From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 10963 invoked by alias); 10 Mar 2005 04:24:41 -0000 Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org Received: (qmail 10865 invoked by uid 48); 10 Mar 2005 04:24:32 -0000 Date: Thu, 10 Mar 2005 04:24:00 -0000 Message-ID: <20050310042432.10864.qmail@sourceware.org> From: "amylaar at gcc dot gnu dot org" To: gcc-bugs@gcc.gnu.org In-Reply-To: <20050305081136.20331.kkojima@gcc.gnu.org> References: <20050305081136.20331.kkojima@gcc.gnu.org> Reply-To: gcc-bugzilla@gcc.gnu.org Subject: [Bug rtl-optimization/20331] [3.4/4.0/4.1 Regression] Wrong code generation for the argument of the pure function in PIC X-Bugzilla-Reason: CC X-SW-Source: 2005-03/txt/msg01239.txt.bz2 List-Id: ------- Additional Comments From amylaar at gcc dot gnu dot org 2005-03-10 04:24 ------- (In reply to comment #3) > I'm taking a look at the uses of MEM_READONLY_P, though I'm not > sure which one has the similar issue. It looks that modified_in_p > has the same problem with modified_between_p. Yes, that's also the one I spotted first. But it's simple to find more: grep tells us alias.c has some MEM_READONLY_P. Bingo, the first hit is another instance: true_dependence does: if (MEM_READONLY_P (x)) return 0; without checking if x contains a MEM that is dependent (possible if the target supports memory-indirect addressing, e.g. m68020) -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20331