Skip to content

ci(harness): restore-round-trip uploads no github-pages artifact (#27) - #309

Merged
mmcky merged 1 commit into
mainfrom
ci/27-restore-round-trip-artifact
Nov 10, 2026
Merged

mmcky merged 1 commit into
mainfrom
ci/27-restore-round-trip-artifact

Conversation

@mmcky

@mmcky mmcky commented Nov 10, 2026

Copy link
Copy Markdown
Contributor

The action harness has been red on main since #306 merged (run 42911209128). The failing job was restore-round-trip: "the github-pages artifact holds no ./sub/a page.html". The action is fine. The test is flaky.

Why

restore-round-trip handed the restored site to publish-gh-pages' mode: upload, then downloaded the github-pages artifact back by name. publish-modes does the same in the same run. Both uploads succeeded, a second apart: artifact 13807796278 (290 bytes, publish-modes) and 13807297206 (360 bytes, restore-round-trip). So each download can return the other job's artifact. On main, restore-round-trip downloaded publish-modes' artifact. publish-modes passes on either, since both hold ./index.html, so its check could be met by the wrong upload.

Changes

File Change
.github/workflows/test-actions.yml restore-round-trip no longer uploads or downloads a github-pages artifact. Its diff -r already shows each restore gives back the exact site. The job's comment says why, and its name drops "and publish-gh-pages takes it". publish-modes keeps the run's only github-pages artifact, so its check is about its own upload again
docs/dev/TESTING.md restore-gh-pages' row: the hand-off to publish-gh-pages moves to the canary, beside the real releases API

The hand-off templates/restore.yml makes is tested end to end in the canary at the next release: it restores publish-staged-v0180 after a gate run, as agreed on #27. That run is the release's own, not this PR's. The YAML parses and generate-docs.py --check passes; the gate's actionlint runs on this PR. It is CI-only, so there's no CHANGELOG entry. The maintainer chose this fix on 2026-11-10.

Part of #27.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WT8bMbZoC2dpGt3sPyBjyd

restore-round-trip (#306) uploaded the restored site with publish-gh-pages'
mode: upload and downloaded the github-pages artifact back by name. So does
publish-modes, in the same run, so each download could return the other
job's artifact. On main (run 42911209128) restore-round-trip read
publish-modes' artifact and failed; publish-modes passes on either, since
both hold ./index.html, so its check could be met by the wrong upload.

restore-round-trip now only restores: diff -r already shows each restore
gives back the exact site. publish-modes keeps the run's only github-pages
artifact. The hand-off templates/restore.yml makes to publish-gh-pages is
tested in the canary, at the next release.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT8bMbZoC2dpGt3sPyBjyd
Copilot AI balanced review requested due to automatic review settings November 10, 2026 23:14

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Note

Copilot was unable to complete this review at the selected tier due to model policy or plan restrictions.

🟢 Approval recommended

The change removes the cross-job github-pages artifact round-trip that was causing CI flakiness from ambiguous downloads, while retaining existing validation and updating testing documentation accordingly.

No open issues 🎉

What changed in this PR

Fixes CI harness flakiness caused by ambiguous github-pages artifact downloads when multiple jobs upload artifacts with the same name in a single workflow run.

Changes:

  • Remove the github-pages artifact upload/download round-trip from restore-round-trip, relying on its existing diff -r validation.
  • Keep publish-modes as the run’s only github-pages artifact producer so its validation refers to its own upload.
  • Update developer testing docs to reflect that the restore→publish hand-off is covered by the canary rather than the harness job.
File Description
docs/​dev/​TESTING.md Updates the developer testing matrix to indicate the restore→publish hand-off is validated via the canary rather than the harness job.
.github/​workflows/​test-actions.yml Removes the github-pages artifact round-trip from restore-round-trip to eliminate cross-job artifact name collisions while keeping restore validation.

Review effort: Balanced

@mmcky
mmcky merged commit 2315a84 into main Nov 10, 2026
71 checks passed
@mmcky
mmcky deleted the ci/27-restore-round-trip-artifact branch November 10, 2026 23:22
mmcky added a commit that referenced this pull request Nov 11, 2026
Brings in #308 (PLAN) and #309 (restore-round-trip). No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT8bMYFuqCc4A8Y1tGJoXT
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.

2 participants