From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 2915 invoked by alias); 2 Sep 2005 22:27:58 -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 2771 invoked by uid 48); 2 Sep 2005 22:27:38 -0000 Date: Fri, 02 Sep 2005 22:27:00 -0000 Message-ID: <20050902222738.2770.qmail@sourceware.org> From: "rearnsha at gcc dot gnu dot org" To: gcc-bugs@gcc.gnu.org In-Reply-To: <20040430162847.15231.pb@gcc.gnu.org> References: <20040430162847.15231.pb@gcc.gnu.org> Reply-To: gcc-bugzilla@gcc.gnu.org Subject: [Bug target/15231] [3.4 only] constant pool entries referring to nonexistent labels X-Bugzilla-Reason: CC X-SW-Source: 2005-09/txt/msg00301.txt.bz2 List-Id: ------- Additional Comments From rearnsha at gcc dot gnu dot org 2005-09-02 22:27 ------- I've made a small amount of headway on this. Labels L22 and L21 were (when created) the addresses of objects in the code. However, they are deleted (presumably as unreachable), but the references to the lables persist. This is presumably exactly the case that RTH believed couldn't happen when he mentioned this in his email here: http://gcc.gnu.org/ml/gcc-patches/2000-05/msg00881.html and the reason we don't abort on ARM is that we never call output_constant_pool since we roll our own mini-pools. I've sofar been unsuccessful in reducing the testcase to something more managable. -- What |Removed |Added ---------------------------------------------------------------------------- Last reconfirmed|2005-02-22 04:10:35 |2005-09-02 22:27:33 date| | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15231