Daniel Gustafsson <daniel@yesql.se> writes:
> The deviation between database names on crake found in the follow-up commit is
> a bit concerning. Too much magic hidden away from the codetree makes it hard
> to get tests right on the first try.
Yeah, I didn't understand that at all. I have a non-reporting
buildfarm setup here that I run when it's necessary to debug these
cross-version tests, and it did not show this failure. crake itself
was passing with the pre-16 branches, so I'm mystified why 16 is
different. Anyway, I see it's green now, so all good.
regards, tom lane