Handle sample size mismatch in MP4 extractors

#minor-release

PiperOrigin-RevId: 351756333

RELEASENOTES.md

@@ -106,6 +106,8 @@
     *   Populate codecs string for H.265/HEVC in MP4, Matroska and MPEG-TS
     *   Populate codecs string for H.265/HEVC in MP4, Matroska and MPEG-TS
         streams to allow decoder capability checks based on codec profile/level
         streams to allow decoder capability checks based on codec profile/level
         ([#8393](https://github.com/google/ExoPlayer/issues/8393)).
         ([#8393](https://github.com/google/ExoPlayer/issues/8393)).
+    *   Handle sample size mismatches between raw audio PCM encoding and `stsz`
+        sample size in MP4 extractors.
 *   Track selection:
 *   Track selection:
     *   Allow parallel adaptation for video and audio
     *   Allow parallel adaptation for video and audio
         ([#5111](https://github.com/google/ExoPlayer/issues/5111)).
         ([#5111](https://github.com/google/ExoPlayer/issues/5111)).

library/common/src/main/java/com/google/android/exoplayer2/util/Util.java

@@ -1674,24 +1674,23 @@ public final class Util {
   }
   }
 
 
   /**
   /**
-   * Returns the frame size for audio with {@code channelCount} channels in the specified encoding.
+   * Returns the sample size for audio in the specified encoding.
    *
    *
    * @param pcmEncoding The encoding of the audio data.
    * @param pcmEncoding The encoding of the audio data.
-   * @param channelCount The channel count.
-   * @return The size of one audio frame in bytes.
+   * @return The size of one audio sample in bytes.
    */
    */
-  public static int getPcmFrameSize(@C.PcmEncoding int pcmEncoding, int channelCount) {
+  public static int getPcmSampleSize(@C.PcmEncoding int pcmEncoding) {
     switch (pcmEncoding) {
     switch (pcmEncoding) {
       case C.ENCODING_PCM_8BIT:
       case C.ENCODING_PCM_8BIT:
-        return channelCount;
+        return 1;
       case C.ENCODING_PCM_16BIT:
       case C.ENCODING_PCM_16BIT:
       case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
       case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
-        return channelCount * 2;
+        return 2;
       case C.ENCODING_PCM_24BIT:
       case C.ENCODING_PCM_24BIT:
-        return channelCount * 3;
+        return 3;
       case C.ENCODING_PCM_32BIT:
       case C.ENCODING_PCM_32BIT:
       case C.ENCODING_PCM_FLOAT:
       case C.ENCODING_PCM_FLOAT:
-        return channelCount * 4;
+        return 4;
       case C.ENCODING_INVALID:
       case C.ENCODING_INVALID:
       case Format.NO_VALUE:
       case Format.NO_VALUE:
       default:
       default:
@@ -1700,6 +1699,17 @@ public final class Util {
   }
   }
 
 
   /**
   /**
+   * Returns the frame size for audio with {@code channelCount} channels in the specified encoding.
+   *
+   * @param pcmEncoding The encoding of the audio data.
+   * @param channelCount The channel count.
+   * @return The size of one audio frame in bytes.
+   */
+  public static int getPcmFrameSize(@C.PcmEncoding int pcmEncoding, int channelCount) {
+    return getPcmSampleSize(pcmEncoding) * channelCount;
+  }
+
+  /**
    * Returns the {@link C.AudioUsage} corresponding to the specified {@link C.StreamType}.
    * Returns the {@link C.AudioUsage} corresponding to the specified {@link C.StreamType}.
    */
    */
   @C.AudioUsage
   @C.AudioUsage

library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mp4/AtomParsers.java

@@ -315,7 +315,7 @@ import org.checkerframework.checker.nullness.compatqual.NullableType;
     SampleSizeBox sampleSizeBox;
     SampleSizeBox sampleSizeBox;
     @Nullable Atom.LeafAtom stszAtom = stblAtom.getLeafAtomOfType(Atom.TYPE_stsz);
     @Nullable Atom.LeafAtom stszAtom = stblAtom.getLeafAtomOfType(Atom.TYPE_stsz);
     if (stszAtom != null) {
     if (stszAtom != null) {
-      sampleSizeBox = new StszSampleSizeBox(stszAtom);
+      sampleSizeBox = new StszSampleSizeBox(stszAtom, track.format);
     } else {
     } else {
       @Nullable Atom.LeafAtom stz2Atom = stblAtom.getLeafAtomOfType(Atom.TYPE_stz2);
       @Nullable Atom.LeafAtom stz2Atom = stblAtom.getLeafAtomOfType(Atom.TYPE_stz2);
       if (stz2Atom == null) {
       if (stz2Atom == null) {
@@ -1720,10 +1720,25 @@ import org.checkerframework.checker.nullness.compatqual.NullableType;
     private final int sampleCount;
     private final int sampleCount;
     private final ParsableByteArray data;
     private final ParsableByteArray data;
 
 
-    public StszSampleSizeBox(Atom.LeafAtom stszAtom) {
+    public StszSampleSizeBox(Atom.LeafAtom stszAtom, Format trackFormat) {
       data = stszAtom.data;
       data = stszAtom.data;
       data.setPosition(Atom.FULL_HEADER_SIZE);
       data.setPosition(Atom.FULL_HEADER_SIZE);
       int fixedSampleSize = data.readUnsignedIntToInt();
       int fixedSampleSize = data.readUnsignedIntToInt();
+      if (MimeTypes.AUDIO_RAW.equals(trackFormat.sampleMimeType)) {
+        @C.PcmEncoding int pcmEncoding = trackFormat.pcmEncoding;
+        int pcmSampleSize = Util.getPcmSampleSize(pcmEncoding);
+        if (pcmSampleSize != fixedSampleSize) {
+          // The sample size from the stsz box is inconsistent with the PCM encoding derived from
+          // the stsd box. Choose the PCM encoding as source of truth [Internal ref: b/171627904].
+          Log.w(
+              TAG,
+              "Audio sample size mismatch. PCM encoding: "
+                  + pcmEncoding
+                  + ", stsz sample size = "
+                  + fixedSampleSize);
+          fixedSampleSize = pcmSampleSize;
+        }
+      }
       this.fixedSampleSize = fixedSampleSize == 0 ? C.LENGTH_UNSET : fixedSampleSize;
       this.fixedSampleSize = fixedSampleSize == 0 ? C.LENGTH_UNSET : fixedSampleSize;
       sampleCount = data.readUnsignedIntToInt();
       sampleCount = data.readUnsignedIntToInt();
     }
     }