancels the PR's in-flight run when the merge closes it, and
those cancelled jobs land as failure statuses — the sibling wait then
reads them as a red push and refuses to tag. Only the (push) contexts
speak for the push being released, so filter the wait to those.

forgejo-actions commented

2026-08-16 13:04:13 +00:00

Test report

Suite	Tests	Result	Skipped
Unit	1332	✅ pass	1
Integration	83	✅ pass	—

Coverage: 27.7%

_{Updated by the check workflow · commit db1e0a7ba7}

## Test report | Suite | Tests | Result | Skipped | | --- | --: | --- | --: | | Unit | 1332 | ✅ pass | 1 | | Integration | 83 | ✅ pass | — | **Coverage:** 27.7% <sub>Updated by the check workflow · commit db1e0a7ba7f4cdf7a76b0fa0e3c2e87465b8a06a</sub>

nalum merged commit db1e0a7ba7 into main

2026-08-16 13:05:58 +00:00

nalum deleted branch fix/tag-sibling-wait

2026-08-16 13:05:58 +00:00

Rows
Columns

fix(ci): ignore pull_request statuses in the tag sibling wait #96

Test report