From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 9789 invoked by alias); 6 Aug 2002 07:08:04 -0000 Mailing-List: contact libc-hacker-help@sources.redhat.com; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-hacker-owner@sources.redhat.com Received: (qmail 9773 invoked from network); 6 Aug 2002 07:08:04 -0000 Received: from unknown (HELO perdition.linnaean.org) (66.31.43.253) by sources.redhat.com with SMTP; 6 Aug 2002 07:08:04 -0000 Received: by perdition.linnaean.org (Postfix, from userid 5281) id 54E051BA1B; Tue, 6 Aug 2002 03:08:03 -0400 (EDT) From: Roland McGrath MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit To: Ulrich Drepper Cc: Jakub Jelinek , Glibc hackers Subject: Re: [PATCH] Nuke another 21 .plt slots In-Reply-To: Ulrich Drepper's message of Mon, 5 August 2002 23:39:35 -0700 <3D4F6F27.1060405@redhat.com> Emacs: because Hell was full. Message-Id: <20020806070803.54E051BA1B@perdition.linnaean.org> Date: Tue, 06 Aug 2002 00:08:00 -0000 X-SW-Source: 2002-08/txt/msg00085.txt.bz2 > The problem is not loss of optimization. The semantics f code gets > changed. And incorrectly so. gcc cannot inline a non-static function > unless its visibility != default. It might be possible to add a gcc > option to signal that all functions can be assumed to be non-preemptible. I don't think it is so unreasonable. Anything that is written so that it would work with static linking would work will come out right (i.e. you can only inline in the same module and it would be a multiple definition if there were a competing definition elsewhere), except for special cases defining weak functions.