[FFmpeg-devel] [PATCH] Add additional FFV1 fate tests

[Peter B.]

On 11/13/2013 11:39 PM, Michael Niedermayer wrote:
> do we have other tests that are split between encoder and decoder ?
> if so how do they do it
> if not why do you want to do it ?

I'll look for existing other encoder+decoder tests.
I've mainly looked at the existing tests for lossless or "pro" codecs,
and the ones I've seen were decoding-only... But I wouldn't assure that
I didn't overlook something.

I want to do it, because FFV1's implementation in ffmpeg is the
reference implementation which all handling of FFV1 is done with by
every application in existence. Therefore I think it is good and
necessary to have both, encoding and decoding tested thoroughly.

For example:
I've ran the ffv1.mak testset on avconv and found out that they are able
to encode gray16 properly, but it gets decoded incorrectly.
Or: FFV1.1 encoding and decoding work fine in avconv. But decoding of v1
is still running on just a single CPU core. I know that this difference
wouldn't be visible as FATE result, but it allows to conveniently
execute, observe and verify FFV1's implementation in action.


Regards,
Pb