[FFmpeg-devel] [PATCH] lavfi/buffersrc: push the frame deeper if requested.

[Michael Niedermayer]

On Sat, Jun 24, 2017 at 05:12:34PM +0200, Paul B Mahol wrote:
> On 6/24/17, Michael Niedermayer <michael at niedermayer.cc> wrote:
> > On Sat, Jun 24, 2017 at 04:54:38PM +0200, Michael Niedermayer wrote:
> >> On Tue, Jun 20, 2017 at 07:59:26PM +0200, Paul B Mahol wrote:
> >> > On 12/24/16, Nicolas George <george at nsup.org> wrote:
> >> > > Le quartidi 4 nivose, an CCXXV, Michael Niedermayer a ecrit :
> >> > >> theres a change with "-af apad=pad_len=10000 -shortest" (no
> >> > >> -vframes)
> >> > >> the file is not sharable, this change undoes the effect of the
> >> > >> non-recursive patch
> >> > >>
> >> > >> use of vframes generally shows differences which seem to return to
> >> > >> what
> >> > >> things were before the "non recursive patch"
> >> > >
> >> > > Thanks for checking. Then I shall push shortly.
> >> >
> >> > This breaks :shortest=1 in all framesync filters, so I'm for reverting
> >> > this.
> >>
> >> The revert broke alot
> >> I would suggest to undo the revert and go through the normal review
> >> process
> >>
> >> examples of breakages:
> >> ./ffmpeg -i bgc.sub.dub.ogm -vframes 3 subs.webm
> >>
> >> (produces 1 instead of 3 frames in the output)
> >>
> >>
> >> ./ffmpeg -i 2014-10-17\ 11.31\ i95Dev\ -\ Carlo\ Pazolini\ _\ KWI\ -\
> >> Meeting.g2m  -max_muxing_queue_size 8000 -vframes 2 g2m5.avi
> >>
> >> fails:
> >> [libmp3lame @ 0x3a5aee0] Queue input is backward in time
> >> [avi @ 0x3a5e280] Too large number of skipped frames 194211 > 60000
> >> 54.1kbits/s speed= 129x
> >> av_interleaved_write_frame(): Invalid argument
> >> [avi @ 0x3a5e280] Too large number of skipped frames 194111 > 60000
> >> frame=    2 fps=1.1 q=2.0 Lsize=    1857kB time=00:03:14.35 bitrate=
> >> 78.3kbits/s speed= 103x
> >> video:149kB audio:1519kB subtitle:0kB other streams:0kB global headers:0kB
> >> muxing overhead: 11.380948%
> >> Conversion failed!
> >> (iam not sure i can share this one)
> >>
> >>
> >> ./ffmpeg -i ~/videos/matrixbench_mpeg2.mpg -vframes 20 -vcodec libx265
> >> -omit_video_pes_length 0 -pix_fmt yuv420p file2.ts
> >>
> >> the change results in the aditional warnings on playback of the resulting
> >> file:
> >> [mpegts @ 0x3d10100] start time for stream 1 is not set in
> >> estimate_timings_from_pts
> >> [mpegts @ 0x3d10100] Could not find codec parameters for stream 1 (Audio:
> >> mp3 ([3][0][0][0] / 0x0003), 0 channels, s16p): unspecified frame size
> >> Consider increasing the value for the 'analyzeduration' and 'probesize'
> >> options
> >
> > For some files the code as in git works better
> >
> > for example:
> > ffmpeg -i
> > ~/tickets/3872/MJ._.Hold.My.Hand._with.Akon__HEVC.1080p.Aex_A__1__001.mkv
> > -vframes 4 file.avi
> >
> > produces a better matching audio stream length now
> 
> The reverted commit breaks lot of filters and I failed to find fix,
> the only logical solution
> was to revert commmit.

I understand, but it kind of sucks as solution because either way
some things are broken. It would be better if there was some change
that would make all or most things work. Not fix some and break others


> 
> Also I'm not happy with current lavfi core state, its left in chaos.

I wish i had more time (and motivation) to work on lavfi, i must admit
i did not follow lavfi core changes as close as i ideally should have
so iam not entirely aware of how exactly things look ...


[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

While the State exists there can be no freedom; when there is freedom there
will be no State. -- Vladimir Lenin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20170624/29f4583b/attachment.sig>