From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 6284 invoked by alias); 8 Jan 2007 05:14:07 -0000 Received: (qmail 6276 invoked by uid 22791); 8 Jan 2007 05:14:06 -0000 X-Spam-Check-By: sourceware.org Received: from wx-out-0506.google.com (HELO wx-out-0506.google.com) (66.249.82.225) by sourceware.org (qpsmtpd/0.31) with ESMTP; Mon, 08 Jan 2007 05:14:00 +0000 Received: by wx-out-0506.google.com with SMTP id h31so7684726wxd for ; Sun, 07 Jan 2007 21:13:58 -0800 (PST) Received: by 10.70.42.16 with SMTP id p16mr47909784wxp.1168233237639; Sun, 07 Jan 2007 21:13:57 -0800 (PST) Received: from ?192.168.0.102? ( [68.121.161.160]) by mx.google.com with ESMTP id 67sm46531194wra.2007.01.07.21.13.35; Sun, 07 Jan 2007 21:13:55 -0800 (PST) Subject: Re: [PATCH, C++] Make Canonical ICE instead of just warn when mismatching To: Mike Stump Cc: Richard Guenther , Doug Gregor , Brooks Moses , gcc-patches@gcc.gnu.org In-Reply-To: References: <459DC221.4030201@codesourcery.com> <24b520d20701050613h10bddaceu567ee42e6b4e111f@mail.gmail.com> <84fc9c000701050620v647343e2n30074cd15d47f02b@mail.gmail.com> Content-Type: text/plain Date: Mon, 08 Jan 2007 05:14:00 -0000 Message-Id: <1168233222.9076.22.camel@celery.andrew.com> Mime-Version: 1.0 X-Mailer: Evolution 2.6.3 (2.6.3-1.fc5.5) Content-Transfer-Encoding: 7bit From: Andrew Pinski Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org X-SW-Source: 2007-01/txt/msg00544.txt.bz2 On Fri, 2007-01-05 at 11:02 -0800, Mike Stump wrote: > If there is a bug in the compiler, the right answer is always > obtained with checking. The wrong answer is obtained without > checking, and there is no warning/error and also no way to know that > the the bug existed. Right which is why this should not be just a warning for checking. This case is that we can recover from but only at expense of compile time speed. If we had it as an ICE before, we would have noticed the problem, the same day the patch went in. So we get only a warning if checking defaults or the user explicitly turns on the checking. So really most people will not see the ICE at all unless someone tells them to turn on the checking. I don't see why this case is special compared to all other checking code inside GCC, some of which we can recover from if we wanted to but we don't currently. Thanks, Andrew Pinski