Received: from sss.sss.pgh.pa.us (sss.pgh.pa.us [209.114.166.2])
	by hub.org (8.9.3/8.9.3) with ESMTP id BAA58757
	for <pgsql-interfaces@postgreSQL.org>;
	Mon, 6 Dec 1999 01:09:43 -0500 (EST)
	(envelope-from tgl@sss.pgh.pa.us)
Received: from sss.sss.pgh.pa.us (localhost [127.0.0.1])
	by sss.sss.pgh.pa.us (8.9.1/8.9.1) with ESMTP id BAA15299;
	Mon, 6 Dec 1999 01:08:33 -0500 (EST)
To: Charles Randle <caspanis@cybervale.com>
cc: pgsql-interfaces@postgreSQL.org
Subject: Re: [INTERFACES] Problems with postgres V6.5.3 large objects 
In-reply-to: Your message of Mon, 06 Dec 1999 00:44:43 -0500
	<384B4D4B.BF6EE9D6@cybervale.com>
Date: Mon, 06 Dec 1999 01:08:33 -0500
Message-ID: <15297.944460513@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Charles Randle <caspanis@cybervale.com> writes:
> In terms of access to the large object, my class maintains an array of
> objects which are currently open (Oid,name,file descriptor etc ..).
> Whenever access is needed to any large object an accessor method
> simply executes 'lo_read'/lo_write' or lo_seek as is desired . If the
> object is not on the active list then another method opens the object
> (using lo_open ).As is to be expected the destructor of the class
> closes all open objects .

Sounds reasonable as far as it goes.  My best guess at this point is a
client-side memory stomp that is clobbering libpq's record of which
function OID to call for each available LO operation.  See
interfaces/libpq/fe-lobj.c, particularly lo_initialize() and its
callers, and then try tracing through the client with a debugger
to see where the conn->lobjfuncs struct gets clobbered.

Good luck!

			regards, tom lane