On Sat, Oct 14, 2023 at 08:27:58PM -0400, Tom Lane wrote:
> Noah Misch <noah@leadboat.com> writes:
> > On Sat, Oct 14, 2023 at 07:45:21PM -0400, Tom Lane wrote:
> >> Hmm, I'm not sure that that last is a good idea. The upshot of this
> >> (because of the opr_sanity.out change) is that "make installcheck"
> >> will fail against existing back-branch installations. That seems
> >> more likely to cause problems/confusion than the alternative of just
> >> depending on the code change.
>
> > I'm way more worried about amcheck failing on all those indexes than I am
> > about someone who needs to tweak their installcheck rig.
>
> Well, that's indeed going to be a pain for affected people, but
> it doesn't seem like a reason to also break installcheck.
That's right. We don't have a standard that installcheck of v13.N will have
zero diffs on an initdb from v13.0. Zero diffs would be help a tiny user
population, and corrected catalog entries will help a different population. I
don't think either choice does enough harm to reopen the decision. We've got
bigger fish to fry.