Add support for depth/editing file format in Mp4Extractor

PiperOrigin-RevId: 662956209

libraries/container/src/main/java/androidx/media3/container/MdtaMetadataEntry.java

@@ -119,6 +119,21 @@ public final class MdtaMetadataEntry implements Metadata.Entry {
     validateData(key, value, typeIndicator);
   }
 
+  /**
+   * Returns the editable track types from the {@linkplain #KEY_EDITABLE_TRACKS_MAP editable tracks
+   * map} metadata.
+   */
+  public List<Integer> getEditableTrackTypesFromMap() {
+    checkState(key.equals(KEY_EDITABLE_TRACKS_MAP), "Metadata is not an editable tracks map");
+    // Value has 1 byte version, 1 byte track count, n bytes track types.
+    int numberOfTracks = value[1];
+    List<Integer> trackTypes = new ArrayList<>();
+    for (int i = 0; i < numberOfTracks; i++) {
+      trackTypes.add((int) value[i + 2]);
+    }
+    return trackTypes;
+  }
+
   @Override
   public boolean equals(@Nullable Object obj) {
     if (this == obj) {
@@ -235,19 +250,4 @@ public final class MdtaMetadataEntry implements Metadata.Entry {
     Joiner.on(',').appendTo(sb, trackTypes);
     return sb.toString();
   }
-
-  /**
-   * Returns the editable track types from the {@linkplain #KEY_EDITABLE_TRACKS_MAP editable tracks
-   * map} metadata.
-   */
-  private List<Integer> getEditableTrackTypesFromMap() {
-    checkState(key.equals(KEY_EDITABLE_TRACKS_MAP), "Metadata is not an editable tracks map");
-    // Value has 1 byte version, 1 byte track count, n bytes track types.
-    int numberOfTracks = value[1];
-    List<Integer> trackTypes = new ArrayList<>();
-    for (int i = 0; i < numberOfTracks; i++) {
-      trackTypes.add((int) value[i + 2]);
-    }
-    return trackTypes;
-  }
 }

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/Atom.java

@@ -446,6 +446,9 @@ import java.util.List;
   @SuppressWarnings("ConstantCaseForConstants")
   public static final int TYPE_iacb = 0x69616362;
 
+  @SuppressWarnings("ConstantCaseForConstants")
+  public static final int TYPE_edvd = 0x65647664;
+
   public final int type;
 
   public Atom(int type) {

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/MetadataUtil.java

@@ -238,6 +238,27 @@ import com.google.common.collect.ImmutableList;
     return null;
   }
 
+  /**
+   * Returns the {@link MdtaMetadataEntry} for a given key, or {@code null} if the key is not
+   * present.
+   *
+   * @param metadata The {@link Metadata} to retrieve the {@link MdtaMetadataEntry} from.
+   * @param key The metadata key to search.
+   */
+  @Nullable
+  public static MdtaMetadataEntry findMdtaMetadataEntryWithKey(Metadata metadata, String key) {
+    for (int i = 0; i < metadata.length(); i++) {
+      Metadata.Entry entry = metadata.get(i);
+      if (entry instanceof MdtaMetadataEntry) {
+        MdtaMetadataEntry mdtaMetadataEntry = (MdtaMetadataEntry) entry;
+        if (mdtaMetadataEntry.key.equals(key)) {
+          return mdtaMetadataEntry;
+        }
+      }
+    }
+    return null;
+  }
+
   @Nullable
   private static TextInformationFrame parseTextAttribute(
       int type, String id, ParsableByteArray data) {

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/Mp4Extractor.java

@@ -15,9 +15,21 @@
  */
 package androidx.media3.extractor.mp4;
 
+import static androidx.media3.common.C.AUXILIARY_TRACK_TYPE_DEPTH_INVERSE;
+import static androidx.media3.common.C.AUXILIARY_TRACK_TYPE_DEPTH_LINEAR;
+import static androidx.media3.common.C.AUXILIARY_TRACK_TYPE_ORIGINAL;
+import static androidx.media3.common.C.AUXILIARY_TRACK_TYPE_UNDEFINED;
 import static androidx.media3.common.util.Assertions.checkNotNull;
+import static androidx.media3.common.util.Assertions.checkState;
+import static androidx.media3.common.util.Assertions.checkStateNotNull;
 import static androidx.media3.common.util.Util.castNonNull;
+import static androidx.media3.container.MdtaMetadataEntry.EDITABLE_TRACKS_SAMPLES_LOCATION_IN_EDIT_DATA_MP4;
+import static androidx.media3.container.Mp4Util.EDITABLE_TRACK_TYPE_DEPTH_INVERSE;
+import static androidx.media3.container.Mp4Util.EDITABLE_TRACK_TYPE_DEPTH_LINEAR;
+import static androidx.media3.container.Mp4Util.EDITABLE_TRACK_TYPE_DEPTH_METADATA;
+import static androidx.media3.container.Mp4Util.EDITABLE_TRACK_TYPE_SHARP;
 import static androidx.media3.extractor.mp4.AtomParsers.parseTraks;
+import static androidx.media3.extractor.mp4.MetadataUtil.findMdtaMetadataEntryWithKey;
 import static androidx.media3.extractor.mp4.Sniffer.BRAND_HEIC;
 import static androidx.media3.extractor.mp4.Sniffer.BRAND_QUICKTIME;
 import static java.lang.Math.max;
@@ -34,6 +46,7 @@ import androidx.media3.common.ParserException;
 import androidx.media3.common.util.Assertions;
 import androidx.media3.common.util.ParsableByteArray;
 import androidx.media3.common.util.UnstableApi;
+import androidx.media3.container.MdtaMetadataEntry;
 import androidx.media3.container.NalUnitUtil;
 import androidx.media3.extractor.Ac3Util;
 import androidx.media3.extractor.Ac4Util;
@@ -62,6 +75,7 @@ import java.lang.annotation.Target;
 import java.util.ArrayDeque;
 import java.util.ArrayList;
 import java.util.List;
+import java.util.Locale;
 import java.util.Objects;
 import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
 
@@ -93,7 +107,8 @@ public final class Mp4Extractor implements Extractor, SeekMap {
         FLAG_READ_SEF_DATA,
         FLAG_MARK_FIRST_VIDEO_TRACK_WITH_MAIN_ROLE,
         FLAG_EMIT_RAW_SUBTITLE_DATA,
-        FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES
+        FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES,
+        FLAG_READ_EDITABLE_VIDEO_TRACKS
       })
   public @interface Flags {}
 
@@ -143,6 +158,18 @@ public final class Mp4Extractor implements Extractor, SeekMap {
    */
   public static final int FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES = 1 << 5;
 
+  // TODO: b/345219017 - Add depth/editing file format spec link after its published.
+  /**
+   * Flag to extract the editable video tracks.
+   *
+   * <p>Either primary video tracks or editable video tracks (but not both) will be extracted based
+   * on the flag.
+   *
+   * <p>If the flag is set but the editable video tracks are not present, then it fallbacks to
+   * extract primary tracks instead.
+   */
+  public static final int FLAG_READ_EDITABLE_VIDEO_TRACKS = 1 << 6;
+
   /**
    * @deprecated Use {@link #newFactory(SubtitleParser.Factory)} instead.
    */
@@ -219,6 +246,12 @@ public final class Mp4Extractor implements Extractor, SeekMap {
   private int sampleCurrentNalBytesRemaining;
   private boolean isSampleDependedOn;
   private boolean seenFtypAtom;
+  private boolean seekToEdvdAtom;
+  private long edvdAtomOffset;
+  private boolean readingEditableVideoTracks;
+
+  // Used when editable video samples are in the edit data MP4 (inside edvd atom).
+  private long sampleOffsetForEditableVideoTracks;
 
   // Extractor outputs.
   private ExtractorOutput extractorOutput;
@@ -527,9 +560,10 @@ public final class Mp4Extractor implements Extractor, SeekMap {
   }
 
   /**
-   * Processes the atom payload. If {@link #atomData} is null and the size is at or above the
-   * threshold {@link #RELOAD_MINIMUM_SEEK_DISTANCE}, {@code true} is returned and the caller should
-   * restart loading at the position in {@code positionHolder}. Otherwise, the atom is read/skipped.
+   * Processes the atom payload.
+   *
+   * <p>If seek is required, {@code true} is returned and the caller should restart loading at the
+   * position in {@code positionHolder}. Otherwise the atom is read/skipped.
    */
   private boolean readAtomPayload(ExtractorInput input, PositionHolder positionHolder)
       throws IOException {
@@ -560,6 +594,12 @@ public final class Mp4Extractor implements Extractor, SeekMap {
       }
     }
     processAtomEnded(atomEndPosition);
+    if (seekToEdvdAtom) {
+      readingEditableVideoTracks = true;
+      positionHolder.position = edvdAtomOffset;
+      seekRequired = true;
+      seekToEdvdAtom = false;
+    }
     return seekRequired && parserState != STATE_READING_SAMPLE;
   }
 
@@ -579,7 +619,9 @@ public final class Mp4Extractor implements Extractor, SeekMap {
         // We've reached the end of the moov atom. Process it and prepare to read samples.
         processMoovAtom(containerAtom);
         containerAtoms.clear();
-        parserState = STATE_READING_SAMPLE;
+        if (!seekToEdvdAtom) {
+          parserState = STATE_READING_SAMPLE;
+        }
       } else if (!containerAtoms.isEmpty()) {
         containerAtoms.peek().add(containerAtom);
       }
@@ -589,13 +631,35 @@ public final class Mp4Extractor implements Extractor, SeekMap {
     }
   }
 
-  /** Updates the stored track metadata to reflect the contents of the specified moov atom. */
+  /**
+   * Processes moov atom and updates the stored track metadata.
+   *
+   * <p>The processing is aborted if the edvd.moov atom needs to be processed instead.
+   */
   private void processMoovAtom(ContainerAtom moov) throws ParserException {
+    // Process metadata first to determine whether to abort processing and seek to the edvd atom.
+    @Nullable Metadata mdtaMetadata = null;
+    @Nullable Atom.ContainerAtom meta = moov.getContainerAtomOfType(Atom.TYPE_meta);
+    List<@C.AuxiliaryTrackType Integer> auxiliaryTrackTypesForEditableVideoTracks =
+        new ArrayList<>();
+    if (meta != null) {
+      mdtaMetadata = AtomParsers.parseMdtaFromMeta(meta);
+      if (readingEditableVideoTracks) {
+        checkStateNotNull(mdtaMetadata);
+        maybeSetDefaultSampleOffsetForEditableVideoTracks(mdtaMetadata);
+        auxiliaryTrackTypesForEditableVideoTracks =
+            getAuxiliaryTrackTypesForEditableVideoTracks(mdtaMetadata);
+      } else if (shouldSeekToEdvdAtom(mdtaMetadata)) {
+        seekToEdvdAtom = true;
+        return;
+      }
+    }
+
     int firstVideoTrackIndex = C.INDEX_UNSET;
     long durationUs = C.TIME_UNSET;
     List<Mp4Track> tracks = new ArrayList<>();
 
-    // Process metadata.
+    // Process remaining metadata.
     boolean isQuickTime = fileType == FILE_TYPE_QUICKTIME;
     GaplessInfoHolder gaplessInfoHolder = new GaplessInfoHolder();
     @Nullable Metadata udtaMetadata = null;
@@ -604,11 +668,6 @@ public final class Mp4Extractor implements Extractor, SeekMap {
       udtaMetadata = AtomParsers.parseUdta(udta);
       gaplessInfoHolder.setFromMetadata(udtaMetadata);
     }
-    @Nullable Metadata mdtaMetadata = null;
-    @Nullable Atom.ContainerAtom meta = moov.getContainerAtomOfType(Atom.TYPE_meta);
-    if (meta != null) {
-      mdtaMetadata = AtomParsers.parseMdtaFromMeta(meta);
-    }
 
     Metadata mvhdMetadata =
         new Metadata(
@@ -625,6 +684,16 @@ public final class Mp4Extractor implements Extractor, SeekMap {
             isQuickTime,
             /* modifyTrackFunction= */ track -> track);
 
+    if (readingEditableVideoTracks) {
+      checkState(
+          auxiliaryTrackTypesForEditableVideoTracks.size() == trackSampleTables.size(),
+          String.format(
+              Locale.US,
+              "The number of auxiliary track types from metadata (%d) is not same as the number of"
+                  + " editable video tracks (%d)",
+              auxiliaryTrackTypesForEditableVideoTracks.size(),
+              trackSampleTables.size()));
+    }
     int trackIndex = 0;
     for (int i = 0; i < trackSampleTables.size(); i++) {
       TrackSampleTable trackSampleTable = trackSampleTables.get(i);
@@ -651,17 +720,20 @@ public final class Mp4Extractor implements Extractor, SeekMap {
       Format.Builder formatBuilder = track.format.buildUpon();
       formatBuilder.setMaxInputSize(maxInputSize);
       if (track.type == C.TRACK_TYPE_VIDEO) {
+        @C.RoleFlags int roleFlags = track.format.roleFlags;
         if ((flags & FLAG_MARK_FIRST_VIDEO_TRACK_WITH_MAIN_ROLE) != 0) {
-          formatBuilder.setRoleFlags(
-              track.format.roleFlags
-                  | (firstVideoTrackIndex == C.INDEX_UNSET
-                      ? C.ROLE_FLAG_MAIN
-                      : C.ROLE_FLAG_ALTERNATE));
+          roleFlags |=
+              firstVideoTrackIndex == C.INDEX_UNSET ? C.ROLE_FLAG_MAIN : C.ROLE_FLAG_ALTERNATE;
         }
         if (trackDurationUs > 0 && trackSampleTable.sampleCount > 0) {
           float frameRate = trackSampleTable.sampleCount / (trackDurationUs / 1000000f);
           formatBuilder.setFrameRate(frameRate);
         }
+        if (readingEditableVideoTracks) {
+          roleFlags |= C.ROLE_FLAG_AUXILIARY;
+          formatBuilder.setAuxiliaryTrackType(auxiliaryTrackTypesForEditableVideoTracks.get(i));
+        }
+        formatBuilder.setRoleFlags(roleFlags);
       }
 
       MetadataUtil.setFormatGaplessInfo(track.type, gaplessInfoHolder, formatBuilder);
@@ -688,6 +760,72 @@ public final class Mp4Extractor implements Extractor, SeekMap {
     extractorOutput.seekMap(this);
   }
 
+  private boolean shouldSeekToEdvdAtom(@Nullable Metadata mdtaMetadata) {
+    if (mdtaMetadata == null) {
+      return false;
+    }
+    if ((flags & FLAG_READ_EDITABLE_VIDEO_TRACKS) != 0) {
+      @Nullable
+      MdtaMetadataEntry edvdAtomOffsetMetadata =
+          findMdtaMetadataEntryWithKey(mdtaMetadata, MdtaMetadataEntry.KEY_EDITABLE_TRACKS_OFFSET);
+      if (edvdAtomOffsetMetadata != null) {
+        long offset = new ParsableByteArray(edvdAtomOffsetMetadata.value).readUnsignedLongToLong();
+        if (offset > 0) {
+          edvdAtomOffset = offset;
+          return true;
+        }
+      }
+    }
+    return false;
+  }
+
+  /**
+   * Sets the sample offset for the editable video tracks, if the samples are in the edit data MP4
+   * (inside edvd atom).
+   */
+  private void maybeSetDefaultSampleOffsetForEditableVideoTracks(Metadata metadata) {
+    @Nullable
+    MdtaMetadataEntry sampleLocationMetadata =
+        findMdtaMetadataEntryWithKey(
+            metadata, MdtaMetadataEntry.KEY_EDITABLE_TRACKS_SAMPLES_LOCATION);
+    if (sampleLocationMetadata != null) {
+      if (sampleLocationMetadata.value[0] == EDITABLE_TRACKS_SAMPLES_LOCATION_IN_EDIT_DATA_MP4) {
+        sampleOffsetForEditableVideoTracks = edvdAtomOffset + 16; // 16 bits for edvd atom header
+      }
+    }
+  }
+
+  private List<@C.AuxiliaryTrackType Integer> getAuxiliaryTrackTypesForEditableVideoTracks(
+      Metadata metadata) {
+    MdtaMetadataEntry trackTypesMetadata =
+        checkStateNotNull(
+            findMdtaMetadataEntryWithKey(metadata, MdtaMetadataEntry.KEY_EDITABLE_TRACKS_MAP));
+    List<Integer> editableVideoTrackTypes = trackTypesMetadata.getEditableTrackTypesFromMap();
+    List<@C.AuxiliaryTrackType Integer> auxiliaryTrackTypes =
+        new ArrayList<>(editableVideoTrackTypes.size());
+    for (int i = 0; i < editableVideoTrackTypes.size(); i++) {
+      @C.AuxiliaryTrackType int auxiliaryTrackType;
+      switch (editableVideoTrackTypes.get(i)) {
+        case EDITABLE_TRACK_TYPE_SHARP:
+          auxiliaryTrackType = AUXILIARY_TRACK_TYPE_ORIGINAL;
+          break;
+        case EDITABLE_TRACK_TYPE_DEPTH_LINEAR:
+          auxiliaryTrackType = AUXILIARY_TRACK_TYPE_DEPTH_LINEAR;
+          break;
+        case EDITABLE_TRACK_TYPE_DEPTH_INVERSE:
+          auxiliaryTrackType = AUXILIARY_TRACK_TYPE_DEPTH_INVERSE;
+          break;
+        case EDITABLE_TRACK_TYPE_DEPTH_METADATA:
+          auxiliaryTrackType = C.AUXILIARY_TRACK_TYPE_DEPTH_METADATA;
+          break;
+        default:
+          auxiliaryTrackType = AUXILIARY_TRACK_TYPE_UNDEFINED;
+      }
+      auxiliaryTrackTypes.add(auxiliaryTrackType);
+    }
+    return auxiliaryTrackTypes;
+  }
+
   /**
    * Attempts to extract the next sample in the current mdat atom for the specified track.
    *
@@ -714,7 +852,7 @@ public final class Mp4Extractor implements Extractor, SeekMap {
     Mp4Track track = tracks[sampleTrackIndex];
     TrackOutput trackOutput = track.trackOutput;
     int sampleIndex = track.sampleIndex;
-    long position = track.sampleTable.offsets[sampleIndex];
+    long position = track.sampleTable.offsets[sampleIndex] + sampleOffsetForEditableVideoTracks;
     int sampleSize = track.sampleTable.sizes[sampleIndex];
     @Nullable TrueHdSampleRechunker trueHdSampleRechunker = track.trueHdSampleRechunker;
     long skipAmount = position - inputPosition + sampleBytesRead;
@@ -1067,7 +1205,8 @@ public final class Mp4Extractor implements Extractor, SeekMap {
         || atom == Atom.TYPE_minf
         || atom == Atom.TYPE_stbl
         || atom == Atom.TYPE_edts
-        || atom == Atom.TYPE_meta;
+        || atom == Atom.TYPE_meta
+        || atom == Atom.TYPE_edvd;
   }
 
   private static final class Mp4Track {

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/Mp4ExtractorNonParameterizedTest.java

@@ -17,10 +17,13 @@ package androidx.media3.extractor.mp4;
 
 import static com.google.common.truth.Truth.assertThat;
 
+import android.content.Context;
 import androidx.media3.extractor.Extractor;
 import androidx.media3.extractor.PositionHolder;
 import androidx.media3.extractor.SniffFailure;
+import androidx.media3.extractor.text.DefaultSubtitleParserFactory;
 import androidx.media3.extractor.text.SubtitleParser;
+import androidx.media3.test.utils.DumpFileAsserts;
 import androidx.media3.test.utils.FakeExtractorInput;
 import androidx.media3.test.utils.FakeExtractorOutput;
 import androidx.media3.test.utils.FakeTrackOutput;
@@ -102,6 +105,44 @@ public final class Mp4ExtractorNonParameterizedTest {
     assertThat(extractorSeekTimeUs).isIn(trackSeekTimesUs.build());
   }
 
+  @Test
+  public void
+      extract_fileHavingNoEditableVideoTracksWithReadEditableVideoTracksFlag_extractsPrimaryVideoTracks()
+          throws Exception {
+    Context context = ApplicationProvider.getApplicationContext();
+    String inputFilePath = "media/mp4/sample.mp4";
+    Mp4Extractor mp4Extractor =
+        new Mp4Extractor(
+            new DefaultSubtitleParserFactory(), Mp4Extractor.FLAG_READ_EDITABLE_VIDEO_TRACKS);
+
+    FakeExtractorOutput primaryTracksOutput =
+        TestUtil.extractAllSamplesFromFile(mp4Extractor, context, inputFilePath);
+
+    String dumpFilePath = getDumpFilePath(inputFilePath, "_with_flag_read_Editable_video_tracks");
+    DumpFileAsserts.assertOutput(context, primaryTracksOutput, dumpFilePath);
+  }
+
+  @Test
+  public void
+      extract_fileHavingEditableVideoTracksWithReadEditableVideoTracksFlag_extractsEditableVideoTracks()
+          throws Exception {
+    Context context = ApplicationProvider.getApplicationContext();
+    String inputFilePath = "media/mp4/sample_with_fake_editable_video_tracks.mp4";
+    Mp4Extractor mp4Extractor =
+        new Mp4Extractor(
+            new DefaultSubtitleParserFactory(), Mp4Extractor.FLAG_READ_EDITABLE_VIDEO_TRACKS);
+
+    FakeExtractorOutput editableTracksOutput =
+        TestUtil.extractAllSamplesFromFile(mp4Extractor, context, inputFilePath);
+
+    String dumpFilePath = getDumpFilePath(inputFilePath, "_with_flag_read_Editable_video_tracks");
+    DumpFileAsserts.assertOutput(context, editableTracksOutput, dumpFilePath);
+  }
+
+  private static String getDumpFilePath(String inputFilePath, String suffix) {
+    return inputFilePath.replaceFirst("media", "extractordumps") + suffix;
+  }
+
   private static FakeExtractorInput createInputForSample(String sample) throws IOException {
     return new FakeExtractorInput.Builder()
         .setData(

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/Mp4ExtractorParameterizedTest.java

@@ -218,6 +218,11 @@ public final class Mp4ExtractorParameterizedTest {
     assertExtractorBehavior("media/mp4/water_180_mvhevc_5frames.mov");
   }
 
+  @Test
+  public void mp4WithEditableVideoTracks() throws Exception {
+    assertExtractorBehavior("media/mp4/sample_with_fake_editable_video_tracks.mp4");
+  }
+
   private void assertExtractorBehavior(String file) throws IOException {
     ExtractorAsserts.AssertionConfig.Builder assertionConfigBuilder =
         new ExtractorAsserts.AssertionConfig.Builder();

libraries/muxer/src/test/java/androidx/media3/muxer/Mp4MuxerEndToEndTest.java

@@ -25,8 +25,6 @@ import android.content.Context;
 import android.media.MediaCodec.BufferInfo;
 import android.util.Pair;
 import androidx.media3.common.C;
-import androidx.media3.common.util.Assertions;
-import androidx.media3.common.util.ParsableByteArray;
 import androidx.media3.common.util.Util;
 import androidx.media3.container.MdtaMetadataEntry;
 import androidx.media3.container.Mp4LocationData;
@@ -43,7 +41,6 @@ import androidx.media3.test.utils.TestUtil;
 import androidx.test.core.app.ApplicationProvider;
 import androidx.test.ext.junit.runners.AndroidJUnit4;
 import java.io.FileOutputStream;
-import java.io.IOException;
 import java.nio.ByteBuffer;
 import org.junit.Rule;
 import org.junit.Test;
@@ -510,7 +507,7 @@ public class Mp4MuxerEndToEndTest {
     FakeExtractorOutput primaryTracksOutput =
         TestUtil.extractAllSamplesFromFilePath(
             new Mp4Extractor(new DefaultSubtitleParserFactory()), outputFilePath);
-    // The Mp4Extractor can not read edvd box and can only parse primary tracks.
+    // The Mp4Extractor extracts primary tracks by default.
     DumpFileAsserts.assertOutput(
         context,
         primaryTracksOutput,
@@ -556,11 +553,11 @@ public class Mp4MuxerEndToEndTest {
       muxer.close();
     }
 
-    byte[] edvdBoxPayload = getEdvdBoxPayload(outputFilePath);
     FakeExtractorOutput editableTracksOutput =
-        TestUtil.extractAllSamplesFromByteArray(
-            new Mp4Extractor(new DefaultSubtitleParserFactory()), edvdBoxPayload);
-    // The Mp4Extractor can parse the MP4 embedded in the edvd box.
+        TestUtil.extractAllSamplesFromFilePath(
+            new Mp4Extractor(
+                new DefaultSubtitleParserFactory(), Mp4Extractor.FLAG_READ_EDITABLE_VIDEO_TRACKS),
+            outputFilePath);
     DumpFileAsserts.assertOutput(
         context,
         editableTracksOutput,
@@ -575,28 +572,4 @@ public class Mp4MuxerEndToEndTest {
       muxer.writeSampleData(trackToken, sampleAndSampleInfo.first, sampleAndSampleInfo.second);
     }
   }
-
-  private static byte[] getEdvdBoxPayload(String filePath) throws IOException {
-    ParsableByteArray data = new ParsableByteArray(TestUtil.getByteArrayFromFilePath(filePath));
-    while (data.bytesLeft() > 0) {
-      long size = data.readInt();
-      String name = data.readString(/* length= */ 4);
-      long payloadSize = size - 8;
-      if (size == 1) {
-        size = data.readUnsignedLongToLong();
-        // Parsing is not supported for box having size > Integer.MAX_VALUE.
-        Assertions.checkState(size <= Integer.MAX_VALUE);
-        // Subtract 4 bytes (32-bit box size) + 4 bytes (box name) + 8 bytes (64-bit box size).
-        payloadSize = size - 16;
-      }
-      if (name.equals("edvd")) {
-        byte[] payloadData = new byte[(int) payloadSize];
-        data.readBytes(payloadData, /* offset= */ 0, (int) payloadSize);
-        return payloadData;
-      } else {
-        data.skipBytes((int) payloadSize);
-      }
-    }
-    return new byte[0];
-  }
 }

libraries/test_data/src/test/assets/extractordumps/mp4/sample.mp4_with_flag_read_Editable_video_tracks

@@ -0,0 +1,343 @@
+seekMap:
+  isSeekable = true
+  duration = 1024000
+  getPosition(0) = [[timeUs=0, position=2269]]
+  getPosition(1) = [[timeUs=0, position=2269]]
+  getPosition(512000) = [[timeUs=0, position=2269]]
+  getPosition(1024000) = [[timeUs=0, position=2269]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 89876
+  sample count = 30
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.64001F
+    maxInputSize = 36722
+    maxNumReorderSamples = 2
+    width = 1080
+    height = 720
+    frameRate = 29.970028
+    colorInfo:
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[TSSE: description=null: values=[Lavf56.1.0], xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=1000]
+    initializationData:
+      data = length 29, hash 4746B5D9
+      data = length 10, hash 7A0D0F2B
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 36692, hash D216076E
+  sample 1:
+    time = 66733
+    flags = 0
+    data = length 5312, hash D45D3CA0
+  sample 2:
+    time = 33366
+    flags = 0
+    data = length 599, hash 1BE7812D
+  sample 3:
+    time = 200200
+    flags = 0
+    data = length 7735, hash 4490F110
+  sample 4:
+    time = 133466
+    flags = 0
+    data = length 987, hash 560B5036
+  sample 5:
+    time = 100100
+    flags = 0
+    data = length 673, hash ED7CD8C7
+  sample 6:
+    time = 166833
+    flags = 0
+    data = length 523, hash 3020DF50
+  sample 7:
+    time = 333666
+    flags = 0
+    data = length 6061, hash 736C72B2
+  sample 8:
+    time = 266933
+    flags = 0
+    data = length 992, hash FE132F23
+  sample 9:
+    time = 233566
+    flags = 0
+    data = length 623, hash 5B2C1816
+  sample 10:
+    time = 300300
+    flags = 0
+    data = length 421, hash 742E69C1
+  sample 11:
+    time = 433766
+    flags = 0
+    data = length 4899, hash F72F86A1
+  sample 12:
+    time = 400400
+    flags = 0
+    data = length 568, hash 519A8E50
+  sample 13:
+    time = 367033
+    flags = 0
+    data = length 620, hash 3990AA39
+  sample 14:
+    time = 567233
+    flags = 0
+    data = length 5450, hash F06EC4AA
+  sample 15:
+    time = 500500
+    flags = 0
+    data = length 1051, hash 92DFA63A
+  sample 16:
+    time = 467133
+    flags = 0
+    data = length 874, hash 69587FB4
+  sample 17:
+    time = 533866
+    flags = 0
+    data = length 781, hash 36BE495B
+  sample 18:
+    time = 700700
+    flags = 0
+    data = length 4725, hash AC0C8CD3
+  sample 19:
+    time = 633966
+    flags = 0
+    data = length 1022, hash 5D8BFF34
+  sample 20:
+    time = 600600
+    flags = 0
+    data = length 790, hash 99413A99
+  sample 21:
+    time = 667333
+    flags = 0
+    data = length 610, hash 5E129290
+  sample 22:
+    time = 834166
+    flags = 0
+    data = length 2751, hash 769974CB
+  sample 23:
+    time = 767433
+    flags = 0
+    data = length 745, hash B78A477A
+  sample 24:
+    time = 734066
+    flags = 0
+    data = length 621, hash CF741E7A
+  sample 25:
+    time = 800800
+    flags = 0
+    data = length 505, hash 1DB4894E
+  sample 26:
+    time = 967633
+    flags = 0
+    data = length 1268, hash C15348DC
+  sample 27:
+    time = 900900
+    flags = 0
+    data = length 880, hash C2DE85D0
+  sample 28:
+    time = 867533
+    flags = 0
+    data = length 530, hash C98BC6A8
+  sample 29:
+    time = 934266
+    flags = 536870912
+    data = length 568, hash 4FE5C8EA
+track 1:
+  total output bytes = 9529
+  sample count = 45
+  format 0:
+    peakBitrate = 200000
+    id = 2
+    sampleMimeType = audio/mp4a-latm
+    codecs = mp4a.40.2
+    maxInputSize = 294
+    channelCount = 1
+    sampleRate = 44100
+    language = und
+    metadata = entries=[TSSE: description=null: values=[Lavf56.1.0], xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=1000]
+    initializationData:
+      data = length 2, hash 5F7
+  sample 0:
+    time = 44000
+    flags = 1
+    data = length 23, hash 47DE9131
+  sample 1:
+    time = 67219
+    flags = 1
+    data = length 6, hash 31EC5206
+  sample 2:
+    time = 90439
+    flags = 1
+    data = length 148, hash 894A176B
+  sample 3:
+    time = 113659
+    flags = 1
+    data = length 189, hash CEF235A1
+  sample 4:
+    time = 136879
+    flags = 1
+    data = length 205, hash BBF5F7B0
+  sample 5:
+    time = 160099
+    flags = 1
+    data = length 210, hash F278B193
+  sample 6:
+    time = 183319
+    flags = 1
+    data = length 210, hash 82DA1589
+  sample 7:
+    time = 206539
+    flags = 1
+    data = length 207, hash 5BE231DF
+  sample 8:
+    time = 229759
+    flags = 1
+    data = length 225, hash 18819EE1
+  sample 9:
+    time = 252979
+    flags = 1
+    data = length 215, hash CA7FA67B
+  sample 10:
+    time = 276199
+    flags = 1
+    data = length 211, hash 581A1C18
+  sample 11:
+    time = 299419
+    flags = 1
+    data = length 216, hash ADB88187
+  sample 12:
+    time = 322639
+    flags = 1
+    data = length 229, hash 2E8BA4DC
+  sample 13:
+    time = 345859
+    flags = 1
+    data = length 232, hash 22F0C510
+  sample 14:
+    time = 369079
+    flags = 1
+    data = length 235, hash 867AD0DC
+  sample 15:
+    time = 392299
+    flags = 1
+    data = length 231, hash 84E823A8
+  sample 16:
+    time = 415519
+    flags = 1
+    data = length 226, hash 1BEF3A95
+  sample 17:
+    time = 438739
+    flags = 1
+    data = length 216, hash EAA345AE
+  sample 18:
+    time = 461959
+    flags = 1
+    data = length 229, hash 6957411F
+  sample 19:
+    time = 485179
+    flags = 1
+    data = length 219, hash 41275022
+  sample 20:
+    time = 508399
+    flags = 1
+    data = length 241, hash 6495DF96
+  sample 21:
+    time = 531619
+    flags = 1
+    data = length 228, hash 63D95906
+  sample 22:
+    time = 554839
+    flags = 1
+    data = length 238, hash 34F676F9
+  sample 23:
+    time = 578058
+    flags = 1
+    data = length 234, hash E5CBC045
+  sample 24:
+    time = 601278
+    flags = 1
+    data = length 231, hash 5FC43661
+  sample 25:
+    time = 624498
+    flags = 1
+    data = length 217, hash 682708ED
+  sample 26:
+    time = 647718
+    flags = 1
+    data = length 239, hash D43780FC
+  sample 27:
+    time = 670938
+    flags = 1
+    data = length 243, hash C5E17980
+  sample 28:
+    time = 694158
+    flags = 1
+    data = length 231, hash AC5837BA
+  sample 29:
+    time = 717378
+    flags = 1
+    data = length 230, hash 169EE895
+  sample 30:
+    time = 740598
+    flags = 1
+    data = length 238, hash C48FF3F1
+  sample 31:
+    time = 763818
+    flags = 1
+    data = length 225, hash 531E4599
+  sample 32:
+    time = 787038
+    flags = 1
+    data = length 232, hash CB3C6B8D
+  sample 33:
+    time = 810258
+    flags = 1
+    data = length 243, hash F8C94C7
+  sample 34:
+    time = 833478
+    flags = 1
+    data = length 232, hash A646A7D0
+  sample 35:
+    time = 856698
+    flags = 1
+    data = length 237, hash E8B787A5
+  sample 36:
+    time = 879918
+    flags = 1
+    data = length 228, hash 3FA7A29F
+  sample 37:
+    time = 903138
+    flags = 1
+    data = length 235, hash B9B33B0A
+  sample 38:
+    time = 926358
+    flags = 1
+    data = length 264, hash 71A4869E
+  sample 39:
+    time = 949578
+    flags = 1
+    data = length 257, hash D049B54C
+  sample 40:
+    time = 972798
+    flags = 1
+    data = length 227, hash 66757231
+  sample 41:
+    time = 996018
+    flags = 1
+    data = length 227, hash BD374F1B
+  sample 42:
+    time = 1019238
+    flags = 1
+    data = length 235, hash 999477F6
+  sample 43:
+    time = 1042458
+    flags = 1
+    data = length 229, hash FFF98DF0
+  sample 44:
+    time = 1065678
+    flags = 536870913
+    data = length 6, hash 31B22286
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.0.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.1.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.2.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.3.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.reading_within_gop_sample_dependencies.0.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.reading_within_gop_sample_dependencies.1.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.reading_within_gop_sample_dependencies.2.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.reading_within_gop_sample_dependencies.3.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.reading_within_gop_sample_dependencies.unknown_length.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4.unknown_length.dump

@@ -0,0 +1,48 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 1
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[mdta: key=editable.tracks.offset, value=400332, mdta: key=editable.tracks.length, value=400628, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/sample_with_fake_editable_video_tracks.mp4_with_flag_read_Editable_video_tracks

@@ -0,0 +1,91 @@
+seekMap:
+  isSeekable = true
+  duration = 0
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400276]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(0) = [[timeUs=0, position=400052]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    roleFlags = [auxiliary]
+    auxiliaryTrackType = original
+    metadata = entries=[mdta: key=editable.tracks.map, value=track types = 0,1, mdta: key=editable.tracks.samples.location, value=0, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+track 1:
+  total output bytes = 280
+  sample count = 5
+  format 0:
+    id = 2
+    sampleMimeType = video/avc
+    codecs = avc1.F4000A
+    maxInputSize = 86
+    maxNumReorderSamples = 2
+    width = 12
+    height = 10
+    colorInfo:
+      colorRange = 1
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    roleFlags = [auxiliary]
+    auxiliaryTrackType = depth-linear
+    metadata = entries=[mdta: key=editable.tracks.map, value=track types = 0,1, mdta: key=editable.tracks.samples.location, value=0, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
+    initializationData:
+      data = length 28, hash 410B510
+      data = length 9, hash FBADD682
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 1:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 2:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 3:
+    time = 0
+    flags = 1
+    data = length 56, hash C4551A2E
+  sample 4:
+    time = 0
+    flags = 536870913
+    data = length 56, hash C4551A2E
+tracksEnded = true

libraries/test_data/src/test/assets/muxerdumps/mp4_with_editable_video_tracks.mp4.dump

@@ -21,6 +21,8 @@ track 0:
       colorRange = 1
       lumaBitdepth = 8
       chromaBitdepth = 8
+    roleFlags = [auxiliary]
+    auxiliaryTrackType = original
     metadata = entries=[mdta: key=editable.tracks.map, value=track types = 0,1, mdta: key=editable.tracks.samples.location, value=0, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
     initializationData:
       data = length 28, hash 410B510
@@ -60,6 +62,8 @@ track 1:
       colorRange = 1
       lumaBitdepth = 8
       chromaBitdepth = 8
+    roleFlags = [auxiliary]
+    auxiliaryTrackType = depth-linear
     metadata = entries=[mdta: key=editable.tracks.map, value=track types = 0,1, mdta: key=editable.tracks.samples.location, value=0, Mp4Timestamp: creation time=1000000, modification time=5000000, timescale=10000]
     initializationData:
       data = length 28, hash 410B510