Re: [PATCH 2/8] um: Do not use SUBARCH

[Richard Weinberger <richard@nod.at>]

Am 26.09.2013 15:26, schrieb Geert Uytterhoeven:
> On Thu, Sep 26, 2013 at 3:13 PM, Ramkumar Ramachandra
> <artagnon@gmail.com> wrote:
>> Ramkumar Ramachandra wrote:
>>> Richard Weinberger wrote:
>>>> I told you already that "make defconfig ARCH=um SUBARCH=x86" will spuriously
>>>> create a x86_64 config on x86_64.
>>>> This breaks existing setups.
>>>
>>> I'll fix this and resubmit soon.
>>
>> Wait a minute. You're now arguing about whether the generic "x86"
>> means i386 or x86_64. Its meaning is already defined in
>> arch/x86/Kconfig and arch/x86/um/Kconfig: see the config 64BIT. Unless
>> i386 is explicitly specified, the default is to build a 64-bit kernel.
>> That is already defined for a normal Linux kernel, and user-mode Linux
>> should not break that convention. So, in the example you pulled out of
>> your hat:
>>
>>   $ make defconfig ARCH=um SUBARCH=x86
>>
>> the user should expect a 64-bit build, and not an i386 build as you
>> say. Both my patches are correct, and the "regression" that you
>> pointed out is a red herring.
> 
> Sorry for chiming in, but... what about cross compiling?
> SUBARCH=x86 should give you a 32-bit ia32 kernel, right?

Correct.
Users expect from SUBARCH=x86 a i386 32bit UML kernel.

Thanks,
//richard