Skip to content

fix(deps-merge): ignore superseded cancelled checks - #198

Merged
brunogbv merged 1 commit into
mainfrom
claude/740-deps-merge-newest-checks
Oct 20, 2026
Merged

brunogbv merged 1 commit into
mainfrom
claude/740-deps-merge-newest-checks

Conversation

@brunogbv

Copy link
Copy Markdown
Collaborator

Summary — The first scheduled deps-merge runs refused marola-ml#90, a green pip-only minor bump, twice. On 2026-10-19 it said "checks not green (CANCELLED)": the PR workflow started 3 times when Dependabot opened the PR, concurrency cancelled 2, and the rollup still lists them after the newest run passed.

MIP none — not MIP-scoped
Tested ✅ gates · ⬜ e2e · ⬜ live · ⬜ ci-only — deps-merge.sh --self-test ok (the 5 new #415/#416 checks and the 2 re-read checks fail before the fix); every quality line in bash (ruff, shellcheck, actionlint, tests/self-tests.sh, wiring --lint-devkit, agents-check, merge_key_check, docs_lint, skills_vendor, release --check, claude plugin validate): all ok; a read-only --unattended --dry-run on marola-ml makes #90 READY, pinned to 25a884f

What changed

  • fix(deps-merge): ignore superseded cancelled checks, read an UNKNOWN mergeable again (5b93451)

…mergeable again

The first scheduled deps-merge runs refused marola-ml#90, a green pip-only minor bump, twice.
On 2026-10-19 it said "checks not green (CANCELLED)": the PR workflow started 3 times when
Dependabot opened the PR, concurrency cancelled 2, and the rollup still lists them after the
newest run passed. On 2026-10-20 it said "not mergeable (UNKNOWN)": GitHub computes mergeable only
when asked, and a read a few seconds later says MERGEABLE.

judge drops a CANCELLED check only when a newer run of the same check (workflow and name; the
higher job id in detailsUrl, then startedAt) exists. An older FAILURE still counts, so a newer run
that skips its jobs cannot hide it. settle reads an UNKNOWN mergeable again, up to five times
3 seconds apart, before the PR is refused. Both apply in both modes. The fixture gets #90's real
rollup and diff (#415, now merged) and an older FAILURE behind a newer pass (#416, refused); a stub
gh tests the re-read. The tools reference row names both rules.

Part of marola-dev/marola#740

Tested: gates — deps-merge.sh --self-test ok (the 5 new #415/#416 checks and the 2 re-read checks fail before the fix); every quality line in bash (ruff, shellcheck, actionlint, tests/self-tests.sh, wiring --lint-devkit, agents-check, merge_key_check, docs_lint, skills_vendor, release --check, claude plugin validate): all ok; a read-only --unattended --dry-run on marola-ml makes #90 READY, pinned to 25a884f
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WpCVVuPjnM15cpFrytsKY8
@brunogbv
brunogbv merged commit d0e88ec into main Oct 20, 2026
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant