[FFmpeg-devel] Westwood VQA3 demuxer additions

[Lynne]

Sep 20, 2021, 20:43 by "Pekka Väänänen" :

> Greetings!
>
> This small patch is required for high color 15-bit Westwood VQA3 videos
> to be decoded properly. It just saves an offset of a VQFL chunk when it
> sees one and then concatenates it with the packet of a later VQFR chunk
> that contains full frame data. This way the (yet-to-be-merged) VQA3
> decoder can apply the updated codebook before rendering a new frame and
> each packet contains a standalone frame, as expected.
>
> This change was already added by Adam Iglewski as part of GSoC 2009:
>
> https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2009-May/075047.html
>
> However that patch had a memory leak with the dynamically allocated
> memory:
>
> https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2009-May/075154.html
>
> Here that problem is avoided because no heap allocation is done.
>
> One question though: is it a problem that I use avio_seek() to rewind
> the stream?
>

It's fine, but you do want to use ffio_ensure_seekback() to guarantee
success, preferably calling it at just the moment the stream is at the
position you're planning on rewinding back to.