Parse QuickTime variant audio sample entries.

Issue: #958

library/src/androidTest/java/com/google/android/exoplayer/extractor/mp4/Mp4ExtractorTest.java

@@ -70,6 +70,9 @@ public final class Mp4ExtractorTest extends TestCase {
   private static final byte[] AUDIO_MDHD_PAYLOAD =
       getByteArray("00000000cf6c4889cf6c488a0000ac4400a3e40055c40000");
 
+  /** String of hexadecimal bytes for an ftyp payload with major_brand mp41 and minor_version 0. **/
+  private static final byte[] FTYP_PAYLOAD = getByteArray("6d70343100000000");
+
   /** String of hexadecimal bytes containing an mvhd payload from an AVC/AAC video. */
   private static final byte[] MVHD_PAYLOAD = getByteArray(
       "00000000cf6c4888cf6c48880000025800023ad40001000001000000000000000000000000010000000000"
@@ -88,7 +91,7 @@ public final class Mp4ExtractorTest extends TestCase {
   /** Indices of key-frames. */
   private static final boolean[] SAMPLE_IS_SYNC = {true, false, false, false, true, true};
   /** Indices of video frame chunk offsets. */
-  private static final int[] CHUNK_OFFSETS = {1200, 2120, 3120, 4120};
+  private static final int[] CHUNK_OFFSETS = {1208, 2128, 3128, 4128};
   /** Numbers of video frames in each chunk. */
   private static final int[] SAMPLES_IN_CHUNK = {2, 2, 1, 1};
   /** The mdat box must be large enough to avoid reading chunk sample data out of bounds. */
@@ -368,7 +371,7 @@ public final class Mp4ExtractorTest extends TestCase {
   /** Gets a valid MP4 file with audio/video tracks and synchronization data. */
   private static byte[] getTestMp4File(boolean mp4vFormat) {
     return Mp4Atom.serialize(
-        atom(Atom.TYPE_ftyp, EMPTY),
+        atom(Atom.TYPE_ftyp, FTYP_PAYLOAD),
         atom(Atom.TYPE_moov,
             atom(Atom.TYPE_mvhd, MVHD_PAYLOAD),
             atom(Atom.TYPE_trak,
@@ -400,13 +403,13 @@ public final class Mp4ExtractorTest extends TestCase {
                             atom(Atom.TYPE_stsc, getStsc()),
                             atom(Atom.TYPE_stsz, getStsz()),
                             atom(Atom.TYPE_stco, getStco())))))),
-        atom(Atom.TYPE_mdat, getMdat(mp4vFormat ? 1168 : 1158, !mp4vFormat)));
+        atom(Atom.TYPE_mdat, getMdat(mp4vFormat ? 1176 : 1166, !mp4vFormat)));
   }
 
   /** Gets a valid MP4 file with audio/video tracks and without a synchronization table. */
   private static byte[] getTestMp4FileWithoutSynchronizationData(boolean mp4vFormat) {
     return Mp4Atom.serialize(
-        atom(Atom.TYPE_ftyp, EMPTY),
+        atom(Atom.TYPE_ftyp, FTYP_PAYLOAD),
         atom(Atom.TYPE_moov,
             atom(Atom.TYPE_mvhd, MVHD_PAYLOAD),
             atom(Atom.TYPE_trak,
@@ -436,7 +439,7 @@ public final class Mp4ExtractorTest extends TestCase {
                             atom(Atom.TYPE_stsc, getStsc()),
                             atom(Atom.TYPE_stsz, getStsz()),
                             atom(Atom.TYPE_stco, getStco())))))),
-        atom(Atom.TYPE_mdat, getMdat(mp4vFormat ? 1112 : 1102, !mp4vFormat)));
+        atom(Atom.TYPE_mdat, getMdat(mp4vFormat ? 1120 : 1110, !mp4vFormat)));
   }
 
   private static Mp4Atom atom(int type, Mp4Atom... containedMp4Atoms) {

library/src/main/java/com/google/android/exoplayer/extractor/mp4/Atom.java

@@ -53,6 +53,7 @@ import java.util.List;
   public static final int TYPE_d263 = Util.getIntegerCodeForString("d263");
   public static final int TYPE_mdat = Util.getIntegerCodeForString("mdat");
   public static final int TYPE_mp4a = Util.getIntegerCodeForString("mp4a");
+  public static final int TYPE_wave = Util.getIntegerCodeForString("wave");
   public static final int TYPE_ac_3 = Util.getIntegerCodeForString("ac-3");
   public static final int TYPE_dac3 = Util.getIntegerCodeForString("dac3");
   public static final int TYPE_ec_3 = Util.getIntegerCodeForString("ec-3");

library/src/main/java/com/google/android/exoplayer/extractor/mp4/AtomParsers.java

@@ -42,9 +42,10 @@ import java.util.List;
    *
    * @param trak Atom to parse.
    * @param mvhd Movie header atom, used to get the timescale.
+   * @param isQuickTime True for QuickTime media. False otherwise.
    * @return A {@link Track} instance, or {@code null} if the track's type isn't supported.
    */
-  public static Track parseTrak(Atom.ContainerAtom trak, Atom.LeafAtom mvhd) {
+  public static Track parseTrak(Atom.ContainerAtom trak, Atom.LeafAtom mvhd, boolean isQuickTime) {
     Atom.ContainerAtom mdia = trak.getContainerAtomOfType(Atom.TYPE_mdia);
     int trackType = parseHdlr(mdia.getLeafAtomOfType(Atom.TYPE_hdlr).data);
     if (trackType != Track.TYPE_soun && trackType != Track.TYPE_vide && trackType != Track.TYPE_text
@@ -66,7 +67,7 @@ import java.util.List;
 
     Pair<Long, String> mdhdData = parseMdhd(mdia.getLeafAtomOfType(Atom.TYPE_mdhd).data);
     StsdData stsdData = parseStsd(stbl.getLeafAtomOfType(Atom.TYPE_stsd).data, tkhdData.id,
-        durationUs, tkhdData.rotationDegrees, mdhdData.second);
+        durationUs, tkhdData.rotationDegrees, mdhdData.second, isQuickTime);
     Pair<long[], long[]> edtsData = parseEdts(trak.getContainerAtomOfType(Atom.TYPE_edts));
     return stsdData.mediaFormat == null ? null
         : new Track(tkhdData.id, trackType, mdhdData.first, movieTimescale, durationUs,
@@ -429,10 +430,11 @@ import java.util.List;
    * @param durationUs The duration of the track in microseconds.
    * @param rotationDegrees The rotation of the track in degrees.
    * @param language The language of the track.
+   * @param isQuickTime True for QuickTime media. False otherwise.
    * @return An object containing the parsed data.
    */
   private static StsdData parseStsd(ParsableByteArray stsd, int trackId, long durationUs,
-      int rotationDegrees, String language) {
+      int rotationDegrees, String language, boolean isQuickTime) {
     stsd.setPosition(Atom.FULL_HEADER_SIZE);
     int numberOfEntries = stsd.readInt();
     StsdData out = new StsdData(numberOfEntries);
@@ -452,7 +454,7 @@ import java.util.List;
           || childAtomType == Atom.TYPE_dtsc || childAtomType == Atom.TYPE_dtse
           || childAtomType == Atom.TYPE_dtsh || childAtomType == Atom.TYPE_dtsl) {
         parseAudioSampleEntry(stsd, childAtomType, childStartPosition, childAtomSize, trackId,
-            durationUs, language, out, i);
+            durationUs, language, isQuickTime, out, i);
       } else if (childAtomType == Atom.TYPE_TTML) {
         out.mediaFormat = MediaFormat.createTextFormat(Integer.toString(trackId),
             MimeTypes.APPLICATION_TTML, MediaFormat.NO_VALUE, durationUs, language);
@@ -695,14 +697,31 @@ import java.util.List;
   }
 
   private static void parseAudioSampleEntry(ParsableByteArray parent, int atomType, int position,
-      int size, int trackId, long durationUs, String language, StsdData out, int entryIndex) {
+      int size, int trackId, long durationUs, String language, boolean isQuickTime, StsdData out,
+      int entryIndex) {
     parent.setPosition(position + Atom.HEADER_SIZE);
-    parent.skipBytes(16);
+
+    int quickTimeSoundDescriptionVersion = 0;
+    if (isQuickTime) {
+      parent.skipBytes(8);
+      quickTimeSoundDescriptionVersion = parent.readUnsignedShort();
+      parent.skipBytes(6);
+    } else {
+      parent.skipBytes(16);
+    }
+
     int channelCount = parent.readUnsignedShort();
     int sampleSize = parent.readUnsignedShort();
     parent.skipBytes(4);
     int sampleRate = parent.readUnsignedFixedPoint1616();
 
+    if (quickTimeSoundDescriptionVersion > 0) {
+      parent.skipBytes(16);
+      if (quickTimeSoundDescriptionVersion == 2) {
+        parent.skipBytes(20);
+      }
+    }
+
     // If the atom type determines a MIME type, set it immediately.
     String mimeType = null;
     if (atomType == Atom.TYPE_ac_3) {
@@ -716,17 +735,22 @@ import java.util.List;
     }
 
     byte[] initializationData = null;
-    int childPosition = parent.getPosition();
-    while (childPosition - position < size) {
-      parent.setPosition(childPosition);
-      int childStartPosition = parent.getPosition();
+    int childAtomPosition = parent.getPosition();
+    while (childAtomPosition - position < size) {
+      parent.setPosition(childAtomPosition);
       int childAtomSize = parent.readInt();
       Assertions.checkArgument(childAtomSize > 0, "childAtomSize should be positive");
       int childAtomType = parent.readInt();
       if (atomType == Atom.TYPE_mp4a || atomType == Atom.TYPE_enca) {
+        int esdsAtomPosition = -1;
         if (childAtomType == Atom.TYPE_esds) {
+          esdsAtomPosition = childAtomPosition;
+        } else if (isQuickTime && childAtomType == Atom.TYPE_wave) {
+          esdsAtomPosition = findEsdsPosition(parent, childAtomPosition, childAtomSize);
+        }
+        if (esdsAtomPosition != -1) {
           Pair<String, byte[]> mimeTypeAndInitializationData =
-              parseEsdsFromParent(parent, childStartPosition);
+              parseEsdsFromParent(parent, esdsAtomPosition);
           mimeType = mimeTypeAndInitializationData.first;
           initializationData = mimeTypeAndInitializationData.second;
           if (MimeTypes.AUDIO_AAC.equals(mimeType)) {
@@ -738,18 +762,18 @@ import java.util.List;
             channelCount = audioSpecificConfig.second;
           }
         } else if (childAtomType == Atom.TYPE_sinf) {
-          out.trackEncryptionBoxes[entryIndex] = parseSinfFromParent(parent, childStartPosition,
+          out.trackEncryptionBoxes[entryIndex] = parseSinfFromParent(parent, childAtomPosition,
               childAtomSize);
         }
       } else if (atomType == Atom.TYPE_ac_3 && childAtomType == Atom.TYPE_dac3) {
         // TODO: Choose the right AC-3 track based on the contents of dac3/dec3.
         // TODO: Add support for encryption (by setting out.trackEncryptionBoxes).
-        parent.setPosition(Atom.HEADER_SIZE + childStartPosition);
+        parent.setPosition(Atom.HEADER_SIZE + childAtomPosition);
         out.mediaFormat = Ac3Util.parseAnnexFAc3Format(parent, Integer.toString(trackId),
             durationUs, language);
         return;
       } else if (atomType == Atom.TYPE_ec_3 && childAtomType == Atom.TYPE_dec3) {
-        parent.setPosition(Atom.HEADER_SIZE + childStartPosition);
+        parent.setPosition(Atom.HEADER_SIZE + childAtomPosition);
         out.mediaFormat = Ac3Util.parseAnnexFEAc3Format(parent, Integer.toString(trackId),
             durationUs, language);
         return;
@@ -761,7 +785,7 @@ import java.util.List;
             language);
         return;
       }
-      childPosition += childAtomSize;
+      childAtomPosition += childAtomSize;
     }
 
     // If the media type was not recognized, ignore the track.
@@ -775,6 +799,22 @@ import java.util.List;
         language);
   }
 
+  /** Returns the position of the esds box within a parent, or -1 if no esds box is found */
+  private static int findEsdsPosition(ParsableByteArray parent, int position, int size) {
+    int childAtomPosition = parent.getPosition();
+    while (childAtomPosition - position < size) {
+      parent.setPosition(childAtomPosition);
+      int childAtomSize = parent.readInt();
+      Assertions.checkArgument(childAtomSize > 0, "childAtomSize should be positive");
+      int childType = parent.readInt();
+      if (childType == Atom.TYPE_esds) {
+        return childAtomPosition;
+      }
+      childAtomPosition += childAtomSize;
+    }
+    return -1;
+  }
+
   /** Returns codec-specific initialization data contained in an esds box. */
   private static Pair<String, byte[]> parseEsdsFromParent(ParsableByteArray parent, int position) {
     parent.setPosition(position + Atom.HEADER_SIZE + 4);

library/src/main/java/com/google/android/exoplayer/extractor/mp4/FragmentedMp4Extractor.java

@@ -305,7 +305,7 @@ public final class FragmentedMp4Extractor implements Extractor {
     ContainerAtom mvex = moov.getContainerAtomOfType(Atom.TYPE_mvex);
     extendsDefaults = parseTrex(mvex.getLeafAtomOfType(Atom.TYPE_trex).data);
     track = AtomParsers.parseTrak(moov.getContainerAtomOfType(Atom.TYPE_trak),
-        moov.getLeafAtomOfType(Atom.TYPE_mvhd));
+        moov.getLeafAtomOfType(Atom.TYPE_mvhd), false);
     checkState(track != null);
     trackOutput.format(track.mediaFormat);
   }

library/src/main/java/com/google/android/exoplayer/extractor/mp4/Mp4Extractor.java

@@ -25,6 +25,7 @@ import com.google.android.exoplayer.extractor.mp4.Atom.ContainerAtom;
 import com.google.android.exoplayer.util.Assertions;
 import com.google.android.exoplayer.util.NalUnitUtil;
 import com.google.android.exoplayer.util.ParsableByteArray;
+import com.google.android.exoplayer.util.Util;
 
 import java.io.IOException;
 import java.util.ArrayList;
@@ -42,6 +43,9 @@ public final class Mp4Extractor implements Extractor, SeekMap {
   private static final int STATE_READING_ATOM_PAYLOAD = 2;
   private static final int STATE_READING_SAMPLE = 3;
 
+  // Brand stored in the ftyp atom for QuickTime media.
+  private static final int BRAND_QUICKTIME = Util.getIntegerCodeForString("qt  ");
+
   /**
    * When seeking within the source, if the offset is greater than or equal to this value (or the
    * offset is negative), the source will be reloaded.
@@ -68,6 +72,7 @@ public final class Mp4Extractor implements Extractor, SeekMap {
   // Extractor outputs.
   private ExtractorOutput extractorOutput;
   private Mp4Track[] tracks;
+  private boolean isQuickTime;
 
   public Mp4Extractor() {
     atomHeader = new ParsableByteArray(Atom.LONG_HEADER_SIZE);
@@ -210,7 +215,9 @@ public final class Mp4Extractor implements Extractor, SeekMap {
     boolean seekRequired = false;
     if (atomData != null) {
       input.readFully(atomData.data, atomHeaderBytesRead, (int) atomPayloadSize);
-      if (!containerAtoms.isEmpty()) {
+      if (atomType == Atom.TYPE_ftyp) {
+        isQuickTime = processFtypAtom(atomData);
+      } else if (!containerAtoms.isEmpty()) {
         containerAtoms.peek().add(new Atom.LeafAtom(atomType, atomData));
       }
     } else {
@@ -240,6 +247,27 @@ public final class Mp4Extractor implements Extractor, SeekMap {
     return seekRequired;
   }
 
+  /**
+   * Process an ftyp atom to determine whether the media is QuickTime.
+   *
+   * @param atomData The ftyp atom data.
+   * @return True if the media is QuickTime. False otherwise.
+   */
+  private static boolean processFtypAtom(ParsableByteArray atomData) {
+    atomData.setPosition(Atom.HEADER_SIZE);
+    int majorBrand = atomData.readInt();
+    if (majorBrand == BRAND_QUICKTIME) {
+      return true;
+    }
+    atomData.skipBytes(4); // minor_version
+    while (atomData.bytesLeft() > 0) {
+      if (atomData.readInt() == BRAND_QUICKTIME) {
+        return true;
+      }
+    }
+    return false;
+  }
+
   /** Updates the stored track metadata to reflect the contents of the specified moov atom. */
   private void processMoovAtom(ContainerAtom moov) {
     List<Mp4Track> tracks = new ArrayList<>();
@@ -250,7 +278,8 @@ public final class Mp4Extractor implements Extractor, SeekMap {
         continue;
       }
 
-      Track track = AtomParsers.parseTrak(atom, moov.getLeafAtomOfType(Atom.TYPE_mvhd));
+      Track track = AtomParsers.parseTrak(atom, moov.getLeafAtomOfType(Atom.TYPE_mvhd),
+          isQuickTime);
       if (track == null) {
         continue;
       }
@@ -387,7 +416,7 @@ public final class Mp4Extractor implements Extractor, SeekMap {
         || atom == Atom.TYPE_stsd || atom == Atom.TYPE_stts || atom == Atom.TYPE_stss
         || atom == Atom.TYPE_ctts || atom == Atom.TYPE_elst || atom == Atom.TYPE_stsc
         || atom == Atom.TYPE_stsz || atom == Atom.TYPE_stco || atom == Atom.TYPE_co64
-        || atom == Atom.TYPE_tkhd;
+        || atom == Atom.TYPE_tkhd || atom == Atom.TYPE_ftyp;
   }
 
   /** Returns whether the extractor should parse a container atom with type {@code atom}. */