Palle Girgensohn <girgen@pingpong.net> writes:
> Hmm, yes, there is indeed one thing I didn't tell you... the server is
> patches with my ICU patch. That might be a problem? The patch is at
> http://people.freebsd.org/~girgen/postgresql-icu
Hmm ... that does fool around with the behavior of string comparisons,
doesn't it?
I know nothing about ICU so can't really critique the patch, but I'd
say the next step is to see if you can reproduce the failure without
ICU in place.
Another thing to think about is whether you maybe need to reindex
whatever indexes exist on participant.userid. If you've changed
the collation ordering by installing or bug-fixing ICU, then pre-
existing textual indexes are corrupt. Not sure why psql wouldn't
show the same failure, but maybe a different plan is being chosen
in the two cases? It'd be worth checking to see if you get the
same plans for
explain analyze
select lang from participant where userid = 'mango' and course_id = 38;
prepare foo(varchar,int4) as
select lang from participant where userid = $1 and course_id = $2;
explain analyze execute foo('mango',38);
regards, tom lane