[FFmpeg-devel] [PATCH] configure: libvmaf requires pthreads

[James Almer]

On 11/13/2020 11:27 AM, lance.lmwang at gmail.com wrote:
> On Fri, Nov 13, 2020 at 01:20:57PM +0100, Timo Rothenpieler wrote:
>> On 13.11.2020 02:23, lance.lmwang at gmail.com wrote:
>>> On Fri, Nov 13, 2020 at 02:14:05AM +0100, Timo Rothenpieler wrote:
>>>> On 13.11.2020 02:05, lance.lmwang at gmail.com wrote:
>>>>> On Thu, Nov 12, 2020 at 05:56:57PM +0100, Timo Rothenpieler wrote:
>>>>>> Technically, libvmaf itself does not, but our filter does, and there is
>>>>>> no other sensible way to prevent a build with --enable-libvmaf from
>>>>>> succeeding while not actually enabling the filter.
>>>>>
>>>>> If it's private filter, I think you it's better to add the pthread depends for your filter
>>>>> only, search for libvmaf_filter_deps
>>>>
>>>> The filter already depends on pthreads correctly.
>>>> This is about the issue where you can produce a build that is configured
>>>> with --enable-libvmaf, but which does not have the libvmaf filter.
>>>>
>>>> Happens for example on Win32, where w32threads are used by default.
>>>
>>> so it doesn't work with  --enable-pthreads also?
>>
>> Of course it does, but there is no indication for that to the user
>> whatsoever.
>> And given that there is only that one consumer of --enable-libvmaf, the
>> libvmaf filter, which also depends on pthreads, adding an error like this
>> seems appropriate to inform users about the pthread requirement.
> 
> But libvmaf library itself does not depend on pthread, so if libvmaf filter
> is used, it'll depend pthread by libvmaf_filter_deps. I'm not sure why the
> deps doesn't working as expected as I can't test win32 system.

The point here is that, if a user configures with --enable-libvmaf but 
doesn't have pthreads, the library will be enabled but the filter will 
not. This results in a libavfilter binary that links to libvmaf for no 
reason, potentially bloating it if it was linked statically.

Since we have no other module using libvmaf, we can simply make libvmaf 
itself dependent on pthreads. This way --enable-libvmaf will fail if 
pthreads is not present, which is better than succeeding but ultimately 
not compiling the filter as the user clearly wanted by passing the above 
option.

> 
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>
>> To unsubscribe, visit link above, or email
>> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>