mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-23 12:43:46 +02:00
Timestamp patch by Gael Chardon <gael DOT ffmpeg AT 4now DOT net>
Originally committed as revision 3865 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
parent
7dbab4a9b4
commit
7fe5a3c0c7
@ -231,6 +231,9 @@ typedef struct MOVStreamContext {
|
||||
long sample_to_chunk_sz;
|
||||
MOV_sample_to_chunk_tbl *sample_to_chunk;
|
||||
long sample_to_chunk_index;
|
||||
int sample_to_time_index;
|
||||
long sample_to_time_sample;
|
||||
uint64_t sample_to_time_time;
|
||||
long sample_size;
|
||||
long sample_count;
|
||||
long *sample_sizes;
|
||||
@ -1895,6 +1898,40 @@ readchunk:
|
||||
#endif
|
||||
|
||||
mov->next_chunk_offset = offset + size;
|
||||
|
||||
/* find the corresponding dts */
|
||||
if (sc && sc->sample_to_time_index < sc->stts_count && pkt) {
|
||||
uint32_t count;
|
||||
uint64_t dts;
|
||||
uint32_t duration = (uint32_t)(sc->stts_data[sc->sample_to_time_index]&0xffff);
|
||||
count = (uint32_t)(sc->stts_data[sc->sample_to_time_index]>>32);
|
||||
if ((sc->sample_to_time_sample + count) < sc->current_sample) {
|
||||
sc->sample_to_time_sample += count;
|
||||
sc->sample_to_time_time += count*duration;
|
||||
sc->sample_to_time_index ++;
|
||||
duration = (uint32_t)(sc->stts_data[sc->sample_to_time_index]&0xffff);
|
||||
}
|
||||
dts = sc->sample_to_time_time + (sc->current_sample-1 - sc->sample_to_time_sample) * duration;
|
||||
dts = av_rescale( dts,
|
||||
(int64_t)s->streams[sc->ffindex]->time_base.den,
|
||||
(int64_t)sc->time_scale * (int64_t)s->streams[sc->ffindex]->time_base.num );
|
||||
pkt->dts = dts;
|
||||
// audio pts = dts, true for video only in low_latency mode (FIXME a good way to know if we are in a low latency stream ?)
|
||||
if (s->streams[sc->ffindex]->codec.codec_type == CODEC_TYPE_AUDIO)
|
||||
pkt->pts = dts;
|
||||
else
|
||||
pkt->pts = AV_NOPTS_VALUE;
|
||||
#ifdef DEBUG
|
||||
/* av_log(NULL, AV_LOG_DEBUG, "stream #%d smp #%ld dts = %ld (smp:%ld time:%ld idx:%ld ent:%d count:%ld dur:%ld)\n"
|
||||
, pkt->stream_index, sc->current_sample-1, (long)pkt->dts
|
||||
, (long)sc->sample_to_time_sample
|
||||
, (long)sc->sample_to_time_time
|
||||
, (long)sc->sample_to_time_index
|
||||
, (long)sc->stts_count
|
||||
, count
|
||||
, duration);*/
|
||||
#endif
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
@ -1919,6 +1956,9 @@ static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp
|
||||
int64_t sample_file_offset;
|
||||
int32_t first_chunk_sample;
|
||||
int32_t sample_to_chunk_idx;
|
||||
int sample_to_time_index;
|
||||
long sample_to_time_sample = 0;
|
||||
uint64_t sample_to_time_time = 0;
|
||||
int mov_idx;
|
||||
|
||||
// Find the corresponding mov stream
|
||||
@ -1957,12 +1997,17 @@ static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp
|
||||
//av_log(s, AV_LOG_DEBUG, "> sample_time %lli \n", (long)sample_time);
|
||||
//av_log(s, AV_LOG_DEBUG, "> count=%i duration=%i\n", count, duration);
|
||||
if ((start_time + count*duration) > sample_time) {
|
||||
sample_to_time_time = start_time;
|
||||
sample_to_time_index = i;
|
||||
sample_to_time_sample = sample;
|
||||
sample += (sample_time - start_time) / duration;
|
||||
break;
|
||||
}
|
||||
sample += count;
|
||||
start_time += count * duration;
|
||||
}
|
||||
sample_to_time_time = start_time;
|
||||
sample_to_time_index = i;
|
||||
/* NOTE: despite what qt doc say, the dt value (Display Time in qt vocabulary) computed with the stts atom
|
||||
is a decoding time stamp (dts) not a presentation time stamp. And as usual dts != pts for stream with b frames */
|
||||
|
||||
@ -1989,6 +2034,8 @@ static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp
|
||||
// av_log(s, AV_LOG_DEBUG, "a=%i (%i) b=%i (%i) m=%i (%i) stream #%i\n", a, sc->keyframes[a], b, sc->keyframes[b], m, sc->keyframes[m], mov_idx);
|
||||
#endif
|
||||
}
|
||||
// for low latency prob: always use the previous keyframe, just uncomment the next line
|
||||
// if (a) a--;
|
||||
seek_sample = sc->keyframes[a];
|
||||
}
|
||||
else
|
||||
@ -2084,6 +2131,20 @@ static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp
|
||||
}
|
||||
msc->current_sample += (msc->next_chunk - (msc->sample_to_chunk[msc->sample_to_chunk_index].first - 1)) * sc->sample_to_chunk[msc->sample_to_chunk_index].count;
|
||||
msc->left_in_chunk = msc->sample_to_chunk[msc->sample_to_chunk_index].count - 1;
|
||||
// Find corresponding position in stts (used later to compute dts)
|
||||
sample = 0;
|
||||
start_time = 0;
|
||||
for (msc->sample_to_time_index = 0; msc->sample_to_time_index < msc->stts_count; msc->sample_to_time_index++) {
|
||||
count = (uint32_t)(msc->stts_data[msc->sample_to_time_index]>>32);
|
||||
duration = (uint32_t)(msc->stts_data[msc->sample_to_time_index]&0xffff);
|
||||
if ((sample + count - 1) > msc->current_sample) {
|
||||
msc->sample_to_time_time = start_time;
|
||||
msc->sample_to_time_sample = sample;
|
||||
break;
|
||||
}
|
||||
sample += count;
|
||||
start_time += count * duration;
|
||||
}
|
||||
#ifdef DEBUG
|
||||
av_log(s, AV_LOG_DEBUG, "Next Sample for stream #%i is #%i @%i\n", i, msc->current_sample + 1, msc->sample_to_chunk_index + 1);
|
||||
#endif
|
||||
|
Loading…
Reference in New Issue
Block a user