Read sample dependencies for H.265 in FragmentedMp4

Add a new flag to FragmentedMp4Extractor
FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES_H265
Read two bytes from H.265 videos to determine NAL unit type and
temporal layer id.

PiperOrigin-RevId: 714046987

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

@@ -682,16 +682,22 @@ public final class NalUnitUtil {
   }
 
   /**
-   * Returns the number of bytes needed from the NAL unit to determine whether subsequent NAL units
+   * Returns the number of bytes in the NAL unit header.
-   * can depend on the current NAL unit.
+   *
+   * <p>The NAL unit header can be used to determine the NAL unit type and whether subsequent NAL
+   * units can depend on the current NAL unit.
+   *
+   * <p>This is {@code nalUnitHeaderBytes} from the H.264 spec, or the size of {@code
+   * nal_unit_header()} in H.265.
    *
    * @param format The sample {@link Format}.
    */
-  public static int numberOfBytesToDetermineSampleDependencies(Format format) {
+  public static int numberOfBytesInNalUnitHeader(Format format) {
     if (Objects.equals(format.sampleMimeType, MimeTypes.VIDEO_H264)) {
       return 1;
     }
-    if (Objects.equals(format.sampleMimeType, MimeTypes.VIDEO_H265)) {
+    if (Objects.equals(format.sampleMimeType, MimeTypes.VIDEO_H265)
+        || MimeTypes.containsCodecsCorrespondingToMimeType(format.codecs, MimeTypes.VIDEO_H265)) {
       return 2;
     }
     return 0;

libraries/container/src/test/java/androidx/media3/container/NalUnitUtilTest.java

@@ -16,7 +16,7 @@
 package androidx.media3.container;
 
 import static androidx.media3.container.NalUnitUtil.isDependedOn;
-import static androidx.media3.container.NalUnitUtil.numberOfBytesToDetermineSampleDependencies;
+import static androidx.media3.container.NalUnitUtil.numberOfBytesInNalUnitHeader;
 import static androidx.media3.test.utils.TestUtil.createByteArray;
 import static com.google.common.truth.Truth.assertThat;
 
@@ -499,10 +499,10 @@ public final class NalUnitUtilTest {
   }
 
   @Test
-  public void numberOfBytesToDetermineSampleDependencies_vp8_returnsZero() {
+  public void numberOfBytesInNalUnitHeader_vp8_returnsZero() {
     Format vp8Video = new Format.Builder().setSampleMimeType(MimeTypes.VIDEO_VP8).build();
 
-    assertThat(numberOfBytesToDetermineSampleDependencies(vp8Video)).isEqualTo(0);
+    assertThat(numberOfBytesInNalUnitHeader(vp8Video)).isEqualTo(0);
   }
 
   @Test

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/FragmentedMp4Extractor.java

@@ -149,6 +149,14 @@ public class FragmentedMp4Extractor implements Extractor {
    */
   public static final int FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES = 1 << 6; // 64
 
+  /**
+   * Flag to extract additional sample dependency information, and mark output buffers with {@link
+   * C#BUFFER_FLAG_NOT_DEPENDED_ON} for {@linkplain MimeTypes#VIDEO_H265 H.265} video.
+   *
+   * <p>See {@link #FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES}.
+   */
+  public static final int FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES_H265 = 1 << 7;
+
   /**
    * @deprecated Use {@link #newFactory(SubtitleParser.Factory)} instead.
    */
@@ -193,7 +201,7 @@ public class FragmentedMp4Extractor implements Extractor {
   // Temporary arrays.
   private final ParsableByteArray nalStartCode;
   private final ParsableByteArray nalPrefix;
-  private final ParsableByteArray nalBuffer;
+  private final ParsableByteArray nalUnitWithoutHeaderBuffer;
   private final byte[] scratchBytes;
   private final ParsableByteArray scratch;
 
@@ -400,8 +408,8 @@ public class FragmentedMp4Extractor implements Extractor {
     eventMessageEncoder = new EventMessageEncoder();
     atomHeader = new ParsableByteArray(Mp4Box.LONG_HEADER_SIZE);
     nalStartCode = new ParsableByteArray(NalUnitUtil.NAL_START_CODE);
-    nalPrefix = new ParsableByteArray(5);
+    nalPrefix = new ParsableByteArray(6);
-    nalBuffer = new ParsableByteArray();
+    nalUnitWithoutHeaderBuffer = new ParsableByteArray();
     scratchBytes = new byte[16];
     scratch = new ParsableByteArray(scratchBytes);
     containerAtoms = new ArrayDeque<>();
@@ -1533,12 +1541,10 @@ public class FragmentedMp4Extractor implements Extractor {
     if (parserState == STATE_READING_SAMPLE_START) {
       sampleSize = trackBundle.getCurrentSampleSize();
       // We must check all NAL units in the Fragmented MP4 sample for dependencies.
-      // When FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES is unset, or codec is not supported,
+      // When reading sample dependencies is disabled, or codec is not supported,
       // set isSampleDependedOn = true and skip parsing the payload bytes.
       isSampleDependedOn =
-          (flags & FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES) == 0
+          !canReadWithinGopSampleDependencies(trackBundle.moovSampleTable.track.format);
-              || !Objects.equals(
-                  trackBundle.moovSampleTable.track.format.sampleMimeType, MimeTypes.VIDEO_H264);
 
       if (trackBundle.currentSampleIndex < trackBundle.firstSampleToOutputIndex) {
         input.skipFully(sampleSize);
@@ -1585,57 +1591,74 @@ public class FragmentedMp4Extractor implements Extractor {
       nalPrefixData[0] = 0;
       nalPrefixData[1] = 0;
       nalPrefixData[2] = 0;
-      int nalUnitPrefixLength = track.nalUnitLengthFieldLength + 1;
       int nalUnitLengthFieldLengthDiff = 4 - track.nalUnitLengthFieldLength;
       // NAL units are length delimited, but the decoder requires start code delimited units.
       // Loop until we've written the sample to the track output, replacing length delimiters with
       // start codes as we encounter them.
       while (sampleBytesWritten < sampleSize) {
         if (sampleCurrentNalBytesRemaining == 0) {
+          int numberOfNalUnitHeaderBytesToRead = 0;
+          if (ceaTrackOutputs.length > 0 || !isSampleDependedOn) {
+            // Try to read the NAL unit header if we're parsing captions or sample dependencies.
+            int nalUnitHeaderSize = NalUnitUtil.numberOfBytesInNalUnitHeader(track.format);
+            if (track.nalUnitLengthFieldLength + nalUnitHeaderSize
+                <= sampleSize - sampleBytesWritten) {
+              // In some malformed videos with padding, the NAL unit may be empty.
+              // See b/383201567#comment20
+              // Only try to read the header if there are enough bytes in this sample.
+              numberOfNalUnitHeaderBytesToRead = nalUnitHeaderSize;
+            }
+          }
+          // Read numberOfNalUnitHeaderBytesToRead in the same readFully call that reads the NAL
+          // length. This ensures sampleBytesRead, sampleBytesWritten and isSampleDependedOn remain
+          // in a consistent state if we have read failures.
+          int nalUnitPrefixLength =
+              track.nalUnitLengthFieldLength + numberOfNalUnitHeaderBytesToRead;
           // Read the NAL length so that we know where we find the next one, and its type.
           input.readFully(nalPrefixData, nalUnitLengthFieldLengthDiff, nalUnitPrefixLength);
           nalPrefix.setPosition(0);
           int nalLengthInt = nalPrefix.readInt();
-          if (nalLengthInt < 1) {
+          if (nalLengthInt < 0) {
             throw ParserException.createForMalformedContainer(
                 "Invalid NAL length", /* cause= */ null);
           }
-          sampleCurrentNalBytesRemaining = nalLengthInt - 1;
+          sampleCurrentNalBytesRemaining = nalLengthInt - numberOfNalUnitHeaderBytesToRead;
           // Write a start code for the current NAL unit.
           nalStartCode.setPosition(0);
           output.sampleData(nalStartCode, 4);
-          // Write the NAL unit type byte.
+          sampleBytesWritten += 4;
-          output.sampleData(nalPrefix, 1);
+          sampleSize += nalUnitLengthFieldLengthDiff;
           processSeiNalUnitPayload =
               ceaTrackOutputs.length > 0
+                  && numberOfNalUnitHeaderBytesToRead > 0
                   && NalUnitUtil.isNalUnitSei(track.format, nalPrefixData[4]);
-          sampleBytesWritten += 5;
+          // Write the extra NAL unit bytes to the output.
-          sampleSize += nalUnitLengthFieldLengthDiff;
+          output.sampleData(nalPrefix, numberOfNalUnitHeaderBytesToRead);
-          if (!isSampleDependedOn
+          sampleBytesWritten += numberOfNalUnitHeaderBytesToRead;
-              && Objects.equals(
+          if (numberOfNalUnitHeaderBytesToRead > 0
-                  trackBundle.moovSampleTable.track.format.sampleMimeType, MimeTypes.VIDEO_H264)
+              && !isSampleDependedOn
-              && NalUnitUtil.isH264NalUnitDependedOn(nalPrefixData[4])) {
+              && NalUnitUtil.isDependedOn(
+                  nalPrefixData,
+                  /* offset= */ 4,
+                  /* length= */ numberOfNalUnitHeaderBytesToRead,
+                  track.format)) {
             isSampleDependedOn = true;
           }
         } else {
           int writtenBytes;
           if (processSeiNalUnitPayload) {
-            // Read and write the payload of the SEI NAL unit.
+            // Read and write the remaining payload of the SEI NAL unit.
-            nalBuffer.reset(sampleCurrentNalBytesRemaining);
+            nalUnitWithoutHeaderBuffer.reset(sampleCurrentNalBytesRemaining);
-            input.readFully(nalBuffer.getData(), 0, sampleCurrentNalBytesRemaining);
+            input.readFully(
-            output.sampleData(nalBuffer, sampleCurrentNalBytesRemaining);
+                nalUnitWithoutHeaderBuffer.getData(), 0, sampleCurrentNalBytesRemaining);
+            output.sampleData(nalUnitWithoutHeaderBuffer, sampleCurrentNalBytesRemaining);
             writtenBytes = sampleCurrentNalBytesRemaining;
             // Unescape and process the SEI NAL unit.
             int unescapedLength =
-                NalUnitUtil.unescapeStream(nalBuffer.getData(), nalBuffer.limit());
+                NalUnitUtil.unescapeStream(
-            // If the format is H.265/HEVC the NAL unit header has two bytes so skip one more byte.
+                    nalUnitWithoutHeaderBuffer.getData(), nalUnitWithoutHeaderBuffer.limit());
-            nalBuffer.setPosition(
+            nalUnitWithoutHeaderBuffer.setPosition(0);
-                Objects.equals(track.format.sampleMimeType, MimeTypes.VIDEO_H265)
+            nalUnitWithoutHeaderBuffer.setLimit(unescapedLength);
-                        || MimeTypes.containsCodecsCorrespondingToMimeType(
-                            track.format.codecs, MimeTypes.VIDEO_H265)
-                    ? 1
-                    : 0);
-            nalBuffer.setLimit(unescapedLength);
 
             if (track.format.maxNumReorderSamples == Format.NO_VALUE) {
               if (reorderingSeiMessageQueue.getMaxSize() != 0) {
@@ -1645,7 +1668,7 @@ public class FragmentedMp4Extractor implements Extractor {
                 != track.format.maxNumReorderSamples) {
               reorderingSeiMessageQueue.setMaxSize(track.format.maxNumReorderSamples);
             }
-            reorderingSeiMessageQueue.add(sampleTimeUs, nalBuffer);
+            reorderingSeiMessageQueue.add(sampleTimeUs, nalUnitWithoutHeaderBuffer);
 
             if ((trackBundle.getCurrentSampleFlags() & C.BUFFER_FLAG_END_OF_STREAM) != 0) {
               reorderingSeiMessageQueue.flush();
@@ -1666,7 +1689,7 @@ public class FragmentedMp4Extractor implements Extractor {
     }
 
     @C.BufferFlags int sampleFlags = trackBundle.getCurrentSampleFlags();
-    if ((flags & FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES) != 0 && !isSampleDependedOn) {
+    if (!isSampleDependedOn) {
       sampleFlags |= C.BUFFER_FLAG_NOT_DEPENDED_ON;
     }
 
@@ -1688,6 +1711,20 @@ public class FragmentedMp4Extractor implements Extractor {
     return true;
   }
 
+  /**
+   * Returns whether reading within GOP sample dependencies is enabled for the sample {@link
+   * Format}.
+   */
+  private boolean canReadWithinGopSampleDependencies(Format format) {
+    if (Objects.equals(format.sampleMimeType, MimeTypes.VIDEO_H264)) {
+      return (flags & FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES) != 0;
+    }
+    if (Objects.equals(format.sampleMimeType, MimeTypes.VIDEO_H265)) {
+      return (flags & FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES_H265) != 0;
+    }
+    return false;
+  }
+
   /**
    * Called immediately after outputting a non-metadata sample, to output any pending metadata
    * samples.

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

@@ -902,15 +902,14 @@ public final class Mp4Extractor implements Extractor, SeekMap {
           int nalUnitPrefixLength = track.track.nalUnitLengthFieldLength;
           int numberOfBytesToDetermineSampleDependencies = 0;
           if (!isSampleDependedOn
-              && nalUnitPrefixLength
+              && nalUnitPrefixLength + NalUnitUtil.numberOfBytesInNalUnitHeader(track.track.format)
-                      + NalUnitUtil.numberOfBytesToDetermineSampleDependencies(track.track.format)
                   <= track.sampleTable.sizes[sampleIndex] - sampleBytesRead) {
             // Parsing sample dependencies needs the first few NAL unit bytes. Read them in the same
             // readFully call that reads the NAL length. This ensures sampleBytesRead,
             // sampleBytesWritten and isSampleDependedOn remain in a consistent state if we have
             // read failures.
             numberOfBytesToDetermineSampleDependencies =
-                NalUnitUtil.numberOfBytesToDetermineSampleDependencies(track.track.format);
+                NalUnitUtil.numberOfBytesInNalUnitHeader(track.track.format);
             nalUnitPrefixLength =
                 track.track.nalUnitLengthFieldLength + numberOfBytesToDetermineSampleDependencies;
           }

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/FragmentedMp4ExtractorParameterizedTest.java

@@ -200,6 +200,16 @@ public final class FragmentedMp4ExtractorParameterizedTest {
         /* closedCaptionFormats= */ ImmutableList.of(), "media/mp4/sample_fragmented_iamf.mp4");
   }
 
+  @Test
+  public void sampleWithNonReferenceH265FramesAndCaptions() throws Exception {
+    // Enabling the CEA-608 track enables SEI payload parsing.
+    List<Format> closedCaptions =
+        Collections.singletonList(
+            new Format.Builder().setSampleMimeType(MimeTypes.APPLICATION_CEA608).build());
+
+    assertExtractorBehavior(closedCaptions, "media/mp4/fragmented_captions_h265.mp4");
+  }
+
   private void assertExtractorBehavior(List<Format> closedCaptionFormats, String file)
       throws IOException {
     ExtractorAsserts.AssertionConfig.Builder assertionConfigBuilder =
@@ -232,6 +242,7 @@ public final class FragmentedMp4ExtractorParameterizedTest {
     }
     if (readWithinGopSampleDependencies) {
       flags |= FragmentedMp4Extractor.FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES;
+      flags |= FragmentedMp4Extractor.FLAG_READ_WITHIN_GOP_SAMPLE_DEPENDENCIES_H265;
     }
 
     @FragmentedMp4Extractor.Flags int finalFlags = flags;

libraries/test_data/src/test/assets/extractordumps/mp4/fragmented_captions_h265.mp4.0.dump

@@ -0,0 +1,283 @@
+seekMap:
+  isSeekable = false
+  duration = UNSET TIME
+  getPosition(0) = [[timeUs=0, position=3203]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 24020
+  sample count = 61
+  format 0:
+    id = 1
+    containerMimeType = video/mp4
+    sampleMimeType = video/hevc
+    codecs = hvc1.1.6.L60.90
+    maxNumReorderSamples = 2
+    width = 416
+    height = 234
+    maxSubLayers = 1
+    colorInfo:
+      colorSpace = 2
+      colorRange = 2
+      colorTransfer = 3
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    initializationData:
+      data = length 2426, hash 25737613
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 5220, hash 5F0CCC08
+  sample 1:
+    time = 233566
+    flags = 0
+    data = length 615, hash CBFF4D3C
+  sample 2:
+    time = 166833
+    flags = 0
+    data = length 217, hash B600557D
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 177, hash 99B5D77F
+  sample 4:
+    time = 133466
+    flags = 0
+    data = length 190, hash AF7F85A2
+  sample 5:
+    time = 200200
+    flags = 0
+    data = length 225, hash BA8F2E19
+  sample 6:
+    time = 333666
+    flags = 0
+    data = length 330, hash DCBABAE8
+  sample 7:
+    time = 300300
+    flags = 0
+    data = length 183, hash 15A2DB25
+  sample 8:
+    time = 266933
+    flags = 0
+    data = length 154, hash 81018D0E
+  sample 9:
+    time = 400400
+    flags = 0
+    data = length 576, hash C9C5389F
+  sample 10:
+    time = 367033
+    flags = 0
+    data = length 173, hash 2D567C7E
+  sample 11:
+    time = 500500
+    flags = 0
+    data = length 356, hash A6D37B60
+  sample 12:
+    time = 467133
+    flags = 0
+    data = length 188, hash 11AB1597
+  sample 13:
+    time = 433766
+    flags = 0
+    data = length 174, hash 18C46868
+  sample 14:
+    time = 567233
+    flags = 0
+    data = length 331, hash EEA184AF
+  sample 15:
+    time = 533866
+    flags = 0
+    data = length 213, hash 3FAB17EE
+  sample 16:
+    time = 667333
+    flags = 0
+    data = length 371, hash D0ECAE97
+  sample 17:
+    time = 633966
+    flags = 0
+    data = length 179, hash 796F24AF
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 154, hash FD43F0C6
+  sample 19:
+    time = 734066
+    flags = 0
+    data = length 622, hash 8AD9409C
+  sample 20:
+    time = 700700
+    flags = 0
+    data = length 195, hash 4EFA8FFD
+  sample 21:
+    time = 900900
+    flags = 0
+    data = length 424, hash 5F9D809C
+  sample 22:
+    time = 834166
+    flags = 0
+    data = length 187, hash FB5137FD
+  sample 23:
+    time = 767433
+    flags = 0
+    data = length 162, hash 7B352EA
+  sample 24:
+    time = 800800
+    flags = 0
+    data = length 173, hash 42EBA5AC
+  sample 25:
+    time = 867533
+    flags = 0
+    data = length 205, hash 15D2186B
+  sample 26:
+    time = 1001000
+    flags = 0
+    data = length 342, hash 70E9AD53
+  sample 27:
+    time = 967633
+    flags = 0
+    data = length 200, hash 2C923950
+  sample 28:
+    time = 934266
+    flags = 0
+    data = length 150, hash 242935D6
+  sample 29:
+    time = 1167833
+    flags = 0
+    data = length 802, hash 60213433
+  sample 30:
+    time = 1101100
+    flags = 0
+    data = length 203, hash 21B384E5
+  sample 31:
+    time = 1034366
+    flags = 0
+    data = length 185, hash B2735872
+  sample 32:
+    time = 1067733
+    flags = 0
+    data = length 1310, hash 8D4F7E35
+  sample 33:
+    time = 1134466
+    flags = 0
+    data = length 176, hash D1D91C73
+  sample 34:
+    time = 1334666
+    flags = 0
+    data = length 258, hash 92056B3A
+  sample 35:
+    time = 1267933
+    flags = 0
+    data = length 214, hash 86C47DEC
+  sample 36:
+    time = 1201200
+    flags = 0
+    data = length 201, hash AFA1B6D4
+  sample 37:
+    time = 1234566
+    flags = 0
+    data = length 146, hash 3E1B1C72
+  sample 38:
+    time = 1301300
+    flags = 0
+    data = length 173, hash FC0B911D
+  sample 39:
+    time = 1401400
+    flags = 0
+    data = length 617, hash 35AAFB9
+  sample 40:
+    time = 1368033
+    flags = 0
+    data = length 179, hash E9838582
+  sample 41:
+    time = 1468133
+    flags = 0
+    data = length 395, hash 8F5373AC
+  sample 42:
+    time = 1434766
+    flags = 0
+    data = length 184, hash 10ED9B63
+  sample 43:
+    time = 1634966
+    flags = 0
+    data = length 460, hash ED8ECDE7
+  sample 44:
+    time = 1568233
+    flags = 0
+    data = length 184, hash BF273413
+  sample 45:
+    time = 1501500
+    flags = 0
+    data = length 162, hash FE32A96B
+  sample 46:
+    time = 1534866
+    flags = 0
+    data = length 201, hash BBB627F2
+  sample 47:
+    time = 1601600
+    flags = 0
+    data = length 169, hash AFD9804F
+  sample 48:
+    time = 1735066
+    flags = 0
+    data = length 668, hash 4B9FB50F
+  sample 49:
+    time = 1701700
+    flags = 0
+    data = length 219, hash E2A651A2
+  sample 50:
+    time = 1668333
+    flags = 0
+    data = length 162, hash ECB52FD2
+  sample 51:
+    time = 1901900
+    flags = 0
+    data = length 405, hash 7DA640B4
+  sample 52:
+    time = 1835166
+    flags = 0
+    data = length 188, hash 46D5E9D4
+  sample 53:
+    time = 1768433
+    flags = 0
+    data = length 167, hash 65F87E6A
+  sample 54:
+    time = 1801800
+    flags = 0
+    data = length 161, hash 2E37B5E0
+  sample 55:
+    time = 1868533
+    flags = 0
+    data = length 199, hash D63CF14E
+  sample 56:
+    time = 2002000
+    flags = 0
+    data = length 282, hash E748555D
+  sample 57:
+    time = 1968633
+    flags = 0
+    data = length 199, hash F341EE39
+  sample 58:
+    time = 1935266
+    flags = 0
+    data = length 166, hash 32F07BFF
+  sample 59:
+    time = 2068733
+    flags = 0
+    data = length 2246, hash E07DFCAE
+  sample 60:
+    time = 2035366
+    flags = 0
+    data = length 53, hash D9F70BD5
+track 100:
+  total output bytes = 45
+  sample count = 2
+  format 0:
+    sampleMimeType = application/cea-608
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 36, hash 6E15DFAD
+  sample 1:
+    time = 1935266
+    flags = 1
+    data = length 9, hash 604EC6AA
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/fragmented_captions_h265.mp4.reading_within_gop_sample_dependencies.0.dump

@@ -0,0 +1,283 @@
+seekMap:
+  isSeekable = false
+  duration = UNSET TIME
+  getPosition(0) = [[timeUs=0, position=3203]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 24020
+  sample count = 61
+  format 0:
+    id = 1
+    containerMimeType = video/mp4
+    sampleMimeType = video/hevc
+    codecs = hvc1.1.6.L60.90
+    maxNumReorderSamples = 2
+    width = 416
+    height = 234
+    maxSubLayers = 1
+    colorInfo:
+      colorSpace = 2
+      colorRange = 2
+      colorTransfer = 3
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    initializationData:
+      data = length 2426, hash 25737613
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 5220, hash 5F0CCC08
+  sample 1:
+    time = 233566
+    flags = 0
+    data = length 615, hash CBFF4D3C
+  sample 2:
+    time = 166833
+    flags = 0
+    data = length 217, hash B600557D
+  sample 3:
+    time = 100100
+    flags = 67108864
+    data = length 177, hash 99B5D77F
+  sample 4:
+    time = 133466
+    flags = 67108864
+    data = length 190, hash AF7F85A2
+  sample 5:
+    time = 200200
+    flags = 67108864
+    data = length 225, hash BA8F2E19
+  sample 6:
+    time = 333666
+    flags = 0
+    data = length 330, hash DCBABAE8
+  sample 7:
+    time = 300300
+    flags = 0
+    data = length 183, hash 15A2DB25
+  sample 8:
+    time = 266933
+    flags = 67108864
+    data = length 154, hash 81018D0E
+  sample 9:
+    time = 400400
+    flags = 0
+    data = length 576, hash C9C5389F
+  sample 10:
+    time = 367033
+    flags = 67108864
+    data = length 173, hash 2D567C7E
+  sample 11:
+    time = 500500
+    flags = 0
+    data = length 356, hash A6D37B60
+  sample 12:
+    time = 467133
+    flags = 0
+    data = length 188, hash 11AB1597
+  sample 13:
+    time = 433766
+    flags = 67108864
+    data = length 174, hash 18C46868
+  sample 14:
+    time = 567233
+    flags = 0
+    data = length 331, hash EEA184AF
+  sample 15:
+    time = 533866
+    flags = 67108864
+    data = length 213, hash 3FAB17EE
+  sample 16:
+    time = 667333
+    flags = 0
+    data = length 371, hash D0ECAE97
+  sample 17:
+    time = 633966
+    flags = 0
+    data = length 179, hash 796F24AF
+  sample 18:
+    time = 600600
+    flags = 67108864
+    data = length 154, hash FD43F0C6
+  sample 19:
+    time = 734066
+    flags = 0
+    data = length 622, hash 8AD9409C
+  sample 20:
+    time = 700700
+    flags = 67108864
+    data = length 195, hash 4EFA8FFD
+  sample 21:
+    time = 900900
+    flags = 0
+    data = length 424, hash 5F9D809C
+  sample 22:
+    time = 834166
+    flags = 0
+    data = length 187, hash FB5137FD
+  sample 23:
+    time = 767433
+    flags = 67108864
+    data = length 162, hash 7B352EA
+  sample 24:
+    time = 800800
+    flags = 67108864
+    data = length 173, hash 42EBA5AC
+  sample 25:
+    time = 867533
+    flags = 67108864
+    data = length 205, hash 15D2186B
+  sample 26:
+    time = 1001000
+    flags = 0
+    data = length 342, hash 70E9AD53
+  sample 27:
+    time = 967633
+    flags = 0
+    data = length 200, hash 2C923950
+  sample 28:
+    time = 934266
+    flags = 67108864
+    data = length 150, hash 242935D6
+  sample 29:
+    time = 1167833
+    flags = 0
+    data = length 802, hash 60213433
+  sample 30:
+    time = 1101100
+    flags = 0
+    data = length 203, hash 21B384E5
+  sample 31:
+    time = 1034366
+    flags = 67108864
+    data = length 185, hash B2735872
+  sample 32:
+    time = 1067733
+    flags = 67108864
+    data = length 1310, hash 8D4F7E35
+  sample 33:
+    time = 1134466
+    flags = 67108864
+    data = length 176, hash D1D91C73
+  sample 34:
+    time = 1334666
+    flags = 0
+    data = length 258, hash 92056B3A
+  sample 35:
+    time = 1267933
+    flags = 0
+    data = length 214, hash 86C47DEC
+  sample 36:
+    time = 1201200
+    flags = 67108864
+    data = length 201, hash AFA1B6D4
+  sample 37:
+    time = 1234566
+    flags = 67108864
+    data = length 146, hash 3E1B1C72
+  sample 38:
+    time = 1301300
+    flags = 67108864
+    data = length 173, hash FC0B911D
+  sample 39:
+    time = 1401400
+    flags = 0
+    data = length 617, hash 35AAFB9
+  sample 40:
+    time = 1368033
+    flags = 67108864
+    data = length 179, hash E9838582
+  sample 41:
+    time = 1468133
+    flags = 0
+    data = length 395, hash 8F5373AC
+  sample 42:
+    time = 1434766
+    flags = 67108864
+    data = length 184, hash 10ED9B63
+  sample 43:
+    time = 1634966
+    flags = 0
+    data = length 460, hash ED8ECDE7
+  sample 44:
+    time = 1568233
+    flags = 0
+    data = length 184, hash BF273413
+  sample 45:
+    time = 1501500
+    flags = 67108864
+    data = length 162, hash FE32A96B
+  sample 46:
+    time = 1534866
+    flags = 67108864
+    data = length 201, hash BBB627F2
+  sample 47:
+    time = 1601600
+    flags = 67108864
+    data = length 169, hash AFD9804F
+  sample 48:
+    time = 1735066
+    flags = 0
+    data = length 668, hash 4B9FB50F
+  sample 49:
+    time = 1701700
+    flags = 0
+    data = length 219, hash E2A651A2
+  sample 50:
+    time = 1668333
+    flags = 67108864
+    data = length 162, hash ECB52FD2
+  sample 51:
+    time = 1901900
+    flags = 0
+    data = length 405, hash 7DA640B4
+  sample 52:
+    time = 1835166
+    flags = 0
+    data = length 188, hash 46D5E9D4
+  sample 53:
+    time = 1768433
+    flags = 67108864
+    data = length 167, hash 65F87E6A
+  sample 54:
+    time = 1801800
+    flags = 67108864
+    data = length 161, hash 2E37B5E0
+  sample 55:
+    time = 1868533
+    flags = 67108864
+    data = length 199, hash D63CF14E
+  sample 56:
+    time = 2002000
+    flags = 0
+    data = length 282, hash E748555D
+  sample 57:
+    time = 1968633
+    flags = 0
+    data = length 199, hash F341EE39
+  sample 58:
+    time = 1935266
+    flags = 0
+    data = length 166, hash 32F07BFF
+  sample 59:
+    time = 2068733
+    flags = 0
+    data = length 2246, hash E07DFCAE
+  sample 60:
+    time = 2035366
+    flags = 67108864
+    data = length 53, hash D9F70BD5
+track 100:
+  total output bytes = 45
+  sample count = 2
+  format 0:
+    sampleMimeType = application/cea-608
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 36, hash 6E15DFAD
+  sample 1:
+    time = 1935266
+    flags = 1
+    data = length 9, hash 604EC6AA
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/fragmented_captions_h265.mp4.reading_within_gop_sample_dependencies.unknown_length.dump

@@ -0,0 +1,283 @@
+seekMap:
+  isSeekable = false
+  duration = UNSET TIME
+  getPosition(0) = [[timeUs=0, position=3203]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 24020
+  sample count = 61
+  format 0:
+    id = 1
+    containerMimeType = video/mp4
+    sampleMimeType = video/hevc
+    codecs = hvc1.1.6.L60.90
+    maxNumReorderSamples = 2
+    width = 416
+    height = 234
+    maxSubLayers = 1
+    colorInfo:
+      colorSpace = 2
+      colorRange = 2
+      colorTransfer = 3
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    initializationData:
+      data = length 2426, hash 25737613
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 5220, hash 5F0CCC08
+  sample 1:
+    time = 233566
+    flags = 0
+    data = length 615, hash CBFF4D3C
+  sample 2:
+    time = 166833
+    flags = 0
+    data = length 217, hash B600557D
+  sample 3:
+    time = 100100
+    flags = 67108864
+    data = length 177, hash 99B5D77F
+  sample 4:
+    time = 133466
+    flags = 67108864
+    data = length 190, hash AF7F85A2
+  sample 5:
+    time = 200200
+    flags = 67108864
+    data = length 225, hash BA8F2E19
+  sample 6:
+    time = 333666
+    flags = 0
+    data = length 330, hash DCBABAE8
+  sample 7:
+    time = 300300
+    flags = 0
+    data = length 183, hash 15A2DB25
+  sample 8:
+    time = 266933
+    flags = 67108864
+    data = length 154, hash 81018D0E
+  sample 9:
+    time = 400400
+    flags = 0
+    data = length 576, hash C9C5389F
+  sample 10:
+    time = 367033
+    flags = 67108864
+    data = length 173, hash 2D567C7E
+  sample 11:
+    time = 500500
+    flags = 0
+    data = length 356, hash A6D37B60
+  sample 12:
+    time = 467133
+    flags = 0
+    data = length 188, hash 11AB1597
+  sample 13:
+    time = 433766
+    flags = 67108864
+    data = length 174, hash 18C46868
+  sample 14:
+    time = 567233
+    flags = 0
+    data = length 331, hash EEA184AF
+  sample 15:
+    time = 533866
+    flags = 67108864
+    data = length 213, hash 3FAB17EE
+  sample 16:
+    time = 667333
+    flags = 0
+    data = length 371, hash D0ECAE97
+  sample 17:
+    time = 633966
+    flags = 0
+    data = length 179, hash 796F24AF
+  sample 18:
+    time = 600600
+    flags = 67108864
+    data = length 154, hash FD43F0C6
+  sample 19:
+    time = 734066
+    flags = 0
+    data = length 622, hash 8AD9409C
+  sample 20:
+    time = 700700
+    flags = 67108864
+    data = length 195, hash 4EFA8FFD
+  sample 21:
+    time = 900900
+    flags = 0
+    data = length 424, hash 5F9D809C
+  sample 22:
+    time = 834166
+    flags = 0
+    data = length 187, hash FB5137FD
+  sample 23:
+    time = 767433
+    flags = 67108864
+    data = length 162, hash 7B352EA
+  sample 24:
+    time = 800800
+    flags = 67108864
+    data = length 173, hash 42EBA5AC
+  sample 25:
+    time = 867533
+    flags = 67108864
+    data = length 205, hash 15D2186B
+  sample 26:
+    time = 1001000
+    flags = 0
+    data = length 342, hash 70E9AD53
+  sample 27:
+    time = 967633
+    flags = 0
+    data = length 200, hash 2C923950
+  sample 28:
+    time = 934266
+    flags = 67108864
+    data = length 150, hash 242935D6
+  sample 29:
+    time = 1167833
+    flags = 0
+    data = length 802, hash 60213433
+  sample 30:
+    time = 1101100
+    flags = 0
+    data = length 203, hash 21B384E5
+  sample 31:
+    time = 1034366
+    flags = 67108864
+    data = length 185, hash B2735872
+  sample 32:
+    time = 1067733
+    flags = 67108864
+    data = length 1310, hash 8D4F7E35
+  sample 33:
+    time = 1134466
+    flags = 67108864
+    data = length 176, hash D1D91C73
+  sample 34:
+    time = 1334666
+    flags = 0
+    data = length 258, hash 92056B3A
+  sample 35:
+    time = 1267933
+    flags = 0
+    data = length 214, hash 86C47DEC
+  sample 36:
+    time = 1201200
+    flags = 67108864
+    data = length 201, hash AFA1B6D4
+  sample 37:
+    time = 1234566
+    flags = 67108864
+    data = length 146, hash 3E1B1C72
+  sample 38:
+    time = 1301300
+    flags = 67108864
+    data = length 173, hash FC0B911D
+  sample 39:
+    time = 1401400
+    flags = 0
+    data = length 617, hash 35AAFB9
+  sample 40:
+    time = 1368033
+    flags = 67108864
+    data = length 179, hash E9838582
+  sample 41:
+    time = 1468133
+    flags = 0
+    data = length 395, hash 8F5373AC
+  sample 42:
+    time = 1434766
+    flags = 67108864
+    data = length 184, hash 10ED9B63
+  sample 43:
+    time = 1634966
+    flags = 0
+    data = length 460, hash ED8ECDE7
+  sample 44:
+    time = 1568233
+    flags = 0
+    data = length 184, hash BF273413
+  sample 45:
+    time = 1501500
+    flags = 67108864
+    data = length 162, hash FE32A96B
+  sample 46:
+    time = 1534866
+    flags = 67108864
+    data = length 201, hash BBB627F2
+  sample 47:
+    time = 1601600
+    flags = 67108864
+    data = length 169, hash AFD9804F
+  sample 48:
+    time = 1735066
+    flags = 0
+    data = length 668, hash 4B9FB50F
+  sample 49:
+    time = 1701700
+    flags = 0
+    data = length 219, hash E2A651A2
+  sample 50:
+    time = 1668333
+    flags = 67108864
+    data = length 162, hash ECB52FD2
+  sample 51:
+    time = 1901900
+    flags = 0
+    data = length 405, hash 7DA640B4
+  sample 52:
+    time = 1835166
+    flags = 0
+    data = length 188, hash 46D5E9D4
+  sample 53:
+    time = 1768433
+    flags = 67108864
+    data = length 167, hash 65F87E6A
+  sample 54:
+    time = 1801800
+    flags = 67108864
+    data = length 161, hash 2E37B5E0
+  sample 55:
+    time = 1868533
+    flags = 67108864
+    data = length 199, hash D63CF14E
+  sample 56:
+    time = 2002000
+    flags = 0
+    data = length 282, hash E748555D
+  sample 57:
+    time = 1968633
+    flags = 0
+    data = length 199, hash F341EE39
+  sample 58:
+    time = 1935266
+    flags = 0
+    data = length 166, hash 32F07BFF
+  sample 59:
+    time = 2068733
+    flags = 0
+    data = length 2246, hash E07DFCAE
+  sample 60:
+    time = 2035366
+    flags = 67108864
+    data = length 53, hash D9F70BD5
+track 100:
+  total output bytes = 45
+  sample count = 2
+  format 0:
+    sampleMimeType = application/cea-608
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 36, hash 6E15DFAD
+  sample 1:
+    time = 1935266
+    flags = 1
+    data = length 9, hash 604EC6AA
+tracksEnded = true

libraries/test_data/src/test/assets/extractordumps/mp4/fragmented_captions_h265.mp4.unknown_length.dump

@@ -0,0 +1,283 @@
+seekMap:
+  isSeekable = false
+  duration = UNSET TIME
+  getPosition(0) = [[timeUs=0, position=3203]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 24020
+  sample count = 61
+  format 0:
+    id = 1
+    containerMimeType = video/mp4
+    sampleMimeType = video/hevc
+    codecs = hvc1.1.6.L60.90
+    maxNumReorderSamples = 2
+    width = 416
+    height = 234
+    maxSubLayers = 1
+    colorInfo:
+      colorSpace = 2
+      colorRange = 2
+      colorTransfer = 3
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    initializationData:
+      data = length 2426, hash 25737613
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 5220, hash 5F0CCC08
+  sample 1:
+    time = 233566
+    flags = 0
+    data = length 615, hash CBFF4D3C
+  sample 2:
+    time = 166833
+    flags = 0
+    data = length 217, hash B600557D
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 177, hash 99B5D77F
+  sample 4:
+    time = 133466
+    flags = 0
+    data = length 190, hash AF7F85A2
+  sample 5:
+    time = 200200
+    flags = 0
+    data = length 225, hash BA8F2E19
+  sample 6:
+    time = 333666
+    flags = 0
+    data = length 330, hash DCBABAE8
+  sample 7:
+    time = 300300
+    flags = 0
+    data = length 183, hash 15A2DB25
+  sample 8:
+    time = 266933
+    flags = 0
+    data = length 154, hash 81018D0E
+  sample 9:
+    time = 400400
+    flags = 0
+    data = length 576, hash C9C5389F
+  sample 10:
+    time = 367033
+    flags = 0
+    data = length 173, hash 2D567C7E
+  sample 11:
+    time = 500500
+    flags = 0
+    data = length 356, hash A6D37B60
+  sample 12:
+    time = 467133
+    flags = 0
+    data = length 188, hash 11AB1597
+  sample 13:
+    time = 433766
+    flags = 0
+    data = length 174, hash 18C46868
+  sample 14:
+    time = 567233
+    flags = 0
+    data = length 331, hash EEA184AF
+  sample 15:
+    time = 533866
+    flags = 0
+    data = length 213, hash 3FAB17EE
+  sample 16:
+    time = 667333
+    flags = 0
+    data = length 371, hash D0ECAE97
+  sample 17:
+    time = 633966
+    flags = 0
+    data = length 179, hash 796F24AF
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 154, hash FD43F0C6
+  sample 19:
+    time = 734066
+    flags = 0
+    data = length 622, hash 8AD9409C
+  sample 20:
+    time = 700700
+    flags = 0
+    data = length 195, hash 4EFA8FFD
+  sample 21:
+    time = 900900
+    flags = 0
+    data = length 424, hash 5F9D809C
+  sample 22:
+    time = 834166
+    flags = 0
+    data = length 187, hash FB5137FD
+  sample 23:
+    time = 767433
+    flags = 0
+    data = length 162, hash 7B352EA
+  sample 24:
+    time = 800800
+    flags = 0
+    data = length 173, hash 42EBA5AC
+  sample 25:
+    time = 867533
+    flags = 0
+    data = length 205, hash 15D2186B
+  sample 26:
+    time = 1001000
+    flags = 0
+    data = length 342, hash 70E9AD53
+  sample 27:
+    time = 967633
+    flags = 0
+    data = length 200, hash 2C923950
+  sample 28:
+    time = 934266
+    flags = 0
+    data = length 150, hash 242935D6
+  sample 29:
+    time = 1167833
+    flags = 0
+    data = length 802, hash 60213433
+  sample 30:
+    time = 1101100
+    flags = 0
+    data = length 203, hash 21B384E5
+  sample 31:
+    time = 1034366
+    flags = 0
+    data = length 185, hash B2735872
+  sample 32:
+    time = 1067733
+    flags = 0
+    data = length 1310, hash 8D4F7E35
+  sample 33:
+    time = 1134466
+    flags = 0
+    data = length 176, hash D1D91C73
+  sample 34:
+    time = 1334666
+    flags = 0
+    data = length 258, hash 92056B3A
+  sample 35:
+    time = 1267933
+    flags = 0
+    data = length 214, hash 86C47DEC
+  sample 36:
+    time = 1201200
+    flags = 0
+    data = length 201, hash AFA1B6D4
+  sample 37:
+    time = 1234566
+    flags = 0
+    data = length 146, hash 3E1B1C72
+  sample 38:
+    time = 1301300
+    flags = 0
+    data = length 173, hash FC0B911D
+  sample 39:
+    time = 1401400
+    flags = 0
+    data = length 617, hash 35AAFB9
+  sample 40:
+    time = 1368033
+    flags = 0
+    data = length 179, hash E9838582
+  sample 41:
+    time = 1468133
+    flags = 0
+    data = length 395, hash 8F5373AC
+  sample 42:
+    time = 1434766
+    flags = 0
+    data = length 184, hash 10ED9B63
+  sample 43:
+    time = 1634966
+    flags = 0
+    data = length 460, hash ED8ECDE7
+  sample 44:
+    time = 1568233
+    flags = 0
+    data = length 184, hash BF273413
+  sample 45:
+    time = 1501500
+    flags = 0
+    data = length 162, hash FE32A96B
+  sample 46:
+    time = 1534866
+    flags = 0
+    data = length 201, hash BBB627F2
+  sample 47:
+    time = 1601600
+    flags = 0
+    data = length 169, hash AFD9804F
+  sample 48:
+    time = 1735066
+    flags = 0
+    data = length 668, hash 4B9FB50F
+  sample 49:
+    time = 1701700
+    flags = 0
+    data = length 219, hash E2A651A2
+  sample 50:
+    time = 1668333
+    flags = 0
+    data = length 162, hash ECB52FD2
+  sample 51:
+    time = 1901900
+    flags = 0
+    data = length 405, hash 7DA640B4
+  sample 52:
+    time = 1835166
+    flags = 0
+    data = length 188, hash 46D5E9D4
+  sample 53:
+    time = 1768433
+    flags = 0
+    data = length 167, hash 65F87E6A
+  sample 54:
+    time = 1801800
+    flags = 0
+    data = length 161, hash 2E37B5E0
+  sample 55:
+    time = 1868533
+    flags = 0
+    data = length 199, hash D63CF14E
+  sample 56:
+    time = 2002000
+    flags = 0
+    data = length 282, hash E748555D
+  sample 57:
+    time = 1968633
+    flags = 0
+    data = length 199, hash F341EE39
+  sample 58:
+    time = 1935266
+    flags = 0
+    data = length 166, hash 32F07BFF
+  sample 59:
+    time = 2068733
+    flags = 0
+    data = length 2246, hash E07DFCAE
+  sample 60:
+    time = 2035366
+    flags = 0
+    data = length 53, hash D9F70BD5
+track 100:
+  total output bytes = 45
+  sample count = 2
+  format 0:
+    sampleMimeType = application/cea-608
+  sample 0:
+    time = 66733
+    flags = 1
+    data = length 36, hash 6E15DFAD
+  sample 1:
+    time = 1935266
+    flags = 1
+    data = length 9, hash 604EC6AA
+tracksEnded = true