[MPlayer-dev-eng] [PATCH] build: fix lavfi dependencies (swscale and swresample).

[Reinhard Tartler]

2012/2/7 Nicolas George <nicolas.george at normalesup.org>:
> L'octidi 18 pluviôse, an CCXX, Reinhard Tartler a écrit :
>> I can confirm that this patch fixes the compilation failure with
>> ./configure --enable-vf-lavfi and FFmpeg.
>>
>> Moreover this patch doesn't break compilation against Libav as long as
>> --enable-vf-liavfi is not passed.
>
> Thanks for your testing. I just pushed it.
>
>>                                   With this flag, I get the following
>> errors:
> <snip>
>> libmpcodecs/vf_lavfi.c:67:15: error: 'AVFilterBufferRefVideoProps' has
>> no member named 'sample_aspect_ratio'
> <snip>
>
> As Clément explained, these error are caused by a real API difference
> between lavfi in both branches. It may be possible to hack vf_lavfi.c to
> make it work with libav's API, but it would probably cost lots of work and
> ugly ifdefry.
>
> Another solution would be to get libav to import the enhancements made to
> lavfi. Unfortunately, the way I have seen work there, I expect someone would
> insist on some minor name change or something, and it would make the API
> definitely incompatible.

Agreed. fixing this was definitly out-of-scope of the proposed patch
and as you say, requires work and ideally coordination on the libav
side to incorporate libavfi improvements.

>> libmpcodecs/vf_lavfi.c:272:13: error: implicit declaration of function
>> 'av_rescale' [-Werror=implicit-function-declaration]
>
> This one seems to be caused by this:
> http://thread.gmane.org/gmane.comp.video.libav.devel/4898/focus=4900
>
> It would be easy to fix, but there is not much point anyway (and personally
> I really do not see the point of this patch series at all).

I see. Well, adding the #include "avutil/mathematics.h" is anyway the
right thing to do, imo. You should always include headers that you
require and don't rely on headers that do that indirectly in general
no?

-- 
regards,
    Reinhard