[FFmpeg-devel] [PATCH] avformat/hls: tag as AVFMT_TS_DISCONT

[Aman Gupta]

On Wed, May 16, 2018 at 11:14 AM, wm4 <nfxjfg at googlemail.com> wrote:

> On Wed, 16 May 2018 10:17:58 -0700
> Aman Gupta <ffmpeg at tmm1.net> wrote:
>
> > From: Aman Gupta <aman at tmm1.net>
> >
> > HLS streams can contain discontinuities. Mark the format as such.
> >
> > This triggers various discontinuity fixes in lavf/utils.c and fftools
> >
> > Signed-off-by: Aman Gupta <aman at tmm1.net>
> > ---
> >  libavformat/hls.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/libavformat/hls.c b/libavformat/hls.c
> > index 4ee4be769d..3199b0ac8d 100644
> > --- a/libavformat/hls.c
> > +++ b/libavformat/hls.c
> > @@ -2277,7 +2277,7 @@ AVInputFormat ff_hls_demuxer = {
> >      .long_name      = NULL_IF_CONFIG_SMALL("Apple HTTP Live Streaming"),
> >      .priv_class     = &hls_class,
> >      .priv_data_size = sizeof(HLSContext),
> > -    .flags          = AVFMT_NOGENSEARCH,
> > +    .flags          = AVFMT_NOGENSEARCH | AVFMT_TS_DISCONT,
> >      .read_probe     = hls_probe,
> >      .read_header    = hls_read_header,
> >      .read_packet    = hls_read_packet,
>
> I think I'm against this. HLS streams do not typically contain
> timestamp resets (even if they could). Otherwise you might as well add
>

mpegts timestamps wrap every 26 hours, so any live continuous HLS broadcast
is guaranteed to have a timestamp reset.
Users have reported seeing these resets on various streams (
http://ffmpeg.org/pipermail/ffmpeg-devel/2018-March/226706.html), plus
the HLS spec itself defines a EXT-X-DISCONTINUITY tag to signal
discontinuities. So I don't think it's fair to say that discontinuities
don't typically occur and can be ignored.


> this flag to the Matroska demuxer. Besides, it would break some of my
> code, which uses this flag as a heuristic to detect mpeg-ts style
> non-container formats.
>

I'm working on a patchset with a new -remove_ts_discont option which will
remove discontinuities and normalize timestamps. This
option would only kick in for formats marked as AVFMT_TS_DISCONT, so I
would like to see any formats where timestamp
discontinuities are possible to be marked as such.

Maybe we can add another AVFMT flag that can be used as your heuristic?

Aman




> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>