From: Yegor Yefremov on
On Wed, Jun 23, 2010 at 5:44 AM, Tiago Maluta <tiago.maluta(a)gmail.com> wrote:
> On Mon, Jun 21, 2010 at 4:54 PM, Michal Marek <mmarek(a)suse.cz> wrote:
>> On 21.6.2010 21:06, Andrew Morton wrote:
>>> On Thu, 17 Jun 2010 12:50:52 +0000
>>> Tiago Maluta <tiago.maluta(a)gmail.com> wrote:
>>>
>>>> Linux has an recent option to define CROSS_COMPILE in menuconfig.
>>>>
>>>> Typically, one of the options when building for another target is
>>>> ('arm' is just an example):
>>>>
>>>> # make ARCH=arm CROSS_COMPILE=arm-cc O=../build
>>>>
>>>> If I omit CROSS_COMPILE and use the new way in .config:
>>>> CONFIG_CROSS_COMPILE="arm-cc"
>>>>
>>>> I got:
>>>>
>>>>    cc1: error: unrecognized command line option "-mlittle-endian"
>>>>    cc1: error: unrecognized command line option "-mabi=aapcs-linux"
>>>>    cc1: error: unrecognized command line option "-mno-thumb-interwork"
>>>>    /usr/src/linux/kernel/bounds.c:1: error: bad value (armv5t) for
>>>> -march= switch
>>>>    /usr/src/linux/kernel/bounds.c:1: error: bad value (armv5t) for
>>>> -mtune= switch
>>>>    make[2]: *** [kernel/bounds.s] Error 1
>>>>    make[1]: *** [prepare0] Error 2
>>>>    make: *** [sub-make] Error 2
>>>>
>>>> Verbose output clearly says that I'm using the host compiler instead
>>>> cross compiler.
>>>> This patch fixed my problem:
>>>>
>>>> ---
>>>> __Makefile |______ 2 +-
>>>> __1 files changed, 1 insertions(+), 1 deletions(-)
>>>>
>>>> diff --git a/Makefile b/Makefile
>>>> index d49d96c..5a6c7d5 100644
>>>> --- a/Makefile
>>>> +++ b/Makefile
>>>> @@ -190,7 +190,7 @@ SUBARCH := $(shell uname -m | sed -e s/i.86/i386/
>>>> -e s/sun4u/sparc64/ \
>>>> __export KBUILD_BUILDHOST := $(SUBARCH)
>>>> __ARCH____________________ ?= $(SUBARCH)
>>>> __CROSS_COMPILE__ ?=
>>>> -CROSS_COMPILE__ ?= $(CONFIG_CROSS_COMPILE:"%"=%)
>>>> +CROSS_COMPILE__ := $(shell grep CONFIG_CROSS_COMPILE .config | cut -d'"' -f2)
>>>>
>>>> __# Architecture as present in compile.h
>>>> __UTS_MACHINE______ := $(ARCH)
>>>> --
>>>
>>> Perhaps the "%"=% thing isn't supported with your version of make(1).
>>> Which version are you using?
>>
>> My guess is that something expands $(CROSS_COMPILE) before make
>> silentoldconfig generates include/config/auto.conf (which is where the
>> makefile reads the CONFIG_* variables from). Tiago, can you try changing
>> the $(shell ...) line to
>>
>> CROSS_COMPILE := $(shell ls include/config/auto.conf >&2; \
>>    echo "CONFIG_CROSS_COMPILE $(CONFIG_CROSS_COMPILE) >&2; \
>>    grep CONFIG_CROSS_COMPILE .config | cut -d'"'  -f2)
>>
>> and tell us what it prints?
>
> I'm using GNU Make 3.81.
>
> I think you missed a quote in echo..., I using:
>
> CROSS_COMPILE := $(shell ls include/config/auto.conf >&2; \
>    echo "CONFIG_CROSS_COMPILE" $(CONFIG_CROSS_COMPILE) >&2; \
>    grep CONFIG_CROSS_COMPILE .config | cut -d'"'  -f2)
>
> And the output:
> $ make ARCH=arm O=../build
>
> include/config/auto.conf
> CONFIG_CROSS_COMPILE
>  Using /usr/src/linux as source for kernel
>  GEN     /usr/src/build/Makefile
>  CHK     include/linux/version.h
>  CHK     include/generated/utsrelease.h
> make[2]: `include/generated/mach-types.h' is up to date.
>  CALL    /usr/src/linux/scripts/checksyscalls.sh
>  CHK     include/generated/compile.h
>  CC      arch/arm/mm/init.o
>
> ^C
>
> Means that is nothing CONFIG_CROSS_COMPILE.
>
> I agree with Andrew, directly grepping .config isn't the right thing to do.
> I'd like to discuss a better solution.

What about such a solution:

just remove empty CROSS_COMPILE ?=

Regards,
Yegor

Fix CONFIG_CROSS_COMPILE issue in .config

Signed-off-by: Yegor Yefremov <yegorslists(a)googlemail.com>

Index: b/Makefile
===================================================================
--- a/Makefile 2010-08-02 00:11:14.000000000 +0200
+++ b/Makefile 2010-08-04 14:08:51.000000000 +0200
@@ -189,7 +189,6 @@
# Note: Some architectures assign CROSS_COMPILE in their arch/*/Makefile
export KBUILD_BUILDHOST := $(SUBARCH)
ARCH ?= $(SUBARCH)
-CROSS_COMPILE ?=
CROSS_COMPILE ?= $(CONFIG_CROSS_COMPILE:"%"=%)

# Architecture as present in compile.h
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Roland Dreier on
> What about such a solution:
>
> just remove empty CROSS_COMPILE ?=
>
> Fix CONFIG_CROSS_COMPILE issue in .config
>
> Signed-off-by: Yegor Yefremov <yegorslists(a)googlemail.com>
>
> Index: b/Makefile
> ===================================================================
> --- a/Makefile 2010-08-02 00:11:14.000000000 +0200
> +++ b/Makefile 2010-08-04 14:08:51.000000000 +0200
> @@ -189,7 +189,6 @@
> # Note: Some architectures assign CROSS_COMPILE in their arch/*/Makefile
> export KBUILD_BUILDHOST := $(SUBARCH)
> ARCH ?= $(SUBARCH)
> -CROSS_COMPILE ?=
> CROSS_COMPILE ?= $(CONFIG_CROSS_COMPILE:"%"=%)

Yes, this make sense to me. It's not clear to me why we ever needed
the conditional assignment of an empty CROSS_COMPILE (that code predates
the start of git history) but clearly having two "?=" assignments one
after another cannot work -- the second assignment to CROSS_COMPILE will
never do anything, since the line before makes CROSS_COMPILE defined.

- R.
--
Roland Dreier <rolandd(a)cisco.com> || For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/index.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Roland McGrath on
> Yes, this make sense to me. It's not clear to me why we ever needed
> the conditional assignment of an empty CROSS_COMPILE (that code predates
> the start of git history) [...]

Perhaps someone used make --warn-undefined-variables.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Yegor Yefremov on
On Fri, Aug 6, 2010 at 12:26 AM, Roland McGrath <roland(a)redhat.com> wrote:
>> Yes, this make sense to me.  It's not clear to me why we ever needed
>> the conditional assignment of an empty CROSS_COMPILE (that code predates
>> the start of git history) [...]
>
> Perhaps someone used make --warn-undefined-variables.

Are there any other comments or is the patch valid? Should I resend it
as a new thread?

Regards,
Yegor
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Roland McGrath on
> On Fri, Aug 6, 2010 at 12:26 AM, Roland McGrath <roland(a)redhat.com> wrote:
> >> Yes, this make sense to me. �It's not clear to me why we ever needed
> >> the conditional assignment of an empty CROSS_COMPILE (that code predates
> >> the start of git history) [...]
> >
> > Perhaps someone used make --warn-undefined-variables.
>
> Are there any other comments or is the patch valid? Should I resend it
> as a new thread?

The patch is fine. The comments were just about how it might have gotten
to be how it was before now.

Thanks,
Roland
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/