From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S966228AbXIKGs2 (ORCPT ); Tue, 11 Sep 2007 02:48:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755809AbXIKGsT (ORCPT ); Tue, 11 Sep 2007 02:48:19 -0400 Received: from pasmtpb.tele.dk ([80.160.77.98]:60461 "EHLO pasmtpB.tele.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754119AbXIKGsS (ORCPT ); Tue, 11 Sep 2007 02:48:18 -0400 Date: Tue, 11 Sep 2007 08:49:42 +0200 From: Sam Ravnborg To: Andi Kleen Cc: Thomas Gleixner , LKML , kbuild devel Subject: Re: [RFC] kbuild - introduce vdir to make life easier for x86_64 Message-ID: <20070911064942.GA3663@uranus.ravnborg.org> References: <20070910191138.GA31541@uranus.ravnborg.org> <200709110040.41605.ak@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200709110040.41605.ak@suse.de> User-Agent: Mutt/1.4.2.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Sep 11, 2007 at 12:40:41AM +0200, Andi Kleen wrote: > > > The below are the minimal clean-up - a bit more could be done. > > > > Comments? > > Looks good in principle. My only suggestion would be to name it something > differently than vdir. I know that's what GNU make calls it, but it's still > pretty cryptic. How about just fallback-dir ? fallback-dir is better. vdir came from make's vpath but then very few knows about vpath anyway so that was not the best source of inspiration. > Also what would be nice (I don't know if it's doable in make) would > be a separate variable (e.g. other-obj-... := ) that contains the fallback > names and that is double checked against the fallback directory. My first attempt was along these lines but I dropped it for two reasons: 1) Mixing obj-y and other-obj-y assinment broke all assumptions on linkorder. It would not be possible to link objexts in order a1.o, a2.o, a3.o if a1.o and a3.o were assigned to obj-y and a2.o was assigned to other-obj-y. 2) It needed a decent amouth of unreadable gmake-foo to introduce a second variable. And the learning curve for kbuild internals are alreay to steep so I try to avoid additional complexity. I'm a bit reluctant to just enable the fallback-dir functionality. There is no point in adding this for the sake of x86 if we merge i386 and x86_64 a few days after. Sam