[llvm-commits] [PATCH] x86_64 detection and building on 10.6

Shantonu Sen ssen at apple.com
Wed Sep 2 09:41:31 PDT 2009


Isn't config.guess upstream from FSF? Do we really want to fork this  
file?

It seems wrong to change this file, since it would cause divergence  
from, e.g, llvm-gcc-4.2, which would still auto-detech i686-apple- 
darwin10.0.0. I agree with the general principle, but can this be done  
in configure.ac?

Shantonu Sen
ssen at apple.com

Sent from my Mac Pro

On Sep 2, 2009, at 9:36 AM, Daniel Dunbar wrote:

> Thanks Evan,
>
> I personally think this functionality should go in. If no one else has
> an opinion, I'm will test/apply at some point (read: object now or
> hold your peace).
>
> - Daniel
>
> On Wed, Sep 2, 2009 at 9:11 AM, Evan Phoenix<evan at fallingsnow.net>  
> wrote:
>>
>> On Aug 31, 2009, at 11:32 AM, Mike Stump wrote:
>>
>>> On Aug 30, 2009, at 12:31 PM, Evan Phoenix wrote:
>>>>
>>>> Currently, it's not possible to build LLVM and link it against  
>>>> something
>>>> compiled by default on 10.6, since LLVM forces itself to build  
>>>> 32bit only,
>>>> and 10.6 defaults to compile 64bit by default.
>>>
>>> Not quite right:
>>>
>>> $ gcc -arch ppc64 -E -dM -x c /dev/null -o - | grep LP64
>>> #define __LP64__ 1
>>> #define _LP64 1
>>>
>>> If you check uname -m and if it says i386, then this is safe to  
>>> do.  I
>>> support the general idea however.
>>
>> Ah! Right you are. Having never had access to a ppc64 machine, I  
>> had no
>> idea. Attached fixes this, only using gcc test if uname -m has said  
>> *86.
>>
>> Thanks!
>>
>>  - Evan
>>
>>
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
>>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




More information about the llvm-commits mailing list