To: Randall Wilson <rwilson@earthcomber.com>
cc: "pgsql-admin@postgresql.org" <pgsql-admin@postgresql.org>
Subject: Re: Idle Error invalid byte sequence 
In-reply-to: <48D2B5B0.9020708@earthcomber.com> 
References: <48D273D8.7010603@earthcomber.com>
 <23674.1221761835@sss.pgh.pa.us>
	<48D2B5B0.9020708@earthcomber.com>
Comments: In-reply-to Randall Wilson <rwilson@earthcomber.com>
	message dated "Thu, 18 Sep 2008 15:10:24 -0500"
Date: Thu, 18 Sep 2008 16:57:30 -0400
Message-ID: <2079.1221771450@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Randall Wilson <rwilson@earthcomber.com> writes:
> Yes, the encoding was changed from SQL_ASCII to UTF-8. Does the fact
> that it's an "Idle" error conform your theory that some client is
> causing the error? I thought the "Idle" error was caused by errors found
> during background processes, like vacuums.

You'd have to tell us what your log_line_prefix is before we could
interpret that, but yeah, an encoding error in an incoming query string
would be reported before the backend could even consider going non-idle.

A SQL_ASCII database would have accepted any data whatsoever.  If you
want to change to something where you actually know the encoding, that's
fine, but realize that you're going to have to make your clients be
honest about what encoding they're using, too.

			regards, tom lane