Re: [FFmpeg-devel] [PATCH 1/2] avformat/movenc: initialize pts/dts/duration of timecode packet

[lance.lmwang@gmail.com]

On Fri, Mar 11, 2022 at 05:16:09PM +0100, Andreas Rheinhardt wrote:
> lance.lmwang@gmail.com:
> > On Fri, Mar 11, 2022 at 03:04:32PM +0100, Andreas Rheinhardt wrote:
> >> lance.lmwang@gmail.com:
> >>> On Wed, Mar 02, 2022 at 09:58:31PM +0800, lance.lmwang@gmail.com wrote:
> >>>> From: Limin Wang <lance.lmwang@gmail.com>
> >>>>
> >>>> Fix below error message when timecode packet is written.
> >>>> "Application provided duration: -9223372036854775808 / timestamp: -9223372036854775808 is out of range for mov/mp4 format"
> >>>>
> >>>> try to reproduce by:
> >>>> ffmpeg -y -f lavfi -i color -metadata "timecode=00:00:00:00" -t 1 test.mov
> >>>>
> >>>> Note although error message is printed, the timecode packet will be written anyway. So
> >>>> the patch 2/2 will try to change the log level to warning.
> >>>>
> >>>> The first two test case of fate-lavf-ismv have timecode setting, so the crc of ref data is different.
> >>>> Fixes ticket #9488
> >>>>
> >>>> Signed-off-by: Limin Wang <lance.lmwang@gmail.com>
> >>>> ---
> >>>>  libavformat/movenc.c | 2 ++
> >>>>  tests/ref/lavf/ismv  | 4 ++--
> >>>>  2 files changed, 4 insertions(+), 2 deletions(-)
> >>>>
> >>>> diff --git a/libavformat/movenc.c b/libavformat/movenc.c
> >>>> index 4c86891..74b94cd 100644
> >>>> --- a/libavformat/movenc.c
> >>>> +++ b/libavformat/movenc.c
> >>>> @@ -6383,6 +6383,8 @@ static int mov_create_timecode_track(AVFormatContext *s, int index, int src_inde
> >>>>      pkt->data = data;
> >>>>      pkt->stream_index = index;
> >>>>      pkt->flags = AV_PKT_FLAG_KEY;
> >>>> +    pkt->pts = pkt->dts = av_rescale_q(tc.start, av_inv_q(rate), (AVRational){1,mov->movie_timescale});
> >>>> +    pkt->duration = av_rescale_q(1, av_inv_q(rate), (AVRational){1,mov->movie_timescale});
> >>>>      pkt->size = 4;
> >>>>      AV_WB32(pkt->data, tc.start);
> >>>>      ret = ff_mov_write_packet(s, pkt);
> >>>> diff --git a/tests/ref/lavf/ismv b/tests/ref/lavf/ismv
> >>>> index ac7f72b..723b432 100644
> >>>> --- a/tests/ref/lavf/ismv
> >>>> +++ b/tests/ref/lavf/ismv
> >>>> @@ -1,7 +1,7 @@
> >>>> -48fb8d7a5d19bd60f3a49ccf4b7d6593 *tests/data/lavf/lavf.ismv
> >>>> +7a24b73c096ec0f13f0f7a2d9101c4c1 *tests/data/lavf/lavf.ismv
> >>>>  313169 tests/data/lavf/lavf.ismv
> >>>>  tests/data/lavf/lavf.ismv CRC=0x9d9a638a
> >>>> -d19cd8e310a2e94fe0a0d11c5dc29217 *tests/data/lavf/lavf.ismv
> >>>> +79646383fd099d45ad0d0c2791c601dd *tests/data/lavf/lavf.ismv
> >>>>  322075 tests/data/lavf/lavf.ismv
> >>>>  tests/data/lavf/lavf.ismv CRC=0xe8130120
> >>>>  3b6023766845b51b075aed474c00f73c *tests/data/lavf/lavf.ismv
> >>>> -- 
> >>>> 1.8.3.1
> >>>>
> >>>
> >>> will apply the patch set tomorrow unless there are any objections.
> >>>
> >>
> >> You have not really answered whether the current files or the new files
> >> are spec-incompliant; you have just reported that one byte is different.
> > 
> > Sorry, I think I have said both current and new file is spec-compliant in the last
> > email. 
> > 
> 
> You stated that you think that both files are valid, but you also said
> that you don't even know what this byte that is different actually means.
> 
> > By Quicktime file format specs:
> > Section Timecode Sample Description, all tmcd field isn't used pts/dts.
> > 
> > As for where is the different for one byte, it's caused by pkt->duration. The
> > old is 0(uninitialized), after the patch it's 33(1 frame duration).  
> > 
> 
> The text about Timecode Sample Description reads as follows: "Frame
> duration: A 32-bit integer that indicates how long each frame lasts in
> real time." This implies that only one of the two files can be
> spec-compliant. I am not a mov/ISOBMFF expert, but it seems to me that
> the current way of doing things is wrong. But I wonder about whether
> your patch is correct for vfr content. Doesn't the property of being vfr
> need to be reflected in the timecodes somehow (with different durations
> for different packets)?

No, it's packet duration, not tmcd frame duration, my patch have do nothing 
for that.(see movenc.c:2348). In addition, for timecode, I don't think vfr is
supported.

The tmcd track just contains one packet with the frame number(4byte), so the
packet data is used by start of timecode. So I set the dts/pts is avoid the
following code think it's invalid packet. If you wonder the patch will change
something, I can update the patch keep packet duration to default zero, then
we can the fate data untouched, for the following track_duration will use it
and make the crc of output is different.


> 
> - Andreas
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

-- 
Thanks,
Limin Wang
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".