Fix parsing of language code in mdhd box

rohitjoins · copybara-github · commit 2a4cbc3be47b · 2025-03-19T11:17:33.000-07:00
Language codes in the `mdhd` box are stored as three 5-bit values, each representing a letter ('a' to 'z') using an offset of `0x60`. If the decoded characters are not in this range, the language should be treated as undefined.

PiperOrigin-RevId: 738470544
diff --git a/libraries/extractor/src/main/java/androidx/media3/extractor/Ac3Util.java b/libraries/extractor/src/main/java/androidx/media3/extractor/Ac3Util.java
@@ -169,12 +169,15 @@ private SyncFrameInfo(
    *
    * @param data The AC3SpecificBox to parse.
    * @param trackId The track identifier to set on the format.
-   * @param language The language to set on the format.
+   * @param language The language to set on the format, or {@code null} if unset.
    * @param drmInitData {@link DrmInitData} to be included in the format.
    * @return The AC-3 format parsed from data in the header.
    */
   public static Format parseAc3AnnexFFormat(
-      ParsableByteArray data, String trackId, String language, @Nullable DrmInitData drmInitData) {
+      ParsableByteArray data,
+      String trackId,
+      @Nullable String language,
+      @Nullable DrmInitData drmInitData) {
     ParsableBitArray dataBitArray = new ParsableBitArray();
     dataBitArray.reset(data);
 
@@ -208,12 +211,15 @@ public static Format parseAc3AnnexFFormat(
    *
    * @param data The EC3SpecificBox to parse.
    * @param trackId The track identifier to set on the format.
-   * @param language The language to set on the format.
+   * @param language The language to set on the format, or {@code null} if unset.
    * @param drmInitData {@link DrmInitData} to be included in the format.
    * @return The E-AC-3 format parsed from data in the header.
    */
   public static Format parseEAc3AnnexFFormat(
-      ParsableByteArray data, String trackId, String language, @Nullable DrmInitData drmInitData) {
+      ParsableByteArray data,
+      String trackId,
+      @Nullable String language,
+      @Nullable DrmInitData drmInitData) {
     ParsableBitArray dataBitArray = new ParsableBitArray();
     dataBitArray.reset(data);
 
diff --git a/libraries/extractor/src/main/java/androidx/media3/extractor/Ac4Util.java b/libraries/extractor/src/main/java/androidx/media3/extractor/Ac4Util.java
@@ -163,14 +163,17 @@ private SyncFrameInfo(
    *
    * @param data The AC4SpecificBox to parse.
    * @param trackId The track identifier to set on the format.
-   * @param language The language to set on the format.
+   * @param language The language to set on the format, or {@code null} if unset.
    * @param drmInitData {@link DrmInitData} to be included in the format.
    * @return The AC-4 format parsed from data in the header.
    * @throws ParserException If an unsupported container feature is encountered while parsing AC-4
    *     Annex E.
    */
   public static Format parseAc4AnnexEFormat(
-      ParsableByteArray data, String trackId, String language, @Nullable DrmInitData drmInitData)
+      ParsableByteArray data,
+      String trackId,
+      @Nullable String language,
+      @Nullable DrmInitData drmInitData)
       throws ParserException {
     ParsableBitArray dataBitArray = new ParsableBitArray();
     dataBitArray.reset(data);
diff --git a/libraries/extractor/src/main/java/androidx/media3/extractor/mp4/BoxParser.java b/libraries/extractor/src/main/java/androidx/media3/extractor/mp4/BoxParser.java
@@ -997,22 +997,34 @@ private static MdhdData parseMdhd(ParsableByteArray mdhd) {
         mediaDurationUs = Util.scaleLargeTimestamp(mediaDuration, C.MICROS_PER_SECOND, timescale);
       }
     }
-    int languageCode = mdhd.readUnsignedShort();
-    String language =
-        ""
-            + (char) (((languageCode >> 10) & 0x1F) + 0x60)
-            + (char) (((languageCode >> 5) & 0x1F) + 0x60)
-            + (char) ((languageCode & 0x1F) + 0x60);
+
+    String language = getLanguageFromCode(/* languageCode= */ mdhd.readUnsignedShort());
     return new MdhdData(timescale, mediaDurationUs, language);
   }
 
+  @Nullable
+  private static String getLanguageFromCode(int languageCode) {
+    char[] chars = {
+      (char) (((languageCode >> 10) & 0x1F) + 0x60),
+      (char) (((languageCode >> 5) & 0x1F) + 0x60),
+      (char) ((languageCode & 0x1F) + 0x60)
+    };
+
+    for (char c : chars) {
+      if (c < 'a' || c > 'z') {
+        return null;
+      }
+    }
+    return new String(chars);
+  }
+
   /**
    * Parses a stsd atom (defined in ISO/IEC 14496-12).
    *
    * @param stsd The stsd atom to decode.
    * @param trackId The track's identifier in its container.
    * @param rotationDegrees The rotation of the track in degrees.
-   * @param language The language of the track.
+   * @param language The language of the track, or {@code null} if unset.
    * @param drmInitData {@link DrmInitData} to be included in the format, or {@code null}.
    * @param isQuickTime True for QuickTime media. False otherwise.
    * @return An object containing the parsed data.
@@ -1021,7 +1033,7 @@ private static StsdData parseStsd(
       ParsableByteArray stsd,
       int trackId,
       int rotationDegrees,
-      String language,
+      @Nullable String language,
       @Nullable DrmInitData drmInitData,
       boolean isQuickTime)
       throws ParserException {
@@ -1125,7 +1137,7 @@ private static void parseTextSampleEntry(
       int position,
       int atomSize,
       int trackId,
-      String language,
+      @Nullable String language,
       StsdData out) {
     parent.setPosition(position + Mp4Box.HEADER_SIZE + StsdData.STSD_HEADER_SIZE);
 
@@ -1820,7 +1832,7 @@ private static void parseAudioSampleEntry(
       int position,
       int size,
       int trackId,
-      String language,
+      @Nullable String language,
       boolean isQuickTime,
       @Nullable DrmInitData drmInitData,
       StsdData out,
@@ -2593,9 +2605,9 @@ public EyesData(StriData striData) {
   private static final class MdhdData {
     private final long timescale;
     private final long mediaDurationUs;
-    private final String language;
+    @Nullable private final String language;
 
-    public MdhdData(long timescale, long mediaDurationUs, String language) {
+    public MdhdData(long timescale, long mediaDurationUs, @Nullable String language) {
       this.timescale = timescale;
       this.mediaDurationUs = mediaDurationUs;
       this.language = language;
diff --git a/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.0.dump b/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.0.dump
@@ -12,7 +12,6 @@ track 0:
     sampleMimeType = audio/opus
     channelCount = 2
     sampleRate = 16000
-    language = 
     initializationData:
       data = length 19, hash 4034F23B
       data = length 8, hash 94446F01
diff --git a/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.reading_within_gop_sample_dependencies.0.dump b/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.reading_within_gop_sample_dependencies.0.dump
@@ -12,7 +12,6 @@ track 0:
     sampleMimeType = audio/opus
     channelCount = 2
     sampleRate = 16000
-    language = 
     initializationData:
       data = length 19, hash 4034F23B
       data = length 8, hash 94446F01
diff --git a/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.reading_within_gop_sample_dependencies.unknown_length.dump b/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.reading_within_gop_sample_dependencies.unknown_length.dump
@@ -12,7 +12,6 @@ track 0:
     sampleMimeType = audio/opus
     channelCount = 2
     sampleRate = 16000
-    language = 
     initializationData:
       data = length 19, hash 4034F23B
       data = length 8, hash 94446F01
diff --git a/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.unknown_length.dump b/libraries/test_data/src/test/assets/extractordumps/mp4/sample_opus_fragmented.mp4.unknown_length.dump
@@ -12,7 +12,6 @@ track 0:
     sampleMimeType = audio/opus
     channelCount = 2
     sampleRate = 16000
-    language = 
     initializationData:
       data = length 19, hash 4034F23B
       data = length 8, hash 94446F01
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_1ch_16kHz_q10_vorbis.ogg.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_1ch_16kHz_q10_vorbis.ogg.dump
@@ -18,7 +18,6 @@ track 0:
     maxInputSize = 295
     channelCount = 1
     sampleRate = 16000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 30, hash C22462B1
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_1ch_16kHz_q10_vorbis.ogg_fragmented.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_1ch_16kHz_q10_vorbis.ogg_fragmented.dump
@@ -13,7 +13,6 @@ track 0:
     sampleMimeType = audio/vorbis
     channelCount = 1
     sampleRate = 16000
-    language = ```
     initializationData:
       data = length 30, hash C22462B1
       data = length 3539, hash F8106892
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_2ch_44kHz.wav.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_2ch_44kHz.wav.dump
@@ -18,7 +18,6 @@ track 0:
     channelCount = 2
     sampleRate = 44100
     pcmEncoding = 2
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
   sample 0:
     time = 0
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_2ch_44kHz.wav_fragmented.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_2ch_44kHz.wav_fragmented.dump
@@ -13,7 +13,6 @@ track 0:
     channelCount = 2
     sampleRate = 44100
     pcmEncoding = 2
-    language = ```
   sample 0:
     time = 0
     flags = 1
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_6ch_8kHz_opus.ogg.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_6ch_8kHz_opus.ogg.dump
@@ -17,7 +17,6 @@ track 0:
     maxInputSize = 1185
     channelCount = 6
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 27, hash 9EE6F879
diff --git a/libraries/test_data/src/test/assets/muxerdumps/bbb_6ch_8kHz_opus.ogg_fragmented.dump b/libraries/test_data/src/test/assets/muxerdumps/bbb_6ch_8kHz_opus.ogg_fragmented.dump
@@ -12,7 +12,6 @@ track 0:
     sampleMimeType = audio/opus
     channelCount = 6
     sampleRate = 48000
-    language = ```
     initializationData:
       data = length 27, hash 9EE6F879
       data = length 8, hash CA22068C
diff --git a/libraries/test_data/src/test/assets/muxerdumps/h265_with_metadata_track.mp4.dump b/libraries/test_data/src/test/assets/muxerdumps/h265_with_metadata_track.mp4.dump
@@ -42,7 +42,6 @@ track 1:
     maxInputSize = 627
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
diff --git a/libraries/test_data/src/test/assets/muxerdumps/h265_with_metadata_track.mp4_fragmented.dump b/libraries/test_data/src/test/assets/muxerdumps/h265_with_metadata_track.mp4_fragmented.dump
@@ -34,7 +34,6 @@ track 1:
     codecs = mp4a.40.2
     channelCount = 2
     sampleRate = 48000
-    language = ```
     initializationData:
       data = length 2, hash 560
   sample 0:
diff --git a/libraries/test_data/src/test/assets/muxerdumps/hdr10-720p.mp4.dump b/libraries/test_data/src/test/assets/muxerdumps/hdr10-720p.mp4.dump
@@ -552,7 +552,6 @@ track 1:
     maxInputSize = 877
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
diff --git a/libraries/test_data/src/test/assets/muxerdumps/hdr10-720p.mp4_fragmented.dump b/libraries/test_data/src/test/assets/muxerdumps/hdr10-720p.mp4_fragmented.dump
@@ -543,7 +543,6 @@ track 1:
     codecs = mp4a.40.2
     channelCount = 2
     sampleRate = 48000
-    language = ```
     initializationData:
       data = length 2, hash 560
   sample 0:
diff --git a/libraries/test_data/src/test/assets/muxerdumps/partial_hdr10-720p.mp4.dump b/libraries/test_data/src/test/assets/muxerdumps/partial_hdr10-720p.mp4.dump
@@ -416,7 +416,6 @@ track 1:
     maxInputSize = 877
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
diff --git a/libraries/test_data/src/test/assets/muxerdumps/sample_batching_and_attempt_streamable_output_disabled_hdr10-720p.mp4.dump b/libraries/test_data/src/test/assets/muxerdumps/sample_batching_and_attempt_streamable_output_disabled_hdr10-720p.mp4.dump
@@ -552,7 +552,6 @@ track 1:
     maxInputSize = 877
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
diff --git a/libraries/test_data/src/test/assets/muxerdumps/sample_batching_disabled_hdr10-720p.mp4.dump b/libraries/test_data/src/test/assets/muxerdumps/sample_batching_disabled_hdr10-720p.mp4.dump
@@ -552,7 +552,6 @@ track 1:
     maxInputSize = 877
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=100000000, modification time=500000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/bbb_1ch_16kHz_q10_vorbis.ogg/transmuxed_with_inappmuxer.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/bbb_1ch_16kHz_q10_vorbis.ogg/transmuxed_with_inappmuxer.dump
@@ -18,7 +18,6 @@ track 0:
     maxInputSize = 295
     channelCount = 1
     sampleRate = 16000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=3000000000, modification time=4000000000, timescale=10000]
     initializationData:
       data = length 30, hash C22462B1
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/bbb_6ch_8kHz_opus.ogg/transmuxed_with_inappmuxer.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/bbb_6ch_8kHz_opus.ogg/transmuxed_with_inappmuxer.dump
@@ -17,7 +17,6 @@ track 0:
     maxInputSize = 1185
     channelCount = 6
     sampleRate = 48000
-    language = ```
     metadata = entries=[Mp4Timestamp: creation time=3000000000, modification time=4000000000, timescale=10000]
     initializationData:
       data = length 27, hash 9EE6F879
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/h265_with_metadata_track.mp4/transmuxed_with_inappmuxer.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/h265_with_metadata_track.mp4/transmuxed_with_inappmuxer.dump
@@ -57,7 +57,6 @@ track 1:
     maxInputSize = 627
     channelCount = 2
     sampleRate = 48000
-    language = ```
     metadata = entries=[xyz: latitude=51.5932, longitude=-0.2431, Mp4Timestamp: creation time=3000000000, modification time=4000000000, timescale=10000]
     initializationData:
       data = length 2, hash 560
