From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 1870 invoked by alias); 18 Oct 2013 15:12:39 -0000 Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org Received: (qmail 1854 invoked by uid 89); 18 Oct 2013 15:12:38 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-2.3 required=5.0 tests=AWL,BAYES_00,RP_MATCHES_RCVD,SPF_HELO_PASS,SPF_PASS autolearn=ham version=3.3.2 X-HELO: mx1.redhat.com Received: from mx1.redhat.com (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Fri, 18 Oct 2013 15:12:37 +0000 Received: from int-mx10.intmail.prod.int.phx2.redhat.com (int-mx10.intmail.prod.int.phx2.redhat.com [10.5.11.23]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id r9IFCZvN012753 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Fri, 18 Oct 2013 11:12:35 -0400 Received: from [127.0.0.1] (ovpn01.gateway.prod.ext.ams2.redhat.com [10.39.146.11]) by int-mx10.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id r9IFCXGj002531; Fri, 18 Oct 2013 11:12:34 -0400 Message-ID: <52614FE1.4040404@redhat.com> Date: Fri, 18 Oct 2013 15:12:00 -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: "Maciej W. Rozycki" CC: gdb-patches@sourceware.org Subject: Re: [PATCH] Avoid producing broken non-native core files References: <525EAF0E.3050801@redhat.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-SW-Source: 2013-10/txt/msg00567.txt.bz2 On 10/16/2013 09:09 PM, Maciej W. Rozycki wrote: > On Wed, 16 Oct 2013, Pedro Alves wrote: > >>> The cause of missing register information is elfcore_write_prstatus in BFD >>> that writes no data (and returns NULL) on non-native targets that have no >>> explicit support (bed->elf_backend_write_core_note is NULL), because >>> HAVE_PRSTATUS_T and HAVE_PRSTATUS32_T are both forcibly undefined for >>> non-native BFD configurations. >> >> And if cross debugging, and bed->elf_backend_write_core_note is NULL for the >> current target, but HAVE_PRSTATUS_T/HAVE_PRSTATUS32_T are defined (for the >> native target), then gcore will generate bogus notes. :-/ > > As I say these macros are forcibly undefined for non-native BFD -- see > its configure.in. You seem to forget --enable-targets=all. In that case, the HAVE_... bits will be defined for the native target, but that might not be the target the core is being generated for. > >> It probably will be a long time before bfd's core generation is >> host-independent everywhere, unfortunately. As future improvement, maybe >> we should try _only_ bed->elf_backend_write_core_note, and skip the >> HAVE_... bits, unless debugging with the native target. Anyway, > > This is effectively already the case. I don't think it is. >>> --- gdb-fsf-trunk-quilt.orig/gdb/linux-tdep.c 2013-10-14 22:44:49.868756722 +0100 >>> +++ gdb-fsf-trunk-quilt/gdb/linux-tdep.c 2013-10-14 22:46:21.887601484 +0100 >>> @@ -1211,7 +1211,9 @@ linux_corefile_thread_callback (struct t >>> args->stop_signal); >>> args->num_notes++; >>> >>> - if (siginfo_data != NULL) >>> + /* Don't return anything if we got no register information above, >>> + such a core file is useless. */ >>> + if (args->note_data != NULL && siginfo_data != NULL) >> >> ... I was surprised to find that it took me a bit to grok the flow of >> this change. I'd prefer the more explicit: >> >> args->note_data = args->collect (regcache, info->ptid, args->obfd, >> args->note_data, args->note_size, >> args->stop_signal); >> >> + if (args->note_data == NULL) >> + { >> + /* Don't return anything if we got no register information above, >> + such a core file is useless. */ >> + do_cleanups (old_chain); >> + return 1; >> + } >> >> args->num_notes++; >> >> if (siginfo_data != NULL) >> { >> args->note_data = elfcore_write_note (args->obfd, >> args->note_data, >> args->note_size, >> "CORE", NT_SIGINFO, >> siginfo_data, siginfo_size); >> args->num_notes++; >> } >> >> >> This is OK with that change. > > I don't like the second exit point and the duplicate call to do_cleanups, > such arrangements require more maintenance care and raise the risk of > being missed in future changes around this place. > I could use a `goto' or a nested `if' statement instead if that made you feel > better than my original proposal -- please pick your preference. It's actually a style used throughout GDB, but no use fighting over it. Let's go with nested if then. No goto please. > I'd also prefer to keep the handling of args->num_notes consistent across > the two cases -- currently we increment it if elfcore_write_note fails, > so let's keep them as they are or change them both at once. OK... > We could as well dump the struct member altogether as it doesn't appear used beyond its > preinitialisation and the two incrementations seen here. Yeah, I'd prefer getting rid of it. Thanks, -- Pedro Alves