Re: libgmp: duplicate symbol

[christos%astron.com@localhost (Christos Zoulas)]

In article <50CC3EAA.9050307%gmx.de@localhost>,
Christoph Egger  <Christoph_Egger%gmx.de@localhost> wrote:
>On 14.12.12 16:41, Christoph Egger wrote:
>>>On 14.12.12 14:46, David Laight wrote:
>
>>>
>>>> On Fri, Dec 14, 2012 at 11:11:16AM +0100, Christoph Egger wrote:
>>>>>> On Thu, Dec 13, 2012 at 09:42:29PM +0100, Christoph Egger wrote:
>>>>>>> Do you think upstream would accept this?
>>>>>>
>>>>>> Ask them?
>>>> ...
>>>>>   #ifdef __GNUC__
>>>>> -#if (defined __GNUC_STDC_INLINE__) || (__GNUC__ == 4&&  __GNUC_MINOR__ 
>>>>> == 
>>>>> 2)
>>>>> +#if (defined __GNUC_STDC_INLINE__) || \
>>>>> +    (__GNUC__ == 4 &&  __GNUC_MINOR__>= 2) || (__GNUC__>  4)
>>>>
>>>> Isn't there a define that is (__GNUC__ * 1000 + __GNUC_MINOR__) ?
>>>> or similar - for those sorts of checks.
>> 
>> __GNUC_PREREQ__(M, m)
>
>
>This is not a compiler-define. We define it ourself in sys/sys/cdefs.h.
>Using __GNUC_PREREQ__ and building a toolchain on OSX fails in libgmp's
>configure:
>
>checking size of mp_limb_t... 0
>configure: error: Oops, mp_limb_t doesn't seem to work
>
>But using compiler-define's above only work.

You need compiler-defined-define not just a compiler-define.

christos