From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 1667 invoked by alias); 4 May 2004 07:31:52 -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 1643 invoked from network); 4 May 2004 07:31:50 -0000 Received: from unknown (HELO miranda.se.axis.com) (212.209.10.220) by sources.redhat.com with SMTP; 4 May 2004 07:31:50 -0000 Received: from axis.com (ironmaiden.se.axis.com [10.13.8.120]) by miranda.se.axis.com (8.12.9/8.12.9/Debian-5local0.1) with ESMTP id i447VCUm001749; Tue, 4 May 2004 09:31:12 +0200 Message-ID: <409746BF.7050503@axis.com> Date: Tue, 04 May 2004 07:31:00 -0000 From: Orjan Friberg Organization: Axis Communications User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040113 MIME-Version: 1.0 To: Eli Zaretskii CC: kettenis@chello.nl, gdb-patches@sources.redhat.com, drow@false.org Subject: Re: Display of read/access watchpoints when HAVE_NONSTEPPABLE_WATCHPOINT References: <407282F4.2080602@axis.com> <20040406142228.GA29473@nevyn.them.org> <6654-Thu15Apr2004111217+0300-eliz@gnu.org> <407E8CEF.2050007@axis.com> <407FC69A.1000701@axis.com> <1438-Sat17Apr2004112204+0300-eliz@gnu.org> <4083E930.8040005@axis.com> <4087DFB6.1030801@axis.com> <200405012117.i41LHZSR001291@elgar.kettenis.dyndns.org> <40962BF0.3030707@axis.com> <6654-Mon03May2004204832+0300-eliz@gnu.org> In-Reply-To: <6654-Mon03May2004204832+0300-eliz@gnu.org> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-SW-Source: 2004-05/txt/msg00089.txt.bz2 Eli Zaretskii wrote: > > So this target could relatively easily support an API where the target > itself tells GDB which of watchpoints triggered and why, provided that > we invent a way for GDB to map between its watchpoint numbers and the > target-size hardware data watchpoints. Is that true? Yes (provided the register contents haven't changed since the watchpoint hit, of course - I assume the mapping procedure you're describing should be done before single-stepping over the watchpoint for targets with nonsteppable watchpoints). Another thing, which may or may not be relevant in this discussion, is if the remote protocol should be altered. Right now, the watch/rwatch/awatch part of the stop reply packet when hitting a watchpoint is not used (except for parsing out the stopped data address). In addition, it doesn't play nice with multiple watchpoint hits, since only one of watch/rwatch/awatch is possible. I don't have any practical suggestion on how to make it more useful though :( . >>(In addition there is also one hardware instruction breakpoint, but I >>guess that doesn't matter in this context.) > > > Used for hardware-assisted breakpoints, yes? (Ugh, I always botch the terminology.) Yes, invoked by the "hbreak" command. -- Orjan Friberg Axis Communications