Re: libpq and mysterious "invalid byte sequence for encoding UTF8".

=?UTF-8?B?SmnFmcOtIFBhdmxvdnNrw70=?= <jiri@pavlovsky.eu> writes:
> Client and server encoding is UTF8. Locales are set to C.UTF8.

If lc_messages is C then no translation of error strings would happen,
so that theory goes out the window.  Oh well.

Perhaps you could attach gdb to the backend and get a stack trace
from the call to errfinish?

https://wiki.postgresql.org/wiki/Generating_a_stack_trace_of_a_PostgreSQL_backend

            regards, tom lane