From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 22443 invoked by alias); 21 Jan 2006 21:02:54 -0000 Received: (qmail 22424 invoked by uid 48); 21 Jan 2006 21:02:52 -0000 Date: Sat, 21 Jan 2006 21:02:00 -0000 Message-ID: <20060121210252.22423.qmail@sourceware.org> X-Bugzilla-Reason: CC References: Subject: [Bug c++/23628] Typeinfo comparison code easily breaks shared libs In-Reply-To: Reply-To: gcc-bugzilla@gcc.gnu.org To: gcc-bugs@gcc.gnu.org From: "rjohnson at dogstar-interactive dot com" 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 X-SW-Source: 2006-01/txt/msg02207.txt.bz2 List-Id: ------- Comment #26 from rjohnson at dogstar-interactive dot com 2006-01-21 21:02 ------- I just got bit by this using gcc version 4.0.3 20051201 (prerelease) (Debian 4.0.2-5) (powerpc) It's just my $.02 from the gallery, but that address comparison in the type_info::oprator==() implementation looks suspect. Stroustrup (3rd edition, $15.4.4) specifically says: It is _not_ guaranteed that there is only one type_info object for each type in the system. and then he goes on to call out dynamically linked libraries as a particular case. If folks are interested, I can propose a relatively inexpensive (i.e. non-strcmp) runtime strawman for consideration. -- rjohnson at dogstar-interactive dot com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |rjohnson at dogstar- | |interactive dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23628