Skip to content

CI: give the race step's packages 40 minutes on slow hosted runners (#254) - #255

Merged
kikashy merged 1 commit into
mainfrom
ci-race-timeout-254
Oct 29, 2026
Merged

kikashy merged 1 commit into
mainfrom
ci-race-timeout-254

Conversation

@kikashy

@kikashy kikashy commented Oct 29, 2026

Copy link
Copy Markdown
Contributor

Raises the go-race step's per-package -timeout from 20m to 40m in dev.json. It is one line.

Why. The adapter package needs about 750 s under -race on a normal runner. On the slower hosted runners CI often lands on now, it needs about twice that, which is past the 20-minute per-package limit. The slow runners run every step about 2x slower, including steps no Go change can affect: python-tests 102–106 s against 49–54 s on a normal runner. Main itself needed 1262 s for this step on such a runner. With Go 1.26.10 and the exact command, local timings for main and for agent #251 are equal: 779 s and 801 s. PR #251 has now failed three times on slow runners for this reason alone. The evidence is on #254.

Bounds kept. The step's own timeout is 3000 s and the job's is 90 min, so a hang still fails the run. No test is skipped under -race.

Not fixed here. Splitting or sharding the heaviest adapter tests, so the package has real headroom, stays open in #254.

🤖 Generated with Claude Code

…254)

The adapter package needs about 750 s under -race on a normal runner and about
twice that on the slower hosted runners CI now often lands on, past the
20-minute per-package limit. Main and pull requests measure the same; see #254.
The step's own 50-minute limit and the job's 90 minutes still bound the run.
Splitting the heaviest adapter tests stays open in #254.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: kikashy <35789537+kikashy@users.noreply.github.com>
@kikashy kikashy added this to the v0.12.0 milestone Oct 29, 2026
@kikashy
kikashy merged commit 5de4e1b into main Oct 29, 2026
5 checks passed
@kikashy
kikashy deleted the ci-race-timeout-254 branch October 29, 2026 14:30
@github-actions

Copy link
Copy Markdown
Contributor

🎉 Thank you for your contribution to Judgment Pack!

If you’d like another task, choose a path that interests you:

kikashy added a commit that referenced this pull request Oct 29, 2026
Conflicts resolved: mutation-rows.json keeps main's agent#223 rows, then the
agent#252 rows; both record-shapes files regenerated with the generator.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: kikashy <35789537+kikashy@users.noreply.github.com>
kikashy added a commit that referenced this pull request Oct 29, 2026
* Prepare v0.12.0

Set the package version to 0.12.0 in packages/harness/package.json and
package-lock.json. Add the release notes docs/releases/v0.12.0.md and the
host upgrade guide docs/upgrade/v0.12.0.md for #241, #242, #243/#252,
#223/#253, the CI timeout (#255) and the lockfile advisory fix (#245).
Mark the ADR-0025 lease amendment and the ADR-0027 rehearse amendment
accepted for v0.12.0, and state in ADR-0025 which decisions are built.
Bring the go-race paragraph of docs/releasing.md to the 40-minute
per-package timeout dev.json sets since #255. Regenerate
docs/dependencies.md for the new package-lock.json digest.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: kikashy <35789537+kikashy@users.noreply.github.com>

* Record the v0.12.0 release-gate decisions

The maintainer decided the three release-specific gates (agent#169,
2026-10-30). Add the v0.12.0 scope decision to docs/release-quality.md:
the live comparison is deferred for v0.12.0 only. Extend the Go advisory
exception for GO-2026-5781 and GO-2026-5932 to v0.12.0 and go/v0.12.0 in
the release notes, as v0.11.0 recorded it; the expiry stays 2026-11-07.
Replace the open-gate text in docs/releases/v0.12.0.md with the decisions
and the live-comparison disclosure.

The runner-image exception is not extended. Its configured scope in
.github/ci-evidence.json refuses every tag other than v0.11.0
(image-exception-tag), so remove it to let the evidence gate run with
exact matching. The repository-config test now asserts that no exception
is configured; docs/ci-evidence.md says so.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: kikashy <35789537+kikashy@users.noreply.github.com>

---------

Signed-off-by: kikashy <35789537+kikashy@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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