Skip to content

ci: raise normal tier timeout to 45 minutes - #21

Merged
andrew-codes merged 1 commit into
mainfrom
fm/fm-fork-ci-timeout
Nov 16, 2026
Merged

andrew-codes merged 1 commit into
mainfrom
fm/fm-fork-ci-timeout

Conversation

@andrew-codes

Copy link
Copy Markdown
Owner

Intent

Raise the time limit on the fork's CI jobs that keep timing out.

Context: on the fork andrew-codes/firstmate, three PRs in a row (PR 19, PR 20 twice) had a CI job cancelled at its 30-minute timeout-minutes limit with no test or lint failure, and each passed or got further on rerun. Lint 2 used to take 20-21 minutes; on the latest rerun it linted 283 files steadily from 13:52 to 14:22 and ran out of time (no hang). "Behavior portable serial 9" was also cancelled at the limit twice after its tests had been passing. "Full suite" then fails only because one of those jobs was cancelled. .github/workflows/ci.yml gives these jobs the "Normal tier" 30-minute limit and documents a timeout policy near the top.

What Changed

  • Raised timeout-minutes from 30 to 45 for every normal-tier job in .github/workflows/ci.yml (lint partitions, portable parallel and serial shards, macOS stock Bash). Added a header comment explaining that healthy 20-21 minute lint and shard jobs were being cancelled at the old bound while still making steady progress.
  • Updated the timeout policy table and the guard description in docs/fm-test-portable-shards.md to the 45-minute bound and its rationale.
  • Updated tests/fm-ci-workflow.test.sh so the normal-tier check expects the single shared 45-minute budget, including the comment and failure message.

Risk Assessment

✅ Low: The change raises the five normal-tier job timeouts from 30 to 45 minutes and updates the policy comment, the docs table and the guard test to match, and the Herdr 75-minute backstop and fast-tier limits are unaffected.

Testing

I ran the workflow guard test, which runs the workflow through its consumers and checks the timeout tiers. All checks passed, including "normal tier jobs share one 45 minute budget". I also parsed the YAML: the five normal-tier jobs (lint, the three portable test lanes, macOS) are at 45, the heavy lanes at 75, and the fast jobs at 5. No GitHub-hosted run was driven. Whether the jobs now finish inside 45 minutes is only knowable from a real CI run.

  • Live validation: ✅ go - 2 of 2 scenarios driven live against the product
Scenario Result Live Evidence
Run the CI workflow guard test (tests/fm-ci-workflow.test.sh): every job has a finite timeout, belongs to a known tier, and normal-tier jobs share one 45 minute budget ✅ pass live bash tests/fm-ci-workflow.test.sh exited 0 and printed 'ok - normal tier jobs share one 45 minute budget'
Parse ci.yml and read each job's timeout-minutes: normal-tier jobs are 45, and the heavy and fast tiers are unchanged ✅ pass live Ruby YAML parse output: lint, tests-portable-parallel-1, tests-portable-parallel-2, tests-portable-serial and macos-stock-bash are 45; tests-herdr and tests-hc are 75; the remaining jobs are 5.

Pipeline

Updates from git push no-mistakes

✅ **Intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

  • Live validation: ✅ go - 2 of 2 scenarios driven live against the product
Scenario Result Live Evidence
Run the CI workflow guard test (tests/fm-ci-workflow.test.sh): every job has a finite timeout, belongs to a known tier, and normal-tier jobs share one 45 minute budget ✅ pass live bash tests/fm-ci-workflow.test.sh exited 0 and printed 'ok - normal tier jobs share one 45 minute budget'
Parse ci.yml and read each job's timeout-minutes: normal-tier jobs are 45, and the heavy and fast tiers are unchanged ✅ pass live Ruby YAML parse output: lint, tests-portable-parallel-1, tests-portable-parallel-2, tests-portable-serial and macos-stock-bash are 45; tests-herdr and tests-hc are 75; the remaining jobs are 5.
  • bash tests/fm-ci-workflow.test.sh
  • Parsed .github/workflows/ci.yml with Ruby YAML and listed each job's timeout-minutes
✅ **Document** - passed

✅ No issues found.

⚠️ **Lint** - 1 warning
  • ⚠️ linter found issues (exit code 1)
✅ **Push** - passed

✅ No issues found.

Healthy lint and portable shard jobs that used to take 20-21 minutes now
reach the 30 minute bound on slower runners and get cancelled while still
making steady progress. Raise the shared normal tier to 45 minutes and
update the policy doc and workflow test that pin it.
@andrew-codes
andrew-codes merged commit 3f5378c into main Nov 16, 2026
21 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