To: Marc Cousin <mcousin@sigma.fr>
cc: pgsql-admin@postgresql.org
Subject: Re: cannot restore a view after a dump 
In-reply-to: <200804080911.56042.mcousin@sigma.fr> 
References: <200804041022.31746.mcousin@sigma.fr>
	<200804041521.52137.mcousin@sigma.fr>
	<200804080911.56042.mcousin@sigma.fr>
Comments: In-reply-to Marc Cousin <mcousin@sigma.fr>
	message dated "Tue, 08 Apr 2008 09:11:55 +0200"
Date: Tue, 08 Apr 2008 10:03:58 -0400
Message-ID: <448.1207663438@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Marc Cousin <mcousin@sigma.fr> writes:
> Sorry to post again, but I feel this issue is a bit strange and I'd like to 
> understand it. The problem is that I've got the same query that runs 
> sometimes, and fails with a syntax error at other times... It's the first 
> time I've seen it, and I've been using PostgreSQL for a while now ...

Oh, sorry for not replying directly to you.  This was analyzed over in
-hackers and we now understand the cause, but haven't a fix quite yet:
http://archives.postgresql.org/pgsql-hackers/2008-04/msg00278.php

It's been broken since approximately forever, so if you'd been using
that view definition for awhile I'm surprised you didn't run into the
problem before.

The reason for the erratic behavior is that it depends on what happened
to already be in the chunks of memory used to form the array constants.
In a freshly-started backend the view creation would probably always
succeed, but as soon as memory has gotten dirtied a bit, maybe not.

			regards, tom lane