Re: unicode searches failing that use % and LIKE operators

Benjamin Weaver <benjamin(dot)weaver(at)classics(dot)ox(dot)ac(dot)uk> writes:
> Postgres 8.1.4
> server_encoding UTF8
> lc_collate en_GB.UTF-8
> lc_ctype en_GB.UTF-8

Hmph, nothing strange-looking there. I tried to reproduce the problem
here, without success. Now I was using 8.1.10 on Linux (I gather your
platform is not Linux from the spelling of the locale names), but there
hasn't been any change in the LIKE code in the 8.1 branch.

One thing that would be worth trying is to turn on query logging
(set log_statement = 'all') and verify that the problem query appears in
the postmaster log as you expect; this is just to eliminate any doubt
about whether the string encoding is getting messed up somewhere between
client and server.

Assuming you don't find a smoking gun that way, can you put together a
self-contained test case? There must be some detail of what you're
doing that I failed to reproduce.

regards, tom lane