Skip to content

TEMPORARY: stop drifted snapshots and the audit gate from failing CI - #121

Open
gordlin wants to merge 2 commits into
masterfrom
gordon/ignore-drifted-snapshots
Open

gordlin wants to merge 2 commits into
masterfrom
gordon/ignore-drifted-snapshots

Conversation

@gordlin

@gordlin gordlin commented Sep 21, 2026 •

Copy link
Copy Markdown

Two temporary CI changes so a phase 4 review can tell whether a change under review broke anything. The revert is already open and stacked on this: #122

Why

Measured from the last complete master Playwright run (18 September, run 35380101859):

Shard Failed
1 152
2 24
3 0
4 1
5 1
6 208
Total 386, across 82 of 246 spec files

Classifying the failure bodies: 2403 screenshot-mismatch mentions against 471 timeout mentions, and effectively no structural errors. The reference images have drifted; the editor has not broken.

Separately, npm audit --all --audit-level=critical fails on every branch including master, on four critical advisories that are all dev dependencies (handlebars, next, shell-quote, websocket-driver).

What this changes

  1. --ignore-snapshots on the Playwright command. Screenshot assertions become no-ops; every functional assertion stays live.
  2. continue-on-error: true on the audit step. It still reports.

Quarantining by file would have disabled a third of the suite including its passing tests. Quarantining 386 test titles would conflict at every upstream merge. Disabling the two failure classes is smaller and reverts in one diff.

What this does not fix

The references still need regenerating on chromium-linux, and the timeouts still need investigating. This hides drift; it does not resolve it.

What did I (author) test?

  • The failure counts and the classification above come from the shard logs of run 35380101859, not from recollection.
  • --ignore-snapshots is documented in the pinned Playwright 1.44.1: "Ignore screenshot and snapshot expectations".

gordlin and others added 2 commits September 21, 2026 16:51
The suite fails 386 tests on master, and the failure bodies carry 2403
screenshot-mismatch mentions against 471 timeout mentions, so the reference
images have drifted rather than the editor having broken. That wall of red
hides whether a change under review broke anything.

--ignore-snapshots turns screenshot assertions into no-ops and leaves every
functional assertion live. It is a stopgap: the fix is to regenerate the
references on chromium-linux. The revert is already open.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The four critical advisories are dev dependencies and they fail the job on
every branch, master included. Reporting them keeps the signal without hiding
the build and test results behind a red check. The revert is already open.

Co-Authored-By: Claude Opus 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