Jeff Janes <jeff.janes@gmail.com> writes:
> The regression test you added for this change causes an infinite loop when
> run against an unpatched server with --install-check. That is a bit
> unpleasant. Is there something we can and should do about that? I was
> expecting regression test failures of course but not an infinite loop
> leading towards disk exhaustion.
We very often add regression test cases that will cause unpleasant
failures on unpatched code. I categorically reject the idea that
that's not a good thing, and question why you think that running
known-broken code against a regression suite is an important use case.
regards, tom lane