From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 3118 invoked by alias); 24 Jul 2013 16:25:09 -0000 Mailing-List: contact gdb-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-owner@sourceware.org Received: (qmail 3088 invoked by uid 89); 24 Jul 2013 16:25:08 -0000 X-Spam-SWARE-Status: No, score=-7.0 required=5.0 tests=AWL,BAYES_00,KHOP_THREADED,RCVD_IN_HOSTKARMA_W,RCVD_IN_HOSTKARMA_WL,RDNS_NONE,SPF_HELO_PASS,SPF_PASS autolearn=no version=3.3.1 Received: from Unknown (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.84/v0.84-167-ge50287c) with ESMTP; Wed, 24 Jul 2013 16:25:08 +0000 Received: from int-mx01.intmail.prod.int.phx2.redhat.com (int-mx01.intmail.prod.int.phx2.redhat.com [10.5.11.11]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id r6OGOvG9002983 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Wed, 24 Jul 2013 12:24:58 -0400 Received: from [127.0.0.1] (ovpn01.gateway.prod.ext.ams2.redhat.com [10.39.146.11]) by int-mx01.intmail.prod.int.phx2.redhat.com (8.13.8/8.13.8) with ESMTP id r6OGOuP0001131; Wed, 24 Jul 2013 12:24:56 -0400 Message-ID: <51EFFFD7.4000000@redhat.com> Date: Wed, 24 Jul 2013 16:25:00 -0000 From: Pedro Alves User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130625 Thunderbird/17.0.7 MIME-Version: 1.0 To: Doug Evans CC: Nikita Karetnikov , gdb Subject: Re: Debugging 64-bit programs using 32-bit GDB References: <87r4f8oxo3.fsf@karetnikov.org> <51EFE5C0.60706@redhat.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-SW-Source: 2013-07/txt/msg00087.txt.bz2 On 07/24/2013 05:09 PM, Doug Evans wrote: > On Wed, Jul 24, 2013 at 7:33 AM, Pedro Alves wrote: >> On 07/09/2013 11:06 AM, Nikita Karetnikov wrote: >>> Is it possible to debug a 64-bit executable ('bin/guile' [1]) using >>> 32-bit GDB? (I use GNU/Linux.) >> >> I don't think so. The ptrace registers interfaces allow retrieving >> the right register sets (at least the modern interfaces), but >> things like breakpoints, etc. won't work, as the 32-bit GDB won't >> be able to handle the inferior's 64-bit addresses (due to how ptrace >> works). Last I checked, Power had extra ptrace interfaces for that. >> Don't know about MIPS. >> >> (you can however connect a 32-bit GDB to a 64-bit GDBserver, and use >> that to debug a 64-bit program, though that's more trouble >> than just using a 64-bit GDB.). >> >>> >>> I've tried to configure GDB 7.6 with '--enable-64-bit-bfd' (see [2]), >>> but it still shows "mips-tdep.c:709: internal-error: bad register >>> size". >> >> The x86_64 port catches the case early and bails out (amd64_linux_read_description): >> >> if (sizeof (void *) == 4 && is_64bit && !is_x32) >> error (_("Can't debug 64-bit process with 32-bit GDB")); > > [filed for reference sake] > > Alas basic 32-bit linux builds of gdb don't compile amd64-linux-nat.c. > It would be really nice to print a better error message than the > following for 32-bit gdb debugging 64-bit executable: > > "/tmp/hello.x64": not in executable format: File format not recognized > > It's a not uncommon failure mode. Agree, but not sure how much clearer can it get. Maybe: "/tmp/hello.x64": not in recognized executable file format. ? I'd just suggest building 32-bit GDB with 64-bit support. Even if it can't debug 64-bit programs natively, it can remote debug them, and debug 64-bit cores, in addition to getting rid of that failure mode. > Hmmm, amd64-linux-nat.c is not compiled even if I do > --enable-targets=all --enable-64-bit-bfd. Oh, you're right, of course. Guess I was thinking of gdbserver, where linux-x86.c is shared between 32-bit and 64-bit, and it has the similar handling: if (sizeof (void *) == 4) { if (is_elf64 > 0) error (_("Can't debug 64-bit process with 32-bit GDBserver")); #ifndef __x86_64__ else if (machine == EM_X86_64) error (_("Can't debug x86-64 process with 32-bit GDBserver")); #endif } Two weeks vacation can swap out a lot of context. :-) > When does the above error get triggered? Looks like it doesn't. -- Pedro Alves