[FFmpeg-devel] [PATCH] avformat/movenc: Fix tfdt out of sync
Hu Weiwen
sehuww at mail.scut.edu.cn
Sun Jul 18 13:22:32 EEST 2021
Fix an edge case when auto flushing an empty fragment, and the previous fragment
has inaccurate duration, the track->frag_start would be out of sync from input
dts, and all subsequent tfdt box will have out of sync base_media_decode_time.
This error can accumulate to quite large over long-running streaming.
This can be easily reproduced by remux a variable frame rate source with dash
muxer and set streaming to 1.
When flushing a fragment, we may guess the duration if we have not got the next
package. Then when we get the first packet of the next fragment, we intend to
set the trk->cluster[0].dts to match the previously guessed duration. However,
at this time trk->track_duration may already be updated from
mov_write_single_packet() when auto flushing the empty fragment (e.g. if
FF_MOV_FLAG_FRAG_EVERY_FRAME is set), but no moof was written to reflect this
change. So the trk->cluster[0].dts do NOT actually match the previously written
duration. Then, when flushing the next non-empty fragment, we will get a wrong
"duration", and track->frag_start will be out of sync. And it cannot get back on
track again.
This error is expected to accumulate over time. trk->cluster[0].dts will only
get too large. it will not get too small because we have check_pkt(), so
track->frag_start will fall more and more behind from input dts.
To fix this, we can directly remove the track->frag_start. It can always be
replaced by "track->cluster[0].dts - track->start_dts".
Signed-off-by: Hu Weiwen <sehuww at mail.scut.edu.cn>
---
libavformat/movenc.c | 24 ++++--------------------
libavformat/movenc.h | 1 -
2 files changed, 4 insertions(+), 21 deletions(-)
diff --git a/libavformat/movenc.c b/libavformat/movenc.c
index c85efe87489..48736acab65 100644
--- a/libavformat/movenc.c
+++ b/libavformat/movenc.c
@@ -4482,8 +4482,7 @@ static int mov_write_tfxd_tag(AVIOContext *pb, MOVTrack *track)
avio_write(pb, uuid, sizeof(uuid));
avio_w8(pb, 1);
avio_wb24(pb, 0);
- avio_wb64(pb, track->start_dts + track->frag_start +
- track->cluster[0].cts);
+ avio_wb64(pb, track->cluster[0].dts + track->cluster[0].cts);
avio_wb64(pb, track->end_pts -
(track->cluster[0].dts + track->cluster[0].cts));
@@ -4562,8 +4561,7 @@ static int mov_add_tfra_entries(AVIOContext *pb, MOVMuxContext *mov, int tracks,
info->size = size;
// Try to recreate the original pts for the first packet
// from the fields we have stored
- info->time = track->start_dts + track->frag_start +
- track->cluster[0].cts;
+ info->time = track->cluster[0].dts + track->cluster[0].cts;
info->duration = track->end_pts -
(track->cluster[0].dts + track->cluster[0].cts);
// If the pts is less than zero, we will have trimmed
@@ -4601,7 +4599,7 @@ static int mov_write_tfdt_tag(AVIOContext *pb, MOVTrack *track)
ffio_wfourcc(pb, "tfdt");
avio_w8(pb, 1); /* version */
avio_wb24(pb, 0);
- avio_wb64(pb, track->frag_start);
+ avio_wb64(pb, track->cluster[0].dts - track->start_dts);
return update_size(pb, pos);
}
@@ -4678,8 +4676,7 @@ static int mov_write_sidx_tag(AVIOContext *pb,
if (track->entry) {
entries = 1;
- presentation_time = track->start_dts + track->frag_start +
- track->cluster[0].cts;
+ presentation_time = track->cluster[0].dts + track->cluster[0].cts;
duration = track->end_pts -
(track->cluster[0].dts + track->cluster[0].cts);
starts_with_SAP = track->cluster[0].flags & MOV_SYNC_SAMPLE;
@@ -5358,10 +5355,6 @@ static int mov_flush_fragment(AVFormatContext *s, int force)
mov->moov_written = 1;
mov->mdat_size = 0;
for (i = 0; i < mov->nb_streams; i++) {
- if (mov->tracks[i].entry)
- mov->tracks[i].frag_start += mov->tracks[i].start_dts +
- mov->tracks[i].track_duration -
- mov->tracks[i].cluster[0].dts;
mov->tracks[i].entry = 0;
mov->tracks[i].end_reliable = 0;
}
@@ -5415,11 +5408,7 @@ static int mov_flush_fragment(AVFormatContext *s, int force)
MOVTrack *track = &mov->tracks[i];
int buf_size, write_moof = 1, moof_tracks = -1;
uint8_t *buf;
- int64_t duration = 0;
- if (track->entry)
- duration = track->start_dts + track->track_duration -
- track->cluster[0].dts;
if (mov->flags & FF_MOV_FLAG_SEPARATE_MOOF) {
if (!track->mdat_buf)
continue;
@@ -5439,8 +5428,6 @@ static int mov_flush_fragment(AVFormatContext *s, int force)
ffio_wfourcc(s->pb, "mdat");
}
- if (track->entry)
- track->frag_start += duration;
track->entry = 0;
track->entries_flushed = 0;
track->end_reliable = 0;
@@ -5759,7 +5746,6 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt)
/* New fragment, but discontinuous from previous fragments.
* Pretend the duration sum of the earlier fragments is
* pkt->dts - trk->start_dts. */
- trk->frag_start = pkt->dts - trk->start_dts;
trk->end_pts = AV_NOPTS_VALUE;
trk->frag_discont = 0;
}
@@ -5781,12 +5767,10 @@ int ff_mov_write_packet(AVFormatContext *s, AVPacket *pkt)
/* Pretend the whole stream started at pts=0, with earlier fragments
* already written. If the stream started at pts=0, the duration sum
* of earlier fragments would have been pkt->pts. */
- trk->frag_start = pkt->pts;
trk->start_dts = pkt->dts - pkt->pts;
} else {
/* Pretend the whole stream started at dts=0, with earlier fragments
* already written, with a duration summing up to pkt->dts. */
- trk->frag_start = pkt->dts;
trk->start_dts = 0;
}
trk->frag_discont = 0;
diff --git a/libavformat/movenc.h b/libavformat/movenc.h
index af1ea0bce64..daeaad1cc68 100644
--- a/libavformat/movenc.h
+++ b/libavformat/movenc.h
@@ -138,7 +138,6 @@ typedef struct MOVTrack {
AVIOContext *mdat_buf;
int64_t data_offset;
- int64_t frag_start;
int frag_discont;
int entries_flushed;
--
2.25.1
More information about the ffmpeg-devel
mailing list