* Tom Lane (tgl@sss.pgh.pa.us) wrote:
> Stephen Frost <sfrost@snowman.net> writes:
> > * Stephen Frost (sfrost@snowman.net) wrote:
> >> Looks like the test_pg_dump extension made the Windows builds upset.
> >> I'm guessing that's because I set 'MODULES_big' even though there isn't
> >> a .c component.
> >>
> >> Doing a local build with that commented out, assuming that works and
> >> doesn't generate the .so any more on my Linux box, I'll push the change
> >> up to hopefully fix those buildfarm members.
>
> > Alright, apparently that made other Windows buildfarm members unhappy...
>
> > I guess the next approach will be to add back MODULES_big and add in a
> > .c file for the Windows systems to be happy about. I'm certainly open
> > to other suggestions.
>
> You should not need to do that; cf src/test/modules/test_extensions,
> which has got SQL-only extensions.
test_extensions is also included in the "contrib_excludes" list in
src/tools/msvc/Mkvcbuild.pm that I mentioned in my last email, so I'm
thinking that's what is needed.
> But at this point I think Peter's complaint has some force to it, and that
> what you ought to do is revert the testing patch. You can have another go
> after beta1.
Are you suggesting commiting to still-9.6-HEAD post-beta1? I took
Peter's comment as suggesting that adding the tests would have to wait
til after we branched 9.6, as we do for features.
I'd really like to have these tests included as that will make them
available to others more easily to add on to, and I'm certainly planning
to continue adding tests until I get pg_dump.c's coverage a lot better.
That seems like the perfect kind of effort that should be happening
right now- adding more tests and working to make sure that what's been
committed is correct (and fixing it when it isn't, as discovered by the
test suite with transforms and casts...).
Thanks!
Stephen