From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S267431AbUG2Ssk (ORCPT ); Thu, 29 Jul 2004 14:48:40 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264954AbUG2SnT (ORCPT ); Thu, 29 Jul 2004 14:43:19 -0400 Received: from fed1rmmtao09.cox.net ([68.230.241.30]:5017 "EHLO fed1rmmtao09.cox.net") by vger.kernel.org with ESMTP id S266635AbUG2Onu (ORCPT ); Thu, 29 Jul 2004 10:43:50 -0400 Date: Thu, 29 Jul 2004 07:43:47 -0700 From: Tom Rini To: Giuliano Pochini Cc: Kumar Gala , Sylvain Munaut , linuxppc-dev@lists.linuxppc.org, linux-kernel@vger.kernel.org, olh@suse.de, Andrew Morton Subject: Re: [PATCH][PPC32] Makefile cleanups and gcc-3.4+binutils-2.14 c Message-ID: <20040729144347.GE16468@smtp.west.cox.net> References: <20040728220733.GA16468@smtp.west.cox.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.6+20040523i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Jul 29, 2004 at 10:05:49AM +0200, Giuliano Pochini wrote: > On 28-Jul-2004 Tom Rini wrote: > > > I've taken the binutils-2.14+gcc-3.4 bit out (and none of the other > > cleanups) as it seems like we get 1-2 reports a week from this bad tools > > combination: > > I had no time to do a lot of testing, but it seems that binutils 2.15 + > gcc 3.3.3 is a bad one too. I didn't try to compile the kernel (which > may also break), but at least I couldn't compile gcc 3.4.1 with the > above combination. It seems that as doesn't get the -mxxx parameter > required to compile altivec stuff. Hacking the Makefile to make it > pass -Wa,-m7455 helped a little, but it eventually failed in another > weird way. I hadn't time to investigate further, sorry. Stock gcc-3.3.3 or from the hammer branch? There is, I think, a second problem that was left out. The problem with gcc-3.4 + binutils-2.14 is that -many gets passed, which zeros out previous flags. -many is fine in binutils-2.15 (and 2.13 and 2.12 and 2.12.1 it seems), but 2.15 does require -maltivec to be passed in order to handle altivec instructions. Getting this right was part of the cleanup that conflicted with the mpc52xx changes (Andrew: trying to take care of getting this into Linus' tree now). -- Tom Rini http://gate.crashing.org/~trini/