demux, player: fix playback of sparse video streams (w/ still images)

Fixes several issues playing back mpegts with video streams marked
as having "still images". For example, see this video which has
frames only every 6s: https://s3.amazonaws.com/tmm1/music-choice.ts

Changes include:
- start playback right away, without waiting for first video frame
- do not consider the sparse video stream in demuxer underrun detection
- do not require multiple video frames for the VO
- use audio as the master stream for demuxer metadata events
- use audio stream for playback time

Signed-off-by: Aman Gupta <aman@tmm1.net>

3 files changed, 10 insertions, 2 deletions

diff --git a/demux/demux.c b/demux/demux.c
index afe8d2b51f..40a68eacae 100644
--- a/demux/demux.c
+++ b/demux/demux.c
@@ -282,6 +282,7 @@ struct demux_stream {
     bool eager;             // try to keep at least 1 packet queued
                             // if false, this stream is disabled, or passively
                             // read (like subtitles)
+    bool still_image;       // stream has still video images
     bool refreshing;        // finding old position after track switches
     bool eof;               // end of demuxed stream? (true if no more packets)
 
@@ -703,8 +704,9 @@ static void update_stream_selection_state(struct demux_internal *in,
     for (int n = 0; n < in->num_streams; n++) {
         struct demux_stream *s = in->streams[n]->ds;
 
+        s->still_image = s->sh->still_image;
         s->eager = s->selected && !s->sh->attached_picture;
-        if (s->eager) {
+        if (s->eager && !s->still_image) {
             any_av_streams |= s->type != STREAM_SUB;
             if (!master ||
                 (master->type == STREAM_VIDEO && s->type == STREAM_AUDIO))
@@ -2994,7 +2996,7 @@ static int cached_demux_control(struct demux_internal *in, int cmd, void *arg)
             struct demux_stream *ds = in->streams[n]->ds;
             if (ds->eager && !(!ds->queue->head && ds->eof) && !ds->ignore_eof)
             {
-                r->underrun |= !ds->reader_head && !ds->eof;
+                r->underrun |= !ds->reader_head && !ds->eof && !ds->still_image;
                 r->ts_reader = MP_PTS_MAX(r->ts_reader, ds->base_ts);
                 r->ts_end = MP_PTS_MAX(r->ts_end, ds->queue->last_ts);
                 any_packets |= !!ds->reader_head;
diff --git a/demux/demux_lavf.c b/demux/demux_lavf.c
index 2049cf6c47..555df9406d 100644
--- a/demux/demux_lavf.c
+++ b/demux/demux_lavf.c
@@ -53,6 +53,9 @@
 #ifndef AV_DISPOSITION_TIMED_THUMBNAILS
 #define AV_DISPOSITION_TIMED_THUMBNAILS 0
 #endif
+#ifndef AV_DISPOSITION_STILL_IMAGE
+#define AV_DISPOSITION_STILL_IMAGE 0
+#endif
 
 #define INITIAL_PROBE_SIZE STREAM_BUFFER_SIZE
 #define PROBE_BUF_SIZE FFMIN(STREAM_MAX_BUFFER_SIZE, 2 * 1024 * 1024)
@@ -717,6 +720,8 @@ static void handle_new_stream(demuxer_t *demuxer, int i)
             sh->forced_track = true;
         if (st->disposition & AV_DISPOSITION_DEPENDENT)
             sh->dependent_track = true;
+        if (st->disposition & AV_DISPOSITION_STILL_IMAGE)
+            sh->still_image = true;
         if (priv->format_hack.use_stream_ids)
             sh->demuxer_id = st->id;
         AVDictionaryEntry *title = av_dict_get(st->metadata, "title", NULL, 0);
diff --git a/demux/stheader.h b/demux/stheader.h
index 700ded89fa..63744487bf 100644
--- a/demux/stheader.h
+++ b/demux/stheader.h
@@ -46,6 +46,7 @@ struct sh_stream {
     bool default_track;         // container default track flag
     bool forced_track;          // container forced track flag
     bool dependent_track;       // container dependent track flag
+    bool still_image;           // video stream contains still images
     int hls_bitrate;
 
     struct mp_tags *tags;