On Tue, Mar 21, 2023 at 9:22 AM Greg Stark <st...@mit.edu> wrote: > Yeah, even aside from flappers there's the problem that it's about as > common for real commits to break some test as it is for patches to > start failing tests. So often it's a real failure but it's nothing to > do with the patch, it has to do with the commit that went into git.
That is true. The best solution to that problem, I think, is to teach cfbot that it should test on top of the most recent master commit that succeeded in CI here https://github.com/postgres/postgres/commits/master . It's on my list...