fix: Fix v1 emsg box start/end times (shaka-project#3198)

The start/end times of emsg event must be in presentation time units.
Fix the bug where times are incorrect when the timestamps do not start at 0.
For emsg v1 the entries in the box are in media time units.
Convert them back to presentation time units by subtracing the offset before raising the event.

Author: duggaraju

lib/media/streaming_engine.js

@@ -1613,7 +1613,7 @@ shaka.media.StreamingEngine = class {
     } else {
       timescale = box.reader.readUint32();
       const pts = box.reader.readUint64();
-      startTime = pts / timescale;
+      startTime = (pts / timescale) + reference.timestampOffset;
       presentationTimeDelta = startTime - reference.startTime;
       eventDuration = box.reader.readUint32();
       id = box.reader.readUint32();

test/media/streaming_engine_unit.js

@@ -120,8 +120,10 @@ describe('StreamingEngine', () => {
     jasmine.clock().mockDate();
   });
 
-  /** @param {boolean=} trickMode */
-  function setupVod(trickMode) {
+  /** @param {boolean=} trickMode
+   * @param {number=} mediaOffset The offset from 0 for the segment start times
+   */
+  function setupVod(trickMode, mediaOffset) {
     // For VOD, we fake a presentation that has 2 Periods of equal duration
     // (20 seconds), where each Period has 1 Variant and 1 text stream.
     //
@@ -132,6 +134,11 @@ describe('StreamingEngine', () => {
     // first Period, and 2 audio, 2 video, and 2 text for the second Period.
     // All media segments are (by default) 10 seconds long.
 
+    const offset = mediaOffset || 0;
+    // timestampOffset is -ve since it is added to bring the timeline to 0.
+    // -0 and 0 are not same so explicitly set to 0.
+    const timestampOffset = offset === 0 ? 0 : -offset;
+
     // Create SegmentData map for FakeMediaSourceEngine.
     const initSegmentSizeAudio = initSegmentRanges[ContentType.AUDIO][1] -
         initSegmentRanges[ContentType.AUDIO][0] + 1;
@@ -151,8 +158,9 @@ describe('StreamingEngine', () => {
           makeBuffer(segmentSizes[ContentType.AUDIO]),
           makeBuffer(segmentSizes[ContentType.AUDIO]),
         ],
-        segmentStartTimes: [0, 10, 20, 30],
+        segmentStartTimes: [offset, offset+10, offset+20, offset+30],
         segmentDuration: 10,
+        timestampOffset: timestampOffset,
       },
       video: {
         initSegments: [
@@ -165,8 +173,9 @@ describe('StreamingEngine', () => {
           makeBuffer(segmentSizes[ContentType.VIDEO]),
           makeBuffer(segmentSizes[ContentType.VIDEO]),
         ],
-        segmentStartTimes: [0, 10, 20, 30],
+        segmentStartTimes: [offset, offset+10, offset+20, offset+30],
         segmentDuration: 10,
+        timestampOffset: timestampOffset,
       },
       text: {
         initSegments: [],
@@ -176,8 +185,9 @@ describe('StreamingEngine', () => {
           makeBuffer(segmentSizes[ContentType.TEXT]),
           makeBuffer(segmentSizes[ContentType.TEXT]),
         ],
-        segmentStartTimes: [0, 10, 20, 30],
+        segmentStartTimes: [offset, offset+10, offset+20, offset+30],
         segmentDuration: 10,
+        timestampOffset: timestampOffset,
       },
     };
     if (trickMode) {
@@ -192,8 +202,9 @@ describe('StreamingEngine', () => {
           makeBuffer(segmentSizes[ContentType.VIDEO]),
           makeBuffer(segmentSizes[ContentType.VIDEO]),
         ],
-        segmentStartTimes: [0, 10, 20, 30],
+        segmentStartTimes: [offset, offset+10, offset+20, offset+30],
         segmentDuration: 10,
+        timestampOffset: timestampOffset,
       };
     }
 
@@ -372,14 +383,24 @@ describe('StreamingEngine', () => {
       video: segmentData[ContentType.VIDEO].segmentDuration,
       text: segmentData[ContentType.TEXT].segmentDuration,
     };
+
+    const timestampOffsets = {
+      audio: segmentData[ContentType.AUDIO].timestampOffset,
+      video: segmentData[ContentType.VIDEO].timestampOffset,
+      text: segmentData[ContentType.TEXT].timestampOffset,
+    };
+
     if (segmentData['trickvideo']) {
       segmentDurations['trickvideo'] =
           segmentData['trickvideo'].segmentDuration;
+      timestampOffsets['trickvideo'] =
+          segmentData['trickvideo'].timestampOffset;
     }
     manifest = shaka.test.StreamingEngineUtil.createManifest(
         /** @type {!shaka.media.PresentationTimeline} */(timeline),
         [firstPeriodStartTime, secondPeriodStartTime],
-        presentationDuration, segmentDurations, initSegmentRanges);
+        presentationDuration, segmentDurations, initSegmentRanges,
+        timestampOffsets);
 
     audioStream = manifest.variants[0].audio;
     videoStream = manifest.variants[0].video;
@@ -2616,7 +2637,7 @@ describe('StreamingEngine', () => {
       expect(event.detail).toEqual(emsgObj);
     });
 
-    it('raises an event for registered embedded v1 emsg boxes ', async () => {
+    it('raises an event for registered embedded v1 emsg boxes', async () => {
       // same event but verison 1.
       const v1EmsgSegment = Uint8ArrayUtils.fromHex(
           '0000003f656d7367010000000000000100000000000000080000ffff00000001' +
@@ -2717,6 +2738,68 @@ describe('StreamingEngine', () => {
     });
   });
 
+  describe('embedded emsg boxes with non zero timestamps', () => {
+    const emsgSegment = Uint8ArrayUtils.fromHex(
+        '0000003b656d736700000000666f6f3a6261723a637573746f6d646174617363' +
+        '68656d6500310000000001000000080000ffff0000000174657374');
+    const emsgObj = {
+      startTime: 8,
+      endTime: 0xffff + 8,
+      schemeIdUri: 'foo:bar:customdatascheme',
+      value: '1',
+      timescale: 1,
+      presentationTimeDelta: 8,
+      eventDuration: 0xffff,
+      id: 1,
+      messageData: new Uint8Array([0x74, 0x65, 0x73, 0x74]),
+    };
+
+    beforeEach(() => {
+      // setup an offset for the timestamps.
+      setupVod(false, 10);
+      mediaSourceEngine = new shaka.test.FakeMediaSourceEngine(segmentData);
+      createStreamingEngine();
+    });
+
+    it('event start matches presentation times for emsg boxes', async () => {
+      segmentData[ContentType.VIDEO].segments[0] = emsgSegment;
+      videoStream.emsgSchemeIdUris = [emsgObj.schemeIdUri];
+
+      // Here we go!
+      streamingEngine.switchVariant(variant);
+      streamingEngine.switchTextStream(textStream);
+      await streamingEngine.start();
+      playing = true;
+      await runTest();
+
+      expect(onEvent).toHaveBeenCalledTimes(1);
+
+      const event = onEvent.calls.argsFor(0)[0];
+      expect(event.detail).toEqual(emsgObj);
+    });
+
+    it('event start matches presentation time for v1 emsg boxes', async () => {
+      // same event but verison 1. start time is 18.
+      const v1EmsgSegment = Uint8ArrayUtils.fromHex(
+          '0000003f656d7367010000000000000100000000000000120000ffff00000001' +
+          '666f6f3a6261723a637573746f6d64617461736368656d6500310074657374');
+      segmentData[ContentType.VIDEO].segments[0] = v1EmsgSegment;
+      videoStream.emsgSchemeIdUris = [emsgObj.schemeIdUri];
+
+      // Here we go!
+      streamingEngine.switchVariant(variant);
+      streamingEngine.switchTextStream(textStream);
+      await streamingEngine.start();
+      playing = true;
+      await runTest();
+
+      expect(onEvent).toHaveBeenCalledTimes(1);
+
+      const event = onEvent.calls.argsFor(0)[0];
+      expect(event.detail).toEqual(emsgObj);
+    });
+  });
+
   describe('network downgrading', () => {
     /** @type {shaka.extern.Variant} */
     let initialVariant;

test/test/util/fake_media_source_engine.js

@@ -52,7 +52,7 @@ shaka.test.FakeMediaSourceEngine = class {
       this.initSegments[type] = data.initSegments.map(() => false);
       this.segments[type] = data.segments.map(() => false);
 
-      this.timestampOffsets_[type] = 0;
+      this.timestampOffsets_[type] = data.timestampOffset || 0;
     }
 
     /** @type {!jasmine.Spy} */
@@ -416,7 +416,8 @@ shaka.test.FakeMediaSourceEngine = class {
  *   initSegments: !Array.<!BufferSource>,
  *   segments: !Array.<!BufferSource>,
  *   segmentStartTimes: !Array.<number>,
- *   segmentDuration: number
+ *   segmentDuration: number,
+ *   timestampOffset: number,
  * }}
  *
  * @property {!Array.<!BufferSource>} initSegments
@@ -429,5 +430,8 @@ shaka.test.FakeMediaSourceEngine = class {
  *   baseMediaDecodeTime (or equivalent) values.
  * @property {number} segmentDuration
  *   The duration of each media segment.
+ * @property {number=} timestampOffset
+ *   The offset to the segment start times that is added to create
+ *   the media timeline.
  */
 shaka.test.FakeMediaSourceEngine.SegmentData;

test/test/util/streaming_engine_util.js

@@ -162,11 +162,13 @@ shaka.test.StreamingEngineUtil = class {
    *   type of segment.
    * @param {!Object.<string, !Array.<number>>} initSegmentRanges The byte
    *   ranges for each type of init segment.
+   * @param {!Object.<string,number>=} timestampOffsets The timestamp offset
+   *  for each type of segment
    * @return {shaka.extern.Manifest}
    */
   static createManifest(
       presentationTimeline, periodStartTimes, presentationDuration,
-      segmentDurations, initSegmentRanges) {
+      segmentDurations, initSegmentRanges, timestampOffsets) {
     const Util = shaka.test.Util;
 
     /**
@@ -263,6 +265,7 @@ shaka.test.StreamingEngineUtil = class {
       const d = segmentDurations[type];
       const getUris = () => [periodIndex + '_' + type + '_' + position];
       const periodStart = periodStartTimes[periodIndex];
+      const timestampOffset = (timestampOffsets && timestampOffsets[type]) || 0;
       const appendWindowStart = periodStartTimes[periodIndex];
       const appendWindowEnd = periodIndex == periodStartTimes.length - 1?
           presentationDuration : periodStartTimes[periodIndex + 1];
@@ -274,7 +277,7 @@ shaka.test.StreamingEngineUtil = class {
           /* startByte= */ 0,
           /* endByte= */ null,
           initSegmentReference,
-          /* timestampOffset= */ 0,
+          timestampOffset,
           appendWindowStart,
           appendWindowEnd);
     };