From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 22834 invoked by alias); 26 May 2005 03:14:51 -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 22735 invoked by uid 22791); 26 May 2005 03:14:49 -0000 Received: from nevyn.them.org (HELO nevyn.them.org) (66.93.172.17) by sourceware.org (qpsmtpd/0.30-dev) with ESMTP; Thu, 26 May 2005 03:14:49 +0000 Received: from drow by nevyn.them.org with local (Exim 4.50) id 1Db8pX-0006hl-Vo; Wed, 25 May 2005 23:14:44 -0400 Date: Thu, 26 May 2005 03:41:00 -0000 From: Daniel Jacobowitz To: Richard Henderson Cc: Joel Brobecker , gdb-patches@gcc.gnu.org Subject: Re: [testsuite/alpha] Add test for step over fbne instruction Message-ID: <20050526031443.GA25712@nevyn.them.org> Mail-Followup-To: Richard Henderson , Joel Brobecker , gdb-patches@gcc.gnu.org References: <20050525064819.GA5406@redhat.com> <20050526023539.GL1525@adacore.com> <20050526031128.GA9264@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050526031128.GA9264@redhat.com> User-Agent: Mutt/1.5.8i X-SW-Source: 2005-05/txt/msg00550.txt.bz2 On Wed, May 25, 2005 at 08:11:28PM -0700, Richard Henderson wrote: > On Thu, May 26, 2005 at 12:35:39PM +1000, Joel Brobecker wrote: > > " cpys $f31,$f31,$f0\n" > > " fbgt $f16,$gt_1\n" /* stop at this instruction. */ > > " cpysn $f16,$f16,$f0\n" > > "$gt_1:\n" > > " ret $31,($26),1\n" > > Hmm, now that I think about it, one potential problem with this form > is that we're not correctly testing for branch taken. The scenerio > is that we expect taken, set the bp on the ret, the branch *isn't* > taken, but we fall through to the ret anyway. > > We'd do better to write this > > " fbgt $f16,$gt_1\n" /* stop at this instruction. */ > " cpysn $f16,$f16,$f0\n" > " ret $31,($26),1\n" > "$gt_1:\n" > " ret $31,($26),1\n" > > so that if we choose wrong, we will in fact lose control. Then the testsuite is likely to get out of sync - either way, we need to do a little dance to recover. Joel, how about checking to see that the instruction we've stopped on is the correct one - how much would that complicate the test? -- Daniel Jacobowitz CodeSourcery, LLC