From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S265668AbUHHPyg (ORCPT ); Sun, 8 Aug 2004 11:54:36 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S265684AbUHHPyg (ORCPT ); Sun, 8 Aug 2004 11:54:36 -0400 Received: from caramon.arm.linux.org.uk ([212.18.232.186]:57092 "EHLO caramon.arm.linux.org.uk") by vger.kernel.org with ESMTP id S265668AbUHHPye (ORCPT ); Sun, 8 Aug 2004 11:54:34 -0400 Date: Sun, 8 Aug 2004 16:54:29 +0100 From: Russell King To: Sam Ravnborg , Linux Kernel List Subject: Re: [PATCH] Fix kallsyms dependency Message-ID: <20040808165429.A17968@flint.arm.linux.org.uk> Mail-Followup-To: Sam Ravnborg , Linux Kernel List References: <20040808123230.B7589@flint.arm.linux.org.uk> <20040808154610.GA18740@mars.ravnborg.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: <20040808154610.GA18740@mars.ravnborg.org>; from sam@ravnborg.org on Sun, Aug 08, 2004 at 05:46:10PM +0200 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sun, Aug 08, 2004 at 05:46:10PM +0200, Sam Ravnborg wrote: > On Sun, Aug 08, 2004 at 12:32:30PM +0100, Russell King wrote: > > Hi, > > > > It appears that kallsyms data is not updated if the kallsyms program is > > changed. The following patch adds an appropriate dependency. > > Added, and implemented your suggestion in scripts/Makefile > with a few other CONFIG selections. > > Btw. any specific reason to hack kallsyms? Just curious if something > shows up in this area. Yes - later ARM binutils adds extra symbols like "$a" and "$d" to the symbol table as "mapping symbols" (binutils terminology). It's not clear whether these will continue to be visible via normal tools or not, and I was considering getting kallsyms to filter them out. Depending on the outcome of the binutils side depends whether or not I end up modifying kallsyms to ignore these symbols and whether we end up saying "binutils earlier than can not be used for ARM." If a release of binutils is imminent which solves both the mapping symbol visibility problem and the undefined symbol issue, then I suspect its just all round easier to prevent the ARM kernel being built with older binutils versions. -- Russell King Linux kernel 2.6 ARM Linux - http://www.arm.linux.org.uk/ maintainer of: 2.6 PCMCIA - http://pcmcia.arm.linux.org.uk/ 2.6 Serial core