ci: retarget the HTML5 self-report and close out #16 #80

Merged

yuris merged 2 commits from ci/close-16 into main

2026-08-21 04:20:49 +00:00

yuris commented

2026-08-21 04:07:27 +00:00

Owner

Closes #16.

#16 asked whether Chromium starts in the job container. It does: runs 163, 167 and several since are green, full six-phase smoke on both loads.

The first version of this PR also dropped continue-on-error and the self-report. That was premature and this one reverses it. Run 716 failed on the PR and run 717 passed on the same commit, so the tier is flaky at roughly 1 in 5, and with the report gone the failure was unreadable. #81 now owns the flake and receives the reports.

What changes here: the report points at #81 instead of #16, and the comments say why each rail is still in place.

Closes #16. #16 asked whether Chromium starts in the job container. It does: runs 163, 167 and several since are green, full six-phase smoke on both loads. The first version of this PR also dropped `continue-on-error` and the self-report. That was premature and this one reverses it. Run 716 failed on the PR and run 717 passed on the same commit, so the tier is flaky at roughly 1 in 5, and with the report gone the failure was unreadable. #81 now owns the flake and receives the reports. What changes here: the report points at #81 instead of #16, and the comments say why each rail is still in place.

yuris added 1 commit

2026-08-21 04:07:28 +00:00

ci: let the HTML5 tier fail loudly and drop its self-report

ci / ci (pull_request) Successful in 21s

Details

conformance / conformance (pull_request) Failing after 2m20s

Details

80f5d86678

#16 asked one question — does Chromium start in the job container — and two
green runs answered it. The reporting step existed only to get that answer
out of a runner whose logs the API will not serve; with the question closed
it becomes a recurring comment on a closed issue, which is the noise 2581e8f
removed.

With no report, continue-on-error would make the tier invisible: a silently
skipped verdict is worse than no verdict. So it goes too, and this tier now
fails the same way the native one does, as a red X read in the web UI.

39 lines down to 9. The self-report and the flag are both one revert away if
the tier turns out to wobble.

yuris referenced this pull request

2026-08-21 04:16:16 +00:00

HTML5 conformance is flaky in CI: same commit fails then passes #81

yuris added 1 commit

2026-08-21 04:16:58 +00:00

ci: keep the HTML5 rails until the flake has a cause (#81 )

ci / ci (pull_request) Successful in 20s

Details

conformance / conformance (pull_request) Successful in 2m33s

Details

01fdcd24d0

Reverses the half of the previous commit that was premature. #16 asked
whether Chromium runs in the job container; it does, and that question is
closed. Whether the tier is steady is a different question, and run 716
answered it badly: failure on a PR, then success on the same commit when
re-dispatched as 717.

Removing the self-report in the same change that made the step gate PRs was
the wrong order — the first gated failure was unreadable, because the API
serves no job logs. So continue-on-error and the report both stay, the
report now feeding #81, which owns the flake.

The tier still runs on every change that touches it. It just cannot fail a
PR over something nobody can explain yet.

yuris changed title from ~~ci: let the HTML5 tier fail loudly and drop its self-report~~ to ci: retarget the HTML5 self-report and close out #16

2026-08-21 04:17:11 +00:00

yuris merged commit e0b337a55d into main

2026-08-21 04:20:49 +00:00

yuris referenced this pull request from a commit

2026-08-21 04:20:51 +00:00

ci: retarget the HTML5 self-report and close out #16 (#80)