fix: Fix handling of non-interleaved multi track FMP4 files (shaka-pr…

…oject#1214)

Do not assume that each fragment contains all tracks. 
Use track id instead of index to pick the correct timestamp.

Fixes shaka-project#1213

AUTHORS

@@ -35,4 +35,5 @@ Richard Eklycke <richard@eklycke.se>
 Sanil Raut <sr1990003@gmail.com>
 Sergio Ammirata <sergio@ammirata.net>
 The Chromium Authors <*@chromium.org>
+Prakash Duggaraju <duggaraju@gmail.com>
 Dennis E. Mungai (Brainiarc7) <dmngaie@gmail.com>

CONTRIBUTORS

@@ -52,4 +52,5 @@ Thomas Inskip <tinskip@google.com>
 Tim Lansen <tim.lansen@gmail.com>
 Vincent Nguyen <nvincen@amazon.com>
 Weiguo Shao <weiguo.shao@dolby.com>
+Prakash Duggaraju <duggaraju@gmail.com>
 Dennis E. Mungai (Brainiarc7) <dmngaie@gmail.com>

packager/media/formats/mp4/mp4_media_parser_unittest.cc

@@ -298,6 +298,13 @@ TEST_F(MP4MediaParserTest, CencWithDecryptionSourceAndSenc) {
   EXPECT_EQ(82u, num_samples_);
 }
 
+TEST_F(MP4MediaParserTest, NonInterleavedFMP4) {
+  // Test small, non-interleaved fragment MP4 with one track per fragment.
+  EXPECT_TRUE(ParseMP4File("BigBuckBunny_10s.ismv", 512));
+  EXPECT_EQ(2u, num_streams_);
+  EXPECT_EQ(770u, num_samples_);
+}
+
 }  // namespace mp4
 }  // namespace media
 }  // namespace shaka

packager/media/formats/mp4/track_run_iterator.cc

@@ -284,10 +284,11 @@ bool TrackRunIterator::Init() {
 bool TrackRunIterator::Init(const MovieFragment& moof) {
   runs_.clear();
 
-  next_fragment_start_dts_.resize(moof.tracks.size(), 0);
+  const auto track_count = std::max(moof.tracks.size(), moov_->tracks.size());
+  next_fragment_start_dts_.resize(track_count, 0);
   for (size_t i = 0; i < moof.tracks.size(); i++) {
     const TrackFragment& traf = moof.tracks[i];
-
+    const auto track_index = traf.header.track_id - 1;
     const Track* trak = NULL;
     for (size_t t = 0; t < moov_->tracks.size(); t++) {
       if (moov_->tracks[t].header.track_id == traf.header.track_id)
@@ -351,7 +352,7 @@ bool TrackRunIterator::Init(const MovieFragment& moof) {
     }
 
     int64_t run_start_dts = traf.decode_time_absent
-                                ? next_fragment_start_dts_[i]
+                                ? next_fragment_start_dts_[track_index]
                                 : traf.decode_time.decode_time;
 
     // dts is directly adjusted, which then propagates to pts as pts is encoded
@@ -426,7 +427,7 @@ bool TrackRunIterator::Init(const MovieFragment& moof) {
       runs_.push_back(tri);
       sample_count_sum += trun.sample_count;
     }
-    next_fragment_start_dts_[i] = run_start_dts;
+    next_fragment_start_dts_[track_index] = run_start_dts;
   }
 
   std::sort(runs_.begin(), runs_.end(), CompareMinTrackRunDataOffset());