A negative start_dts value (eg, delay from edit lists) typically yields a duration larger than end_pts. During edit list processing, the delay is removed again, yielding the correct duration within the elst.
However, other duration-carrying atoms (tkhd, mvhd, mdhd) still have the delay incorporated into their durations. This is incorrect. Fix this by withholding delay from the duration if edit lists are used. This also simplifies edit-list processing a bit, since the delay does not need to be removed from the calculated duration again. --- The mov spec says that the tkhd duration is "derived from the track's edits" [1] and the duratons of the other atoms (mvhd, mdhd) are in turn taken from the longest track. So it seems that incorporating the delay into the track duration is a bug in itself when the edit list has the correct duration, and this propagates out tothe other top-level durations. Unsure of how this change interacts with other modes that may expect negative timestamps such as CMAF, so the patch errs on the side of caution and only takes effect if edit lists are used. Can loosen that up if necessary. [1] https://developer.apple.com/library/archive/documentation/QuickTime/QTFF/QTFFChap2/qtff2.html#//apple_ref/doc/uid/TP40000939-CH204-BBCEIDFA libavformat/movenc.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/libavformat/movenc.c b/libavformat/movenc.c index 7db2e28840..31441a9f6c 100644 --- a/libavformat/movenc.c +++ b/libavformat/movenc.c @@ -2831,7 +2831,14 @@ static int64_t calc_pts_duration(MOVMuxContext *mov, MOVTrack *track) if (track->end_pts != AV_NOPTS_VALUE && track->start_dts != AV_NOPTS_VALUE && track->start_cts != AV_NOPTS_VALUE) { - return track->end_pts - (track->start_dts + track->start_cts); + int64_t dur = track->end_pts, delay = track->start_dts + track->start_cts; + /* Note, this delay is calculated from the pts of the first sample, + * ensuring that we don't reduce the duration for cases with + * dts<0 pts=0. */ + if (delay > 0 || !mov->use_editlist) { + dur -= delay; + } + return dur; } return track->track_duration; } @@ -3118,10 +3125,6 @@ static int mov_write_edts_tag(AVIOContext *pb, MOVMuxContext *mov, * rounded to 0 when represented in MOV_TIMESCALE units. */ av_assert0(av_rescale_rnd(start_dts, MOV_TIMESCALE, track->timescale, AV_ROUND_DOWN) <= 0); start_ct = -FFMIN(start_dts, 0); - /* Note, this delay is calculated from the pts of the first sample, - * ensuring that we don't reduce the duration for cases with - * dts<0 pts=0. */ - duration += delay; } /* For fragmented files, we don't know the full length yet. Setting -- 2.24.3 (Apple Git-128) _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
