On ons, 2011-03-02 at 18:00 +0200, Heikki Linnakangas wrote:
> Hmm, regardless of what's causing this, I wonder if we should check
> that the locale works at CREATE COLLATION time, rather than fail when
> you try to use it. Ie. add a call to pg_newlocale_from_collation() at
> the end of DefineCollation().
Yeah, I was apparently hoping that check_encoding_locale_matches() would
catch invalid locales, but it doesn't, so we should put in an explicit
check.