From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 14107 invoked by alias); 28 Apr 2006 21:42:31 -0000 Received: (qmail 14099 invoked by uid 22791); 28 Apr 2006 21:42:31 -0000 X-Spam-Check-By: sourceware.org Received: from xproxy.gmail.com (HELO xproxy.gmail.com) (66.249.82.199) by sourceware.org (qpsmtpd/0.31) with ESMTP; Fri, 28 Apr 2006 21:42:29 +0000 Received: by xproxy.gmail.com with SMTP id h29so1401516wxd for ; Fri, 28 Apr 2006 14:42:27 -0700 (PDT) Received: by 10.70.78.17 with SMTP id a17mr219513wxb; Fri, 28 Apr 2006 14:42:24 -0700 (PDT) Received: by 10.70.129.13 with HTTP; Fri, 28 Apr 2006 14:42:24 -0700 (PDT) Message-ID: <8f2776cb0604281442m2f514240i9f86b4bfa38b1c33@mail.gmail.com> Date: Fri, 28 Apr 2006 21:42:00 -0000 From: "Jim Blandy" To: "Mark Kettenis" , jimb@red-bean.com, brobecker@adacore.com, gdb-patches@sources.redhat.com Subject: Re: [RFC/RFA/i386] pb reading insns if breakpoints still inserted In-Reply-To: <20060428211258.GA6713@nevyn.them.org> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <20060428171154.GP17613@adacore.com> <8f2776cb0604281054y116acfdavc3649dd8198d80d0@mail.gmail.com> <200604281839.k3SIdfsq030892@elgar.sibelius.xs4all.nl> <8f2776cb0604281358x2f667d00s90e03051f034b91c@mail.gmail.com> <200604282109.k3SL9Jwp020317@elgar.sibelius.xs4all.nl> <20060428211258.GA6713@nevyn.them.org> X-Google-Sender-Auth: 2cf88bc5f1ba742f X-IsSubscribed: yes Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org X-SW-Source: 2006-04/txt/msg00376.txt.bz2 On 4/28/06, Daniel Jacobowitz wrote: > On Fri, Apr 28, 2006 at 11:09:19PM +0200, Mark Kettenis wrote: > > But if we don't have debug information, what's the point in trying to > > skip the prologue in order to put a breakpoint on ... eh what exactly? > > Isn't it better to just punt prologue skipping in that case and place > > the breakpoint on the first instruction of the code? > > Yes, I agree. So you guys would like to push the deprecation boundary further the other way, and deprecate SKIP_PROLOGUE itself? That's an idea. > I'd like to note, though, that the way we skip prologues based on debug > information is completely wrong. It only works with GCC and with other > compilers that have chosen to be bug-compatible with GCC. DWARF does > support a "this is the end of the prologue" flag, but I don't know if > GCC emits it, and I'm pretty positive GDB doesn't know how to consume > it. I think the argument was that, in the long run, GCC should be emitting location lists so that the debugging information accurately describes the location of the arguments even within the prologues. Setting a breakpoint on a function would actually set the breakpoint ... at the function's entry point. We could delete SKIP_PROLOGUE altogether.=20 Then, there'd be no point in producing or consuming an "end of prologue" marker whose meaning is kind of vague anyway.