From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 13108 invoked by alias); 12 Jun 2005 08:50:34 -0000 Mailing-List: contact gdb-patches-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sources.redhat.com Received: (qmail 13099 invoked by uid 22791); 12 Jun 2005 08:50:32 -0000 Received: from legolas.inter.net.il (HELO legolas.inter.net.il) (192.114.186.24) by sourceware.org (qpsmtpd/0.30-dev) with ESMTP; Sun, 12 Jun 2005 08:50:31 +0000 Received: from HOME-C4E4A596F7 (IGLD-83-130-207-5.inter.net.il [83.130.207.5]) by legolas.inter.net.il (MOS 3.5.8-GR) with ESMTP id EOZ61656 (AUTH halo1); Sun, 12 Jun 2005 11:50:28 +0300 (IDT) Date: Sun, 12 Jun 2005 08:50:00 -0000 Message-Id: From: Eli Zaretskii To: Mark Kettenis CC: jmolenda@apple.com, gdb-patches@sources.redhat.com In-reply-to: <200506120707.j5C77TL6005645@elgar.sibelius.xs4all.nl> (message from Mark Kettenis on Sun, 12 Jun 2005 09:07:29 +0200 (CEST)) Subject: Re: The gdb x86 function prologue parser Reply-to: Eli Zaretskii References: <85C775AE-3B05-431E-96D2-49EA9D1413E6@apple.com> <200506120707.j5C77TL6005645@elgar.sibelius.xs4all.nl> X-SW-Source: 2005-06/txt/msg00122.txt.bz2 > Date: Sun, 12 Jun 2005 09:07:29 +0200 (CEST) > From: Mark Kettenis > CC: gdb-patches@sources.redhat.com > > Being able to debug -fomit-frame-pointer code without CFI probably > means that instead of scanning the prologue, we'll have to scan the > complete function up to the current instruction pointer. I really > wonder if that's the way we should go. If this might make sense in some, possibly rare, situations, how about if we do that conditioned on some user option? It is IMHO much better to have some way of dealing with such code than to have none. > assuming that a function > has set up a stack frame isn't right; it makes gdb silently skip > function calls in backtraces. That can be very confusing. As I've > stated before, I'd rather have a backtrace that's obviously wrong than > one that silently omits things. Again, a user option, off by default, that would produce a backtrace with possibly omitted frames rather than a blatantly botched one, could be a Good Thing here. Just my $0.02.