From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 14743 invoked by alias); 20 Jul 2004 10:27:31 -0000 Mailing-List: contact gdb-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-owner@sources.redhat.com Received: (qmail 14726 invoked from network); 20 Jul 2004 10:27:29 -0000 Received: from unknown (HELO holly.csn.ul.ie) (136.201.105.4) by sourceware.org with SMTP; 20 Jul 2004 10:27:29 -0000 Received: from skynet.csn.ul.ie (skynet [136.201.105.2]) by holly.csn.ul.ie (Postfix) with ESMTP id 7DC12937DA for ; Tue, 20 Jul 2004 10:27:29 +0000 (Eire) Received: by skynet.csn.ul.ie (Postfix, from userid 2009) id 60C88EA66; Tue, 20 Jul 2004 11:27:29 +0100 (IST) Date: Tue, 20 Jul 2004 10:35:00 -0000 To: gdb@sources.redhat.com Subject: OpenOffice.org and gdb Message-ID: <20040720102727.GA15676@skynet.ie> Mail-Followup-To: gdb@sources.redhat.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.3.28i From: caolan@csn.ul.ie (Caolan McNamara) X-SW-Source: 2004-07/txt/msg00259.txt.bz2 There are a few little glitches with gdb that make trying to debug OpenOffice quite difficult, I've put together a little set of testcases at http://people.redhat.com/caolanm/mygdbproblems.tar.gz The exact gdb version is 6.1post-1.20040607.8rh from fc3 development but I've struggled with similiar ones in other cvs varients and older releases for some time. 1. make gdbtestc Deferred breakpoints that get resolved in dlopened modules only work the first cycle of debugging, subsequent runs attempt to reset them often break ? 2. make gdbtestc++ The known issues of gdb/1091 & gdb/1193. Not being able to set breakpoints in constructors :-( 3. make gdbtest2c++ Deferred breakpoints don't allow deferring on an unseen c++ member function ?, but do allow breaking on an unseen line i.e. c++test2cmdsA works, while c++test2cmdsB doesn't 4. make gdbtest3c++ A debug build of OOo is extremely large, the default is to build it without debugging symbols and then only build the parts being debugged with symbols enabled. The component model should boils down to effectively this test, with the problem that deferred breakpoints apparently can't be set from non debug code ? though the lib being loaded will have the desired symbols. I can work around 4 for my own needs by building the bit of OOo that contains main with debugging enabled, and I can work around 3 by restricting myself to line numbers, but 1 and 2 are a real curse. Are there any other known workarounds that I can employ, or is there simply a better compile flag incantation I should use :-) I guess I should just log bugs against them. C. - Caolan McNamara | caolan@skynet.ie http://www.skynet.ie/~caolan | +353 86 8161184 I have a wonderful proof of that theorem, but this line is too small to show it