[FFmpeg-devel] [libav-devel] [PATCHv3] avcodec: Cineform HD Decoder

[Vittorio Giovara]

On Thu, Jan 28, 2016 at 1:15 PM, Kieran Kunhya <kieran618 at googlemail.com> wrote:
> On Thu, 28 Jan 2016 at 16:26 Vittorio Giovara <vittorio.giovara at gmail.com>
> wrote:
>
>> On Mon, Jan 25, 2016 at 6:15 PM, Kieran Kunhya <kieran618 at googlemail.com>
>> wrote:
>> >>> +{
>> >>> +    CFHDContext *s = avctx->priv_data;
>> >>> +
>> >>> +    avctx->pix_fmt             = AV_PIX_FMT_YUV422P10;
>> >>
>> >> if the decoder supports multiple pixel formats it's better not to
>> >> initialize the pixel format here, and wait until decode(). Otherwise
>> >> it's going to cause a "parameters changed" warning and reinit any
>> >> previous filter chain.
>> >
>> > There are some samples which don't have a pixel format flagged and are
>> > implicitly AV_PIX_FMT_YUV422P10.
>>
>> I still think this should be put off until you get to decoding a
>> frame: having a filterchain reinitialization can be disastrous under
>> certain conditions.
>>
>>
> Not my problem to handle issues with applications.

sure, but to give another example if you ffprobe/avprobe a chfd sample
it will always report yuv422p10, while sometimes it's not. In cases
like this one I think it's better to have it decode one frame to get
the correct pixel format (until some proper parser apis replace this
way).

>> >>> +
>> >>> +end:
>> >>> +    if (ret < 0)
>> >>> +        return ret;
>> >>> +
>> >>> +    *got_frame = 1;
>> >>> +    return avpkt->size;
>> >>
>> >> avpkt->size - bytestream2_get_bytes_left(&gb) no?
>> >
>> > Guaranteed to read all tags.
>>
>> if all bytes have been read return 0 then
>>
>>
> eh?

sorry, never mind, I misread parts of the doc
-- 
Vittorio