[FFmpeg-devel] [PATCH] fate: Add Canon XF705 demux tests

[Tomas Härdin]

fre 2019-03-08 klockan 20:28 +0100 skrev Carl Eugen Hoyos:
> > 2019-03-08 20:10 GMT+01:00, Matthew McKenna <mmckenna at idsys.net>:
> > On 3/8/2019 1:36 PM, Tomas Härdin wrote:
> > > fre 2019-03-08 klockan 19:04 +0100 skrev Carl Eugen Hoyos:
> > > > > > > > 2019-03-08 14:20 GMT+01:00, Tomas Härdin <tjoppen at acc.umu.se>:
> > > > 
> > > > > This adds some HEVC demux tests for mxfdec.c for the
> > > > > Canon XF705 samples provided by Matthew McKenna.
> > > > > I suspect the thing that's most valuable here is that the
> > > > > codec probing works correctly, which it seems to do.
> > > > 
> > > > This email is missing the additional size in the fate repo.
> > > > 
> > > > I would prefer not to be the one arguing against an additional
> > > > test but afaict this is testing one line in a static array: Is that
> > > > really useful and does it justify a large additional file?
> > > 
> > > That's a fair point. Maybe if they were a single frame each.. Matthew?
> > 
> > The "verysmall" clips are as small as I can make them -- 1.3 megabytes
> > -- 1 second of video + audio w/ bars+tone.  As this is coming off the
> > camera there is not much editing capability, and the camera has a
> > minimum record duration.
> > 
> > Would it suffice to test only a subset or sample of the clips?
> 
> Allow me to repeat my question:
> Is this test meant to check a single entry in a static table?
> If this test is necessary, we should patch the mxf muxer to
> create a very low-res hevc mxf file that uses this uid.

Well, it'd be a functional test. But more than likely the email chain
that goes with the original patch would provide enough justification
for the UL being there. And there already tests for HEVC itself. So no,
it's probably not necessary. The files are very well behaved

/Tomas