Skip to content

ci(release): wait for the tagged commit's CI instead of reading it as failed #120

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kingchenc merged 2 commits into from
Aug 31, 2026
Merged

ci(release): wait for the tagged commit's CI instead of reading it as failed #120

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b932a47
ci(release): wait for the tagged commit's CI instead of reading it as…
kingchenc Aug 31, 2026
1e01c54
ci(release): restore the escapes the previous commit flattened
kingchenc Aug 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
72 changes: 49 additions & 23 deletions .github/workflows/release.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@ jobs:
name: Gate the irreversible publish
needs: [python-wheels, python-sdist, node-build, wasm-build, c-abi-build]
runs-on: ubuntu-latest
timeout-minutes: 10 # two API calls; if it is not done in ten minutes it is stuck
timeout-minutes: 60 # waits for ci.yml to settle; the wait itself is capped at 45 minutes
permissions:
actions: read # list the workflow runs for this commit
contents: read
Expand All @@ -78,32 +78,58 @@ jobs:
THIS_RUN: ${{ github.run_id }}
run: |
set -euo pipefail
runs=$(gh api --paginate \
"repos/$REPO/actions/runs?head_sha=$SHA&per_page=100" \
--jq '.workflow_runs[] | [.id, .name, .path, .status, (.conclusion // "pending")] | @tsv')

if [ -z "$runs" ]; then
echo "::error::no workflow runs recorded for $SHA -- nothing proves this commit was ever built"
exit 1
fi
# ci.yml that has not finished yet reports conclusion `pending`, which
# is not the same as a failure -- and a tag pushed shortly after its
# merge lands while ci.yml is still queued behind this run's own build
# jobs. Refusing there turns "tagged a minute early" into a failed
# release, so wait for the verdict rather than reading its absence as
# one. Bounded to 45 minutes; anything actually red ends this on the
# spot, so only the undecided case waits.
deadline=$(( $(date +%s) + 45 * 60 ))
while :; do
runs=$(gh api --paginate \
"repos/$REPO/actions/runs?head_sha=$SHA&per_page=100" \
--jq '.workflow_runs[] | [.id, .name, .path, .status, (.conclusion // "pending")] | @tsv')

echo "runs on $SHA:"
printf '%s\n' "$runs" | cut -f2,4,5
if [ -z "$runs" ]; then
echo "::error::no workflow runs recorded for $SHA -- nothing proves this commit was ever built"
exit 1
fi

# This release run is itself in progress on this SHA; so is anything
# else still going. Neither is evidence of failure.
red=$(printf '%s\n' "$runs" | awk -F'\t' -v this="$THIS_RUN" \
'$1 != this && ($5 == "failure" || $5 == "timed_out" || $5 == "action_required") { print " " $2 " (" $5 ")" }')
if [ -n "$red" ]; then
echo "::error::this commit has failing checks; refusing to publish"
printf '%s\n' "$red"
exit 1
fi
# This release run is itself in progress on this SHA; so is anything
# else still going. Neither is evidence of failure.
red=$(printf '%s\n' "$runs" | awk -F'\t' -v this="$THIS_RUN" \
'$1 != this && ($5 == "failure" || $5 == "timed_out" || $5 == "action_required") { print " " $2 " (" $5 ")" }')
if [ -n "$red" ]; then
echo "runs on $SHA:"
printf '%s\n' "$runs" | cut -f2,4,5
echo "::error::this commit has failing checks; refusing to publish"
printf '%s\n' "$red"
exit 1
fi

if ! printf '%s\n' "$runs" | awk -F'\t' '$3 == ".github/workflows/ci.yml" && $5 == "success"' | grep -q .; then
echo "::error::no successful ci.yml run for $SHA -- tag a commit that has passed CI on main"
exit 1
fi
if printf '%s\n' "$runs" | awk -F'\t' '$3 == ".github/workflows/ci.yml" && $5 == "success"' | grep -q .; then
break
fi

if ! printf '%s\n' "$runs" | awk -F'\t' '$3 == ".github/workflows/ci.yml" && $4 != "completed"' | grep -q .; then
echo "runs on $SHA:"
printf '%s\n' "$runs" | cut -f2,4,5
echo "::error::no successful ci.yml run for $SHA -- tag a commit that has passed CI on main"
exit 1
fi

if [ "$(date +%s)" -ge "$deadline" ]; then
echo "::error::ci.yml on $SHA did not finish within 45 minutes; refusing to publish"
exit 1
fi
echo "ci.yml on $SHA has not finished yet -- waiting"
sleep 30
done

echo "runs on $SHA:"
printf '%s\n' "$runs" | cut -f2,4,5
echo "ci.yml is green on $SHA and nothing else is red"

# --------------------------------------------------------------------------
Expand Down
16 changes: 16 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,22 @@ to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [Unreleased]

### Fixed

- **The publish gate refused a tag whose CI had not finished yet.** It required a
`ci.yml` run on the tagged commit with conclusion `success`; a run still in
progress reports `pending`, which the check treated exactly like a failure. A
tag pushed shortly after its merge -- with CI still queued behind this run's own
build jobs -- was therefore refused for having "no successful ci.yml run", on a
commit that was green half an hour later. v0.1.2 hit this on its first attempt
and published unchanged on a rerun.

The step now waits for that run to settle, bounded to 45 minutes, and says what
it is waiting on. Anything genuinely red still fails on the spot, so a broken
commit is refused as fast as before -- only the undecided case waits.
`timeout-minutes` on the job moves from 10 to 60 to outlive that wait.


## [0.1.2] - 2026-09-01

### Fixed
Expand Down
Loading