From: Jeremy Fitzhardinge <jeremy@goop.org>
Subject: Re: [RFC PATCH 05/33] Makefile support to build Xen subarch
Date: Tue, 18 Jul 2006 13:41:48 -0700
Message-ID: <44BD478C.9040509@goop.org>
References: <20060718091807.467468000@sous-sol.org>	<20060718091949.842251000@sous-sol.org>	<1153216813.3038.22.camel@laptopd505.fenrus.org> <20060718044007.74324d93.akpm@osdl.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Return-path: <linux-kernel-owner+glk-linux-kernel-3=40m.gmane.org-S932397AbWGRWz7@vger.kernel.org>
In-Reply-To: <20060718044007.74324d93.akpm@osdl.org>
Sender: linux-kernel-owner@vger.kernel.org
To: Andrew Morton <akpm@osdl.org>
Cc: Arjan van de Ven <arjan@infradead.org>, chrisw@sous-sol.org, linux-kernel@vger.kernel.org, virtualization@lists.osdl.org, xen-devel@lists.xensource.com, ak@suse.de, rusty@rustcorp.com.au, zach@vmware.com, ian.pratt@xensource.com, Christian.Limpach@cl.cam.ac.uk
List-Id: xen-devel@lists.xenproject.org

Andrew Morton wrote:
> @@ -99,6 +103,7 @@ drivers-$(CONFIG_PM)			+= arch/i386/powe
>  
>  CFLAGS += $(mflags-y)
>  AFLAGS += $(mflags-y)
> +CPPFLAGS += $(mflags-y)
>
> This change affects _all_ subarchitectures (by potentially altering their
> CPPFLAGS) and it's rather a mystery why one subarch needs the -Ifoo in its
> CPPFLAGS whereas all the others do not.
>   

The reason is that arch-i386/kernel/vmlinux.lds.S is run through CPP, 
and it includes asm/thread_info.h and asm/page.h, which end up including 
"mach_page.h" (which this patch series introduces).  There is a version 
in both mach-default/mach_page.h and mach-xen/mach_page.h, so the -I is 
necessary for non-Xen sub-arches as well. 

I guess the conservative approach would be to only add this -I for the 
vmlinux.lds target, assuming there are no later compile problems.  On 
the flip-side, would you want C and Asm code getting a different set of 
includes from "manually" preprocessed-files?  I would think you'd want 
either defines/includes at all, or to have them identical.

The CPPFLAGS assignment also appears to make the previous two lines 
redundant, since a_flags and c_flags (which is what actually gets used 
for compilation) end up having CPPFLAGS incorporated into them.

    J