Skip to content

ci: give the two full-history checkout jobs a 15-minute budget - #13126

Merged
iamwhatever merged 1 commit into
kirodotdev:mainfrom
billygerhard:fix/ci-checkout-timeouts
Sep 23, 2026
Merged

iamwhatever merged 1 commit into
kirodotdev:mainfrom
billygerhard:fix/ci-checkout-timeouts

Conversation

@billygerhard

Copy link
Copy Markdown
Contributor

Problem / Motivation

PR Scope and CI's Detect changed surface both declare timeout-minutes: 5 and both check out with fetch-depth: 0. On a slow hosted runner the full-history fetch of this repository alone has run past 5 minutes and the job is CANCELLED mid-checkout (The operation was canceled) -- twice on #11827 today, both on the first run after a maintainer released the fork hold (#13111).

Why it matters

Detect changed surface gates every test lane, so its cancel skipped the entire matrix and failed Coverage Gate with nothing to read; PR Scope is a required gate. Neither failure is retried, neither leaves a diagnostic beyond the cancel, and on a fork PR each costs a maintainer another workflow release to re-run.

What changed

Symptom: job cancelled during actions/checkout at exactly the 5-minute mark, last command the full git fetch → root cause: a 5-minute job budget sized for the job's own work (seconds), not for a full-history fetch of a large repository on a slow runner → change: timeout-minutes: 15 on both jobs, each with a comment recording why. The alternative -- a shallow checkout -- would need a separate merge-base fetch in both jobs (BASE...HEAD diffs, dorny/paths-filter with local git, the leaf resolver) and is a larger change than the incident calls for; the budget covers a slow fetch and still ends a hung one.

This PR changes workflow files from a fork, so the Fork Workflow-Change Guard will flag it for a maintainer's eyes; the diff is two timeout-minutes lines plus comments.

Backwards compatibility

Compatible: no step, permission or output changes; only the per-job budget grows.

Tests

The eight test files that read ci.yml / pr-scope.yml contracts pass locally (386 tests). No test asserts the 5-minute value.

Manual verification

Observe the two jobs on this PR's own run; both should complete well inside the new budget, and the run must not be cancelled at checkout.

Screenshots / video

N/A — no UI change.

Closes #13111

PR Scope and CI's 'Detect changed surface' both declare timeout-minutes: 5
and both check out with fetch-depth: 0 because they diff against the merge
base (BASE...HEAD, dorny/paths-filter with local git, the leaf resolver). On
a slow hosted runner the full-history fetch of this repository alone has run
past 5 minutes, and the job is CANCELLED mid-checkout ('The operation was
canceled'): nothing to read, no retry, and on a fork PR another maintainer
release to re-run. Detect changed surface gates every test lane, so its
cancel skipped the whole matrix and failed Coverage Gate (kirodotdev#13111, twice on
kirodotdev#11827 today). 15 minutes covers a slow fetch and still ends a hung one.
The alternative -- a shallow checkout -- would need a separate merge-base
fetch in both jobs and is a larger change than the incident calls for.

Closes kirodotdev#13111
@billygerhard
billygerhard requested review from a team and krishdhasmana September 23, 2026 17:07
@github-actions github-actions Bot added fork Pull request from a fork (external contributor) readiness: action required A blocking check or review needs attention labels Sep 23, 2026
@iamwhatever
iamwhatever merged commit 0bb65a4 into kirodotdev:main Sep 23, 2026
4 of 5 checks passed
@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Sep 23, 2026
@billygerhard
billygerhard deleted the fix/ci-checkout-timeouts branch September 23, 2026 18:30
wknight1 pushed a commit to wknight1/KiroCrew that referenced this pull request Oct 5, 2026
…otdev#13126)

PR Scope and CI's 'Detect changed surface' both declare timeout-minutes: 5
and both check out with fetch-depth: 0 because they diff against the merge
base (BASE...HEAD, dorny/paths-filter with local git, the leaf resolver). On
a slow hosted runner the full-history fetch of this repository alone has run
past 5 minutes, and the job is CANCELLED mid-checkout ('The operation was
canceled'): nothing to read, no retry, and on a fork PR another maintainer
release to re-run. Detect changed surface gates every test lane, so its
cancel skipped the whole matrix and failed Coverage Gate (kirodotdev#13111, twice on
kirodotdev#11827 today). 15 minutes covers a slow fetch and still ends a hung one.
The alternative -- a shallow checkout -- would need a separate merge-base
fetch in both jobs and is a larger change than the incident calls for.

Closes kirodotdev#13111
@bolichen97 bolichen97 mentioned this pull request Oct 6, 2026
5 tasks done
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor)

Projects

None yet

2 participants