* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> Stephen Frost <sfrost@snowman.net> writes:
> > * Tom Lane (tgl@sss.pgh.pa.us) wrote:
> >> I'm not entirely sure what to do about this. We could back-patch that
> >> patch into 9.0 and 9.1, but it's conceivable somebody would squawk about
> >> planner behavioral changes. The only other idea that seems practical is
> >> to remove regression test cases that have platform-specific results in
> >> those branches. Probably that wouldn't result in a real reduction in the
> >> quality of the test coverage for those branches (we could still execute
> >> the query, just not EXPLAIN it). But it seems like a pretty ad-hoc
> >> answer, and the next case might be one that hurts more not to test.
> >>
> >> Thoughts?
>
> > Have an alternate file for those other cases, rather than remove the
> > test? The complaint was about one buildfarm member, so hopefully that's
> > practical and doesn't require a lot of different permutations.
>
> I considered that but don't find it practical or attractive, especially
> not if the only way to keep such a file updated is to wait and see whether
> the buildfarm complains.
I agree, that's a bit unfortunate, but it strikes me as pretty unlikely
that we're ever going to change those tests or that a code change would
end up causing yet another different plan before 9.1 is completely out
of support in the next couple years.
> On the whole I'm leaning towards back-patching 33e99153e. While the case
> of exactly equal plan costs does come up in the regression tests (which
> tend to inspect plans for queries on small simple tables), I think it's
> relatively unlikely to happen with real-world data.
I agree it's unlikely, but I don't particularly like changing our mind
on a back-patching decision 3 years later to satisfy our regression
tests.
Still, I don't feel particularly strongly about either side of this, so
I'm happy with you making the decision.
Thanks! Stephen