shaka-player 4.4.0 → 4.4.1

package/lib/util/ts_parser.js

@@ -6,7 +6,6 @@
 
 goog.provide('shaka.util.TsParser');
 
-goog.require('goog.asserts');
 goog.require('shaka.log');
 goog.require('shaka.util.ExpGolomb');
 goog.require('shaka.util.Id3Utils');
@@ -26,9 +25,6 @@ shaka.util.TsParser = class {
     /** @private {boolean} */
     this.pmtParsed_ = false;
 
-    /** @private {?number} */
-    this.videoStartTime_ = null;
-
     /** @private {?number} */
     this.videoPid_ = null;
 
@@ -38,9 +34,6 @@ shaka.util.TsParser = class {
     /** @private {!Array.<Uint8Array>} */
     this.videoData_ = [];
 
-    /** @private {?number} */
-    this.audioStartTime_ = null;
-
     /** @private {?number} */
     this.audioPid_ = null;
 
@@ -63,9 +56,7 @@ shaka.util.TsParser = class {
    * @export
    */
   clearData() {
-    this.videoStartTime_ = null;
     this.videoData_ = [];
-    this.audioStartTime_ = null;
     this.audioData_ = [];
     this.id3Data_ = [];
   }
@@ -78,8 +69,8 @@ shaka.util.TsParser = class {
    * @export
    */
   parse(data) {
-    const timescale = shaka.util.TsParser.Timescale;
     const packetLength = shaka.util.TsParser.PacketLength_;
+    const Uint8ArrayUtils = shaka.util.Uint8ArrayUtils;
 
     // A TS fragment should contain at least 3 TS packets, a PAT, a PMT, and
     // one PID.
@@ -161,33 +152,43 @@ shaka.util.TsParser = class {
           }
           case this.videoPid_: {
             const videoData = data.subarray(offset, start + packetLength);
-            const pes = this.parsePES_(videoData, /* isVideo= */ true);
-            if (pes && pes.pts != null) {
-              const startTime = Math.min(pes.dts, pes.pts) / timescale;
-              if (this.videoStartTime_ == null ||
-                  this.videoStartTime_ > startTime) {
-                this.videoStartTime_ = startTime;
+            if (payloadUnitStartIndicator) {
+              this.videoData_.push(videoData);
+            } else if (this.videoData_.length) {
+              const prevVideoData = this.videoData_[this.videoData_.length - 1];
+              if (prevVideoData) {
+                this.videoData_[this.videoData_.length - 1] =
+                    Uint8ArrayUtils.concat(prevVideoData, videoData);
               }
             }
-            this.videoData_.push(videoData);
             break;
           }
           case this.audioPid_: {
             const audioData = data.subarray(offset, start + packetLength);
-            const pes = this.parsePES_(audioData);
-            if (pes && pes.pts != null) {
-              const startTime = Math.min(pes.dts, pes.pts) / timescale;
-              if (this.audioStartTime_ == null ||
-                  this.audioStartTime_ > startTime) {
-                this.audioStartTime_ = startTime;
+            if (payloadUnitStartIndicator) {
+              this.audioData_.push(audioData);
+            } else if (this.audioData_.length) {
+              const prevAudioData = this.audioData_[this.audioData_.length - 1];
+              if (prevAudioData) {
+                this.audioData_[this.audioData_.length - 1] =
+                    Uint8ArrayUtils.concat(prevAudioData, audioData);
               }
             }
-            this.audioData_.push(audioData);
             break;
           }
-          case this.id3Pid_:
-            this.id3Data_.push(data.subarray(offset, start + packetLength));
+          case this.id3Pid_: {
+            const id3Data = data.subarray(offset, start + packetLength);
+            if (payloadUnitStartIndicator) {
+              this.id3Data_.push(id3Data);
+            } else if (this.id3Data_.length) {
+              const prevId3Data = this.id3Data_[this.id3Data_.length - 1];
+              if (prevId3Data) {
+                this.id3Data_[this.id3Data_.length - 1] =
+                    Uint8ArrayUtils.concat(prevId3Data, id3Data);
+              }
+            }
             break;
+          }
           default:
             unknownPIDs = true;
             break;
@@ -340,11 +341,10 @@ shaka.util.TsParser = class {
    * Parse PES
    *
    * @param {Uint8Array} data
-   * @param {boolean=} isVideo
    * @return {?shaka.extern.MPEG_PES}
    * @private
    */
-  parsePES_(data, isVideo = false) {
+  parsePES_(data) {
     const startPrefix = (data[0] << 16) | (data[1] << 8) | data[2];
     // In certain live streams, the start of a TS fragment has ts packets
     // that are frame data that is continuing from the previous fragment. This
@@ -361,12 +361,6 @@ shaka.util.TsParser = class {
       dts: null,
     };
 
-    // If it is video, we expect the packet length to be 0. When the PES packet
-    // length is set to zero, the PES packet can be any length.
-    if (isVideo && pes.packetLength != 0) {
-      return null;
-    }
-
     // PES packets may be annotated with a PTS value, or a PTS value
     // and a DTS value. Determine what combination of values is
     // available to work with.
@@ -454,13 +448,12 @@ shaka.util.TsParser = class {
       } else if (numZeros >= 2 && value == 1 && tryToFinishLastNalu) {
         // If we are scanning the next PES, we need append the data to the
         // previous Nalu and don't scan for more nalus.
-        if (i <= 2) {
-          break;
-        }
+        const startCodeSize = numZeros > 3 ? 3 : numZeros;
+        const lastByteToKeep = i - startCodeSize;
         infoOfLastNalu.data = shaka.util.Uint8ArrayUtils.concat(
-            infoOfLastNalu.data, data.subarray(0, i - 2));
+            infoOfLastNalu.data, data.subarray(0, lastByteToKeep));
         infoOfLastNalu.fullData = shaka.util.Uint8ArrayUtils.concat(
-            infoOfLastNalu.fullData, data.subarray(0, i - 2));
+            infoOfLastNalu.fullData, data.subarray(0, lastByteToKeep));
         break;
       } else if (numZeros >= 2 && value == 1) {
         // We just read a start code.  Consume the NALU we passed, if any.
@@ -519,7 +512,7 @@ shaka.util.TsParser = class {
             type: lastNaluType,
             time: time,
           };
-          if (nextPes) {
+          if (nextPes && pes.packetLength == 0) {
             data = nextPes.data;
             len = data.byteLength;
             i = -1;
@@ -542,30 +535,17 @@ shaka.util.TsParser = class {
    */
   getMetadata() {
     const timescale = shaka.util.TsParser.Timescale;
-    const Uint8ArrayUtils = shaka.util.Uint8ArrayUtils;
     const metadata = [];
-    let prevId3Data = new Uint8Array(0);
-    // parsePES_() only works if the data begins on a PES boundary.
-    // Try the last data blob first, and if it doesn't begin on a
-    // PES boundary, prepend the previous blob and try again.
-    // This way, a successful parse will always begin and end on
-    // the correct boundary, and no data will be skipped.
-    for (let i = this.id3Data_.length - 1; i >= 0; i--) {
-      const data = this.id3Data_[i];
-      goog.asserts.assert(data, 'We should have a data');
-      const id3Data = Uint8ArrayUtils.concat(data, prevId3Data);
+    for (const id3Data of this.id3Data_) {
       const pes = this.parsePES_(id3Data);
       if (pes) {
-        metadata.unshift({
+        metadata.push({
           cueTime: pes.pts ? pes.pts / timescale : null,
           data: pes.data,
           frames: shaka.util.Id3Utils.getID3Frames(pes.data),
           dts: pes.dts,
           pts: pes.pts,
         });
-        prevId3Data = new Uint8Array(0);
-      } else {
-        prevId3Data = id3Data;
       }
     }
     return metadata;
@@ -578,24 +558,11 @@ shaka.util.TsParser = class {
    * @export
    */
   getAudioData() {
-    const Uint8ArrayUtils = shaka.util.Uint8ArrayUtils;
     const audio = [];
-    let prevAudioData = new Uint8Array(0);
-    // parsePES_() only works if the data begins on a PES boundary.
-    // Try the last data blob first, and if it doesn't begin on a
-    // PES boundary, prepend the previous blob and try again.
-    // This way, a successful parse will always begin and end on
-    // the correct boundary, and no data will be skipped.
-    for (let i = this.audioData_.length - 1; i >= 0; i--) {
-      const data = this.audioData_[i];
-      goog.asserts.assert(data, 'We should have a data');
-      const audioData = Uint8ArrayUtils.concat(data, prevAudioData);
+    for (const audioData of this.audioData_) {
       const pes = this.parsePES_(audioData);
       if (pes) {
-        audio.unshift(pes);
-        prevAudioData = new Uint8Array(0);
-      } else {
-        prevAudioData = audioData;
+        audio.push(pes);
       }
     }
     return audio;
@@ -608,24 +575,11 @@ shaka.util.TsParser = class {
    * @export
    */
   getVideoData() {
-    const Uint8ArrayUtils = shaka.util.Uint8ArrayUtils;
     const video = [];
-    let prevVideoData = new Uint8Array(0);
-    // parsePES_() only works if the data begins on a PES boundary.
-    // Try the last data blob first, and if it doesn't begin on a
-    // PES boundary, prepend the previous blob and try again.
-    // This way, a successful parse will always begin and end on
-    // the correct boundary, and no data will be skipped.
-    for (let i = this.videoData_.length - 1; i >= 0; i--) {
-      const data = this.videoData_[i];
-      goog.asserts.assert(data, 'We should have a data');
-      const videoData = Uint8ArrayUtils.concat(data, prevVideoData);
-      const pes = this.parsePES_(videoData, /* isVideo= */ true);
+    for (const videoData of this.videoData_) {
+      const pes = this.parsePES_(videoData);
       if (pes) {
-        video.unshift(pes);
-        prevVideoData = new Uint8Array(0);
-      } else {
-        prevVideoData = videoData;
+        video.push(pes);
       }
     }
     return video;
@@ -638,9 +592,28 @@ shaka.util.TsParser = class {
    * @export
    */
   getStartTime() {
+    const timescale = shaka.util.TsParser.Timescale;
+    let audioStartTime = null;
+    for (const pes of this.getAudioData()) {
+      if (pes && pes.pts != null) {
+        const startTime = Math.min(pes.dts, pes.pts) / timescale;
+        if (audioStartTime == null || audioStartTime > startTime) {
+          audioStartTime = startTime;
+        }
+      }
+    }
+    let videoStartTime = null;
+    for (const pes of this.getVideoData()) {
+      if (pes && pes.pts != null) {
+        const startTime = Math.min(pes.dts, pes.pts) / timescale;
+        if (videoStartTime == null || videoStartTime > startTime) {
+          videoStartTime = startTime;
+        }
+      }
+    }
     return {
-      audio: this.audioStartTime_,
-      video: this.videoStartTime_,
+      audio: audioStartTime,
+      video: videoStartTime,
     };
   }
 
@@ -664,30 +637,20 @@ shaka.util.TsParser = class {
    * @export
    */
   getVideoNalus() {
-    const Uint8ArrayUtils = shaka.util.Uint8ArrayUtils;
-    let nalus = [];
-    let prevVideoData = new Uint8Array(0);
-    // parsePES_() only works if the data begins on a PES boundary.
-    // Try the last data blob first, and if it doesn't begin on a
-    // PES boundary, prepend the previous blob and try again.
-    // This way, a successful parse will always begin and end on
-    // the correct boundary, and no data will be skipped.
-    for (let i = this.videoData_.length - 1; i >= 0; i--) {
-      const data = this.videoData_[i];
-      goog.asserts.assert(data, 'We should have a data');
-      const videoData = Uint8ArrayUtils.concat(data, prevVideoData);
-      const pes = this.parsePES_(videoData);
-      if (pes) {
-        if (this.videoCodec_ == 'avc') {
-          nalus = nalus.concat(this.parseAvcNalus(pes));
-        }
-        prevVideoData = new Uint8Array(0);
-      } else {
-        prevVideoData = videoData;
+    const nalus = [];
+    if (this.videoCodec_ != 'avc') {
+      return nalus;
+    }
+    const videoData = this.getVideoData();
+    for (let i = 0; i < videoData.length; i++) {
+      const pes = videoData[i];
+      let nextPes;
+      if (i + 1 < videoData.length) {
+        nextPes = videoData[i + 1];
       }
+      nalus.push(...this.parseAvcNalus(pes, nextPes));
     }
-    // We need to invert the array to return it in the correct order.
-    return nalus.reverse();
+    return nalus;
   }
 
   /**

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "shaka-player",
   "description": "DASH/EME video player library",
-  "version": "4.4.0",
+  "version": "4.4.1",
   "homepage": "https://github.com/shaka-project/shaka-player",
   "author": "Google",
   "maintainers": [

package/test/hls/hls_parser_unit.js

@@ -4180,19 +4180,11 @@ describe('HlsParser', () => {
     expect(videoSegments2[2].endTime).toBe(10);
   });
 
-  // Issue #1875
-  it('ignores audio groups on audio-only content', async () => {
-    // NOTE: To reproduce the original issue accurately, the two audio playlist
-    // URIs must differ.  When the issue occurred, the audio-only variant would
-    // be detected as a video stream and combined with the audio group, leading
-    // the player to buffer "video" that was really audio, resulting in
-    // audio-only playback to the exclusion of any other streams.  Since the
-    // root cause of that was the mis-detection, this repro case does not need
-    // to include any audio+video variants.
+  it('allow audio groups on audio-only content', async () => {
     const master = [
       '#EXTM3U\n',
-      '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",LANG="en",URI="audio1"\n',
-      '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",LANG="eo",URI="audio2"\n',
+      '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",LANGUAGE="en",URI="audio1"\n',
+      '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud",LANGUAGE="eo",URI="audio2"\n',
       '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="mp4a",AUDIO="aud"\n',
       'audio3\n',
     ].join('');
@@ -4210,7 +4202,14 @@ describe('HlsParser', () => {
       manifest.anyTimeline();
       manifest.addPartialVariant((variant) => {
         variant.bandwidth = 200;
-        variant.language = 'und';
+        variant.language = 'en';
+        variant.addPartialStream(ContentType.AUDIO, (stream) => {
+          stream.mime('audio/mp4', 'mp4a');
+        });
+      });
+      manifest.addPartialVariant((variant) => {
+        variant.bandwidth = 200;
+        variant.language = 'eo';
         variant.addPartialStream(ContentType.AUDIO, (stream) => {
           stream.mime('audio/mp4', 'mp4a');
         });
@@ -4229,7 +4228,7 @@ describe('HlsParser', () => {
 
     const actual = await parser.start('test:/master', playerInterface);
     await loadAllStreamsFor(actual);
-    expect(actual.variants.length).toBe(1);
+    expect(actual.variants.length).toBe(2);
     expect(actual).toEqual(manifest);
   });
 

package/test/media/media_source_engine_integration.js

@@ -656,7 +656,7 @@ describe('MediaSourceEngine', () => {
   it('extracts ID3 metadata from AAC', async () => {
     if (!MediaSource.isTypeSupported('audio/aac') ||
         !shaka.util.Platform.supportsSequenceMode()) {
-      return;
+      pending('Raw AAC codec is not supported by the platform.');
     }
     metadata = shaka.test.TestScheme.DATA['id3-metadata_aac'];
     generators = shaka.test.TestScheme.GENERATORS['id3-metadata_aac'];
@@ -672,7 +672,7 @@ describe('MediaSourceEngine', () => {
 
   it('extracts ID3 metadata from AAC when transmuxing', async () => {
     if (!MediaSource.isTypeSupported('audio/aac')) {
-      return;
+      pending('Raw AAC codec is not supported by the platform.');
     }
     metadata = shaka.test.TestScheme.DATA['id3-metadata_aac'];
     generators = shaka.test.TestScheme.GENERATORS['id3-metadata_aac'];

package/test/mss/mss_player_integration.js

@@ -94,7 +94,7 @@ describe('MSS Player', () => {
   it('MSS VoD PlayReady', async () => {
     const support = await shaka.media.DrmEngine.probeSupport();
     if (!support['com.microsoft.playready']) {
-      return;
+      pending('PlayReady is not supported by the platform.');
     }
 
     player.configure({