[FFmpeg-devel] [PATCH] avformat/webvttdec, enc: correctly process files containing STYLE, REGION blocks

[Dave Evans]

On Mon, Oct 12, 2020 at 11:20 PM Jan Ekström <jeebjp at gmail.com> wrote:
>
> On Tue, Oct 13, 2020 at 12:07 AM Dave Evans <dave.evans at m2amedia.tv> wrote:
> >
> > This patch fixes the total failure to parse cues when style and region
> > definition blocks are contained in the input file, and ensures those blocks
> > are written to the output when copying.
> >
>
> Thank you for taking time to add a FATE test, but unfortunately I am
> not sure if this test tests the functionality you have added.
>
> You have added what appears to be parsing of webvtt styles into
> extradata in the demuxer, and then writing it out in the muxer. And
> the test just outputs this webvtt into ASS, which doesn't keep any of
> this since the decoder doesn't translate the newly inserted things
> into ASS styles :)
>
> So, I think a more fitting test would be to add this to the webvtt
> tests with -c copy and pushing it out as webvtt, I would think?
>
> Jan

Hi Jan,

Thanks for taking the time to review.

In fact, this was the test I intended to add. The test case was
developed before the fix and serves to show the original issue we saw
(total failure to extract any cues under the circumstances of the
test) is fixed and that there should be no regression with that
content, and I believe it serves those purposes. Apologies for not
making this clearer in the original submission.

That said, I agree that it may also be worth adding test coverage for
the encoder change and I will take the time to add that as well.

Cheers,
Dave