handle meta atom as a full box when parsing mp4

Issues: #5698, #5694
PiperOrigin-RevId: 241762106

RELEASENOTES.md

@@ -96,6 +96,9 @@
 * Use full BCP 47 language tags in `Format`.
 * Take byte offset into account when unsynchronizing an id3 frame
   ([#5673](https://github.com/google/ExoPlayer/issues/5673)).
+* Handle meta atom as a full box when parsing mp4
+  ([#5698](https://github.com/google/ExoPlayer/issues/5698),
+  [#5694](https://github.com/google/ExoPlayer/issues/5694)).
 
 ### 2.9.6 ###
 

library/core/src/main/java/com/google/android/exoplayer2/extractor/mp4/Mp4Extractor.java

@@ -309,6 +309,9 @@ private boolean readAtomHeader(ExtractorInput input) throws IOException, Interru
       if (atomSize == atomHeaderBytesRead) {
         processAtomEnded(endPosition);
       } else {
+        if (atomType == Atom.TYPE_meta) {
+          maybeSkipRemainingMetaAtomHeaderBytes(input);
+        }
         // Start reading the first child atom.
         enterReadingAtomHeaderState();
       }
@@ -643,6 +646,32 @@ private void updateSampleIndices(long timeUs) {
     }
   }
 
+  /**
+   * Possibly skips the version and flags fields (1+3 byte) of a full meta atom of the {@code
+   * input}.
+   *
+   * <p>Atoms of type {@link Atom#TYPE_meta} are defined to be full atoms which have four additional
+   * bytes for a version and a flags field (see 4.2 'Object Structure' in ISO/IEC 14496-12:2005).
+   * QuickTime do not have such a full box structure. Since some of these files are encoded wrongly,
+   * we can't rely on the file type though. Instead we must check the 8 bytes after the common
+   * header bytes ourselves.
+   */
+  private void maybeSkipRemainingMetaAtomHeaderBytes(ExtractorInput input)
+      throws IOException, InterruptedException {
+    scratch.reset(8);
+    // Peek the next 8 bytes which can be either
+    // (iso) [1 byte version + 3 bytes flags][4 byte size of next atom]
+    // (qt)  [4 byte size of next atom      ][4 byte hdlr atom type   ]
+    // In case of (iso) we need to skip the next 4 bytes.
+    input.peekFully(scratch.data, 0, 8);
+    scratch.skipBytes(4);
+    if (scratch.readInt() == Atom.TYPE_hdlr) {
+      input.resetPeekPosition();
+    } else {
+      input.skipFully(4);
+    }
+  }
+
   /**
    * For each sample of each track, calculates accumulated size of all samples which need to be read
    * before this sample can be used.