"Albe Laurenz" <laurenz.albe@wien.gv.at> writes:
> Leonardo M. Ram� wrote:
>> I did what you suggested, and it responds with a 63 when the string is
>> "NU?NEZ" and 209 when it's "NU�EZ".
> 63 is indeed a question mark. Since such a conversion would not be
> done by PostgreSQL, "something else" must convert � to ?N *before*
> it gets into PostgreSQL...
Yeah, I think this destroys the theory that it's due to a wrong choice
of client_encoding setting. What you'd be likely to get from that is
a "character can't be translated" kind of error, not silent substitution
of a question mark. The damage must be getting done on the client side.
regards, tom lane
