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 }}
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
When the unittests are run, the built repo under test is not the PR's HEAD commit, it's a merge of master into the PR. One nice thing about this is that if the PR is accepted, the result of merging the accepted PR into master is exactly what was tested. Also, if a different PR is merged into master that fixes test failures, it immediately helps all PR's in development- they don't need to merge the new master into the PR to pick up the fix. Sadly, there's a dark side to this: an accepted PR that introduces a flakey test will hamper in-development PR's immediately.
We recently had a linkcheck failure fixed on master by PR #20879, but the linkcheck failure was not eliminated for in-development PRs. Only after remerging master into the PR did the failure disappear. Shouldn't the same approach of merging master into the PR be used with linkcheck, as it is with the unittests?
The text was updated successfully, but these errors were encountered:
Description
When the unittests are run, the built repo under test is not the PR's HEAD commit, it's a merge of master into the PR. One nice thing about this is that if the PR is accepted, the result of merging the accepted PR into master is exactly what was tested. Also, if a different PR is merged into master that fixes test failures, it immediately helps all PR's in development- they don't need to merge the new master into the PR to pick up the fix. Sadly, there's a dark side to this: an accepted PR that introduces a flakey test will hamper in-development PR's immediately.
We recently had a linkcheck failure fixed on master by PR #20879, but the linkcheck failure was not eliminated for in-development PRs. Only after remerging master into the PR did the failure disappear. Shouldn't the same approach of merging master into the PR be used with linkcheck, as it is with the unittests?
The text was updated successfully, but these errors were encountered: