You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
display-redaction: linear-time mask test times out at 5000 ms in CI #7273
plugins/display-redaction/hooks/mask.test.ts:256 ("runs each of the worst accepted extra patterns in time linear in the text length") failed in merge-queue run 38185144497 (check-plugins). It took 5438 ms against the test runner's 5000 ms default timeout. The queue then removed #7172, which does not touch display-redaction (failed_checks, 2026-10-11T23:43:48Z).
The test measures wall time on a shared hosted runner, so a slow runner fails it even when the patterns are linear. Possible fixes: compare against a measured baseline in the same run rather than a fixed budget, give the test an explicit, larger timeout, or measure step counts instead of time.
Thanks for the detailed report. I understand this as: the linear-time pattern-matching test in plugins/display-redaction/hooks/mask.test.ts asserts against a fixed 5000 ms wall-clock budget, and it exceeded that (5438 ms) on a shared hosted runner during merge-queue run 38185144497, causing the queue to fail and remove an unrelated PR (#7172). The three suggested fixes (compare against a measured in-run baseline, use a larger explicit timeout, or assert on step counts instead of wall time) all sound reasonable; a maintainer will need to pick the approach.
No duplicates found for this specific test/timeout combination. It references #7221, which introduced the mask.ts module and its test suite.
Activity
github-actions commentedon Oct 11, 2026
Thanks for the detailed report. I understand this as: the linear-time pattern-matching test in
plugins/display-redaction/hooks/mask.test.tsasserts against a fixed 5000 ms wall-clock budget, and it exceeded that (5438 ms) on a shared hosted runner during merge-queue run 38185144497, causing the queue to fail and remove an unrelated PR (#7172). The three suggested fixes (compare against a measured in-run baseline, use a larger explicit timeout, or assert on step counts instead of wall time) all sound reasonable; a maintainer will need to pick the approach.No duplicates found for this specific test/timeout combination. It references #7221, which introduced the
mask.tsmodule and its test suite.Automated triage. A maintainer reviews it.