David E. Wheeler wrote:
>>
>> (If we get to having per-database collations in 8.4 then integrating a
>> test with a non-default collation would get a lot easier, of course;
>> but for the moment I'm afraid you've got to work with what's there.)
>
> Could I supply two comparison files, one for Mac OS X with en_US.UTF-8
> and one for everything else, as described in the last three paragraphs
> here?
>
> http://www.postgresql.org/docs/current/static/regress-variant.html
>
> That way, I can at least make sure that the multibyte stuff *does*
> work. Even if it's tested on only one platform, the purpose is not to
> test a particular collation, but to test that CITEXT is actually
> sensitive to locale.
>
>
You can certainly add any tests you like. But the buildfarm does all its
regression tests using an install done with --no-locale. Any test that
requires some locale to work, or make sense, should probably not be in
your Makefile's REGRESS target.
Some time ago I raised the question of doing locale- and encoding-aware
regression tests, and IIRC the consensus seemed to be that it was not
worth the effort, but maybe we need to revisit that. We certainly seem
to be doing a lot more work now to get Postgres to where it needs to be
w.r.t. locale support, and we've cleaned up a lot of the encoding issues
we had, so maybe we need to reflect that in our testing regime more.
cheers
andrew