From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 1802 invoked by alias); 3 Dec 2008 15:34:56 -0000 Received: (qmail 1600 invoked by uid 48); 3 Dec 2008 15:33:27 -0000 Date: Wed, 03 Dec 2008 15:34:00 -0000 Message-ID: <20081203153327.1599.qmail@sourceware.org> X-Bugzilla-Reason: CC References: Subject: [Bug middle-end/36143] [4.4 Regression]: FAIL: g++.dg/tree-ssa/pr19637.C In-Reply-To: Reply-To: gcc-bugzilla@gcc.gnu.org To: gcc-bugs@gcc.gnu.org From: "rguenth at gcc dot gnu dot org" Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org X-SW-Source: 2008-12/txt/msg00289.txt.bz2 ------- Comment #13 from rguenth at gcc dot gnu dot org 2008-12-03 15:33 ------- We should be able to go via a VIEW_CONVERT_EXPR when propagating (struct Foo *) &i into the LHS dereference D.2279_5->i[0]. That is, convert that to VIEW_CONVERT_EXPR (i)->i[0] and further fold that by noting that the final access is at offset zero and of the same type as i. (or more generally, for a final offset zero access always strip all component-refs and fold to VIEW_CONVERT_EXPR (i)). -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36143