[FFmpeg-devel] [PATCH 4/4] lavf/mov: Add support for edit list parsing.

wm4 nfxjfg at googlemail.com
Mon Sep 19 21:37:16 EEST 2016


On Tue,  9 Aug 2016 18:48:22 -0700
Sasi Inguva <isasi-at-google.com at ffmpeg.org> wrote:


> diff --git a/tests/ref/fate/gaplessenc-itunes-to-ipod-aac b/tests/ref/fate/gaplessenc-itunes-to-ipod-aac
> index aacb058..145bd09 100644
> --- a/tests/ref/fate/gaplessenc-itunes-to-ipod-aac
> +++ b/tests/ref/fate/gaplessenc-itunes-to-ipod-aac


> -packet|pts=-1024|dts=-1024|duration=1024
> +packet|pts=0|dts=0|duration=N/A
>  packet|pts=0|dts=0|duration=1024

You've just got to be kidding me. Things wrong here:
- demuxer returns packets with broken timestamps (dts not strictly
  monotonic, decoded pts won't be strictly monotonic either)
- the entirely correct pts is broken and will silently break gapless
  decoding in existing API users (but I guess you don't care and
  neither did whoever pushed this?)
- I know there's supposed to be a "discard frame" flag here (which
  existing API users will ignore, leading to broken decoding and all),
  but it's not tested??? This makes the entire gapless test for
  dumping packet metadata useless, you may as well have deleted the
  tests!

And of course you could just have let the mov demuxer export the edit
list explicitly, and assembled it after decoding. But I guess you
didn't want to deal with that ffmpeg.c garbage dump (and neither did
whoever reviewed/pushed this), so you moved it into the libs as uaual?

This is all very disappointing.


More information about the ffmpeg-devel mailing list