Re: stand-alone composite types patch (was [HACKERS] Proposal:

Joe Conway <mail(at)joeconway(dot)com> writes:
> Joe Conway wrote:
>> There is also a small adjustment to the expected output file for
>> select-having. I was getting a regression failure based on ordering of
>> the results, so I added ORDER BY clauses.

> I now see that the select-having fails differently on my PC at home from
> the one at work. At home I see in postgresql.conf:

> LC_MESSAGES = 'C'
> LC_MONETARY = 'C'
> LC_NUMERIC = 'C'
> LC_TIME = 'C'

> and at work:

> LC_MESSAGES = 'en_US'
> LC_MONETARY = 'en_US'
> LC_NUMERIC = 'en_US'
> LC_TIME = 'en_US'

> I have been running `make installcheck` instead of `make check`. I
> gather that `make installcheck` does not set LOCALE to 'C' (as make
> check does -- I think). Should it?

The problem is that LC_COLLATE is (presumably) also en_US at work;
"make installcheck" hasn't got any way of changing the collation of
the preinstalled server, because that was frozen at initdb. See
the discussion in the regression-test documentation.

regards, tom lane