[FFmpeg-trac] #7482(ffmpeg:new): ffmpeg.c's discontinuity handling breaks when one of the A/V streams creeps

FFmpeg trac at avcodec.org
Wed Oct 10 00:08:39 EEST 2018


#7482: ffmpeg.c's discontinuity handling breaks when one of the A/V streams creeps
----------------------------------------+----------------------------------
             Reporter:  JEEB            |                    Owner:
                 Type:  defect          |                   Status:  new
             Priority:  important       |                Component:  ffmpeg
              Version:  git-master      |               Resolution:
             Keywords:  dts regression  |               Blocked By:
             Blocking:                  |  Reproduced by developer:  0
Analyzed by developer:  0               |
----------------------------------------+----------------------------------

Comment (by JEEB):

 Thank you for the bisect.

 I think that might only end up being the "breaker" as that was the point
 at which libavformat got discontinuity handling by itself, and thus
 ffmpeg.c started receiving monotonically rising timestamps. The sample
 does also contain an MPEG-TS timestamp discontinuity at around that point
 as well, which would before that point be handled by ffmpeg.c instead of
 "what came before it".

 In any case, this leads me to think how much of the ffmpeg.c discontinuity
 handling could we clean out? Definitely not all of it since the
 timestamps-start-at-zero logic is also under it I think, but some of it
 seems to attempt to do things which lavf is doing without possibly having
 all of the context (like the demuxer-specific discontinuity bits etc,
 which IIRC are private to lavf?).

 I will try to see from any of the cases where I've used ffmpeg.c which
 parts of discontinuity handling were actually useful. And if those could
 actually have been fixed in lavf itself instead.

--
Ticket URL: <https://trac.ffmpeg.org/ticket/7482#comment:3>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker


More information about the FFmpeg-trac mailing list