diff mbox series

[FFmpeg-devel,v3] mov: Compare frag times in correct time base when seeking a stream without a corresponding sidx

Message ID 20220816144850.196109-1-derek.buitenhuis@gmail.com
State Accepted
Commit e1e981c65e9b45c8acd8a907ac67801acadd262c
Headers show
Series [FFmpeg-devel,v3] mov: Compare frag times in correct time base when seeking a stream without a corresponding sidx | expand

Checks

Context Check Description
yinshiyou/make_loongarch64 success Make finished
yinshiyou/make_fate_loongarch64 success Make fate finished
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished

Commit Message

Derek Buitenhuis Aug. 16, 2022, 2:48 p.m. UTC
Some muxers, such as GPAC, create files with only one sidx, but two streams
muxed into the same fragments pointed to by this sidx.

Prevously, in such a case, when we seeked in such files, we fell back
to, for example, using the sidx associated with the video stream, to
seek the audio stream, leaving the seekhead in the wrong place.

We can still do this, but we need to take care to compare timestamps
in the same time base.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
---
 libavformat/mov.c | 38 ++++++++++++++++++++------------------
 1 file changed, 20 insertions(+), 18 deletions(-)

Comments

Derek Buitenhuis Aug. 16, 2022, 4:32 p.m. UTC | #1
On 8/16/2022 5:21 PM, Zhao Zhili wrote:
> get_frag_time() can be called with an mp4 file which has no sidx at all. In that case,
> dst_st should have a higher priority than other streams, even if sc->has_sidx is false.
> And first_tfra_pts might be used here, which makes the check of sc->has_sidx unnatural.
> So in my opinion, the check on sc->has_sidx should be removed.

This seems like it should be in a separate patch, though - it is changing a different
behavior than what this patch does.

> +    frag_stream_info = get_frag_stream_info(frag_index, index, dst_st->id);
> +    timestamp = get_stream_info_time(frag_stream_info);
> +    if (timestamp != AV_NOPTS_VALUE)
> +        return timestamp;

I did look at that, but I do not think it can be.

get_stream_info_time is not equivalent to what is here. get_stream_info_time will
eventually fall back to frag_stream_info->tfdt_dts, where as this code falls back
to frag_stream_info->sidx_pts even if it is AV_NOPTS_VALUE. It would be a behavior
change do use get_stream_info_time here.

- Derek
Zhao Zhili Aug. 17, 2022, 10:29 a.m. UTC | #2
On Tue, 2022-08-16 at 17:32 +0100, Derek Buitenhuis wrote:
> On 8/16/2022 5:21 PM, Zhao Zhili wrote:
> > get_frag_time() can be called with an mp4 file which has no sidx at
> > all. In that case,
> > dst_st should have a higher priority than other streams, even if
> > sc->has_sidx is false.
> > And first_tfra_pts might be used here, which makes the check of sc-
> > >has_sidx unnatural.
> > So in my opinion, the check on sc->has_sidx should be removed.
> 
> This seems like it should be in a separate patch, though - it is
> changing a different
> behavior than what this patch does.

OK.

> 
> > +    frag_stream_info = get_frag_stream_info(frag_index, index,
> > dst_st->id);
> > +    timestamp = get_stream_info_time(frag_stream_info);
> > +    if (timestamp != AV_NOPTS_VALUE)
> > +        return timestamp;
> 
> I did look at that, but I do not think it can be.
> 
> get_stream_info_time is not equivalent to what is here.
> get_stream_info_time will
> eventually fall back to frag_stream_info->tfdt_dts, where as this
> code falls back
> to frag_stream_info->sidx_pts even if it is AV_NOPTS_VALUE. It would
> be a behavior
> change do use get_stream_info_time here.

OK, it make sense together with `if (sc->has_sidx)`.

No more comments from me.

> 
> - Derek
> _______________________________________________
> 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".
Derek Buitenhuis Aug. 19, 2022, 4:38 p.m. UTC | #3
On 8/17/2022 11:29 AM, Zhao Zhili wrote:
>> This seems like it should be in a separate patch, though - it is
>> changing a different
>> behavior than what this patch does.
> 
> OK.

Will send a 2nd patch with this in the next few days.


>> I did look at that, but I do not think it can be.
>>
>> get_stream_info_time is not equivalent to what is here.
>> get_stream_info_time will
>> eventually fall back to frag_stream_info->tfdt_dts, where as this
>> code falls back
>> to frag_stream_info->sidx_pts even if it is AV_NOPTS_VALUE. It would
>> be a behavior
>> change do use get_stream_info_time here.
> 
> OK, it make sense together with `if (sc->has_sidx)`.
> 
> No more comments from me.

Applied. Thank you for your reviews :).

- Derek
diff mbox series

Patch

diff --git a/libavformat/mov.c b/libavformat/mov.c
index 31f3249ca6..1d8c5b2904 100644
--- a/libavformat/mov.c
+++ b/libavformat/mov.c
@@ -1271,15 +1271,18 @@  static int64_t get_stream_info_time(MOVFragmentStreamInfo * frag_stream_info)
     return frag_stream_info->tfdt_dts;
 }
 
-static int64_t get_frag_time(MOVFragmentIndex *frag_index,
-                             int index, int track_id)
+static int64_t get_frag_time(AVFormatContext *s, AVStream *dst_st,
+                             MOVFragmentIndex *frag_index, int index)
 {
     MOVFragmentStreamInfo * frag_stream_info;
+    MOVStreamContext *sc = dst_st->priv_data;
     int64_t timestamp;
-    int i;
+    int i, j;
 
-    if (track_id >= 0) {
-        frag_stream_info = get_frag_stream_info(frag_index, index, track_id);
+    // If the stream is referenced by any sidx, limit the search
+    // to fragments that referenced this stream in the sidx
+    if (sc->has_sidx) {
+        frag_stream_info = get_frag_stream_info(frag_index, index, dst_st->id);
         if (frag_stream_info->sidx_pts != AV_NOPTS_VALUE)
             return frag_stream_info->sidx_pts;
         if (frag_stream_info->first_tfra_pts != AV_NOPTS_VALUE)
@@ -1288,28 +1291,27 @@  static int64_t get_frag_time(MOVFragmentIndex *frag_index,
     }
 
     for (i = 0; i < frag_index->item[index].nb_stream_info; i++) {
+        AVStream *frag_stream = NULL;
         frag_stream_info = &frag_index->item[index].stream_info[i];
+        for (j = 0; j < s->nb_streams; j++)
+            if (s->streams[j]->id == frag_stream_info->id)
+                frag_stream = s->streams[j];
+        if (!frag_stream) {
+            av_log(s, AV_LOG_WARNING, "No stream matching sidx ID found.\n");
+            continue;
+        }
         timestamp = get_stream_info_time(frag_stream_info);
         if (timestamp != AV_NOPTS_VALUE)
-            return timestamp;
+            return av_rescale_q(timestamp, frag_stream->time_base, dst_st->time_base);
     }
     return AV_NOPTS_VALUE;
 }
 
-static int search_frag_timestamp(MOVFragmentIndex *frag_index,
+static int search_frag_timestamp(AVFormatContext *s, MOVFragmentIndex *frag_index,
                                  AVStream *st, int64_t timestamp)
 {
     int a, b, m, m0;
     int64_t frag_time;
-    int id = -1;
-
-    if (st) {
-        // If the stream is referenced by any sidx, limit the search
-        // to fragments that referenced this stream in the sidx
-        MOVStreamContext *sc = st->priv_data;
-        if (sc->has_sidx)
-            id = st->id;
-    }
 
     a = -1;
     b = frag_index->nb_items;
@@ -1318,7 +1320,7 @@  static int search_frag_timestamp(MOVFragmentIndex *frag_index,
         m0 = m = (a + b) >> 1;
 
         while (m < b &&
-               (frag_time = get_frag_time(frag_index, m, id)) == AV_NOPTS_VALUE)
+               (frag_time = get_frag_time(s, st, frag_index, m)) == AV_NOPTS_VALUE)
             m++;
 
         if (m < b && frag_time <= timestamp)
@@ -8862,7 +8864,7 @@  static int mov_seek_fragment(AVFormatContext *s, AVStream *st, int64_t timestamp
     if (!mov->frag_index.complete)
         return 0;
 
-    index = search_frag_timestamp(&mov->frag_index, st, timestamp);
+    index = search_frag_timestamp(s, &mov->frag_index, st, timestamp);
     if (index < 0)
         index = 0;
     if (!mov->frag_index.item[index].headers_read)