[theora] bitstream test suite

[Gregory Maxwell]

On Wed, Jan 27, 2010 at 7:59 AM, dos386 <dos386 at gmail.com> wrote:
> - The "offset" movie is "frozen" ... deliberate ??? Most players I
> tested so far fail the test anyway :-( A movie should show something
> moving, otherwise you can't find out whether it plays or is just
> frozen (I have trouble with Opera 10.5 here ...)

The only 'players' that I'm aware of failing this is absolutely
everything based on ffmpeg. As far as I know everything else does the
offsetting fine.  If you're aware of non-ffmpeg based things that
offset incorrectly, please let me know... it usually only takes a
couple of lines of code to fix it, I submitted patches to several
places last year in order to squash that one wherever I found it.

Some players have had problems with still test material, so you can
think of this as testing an additional case as well. There should be
no important interaction with offset and motion, so I don't see any
reason that particular test should move. There is another clip with
both offset and motion, in any case.

> 2. some players do screen switches and a recovery from such a
> switch can take up to 3 s

Short videos are a pretty valid use case.  ::shurgs:: Long videos run
into the concerns you raised about size.

> - There should be some description what the movie is supposed to show,
> and possibly a shot (or one good and one or multiple evil), for all
> movies, not only the "offset" test.

Yes, thats a good point. Although it can be difficult to enumerate all
the ways things can fail.

Without postprocessing Theora decode must be bit accurate, we could
provide correctly decoded examples, I suppose.