[FFmpeg-devel] compiling ffmpeg with cygwin

[Luca Barbato]

On 03/10/2011 08:10 PM, V?ctor Paesa wrote:
> Hi
> 
> 2011/3/6 V?ctor Paesa:
>> Hi,
>>
>> On Fri, Mar 4, 2011 at 00:42, M?ns Rullg?rd  wrote:
>>> V?ctor Paesa  writes:
>>>
>>>> Hi,
>>>>
>>>> 2011/3/2 V?ctor Paesa :
>>>>> Hi,
>>>>>
>>>>> On Wed, Mar 2, 2011 at 15:36, Luca Barbato wrote:
>>>>>> On 03/02/2011 02:40 PM, alexander ehlert wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> I tried to compile the git trunk with cygwin 1.7.8. It failed
>>>>>>> compiling libavcodec/aaccoder.c with
>>>>>>> "float_t undeclared type".
>>>>>>>
>>>>>>> Adding "#include <math.h>" fixed this problem.
>>>>>>
>>>>>> Could you please try to track which header inject a float_t redefinition
>>>>>> and report cygwin the issue?
>>>>>>
>>>>>> inspecting the output of the compile line with -E would help.
>>>>>
>>>>> It seems a typo in <math.h>
>>>>> I reported it to Cygwin maillist.
>>>>
>>>> Besides the typo in Cygwin's header, I think we should include <math.h>
>>>> in libavcodec/aaccoder.c and libavcodec/aacsbr.c, because they use log2f()
>>>> (same as the configure "check_matfunc" includes <math.h>)
>>>
>>> If they use any of the math functions we check for, they should include
>>> libavutil/libm.h, possibly math.h as well for clarity.
>>
>> A patch including libavutil/libm.h
>>
>> Fair warning: you'll need to upgrade from 1.7.8 to the latest Cygwin
>> developer snapshot, to avoid
>> a __xpg_strerror_r() bug.
> 
> A user reported the patch solved his problem:
> http://lists.mplayerhq.hu/pipermail/ffmpeg-user/2011-March/029301.html
> 
> Any objections to apply it?

The patch seems fine, if you could send as git-format-patch would be
great ^^;

lu

-- 

Luca Barbato
Gentoo/linux
http://dev.gentoo.org/~lu_zero