Repository navigation
CI: retry the LSQB runner's wall-clock tests; document the flake - #14
Merged
Merged
Conversation
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
main'sworkspacerun failed on a documentation-only commit (5507be0): three of the LSQB runner's process-supervision tests failed together withcould not be reaped after SIGKILLanddid not become READY within the configured timeout. Those tests assert wall-clock windows as short as 50 ms, and a starved hosted runner can stall a process for longer.Graph benchmark testsstep makes up to three attempts and emits a warning per failed attempt. A real defect fails all three, so a red build still means something.docs/LSQB_RUNNER_TIMING_FLAKES.mdrecords the observations, the reproduction, and the fix that belongs in the harness.Reproduction: none on macOS with all cores saturated, none on Linux limited to one CPU (0 of 27 runs). On Linux limited to 0.4 CPU: 1 of 12 runs failed in parallel and 2 of 12 with
--test-threads=1, so serialising the tests is not a fix.This is a workaround. The tests are not mine and I have not changed them; widening their windows is left to the harness's owner, with the evidence.
🤖 Generated with Claude Code