mikebridge commented on PR #45548: URL: https://github.com/apache/superset/pull/45548#issuecomment-6376908876
_AI-assisted comment on Mike's behalf._ Closing this one. @fitzee, thanks for catching that reusing the full run's coverage let other packages' tests satisfy a package's 100% gate. The fix in 81bdaeec (per-test coverage contexts, with each gate filtered to its own test directory) is correct, but it uses up most of the saving: - A 1,631-test local sample ran about 12.7% slower with contexts, and its coverage file grew from 3.3 MB to 11.9 MB. Extrapolated to the full run, that's roughly +160 s and +156 MB, against about 211 s saved by dropping the isolated reruns. - In CI, this PR's unit job took 36m45s. Today's master runs took between 24m and 38m, so any gain is lost in normal variation. - Dropping the serial reruns would also lose their incidental check for test-order and shared-state dependence. The remaining gain is too small to justify the extra complexity, so the isolated package gates stay as they are. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
