Skip to content

ci: give the test job 60 minutes - #457

Merged
hellices merged 2 commits into
mainfrom
fix/454-test-timeout
Oct 23, 2026
Merged

hellices merged 2 commits into
mainfrom
fix/454-test-timeout

Conversation

@hellices

@hellices hellices commented Oct 22, 2026 •

Copy link
Copy Markdown
Owner

Closes #454
Closes #463

Summary

The test job runs lint, type checks, and the full suite under coverage, all within timeout-minutes: 45. On Python 3.12 the job takes 39–46 minutes. In the last 25 CI runs, 3.12 took 43 minutes or more ten times. It was cut off at the limit with no test failures on:

A cancelled required check blocks a pull request until someone reruns it, and an outside contributor cannot rerun it.

This PR raises the test job to 60 minutes and updates the pinned value in tests/test_ci_workflow.py. It also updates the "10–45 minute" range in docs/dev/quality-gates.md to 10–60. windows-test stays at 45 minutes because it finishes in about 30.

docs/superpowers/specs/2026-09-11-repository-stabilization-design.md still says 45 minutes. That is the dated design record of the original sizing ("sized above measured healthy durations"), so this PR leaves it as written. This change applies the same rule to newer measurements.

It also carries build: relock opentelemetry-api for #463, cherry-picked from the Relock workflow's relock/20261023-033350 (run 39892989114). Since 2026-10-22 the required security check fails on every branch: pip-audit reports GHSA-xcq7-p3pf-p88h (CVE-2026-114603) against the locked opentelemetry-api 1.44.0. The relock moves only that package to 1.45.1, so uv.lock changes by three lines and still resolves from PyPI only.

Verification

tests/test_ci_workflow.py::test_ci_jobs_have_explicit_bounded_timeouts pins every job's timeout. Setting ci.yml back to 45 while the pin says 60 makes it fail; restoring 60 makes it pass.

Run locally on macOS with Python 3.13 and the frozen lock. make check runs each step through uv run, which in a fresh worktree tries to sync from PyPI that this machine cannot reach, so each step ran directly with the project's environment:

Command Result
pytest -p no:tach tests/test_ci_workflow.py tests/test_docs_links.py tests/test_docs_workflow.py tests/test_docs_build_config.py tests/test_platforms.py 170 passed
python scripts/check_source_size.py passed
ruff check src/ tests/ and ruff format --check src/ tests/ passed; 659 files already formatted
pre-commit run --files <changed files> (ruff, ruff format, typos, bare type: ignore) passed
mypy no issues in 659 source files
tach check all modules validated
pytest -x -q 13,694 passed, 25 skipped

The Relock job verified the new lock on Linux/3.12 before its branch existed: PyPI-only hosts, uv sync --locked, ruff, the format check, mypy, tach, and pytest -q (13,692 passed, 27 skipped).

After cherry-picking it, with opentelemetry-api 1.45.1 installed in the same environment. The mypy and source-size hooks run through uv run, which re-locks through this machine's mirror, so they ran directly:

Command Result
python scripts/check_lock_hosts.py uv.lock uv.lock resolves from PyPI only
SKIP=mypy,source-size pre-commit run --files <changed files> passed, including "No private index URLs in uv.lock"
mypy, python scripts/check_source_size.py, tach check no issues in 659 source files; passed; all modules validated
pytest -p no:tach tests/test_lockfile.py tests/test_ci_workflow.py tests/test_docs_links.py tests/test_docs_build_config.py 106 passed
pytest -n 8 -q 13,694 passed, 25 skipped
uv export --frozen --all-extras --no-emit-project --no-dev, then uvx pip-audit -r <export> --disable-pip no known vulnerabilities found

Checklist

  • This closes an accepted issue in this repository that is assigned to me.
  • Tests cover the change, written to fail first where behavior changed.
  • make check passes locally.
  • User-facing docs and the README match the change, or nothing user-facing changed.
  • No security invariant is weakened: writes still need a keystroke approval
    and a fail-closed audit entry, and sensitive reads are still masked.

Copilot AI balanced review requested due to automatic review settings October 22, 2026 12:35

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

No code issues were identified, but successful completion of the required CI checks under the revised timeout remains unconfirmed.

No open issues 🎉

What changed in this PR

Raises the main CI test job’s timeout to give Python 3.12 coverage runs more headroom, addressing the cancellations reported in #454.

Changes:

  • Increase the test job timeout from 45 to 60 minutes; Windows remains at 45 minutes.
  • Update the timeout regression test and document the new deadline range.
File Description
tests/​test_ci_workflow.py Pin the test job timeout to 60 minutes and explain the measured runtime.
docs/​dev/​quality-gates.md Update the documented deadline range to 10–60 minutes.
.github/​workflows/​ci.yml Set the test matrix’s timeout to 60 minutes.

Review effort: Balanced

@my-reviewer-agent my-reviewer-agent Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

검토 범위에서 보고할 이슈가 없습니다.

@hellices

Copy link
Copy Markdown
Owner Author

This PR's own test (3.12) job took 46 minutes 30 seconds (12:35:14 → 13:21:44 UTC) and passed. Under the old 45-minute limit it would have been cancelled. On the same run, 3.11 took 30 minutes and 3.13 took 31.

The test job runs the full suite under coverage after lint and type
checks, all within 45 minutes. On Python 3.12 that took 39-46 minutes:
in the last 25 CI runs the 3.12 job took 43 minutes or more ten times,
and it was cut off at the limit with no failures on 9793c5c, on #446's
3dce28c (twice, at 99%), and on main at 0648d1b (at 93%, on
korvid-runners). A cancelled required check blocks a pull request until
someone reruns it, which an outside contributor cannot do. Raise the
limit to 60 minutes and update the pinned value and the quality-gates
note. Windows keeps 45; it finishes in about 30.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@hellices
hellices force-pushed the fix/454-test-timeout branch from 753a033 to b3d9b80 Compare October 23, 2026 02:47

@my-reviewer-agent my-reviewer-agent Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

검토 범위에서 보고할 이슈가 없습니다.

Regenerated by the Relock workflow, which runs on a runner with
direct PyPI access. Locking is not possible from a machine behind
a TLS-intercepting mirror: the resulting lock names the mirror,
and the repository rejects that.

(cherry picked from commit c5e3a4e)

@my-reviewer-agent my-reviewer-agent Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

검토 범위에서 보고할 이슈가 없습니다.

리뷰 범위: 변경사항 일부만 검토했습니다.

@hellices
hellices merged commit 24b6d9c into main Oct 23, 2026
14 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.

Update opentelemetry-api past GHSA-xcq7-p3pf-p88h (unbounded tracestate parsing) CI: test (3.12) on GitHub-hosted runners hits its 45-minute timeout

2 participants