Record AAC sample offsets

The generated fragmented mp4 for the audio track can now be appended to a source buffer in Chrome 43 without errors but the buffered duration remains zero after the operation completes. All current tests pass. Sample table entries and mdat contents look consistent with an audio-only, fragmented mp4 generated by ffmpeg though some discrepancies exists in track metadata still.

@@ -333,17 +333,18 @@ sdtp = function(track) {
   var
   var
     samples = track.samples || [],
     samples = track.samples || [],
     bytes = new Uint8Array(4 + samples.length),
     bytes = new Uint8Array(4 + samples.length),
-    sample,
+    flags,
     i;
     i;
 
 
   // leave the full box header (4 bytes) all zero
   // leave the full box header (4 bytes) all zero
 
 
   // write the sample table
   // write the sample table
   for (i = 0; i < samples.length; i++) {
   for (i = 0; i < samples.length; i++) {
-    sample = samples[i];
-    bytes[i + 4] = (sample.flags.dependsOn << 4) |
-      (sample.flags.isDependedOn << 2) |
-      (sample.flags.hasRedundancy);
+    flags = samples[i].flags;
+
+    bytes[i + 4] = (flags.dependsOn << 4) |
+      (flags.isDependedOn << 2) |
+      (flags.hasRedundancy);
   }
   }
 
 
   return box(types.sdtp,
   return box(types.sdtp,
@@ -508,30 +509,58 @@ tkhd = function(track) {
   return box(types.tkhd, result);
   return box(types.tkhd, result);
 };
 };
 
 
+/**
+ * Generate a track fragment (traf) box. A traf box collects metadata
+ * about tracks in a movie fragment (moof) box.
+ */
 traf = function(track) {
 traf = function(track) {
-  var sampleDependencyTable = sdtp(track);
-  return box(types.traf,
-             box(types.tfhd, new Uint8Array([
+  var trackFragmentHeader, trackFragmentDecodeTime,
+      trackFragmentRun, sampleDependencyTable, dataOffset;
+
+  trackFragmentHeader = box(types.tfhd, new Uint8Array([
     0x00, // version 0
     0x00, // version 0
     0x00, 0x00, 0x00, // flags
     0x00, 0x00, 0x00, // flags
     (track.id & 0xFF000000) >> 24,
     (track.id & 0xFF000000) >> 24,
     (track.id & 0xFF0000) >> 16,
     (track.id & 0xFF0000) >> 16,
     (track.id & 0xFF00) >> 8,
     (track.id & 0xFF00) >> 8,
     (track.id & 0xFF) // track_ID
     (track.id & 0xFF) // track_ID
-             ])),
-             box(types.tfdt, new Uint8Array([
+  ]));
+
+  trackFragmentDecodeTime = box(types.tfdt, new Uint8Array([
     0x00, // version 0
     0x00, // version 0
     0x00, 0x00, 0x00, // flags
     0x00, 0x00, 0x00, // flags
     0x00, 0x00, 0x00, 0x00 // baseMediaDecodeTime
     0x00, 0x00, 0x00, 0x00 // baseMediaDecodeTime
-             ])),
-             trun(track,
-                  sampleDependencyTable.length +
-                  16 + // tfhd
+  ]));
+
+  // the data offset specifies the number of bytes from the start of
+  // the containing moof to the first payload byte of the associated
+  // mdat
+  dataOffset = (16 + // tfhd
                 16 + // tfdt
                 16 + // tfdt
                 8 +  // traf header
                 8 +  // traf header
                 16 + // mfhd
                 16 + // mfhd
                 8 +  // moof header
                 8 +  // moof header
-                  8),  // mdat header
+                8);  // mdat header
+
+  // audio tracks require less metadata
+  if (track.type === 'audio') {
+    trackFragmentRun = trun(track, dataOffset);
+    return box(types.traf,
+               trackFragmentHeader,
+               trackFragmentDecodeTime,
+               trackFragmentRun);
+  }
+
+  // video tracks should contain an independent and disposable samples
+  // box (sdtp)
+  // generate one and adjust offsets to match
+  sampleDependencyTable = sdtp(track);
+  trackFragmentRun = trun(track,
+                          sampleDependencyTable.length + dataOffset);
+  return box(types.traf,
+             trackFragmentHeader,
+             trackFragmentDecodeTime,
+             trackFragmentRun,
              sampleDependencyTable);
              sampleDependencyTable);
 };
 };
 
 
@@ -571,15 +600,38 @@ trex = function(track) {
   return box(types.trex, result);
   return box(types.trex, result);
 };
 };
 
 
-trun = function(track, offset) {
-  var bytes, samples, sample, i;
-
-  samples = track.samples || [];
-  offset += 8 + 12 + (16 * samples.length);
+(function() {
+  var audioTrun, videoTrun, trunHeader;
+
+  // This method assumes all samples are uniform. That is, if a
+  // duration is present for the first sample, it will be present for
+  // all subsequent samples.
+  // see ISO/IEC 14496-12:2012, Section 8.8.8.1
+  trunHeader = function(samples, offset) {
+    var durationPresent = 0, sizePresent = 0,
+        flagsPresent = 0, compositionTimeOffset = 0;
+
+    // trun flag constants
+    if (samples.length) {
+      if (samples[0].duration !== undefined) {
+        durationPresent = 0x1;
+      }
+      if (samples[0].size !== undefined) {
+        sizePresent = 0x2;
+      }
+      if (samples[0].flags !== undefined) {
+        flagsPresent = 0x4;
+      }
+      if (samples[0].compositionTimeOffset !== undefined) {
+        compositionTimeOffset = 0x8;
+      }
+    }
 
 
-  bytes = [
+    return [
       0x00, // version 0
       0x00, // version 0
-    0x00, 0x0f, 0x01, // flags
+      0x00,
+      durationPresent | sizePresent | flagsPresent | compositionTimeOffset,
+      0x01, // flags
       (samples.length & 0xFF000000) >>> 24,
       (samples.length & 0xFF000000) >>> 24,
       (samples.length & 0xFF0000) >>> 16,
       (samples.length & 0xFF0000) >>> 16,
       (samples.length & 0xFF00) >>> 8,
       (samples.length & 0xFF00) >>> 8,
@@ -589,6 +641,15 @@ trun = function(track, offset) {
       (offset & 0xFF00) >>> 8,
       (offset & 0xFF00) >>> 8,
       offset & 0xFF // data_offset
       offset & 0xFF // data_offset
     ];
     ];
+  };
+
+  videoTrun = function(track, offset) {
+    var bytes, samples, sample, i;
+
+    samples = track.samples || [];
+    offset += 8 + 12 + (16 * samples.length);
+
+    bytes = trunHeader(samples, offset);
 
 
     for (i = 0; i < samples.length; i++) {
     for (i = 0; i < samples.length; i++) {
       sample = samples[i];
       sample = samples[i];
@@ -615,7 +676,40 @@ trun = function(track, offset) {
       ]);
       ]);
     }
     }
     return box(types.trun, new Uint8Array(bytes));
     return box(types.trun, new Uint8Array(bytes));
-};
+  };
+
+  audioTrun = function(track, offset) {
+    var bytes, samples, sample, i;
+
+    samples = track.samples || [];
+    offset += 8 + 12 + (8 * samples.length);
+
+    bytes = trunHeader(samples, offset);
+
+    for (i = 0; i < samples.length; i++) {
+      sample = samples[i];
+      bytes = bytes.concat([
+        (sample.duration & 0xFF000000) >>> 24,
+        (sample.duration & 0xFF0000) >>> 16,
+        (sample.duration & 0xFF00) >>> 8,
+        sample.duration & 0xFF, // sample_duration
+        (sample.size & 0xFF000000) >>> 24,
+        (sample.size & 0xFF0000) >>> 16,
+        (sample.size & 0xFF00) >>> 8,
+        sample.size & 0xFF]); // sample_size
+    }
+
+    return box(types.trun, new Uint8Array(bytes));
+  };
+
+  trun = function(track, offset) {
+    if (track.type === 'audio') {
+      return audioTrun(track, offset);
+    } else {
+      return videoTrun(track, offset);
+    }
+  };
+})();
 
 
 window.videojs.mp4 = {
 window.videojs.mp4 = {
   ftyp: ftyp,
   ftyp: ftyp,

@@ -486,7 +486,8 @@ AudioSegmentStream = function(track) {
     // concatenate the audio data to constuct the mdat
     // concatenate the audio data to constuct the mdat
     data = new Uint8Array(aacFramesLength);
     data = new Uint8Array(aacFramesLength);
     track.samples = [];
     track.samples = [];
-    while (aacFramesLength.length) {
+    i = 0;
+    while (aacFrames.length) {
       currentFrame = aacFrames[0];
       currentFrame = aacFrames[0];
       sample = {
       sample = {
         size: currentFrame.data.byteLength,
         size: currentFrame.data.byteLength,

@@ -518,6 +518,35 @@ test('generates a minimal moof', function() {
   }, 'wrote the sample data table');
   }, 'wrote the sample data table');
 });
 });
 
 
+test('generates a moof for audio', function() {
+  var
+    data = mp4.moof(7, [{
+      id: 17,
+      type: 'audio',
+      samples: [{
+        duration: 9000,
+        size: 10
+      }, {
+        duration: 10000,
+        size: 11
+      }]
+    }]),
+    moof = videojs.inspectMp4(data),
+    trun;
+
+  deepEqual(moof[0].boxes[1].boxes.length, 3, 'generated three traf children');
+  trun = moof[0].boxes[1].boxes[2];
+  ok(trun, 'generated a trun');
+  deepEqual(trun.dataOffset, data.byteLength + 8, 'calculated the data offset');
+  deepEqual(trun.samples, [{
+    duration: 9000,
+    size: 10
+  }, {
+    duration: 10000,
+    size: 11
+  }], 'wrote simple audio samples');
+});
+
 test('can generate a traf without samples', function() {
 test('can generate a traf without samples', function() {
   var
   var
     data = mp4.moof(8, [{
     data = mp4.moof(8, [{

@@ -39,6 +39,12 @@ var
     for (i = 0; i < avcStream.length; i += length) {
     for (i = 0; i < avcStream.length; i += length) {
       length = avcView.getUint32(i);
       length = avcView.getUint32(i);
       i += 4;
       i += 4;
+
+      // bail if this doesn't appear to be an H264 stream
+      if (length <= 0) {
+        return;
+      }
+
       switch(avcStream[i] & 0x1F) {
       switch(avcStream[i] & 0x1F) {
       case 0x01:
       case 0x01:
         result.push('slice_layer_without_partitioning_rbsp');
         result.push('slice_layer_without_partitioning_rbsp');