Relationships
#2909 Flaky: usecase_sync_characterization_data_test 'data gc (serve)' asserts markDirty order that varies under parallel load
Opened by hammz · 10/1/2026· Shipped 10/2/2026
Description
integration/usecase_sync_characterization_data_test.ts, step 'data gc (serve)', failed in a full verify-build run, with the two markDirty ops in the opposite order to the pinned fixture:
Actual: markDirty data///state/2, markDirty data///state/1, push[0 del 4] Expected: markDirty data///state/1, markDirty data///state/2, push[0 del 4]
The remote removals are identical; only the order of the markDirty calls differs. The step was added with swamp-club#2856 (8456bba0).
Steps to reproduce
Run the full suite (deno run test, --parallel), as verify-build does. The failure is intermittent: the file passed 15 out of 15 runs on its own, and it passed in an earlier full run of the same tree. Seen in verify-build run 97512354-2b4a-452e-99c1-5d75b3ef229d at commit 0f52cd95 on branch issue-2897, whose changes do not touch data gc.
Expected
The characterisation compares markDirty ops as an unordered set, or gc marks versions in a deterministic order (for example, sorted by version) so that the pinned order holds.
Shipped
Click a lifecycle step above to view its details.
Sign in to post a ripple.