[FFmpeg-devel] [PATCH 02/10] crystalhd: Switch to new decode API and remove the insanity

[Hendrik Leppkes]

On Thu, Oct 27, 2016 at 10:59 PM, Philip Langdale <philipl at overt.org> wrote:
> On 2016-10-27 13:47, Hendrik Leppkes wrote:
>>
>> On Thu, Oct 27, 2016 at 5:09 PM, Philip Langdale <philipl at overt.org>
>> wrote:
>>>
>>> On Thu, 27 Oct 2016 16:01:42 +0200
>>> Timo Rothenpieler <timo at rothenpieler.org> wrote:
>>>
>>>> >
>>>> > breaks fate:
>>>> >
>>>> > ./configure && make -j12 fate-vsynth1-msmpeg4
>>>> > ...
>>>> > TEST    vsynth1-msmpeg4
>>>> > --- ./tests/ref/vsynth/vsynth1-msmpeg4  2016-10-27
>>>> > 03:11:18.675647981 +0200 +++ tests/data/fate/vsynth1-msmpeg4
>>>> > 2016-10-27 15:01:15.397863504 +0200
>>>>
>>>> Does fate seriously test crystalhd hardware?
>>>> I doubt this patch/series is related to that.
>>>
>>>
>>> Yeah, that's failing because it's attempting to load the decoder and
>>> the library spams stdout (which sucks, but what can you do) with the
>>> error messages.
>>>
>>> I've already marked the codec as AV_CODEC_CAP_AVOID_PROBING so I don't
>>> think there's anything else I can do.
>>>
>>
>> If the decoder is being used without requesting it explicitly, then
>> something is wrong somewhere.
>> All these hardware decoders should only ever get used when the user
>> asks for it directly.
>
>
> I suspect the answer is this:
>
>     REGISTER_DECODER(MSMPEG4_CRYSTALHD, msmpeg4_crystalhd);
>     REGISTER_DECODER(MSMPEG4V1,         msmpeg4v1);
>     REGISTER_ENCDEC (MSMPEG4V2,         msmpeg4v2);
>     REGISTER_ENCDEC (MSMPEG4V3,         msmpeg4v3);
>
> It's getting first priority due to declaration order. I imagine that
> moving it last will avoid this test failure.
>
>

Indeed. Hardware codecs should appear after the software.

- Hendrik