From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 24575 invoked by alias); 5 Dec 2003 01:04:21 -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 24544 invoked from network); 5 Dec 2003 01:04:20 -0000 Received: from unknown (HELO localhost.redhat.com) (207.219.125.105) by sources.redhat.com with SMTP; 5 Dec 2003 01:04:20 -0000 Received: from gnu.org (localhost [127.0.0.1]) by localhost.redhat.com (Postfix) with ESMTP id B27832B8F; Thu, 4 Dec 2003 20:04:20 -0500 (EST) Message-ID: <3FCFD994.6010409@gnu.org> Date: Fri, 05 Dec 2003 01:04:00 -0000 From: Andrew Cagney User-Agent: Mozilla/5.0 (X11; U; NetBSD macppc; en-US; rv:1.0.2) Gecko/20030820 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Richard Henderson Cc: Ulrich Weigand , gdb-patches@sources.redhat.com, uweigand@de.ibm.com Subject: Re: [PATCH] S/390 DWARF-2 CFI frame support References: <200312042009.VAA07733@faui1d.informatik.uni-erlangen.de> <20031205004756.GA18170@redhat.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-SW-Source: 2003-12/txt/msg00166.txt.bz2 > On Thu, Dec 04, 2003 at 09:09:12PM +0100, Ulrich Weigand wrote: > >> To fix this, I suggest the following. What GCC assumes to happen >> when it leaves a register unspecified in the CFI depends on whether >> the register is call-saved or call-clobbered according to the >> target's ABI. If it is call-saved (and unspecified), the function >> doesn't save/restore it because it does not in fact ever modify it. >> Thus, in this case the debugger should copy the value from the >> inner frame. If it is call-clobbered (those will always be left >> unspecified), it should be assumed undefined. > > > This is wrong. The debugger should just assume *all* registers > that are not explicitly saved are preserved. In the case of > call-clobbered registers, you just won't *know* that they are > actually dead. But so what? This is no worse than not having > location list information that tells you that a value is dead > after its register gets re-used for something else. > > However, when I made this argument before, it wasn't good enough > for some people, and they added the annoying warning anyway. From what I've seen of the PPC64, Ulrich's change and even more is likely needed. PPC64, for instance, appears to point the return-address column and FPSCR and then forget to specify that it is initally found in LR. Andrew