[FFmpeg-devel] Change 9963 introduces regression - start_time in asf.c

[Neil Brown]

On Saturday August 11, michaelni at gmx.at wrote:
> > 
> > > Then because st->start_time == AV_NOPTS_VALUE, it sets st->start_time
> > > to 0 
> > 
> > this is a bug
> > 
> > 
> > > and st->duration to the estimate of 4:41 even though
> > > st->duration already has the correct value of 300837 (5:00).
> > 
> > this is a bug as well
> 
> btw, patches welcome for these :)

I'd love to submit a patch, but I'm not certain that I understand all
the issues.  At all.

e.g. what does start_time *mean* for a video file?  Doesn't it just
start at the beginning, and end at the end?
And what does it mean for start_time to be AV_NOPTS_VALUE.  Does that
just mean that the format doesn't contain any explicit start_time
value.

The documentation says:
   decoding: pts of the first frame of the stream, in stream time
   base.

and while I guess that is better than nothing, it isn't a lot better
than nothing.

The current code seems to treat start_time and duration as a pair -
either they are both set or neither are set.  Why does it do that?
And what will break if break that pairing and have asf set one but not
the other?

And how can I be sure that any given patch won't break anything given
how shallow my understanding still is.  If there a regression test
suite any where?  When there a so many format, options, variations
etc - such a thing would seem very useful.

So anyway, if you can just tell me how start_time is used, it might
give me enough confidence to submit a patch.

Thanks,
NeilBrown