From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 23693 invoked by alias); 26 Apr 2011 08:30:25 -0000 Received: (qmail 23671 invoked by uid 22791); 26 Apr 2011 08:30:19 -0000 X-SWARE-Spam-Status: No, hits=-2.3 required=5.0 tests=AWL,BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,RFC_ABUSE_POST,TW_BJ,TW_JC X-Spam-Check-By: sourceware.org Received: from mail-ww0-f43.google.com (HELO mail-ww0-f43.google.com) (74.125.82.43) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Tue, 26 Apr 2011 08:30:05 +0000 Received: by wwb17 with SMTP id 17so309340wwb.12 for ; Tue, 26 Apr 2011 01:30:04 -0700 (PDT) Received: by 10.227.207.196 with SMTP id fz4mr473585wbb.42.1303806604213; Tue, 26 Apr 2011 01:30:04 -0700 (PDT) Received: from richards-thinkpad (gbibp9ph1--blueice2n1.emea.ibm.com [195.212.29.75]) by mx.google.com with ESMTPS id e13sm3692603wbi.6.2011.04.26.01.30.02 (version=TLSv1/SSLv3 cipher=OTHER); Tue, 26 Apr 2011 01:30:03 -0700 (PDT) From: Richard Sandiford To: Tristan Gingold Mail-Followup-To: Tristan Gingold ,Dave Korn , binutils , rdsandiford@googlemail.com Cc: Dave Korn , binutils Subject: Re: Ping: RFC: add dwarf support for xcoff/AIX References: <482268C0-87C5-466E-BB37-59F1DFB341F9@adacore.com> <340C0197-8B8D-43BB-8B64-D2CC55EA2482@adacore.com> Date: Tue, 26 Apr 2011 08:30:00 -0000 In-Reply-To: <340C0197-8B8D-43BB-8B64-D2CC55EA2482@adacore.com> (Tristan Gingold's message of "Tue, 26 Apr 2011 09:43:35 +0200") Message-ID: User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Mailing-List: contact binutils-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: binutils-owner@sourceware.org X-SW-Source: 2011-04/txt/msg00364.txt.bz2 Tristan Gingold writes: >> * because section names are limited to 8 characters, the standard >> dwarf names cannot be used. XCOFF defines alternate names (.dwinfo, >> .dwline...) But to simplify the interface with gdb and objdump, the >> xcoff back-end automatically canonicalize the names to the standard >> dwarf names. To do this I added a coff hook (expand_section_name) >> and a new coff macro (adjust_scnhdr_before_swap). Not sure that this >> conversion is a good idea, so comments are welcome here. I can see the attraction of this, but does it mean that objcopy and objdump operations on the original ("real") sections won't work? That might be a bit confusing for human (rather than scripted) users of the tools. Especially if you use the AIX and GNU tools together. An alternative might be to have a pointer to a table of section names in the bfd_target structure. We could then replace direct uses of ".debug_frame" etc. with uses of accessor macros. What do others think? >> * AIX as has a special pseudo for declaring dwarf sections: .dwsect >> However this pseudo does prepend the size. As this is very unusual >> (and would require more work in gcc), I choose to not implement it >> and to use the standard .section semantic (which was already present >> for convenience). Of course, nothing prevents from adding .dwsect >> later (the difficult part would be in gcc) That's fair enough, but it could also be added to gas with the gcc side left as future work. Richard