fix: allow audio fmp4 emsg probe (#1385)

* fix: allow audio fmp4 emsg probe

* add audio only test

src/media-segment-request.js

@@ -497,38 +497,37 @@ const handleSegmentBytes = ({
         if (trackInfo.hasVideo) {
           timingInfoFn(segment, 'video', 'start', startTime);
         }
-
-        // Run through the CaptionParser in case there are captions.
-        // Initialize CaptionParser if it hasn't been yet
-        if (!tracks.video || !data.byteLength || !segment.transmuxer) {
-          finishLoading();
-          return;
-        }
-
         workerCallback({
-          action: 'pushMp4Captions',
-          endAction: 'mp4Captions',
-          transmuxer: segment.transmuxer,
+          action: 'probeEmsgID3',
           data: bytesAsUint8Array,
-          timescales: segment.map.timescales,
-          trackIds: [tracks.video.id],
-          callback: (message) => {
+          transmuxer: segment.transmuxer,
+          offset: startTime,
+          callback: ({emsgData, id3Frames}) => {
             // transfer bytes back to us
-            bytes = message.data.buffer;
-            segment.bytes = bytesAsUint8Array = message.data;
-            message.logs.forEach(function(log) {
-              onTransmuxerLog(merge(log, {stream: 'mp4CaptionParser'}));
-            });
+            bytes = emsgData.buffer;
+            segment.bytes = bytesAsUint8Array = emsgData;
+
+            // Run through the CaptionParser in case there are captions.
+            // Initialize CaptionParser if it hasn't been yet
+            if (!tracks.video || !data.byteLength || !segment.transmuxer) {
+              finishLoading(undefined, id3Frames);
+              return;
+            }
 
             workerCallback({
-              action: 'probeEmsgID3',
-              data: bytesAsUint8Array,
+              action: 'pushMp4Captions',
+              endAction: 'mp4Captions',
               transmuxer: segment.transmuxer,
-              offset: startTime,
-              callback: ({emsgData, id3Frames}) => {
+              data: bytesAsUint8Array,
+              timescales: segment.map.timescales,
+              trackIds: [tracks.video.id],
+              callback: (message) => {
                 // transfer bytes back to us
-                bytes = emsgData.buffer;
-                segment.bytes = bytesAsUint8Array = emsgData;
+                bytes = message.data.buffer;
+                segment.bytes = bytesAsUint8Array = message.data;
+                message.logs.forEach(function(log) {
+                  onTransmuxerLog(merge(log, {stream: 'mp4CaptionParser'}));
+                });
                 finishLoading(message.captions, id3Frames);
               }
             });

test/media-segment-request.test.js

@@ -15,6 +15,8 @@ import {
   ac3WithoutId3 as ac3WithoutId3Segment,
   video as videoSegment,
   audio as audioSegment,
+  mp4Audio,
+  mp4AudioInit,
   mp4Video,
   mp4VideoInit,
   muxed as muxedSegment,
@@ -1576,7 +1578,7 @@ QUnit.test('non-TS segment will get parsed for captions on next segment request
   this.standardXHRResponse(initReq, mp4VideoInit());
 });
 
-QUnit.test('can get emsg ID3 frames from fmp4 segment', function(assert) {
+QUnit.test('can get emsg ID3 frames from fmp4 video segment', function(assert) {
   const done = assert.async();
   let gotEmsgId3 = 0;
   let gotData = 0;
@@ -1705,3 +1707,133 @@ QUnit.test('can get emsg ID3 frames from fmp4 segment', function(assert) {
   // Simulate receiving the init segment after the media
   this.standardXHRResponse(initReq, mp4VideoInit());
 });
+
+QUnit.test('can get emsg ID3 frames from fmp4 audio segment', function(assert) {
+  const done = assert.async();
+  let gotEmsgId3 = 0;
+  let gotData = 0;
+  // expected frame data
+  const id3Frames = [{
+    cueTime: 1,
+    duration: 0,
+    frames: [{
+      id: 'TXXX',
+      description: 'foo bar',
+      data: { key: 'value' }
+    },
+    {
+      id: 'PRIV',
+      owner: '[email protected]',
+      // 'foo'
+      data: new Uint8Array([0x66, 0x6F, 0x6F])
+    }]
+  },
+  {
+    cueTime: 3,
+    duration: 0,
+    frames: [{
+      id: 'PRIV',
+      owner: '[email protected]',
+      // 'bar'
+      data: new Uint8Array([0x62, 0x61, 0x72])
+    },
+    {
+      id: 'TXXX',
+      description: 'bar foo',
+      data: { key: 'value' }
+    }]
+  }];
+  const transmuxer = new videojs.EventTarget();
+
+  transmuxer.postMessage = (event) => {
+    if (event.action === 'pushMp4Captions') {
+      transmuxer.trigger({
+        type: 'message',
+        data: {
+          action: 'mp4Captions',
+          data: event.data,
+          captions: 'foo bar',
+          logs: []
+        }
+      });
+    }
+
+    if (event.action === 'probeMp4StartTime') {
+      transmuxer.trigger({
+        type: 'message',
+        data: {
+          action: 'probeMp4StartTime',
+          data: event.data,
+          timingInfo: {}
+        }
+      });
+    }
+
+    if (event.action === 'probeMp4Tracks') {
+      transmuxer.trigger({
+        type: 'message',
+        data: {
+          action: 'probeMp4Tracks',
+          data: event.data,
+          tracks: [{type: 'audio', codec: 'mp4a.40.2'}]
+        }
+      });
+    }
+
+    if (event.action === 'probeEmsgID3') {
+      transmuxer.trigger({
+        type: 'message',
+        data: {
+          action: 'probeEmsgID3',
+          emsgData: event.data,
+          id3Frames
+        }
+      });
+    }
+  };
+
+  mediaSegmentRequest({
+    xhr: this.xhr,
+    xhrOptions: this.xhrOptions,
+    decryptionWorker: this.mockDecrypter,
+    segment: {
+      transmuxer,
+      resolvedUri: 'mp4Audio.mp4',
+      map: {
+        resolvedUri: 'mp4AudioInit.mp4'
+      }
+    },
+    progressFn: this.noop,
+    trackInfoFn: this.noop,
+    timingInfoFn: this.noop,
+    id3Fn: (segment, _id3Frames) => {
+      gotEmsgId3++;
+      assert.deepEqual(_id3Frames, id3Frames, 'got expected emsg id3 data.');
+    },
+    captionsFn: this.noop,
+    dataFn: (segment, segmentData) => {
+      gotData++;
+      assert.ok(segmentData, 'init segment bytes in map');
+      assert.ok(segment.map.tracks, 'added tracks');
+      assert.ok(segment.map.tracks.audio, 'added audio track');
+    },
+    doneFn: () => {
+      assert.equal(gotEmsgId3, 1, 'received emsg ID3 event');
+      assert.equal(gotData, 1, 'received data event');
+      transmuxer.off();
+      done();
+    }
+  });
+  assert.equal(this.requests.length, 2, 'there are two requests');
+
+  const initReq = this.requests.shift();
+  const segmentReq = this.requests.shift();
+
+  assert.equal(initReq.uri, 'mp4AudioInit.mp4', 'the first request is for the init segment');
+  assert.equal(segmentReq.uri, 'mp4Audio.mp4', 'the second request is for a segment');
+
+  // Simulate receiving the media first
+  this.standardXHRResponse(segmentReq, mp4Audio());
+  // Simulate receiving the init segment after the media
+  this.standardXHRResponse(initReq, mp4AudioInit());
+});