JpegExtractor: support JFIF segment preceding Exif segment

#minor-release

PiperOrigin-RevId: 364561115

RELEASENOTES.md

@@ -43,6 +43,8 @@
 *   Extractors:
     *   Add support for `GContainer` and `GContainerItem` XMP namespace prefixes
         in JPEG motion photo parsing.
+    *   Allow JFIF APP0 marker segment preceding Exif APP1 segment in
+        `JpegExtractor`.
 *   Remove deprecated symbols:
     *   Remove `Player.DefaultEventListener`. Use `Player.EventListener`
         instead.
@@ -77,9 +79,8 @@
         media item and so that it is not triggered after a timeline change.
 *   IMA extension:
     *   Fix error caused by `AdPlaybackState` ad group times being cleared,
-        which can occur if the `ImaAdsLoader` is released while an ad is
-        pending loading
-        ([#8693](https://github.com/google/ExoPlayer/issues/8693)).
+        which can occur if the `ImaAdsLoader` is released while an ad is pending
+        loading ([#8693](https://github.com/google/ExoPlayer/issues/8693)).
     *   Upgrade IMA SDK dependency to 3.22.3, fixing an issue with
         `NullPointerExceptions` within `WebView` callbacks
         ([#8447](https://github.com/google/ExoPlayer/issues/8447)).

library/extractor/src/main/java/com/google/android/exoplayer2/extractor/jpeg/JpegExtractor.java

@@ -60,10 +60,11 @@ public final class JpegExtractor implements Extractor {
   private static final int STATE_READING_MOTION_PHOTO_VIDEO = 5;
   private static final int STATE_ENDED = 6;
 
-  private static final int JPEG_EXIF_HEADER_LENGTH = 12;
+  private static final int EXIF_ID_CODE_LENGTH = 6;
   private static final long EXIF_HEADER = 0x45786966; // Exif
   private static final int MARKER_SOI = 0xFFD8; // Start of image marker
   private static final int MARKER_SOS = 0xFFDA; // Start of scan (image data) marker
+  private static final int MARKER_APP0 = 0xFFE0; // Application data 0 marker
   private static final int MARKER_APP1 = 0xFFE1; // Application data 1 marker
   private static final String HEADER_XMP_APP1 = "http://ns.adobe.com/xap/1.0/";
 
@@ -85,21 +86,33 @@ public final class JpegExtractor implements Extractor {
   @Nullable private MotionPhotoMetadata motionPhotoMetadata;
   private @MonotonicNonNull ExtractorInput lastExtractorInput;
   private @MonotonicNonNull StartOffsetExtractorInput mp4ExtractorStartOffsetExtractorInput;
-  private @MonotonicNonNull Mp4Extractor mp4Extractor;
+  @Nullable private Mp4Extractor mp4Extractor;
 
   public JpegExtractor() {
-    scratch = new ParsableByteArray(JPEG_EXIF_HEADER_LENGTH);
+    scratch = new ParsableByteArray(EXIF_ID_CODE_LENGTH);
     mp4StartPosition = C.POSITION_UNSET;
   }
 
   @Override
   public boolean sniff(ExtractorInput input) throws IOException {
     // See ITU-T.81 (1992) subsection B.1.1.3 and Exif version 2.2 (2002) subsection 4.5.4.
-    input.peekFully(scratch.getData(), /* offset= */ 0, JPEG_EXIF_HEADER_LENGTH);
-    if (scratch.readUnsignedShort() != MARKER_SOI || scratch.readUnsignedShort() != MARKER_APP1) {
+    if (peekMarker(input) != MARKER_SOI) {
       return false;
     }
-    scratch.skipBytes(2); // Unused segment length
+    marker = peekMarker(input);
+    // Even though JFIF and Exif standards are incompatible in theory, Exif files often contain a
+    // JFIF APP0 marker segment preceding the Exif APP1 marker segment. Skip the JFIF segment if
+    // present.
+    if (marker == MARKER_APP0) {
+      advancePeekPositionToNextSegment(input);
+      marker = peekMarker(input);
+    }
+    if (marker != MARKER_APP1) {
+      return false;
+    }
+    input.advancePeekPosition(2); // Unused segment length
+    scratch.reset(/* limit= */ EXIF_ID_CODE_LENGTH);
+    input.peekFully(scratch.getData(), /* offset= */ 0, EXIF_ID_CODE_LENGTH);
     return scratch.readUnsignedInt() == EXIF_HEADER && scratch.readUnsignedShort() == 0; // Exif\0\0
   }
 
@@ -152,6 +165,7 @@ public final class JpegExtractor implements Extractor {
   public void seek(long position, long timeUs) {
     if (position == 0) {
       state = STATE_READING_MARKER;
+      mp4Extractor = null;
     } else if (state == STATE_READING_MOTION_PHOTO_VIDEO) {
       checkNotNull(mp4Extractor).seek(position, timeUs);
     }
@@ -164,6 +178,19 @@ public final class JpegExtractor implements Extractor {
     }
   }
 
+  private int peekMarker(ExtractorInput input) throws IOException {
+    scratch.reset(/* limit= */ 2);
+    input.peekFully(scratch.getData(), /* offset= */ 0, /* length= */ 2);
+    return scratch.readUnsignedShort();
+  }
+
+  private void advancePeekPositionToNextSegment(ExtractorInput input) throws IOException {
+    scratch.reset(/* limit= */ 2);
+    input.peekFully(scratch.getData(), /* offset= */ 0, /* length= */ 2);
+    int segmentLength = scratch.readUnsignedShort() - 2;
+    input.advancePeekPosition(segmentLength);
+  }
+
   private void readMarker(ExtractorInput input) throws IOException {
     scratch.reset(/* limit= */ 2);
     input.readFully(scratch.getData(), /* offset= */ 0, /* length= */ 2);

library/extractor/src/test/java/com/google/android/exoplayer2/extractor/jpeg/JpegExtractorTest.java

@@ -46,6 +46,14 @@ public final class JpegExtractorTest {
   }
 
   @Test
+  public void samplePixelMotionPhotoJfifSegmentShortened() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        JpegExtractor::new,
+        "media/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg",
+        simulationConfig);
+  }
+
+  @Test
   public void samplePixelMotionPhotoVideoRemovedShortened() throws Exception {
     ExtractorAsserts.assertBehavior(
         JpegExtractor::new,

testdata/src/test/assets/extractordumps/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg.0.dump

+seekMap:
+  isSeekable = true
+  duration = 867000
+  getPosition(0) = [[timeUs=0, position=6425]]
+  getPosition(1) = [[timeUs=0, position=6425]]
+  getPosition(433500) = [[timeUs=0, position=6425]]
+  getPosition(867000) = [[timeUs=0, position=6425]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 3865
+  sample count = 1
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.64000A
+    maxInputSize = 3895
+    width = 180
+    height = 120
+    pixelWidthHeightRatio = 0.5
+    initializationData:
+      data = length 32, hash 1F3D6E87
+      data = length 10, hash 7A0D0F2B
+  sample 0:
+    time = 0
+    flags = 536870913
+    data = length 3865, hash 5B0DEEC7
+track 1024:
+  total output bytes = 0
+  sample count = 0
+  format 0:
+    metadata = entries=[Motion photo metadata: photoStartPosition=0, photoSize=6377, photoPresentationTimestampUs=1232840, videoStartPosition=6377, videoSize=4686]
+tracksEnded = true

testdata/src/test/assets/extractordumps/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg.1.dump

+seekMap:
+  isSeekable = true
+  duration = 867000
+  getPosition(0) = [[timeUs=0, position=6425]]
+  getPosition(1) = [[timeUs=0, position=6425]]
+  getPosition(433500) = [[timeUs=0, position=6425]]
+  getPosition(867000) = [[timeUs=0, position=6425]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 3865
+  sample count = 1
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.64000A
+    maxInputSize = 3895
+    width = 180
+    height = 120
+    pixelWidthHeightRatio = 0.5
+    initializationData:
+      data = length 32, hash 1F3D6E87
+      data = length 10, hash 7A0D0F2B
+  sample 0:
+    time = 0
+    flags = 536870913
+    data = length 3865, hash 5B0DEEC7
+track 1024:
+  total output bytes = 0
+  sample count = 0
+  format 0:
+    metadata = entries=[Motion photo metadata: photoStartPosition=0, photoSize=6377, photoPresentationTimestampUs=1232840, videoStartPosition=6377, videoSize=4686]
+tracksEnded = true

testdata/src/test/assets/extractordumps/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg.2.dump

+seekMap:
+  isSeekable = true
+  duration = 867000
+  getPosition(0) = [[timeUs=0, position=6425]]
+  getPosition(1) = [[timeUs=0, position=6425]]
+  getPosition(433500) = [[timeUs=0, position=6425]]
+  getPosition(867000) = [[timeUs=0, position=6425]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 3865
+  sample count = 1
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.64000A
+    maxInputSize = 3895
+    width = 180
+    height = 120
+    pixelWidthHeightRatio = 0.5
+    initializationData:
+      data = length 32, hash 1F3D6E87
+      data = length 10, hash 7A0D0F2B
+  sample 0:
+    time = 0
+    flags = 536870913
+    data = length 3865, hash 5B0DEEC7
+track 1024:
+  total output bytes = 0
+  sample count = 0
+  format 0:
+    metadata = entries=[Motion photo metadata: photoStartPosition=0, photoSize=6377, photoPresentationTimestampUs=1232840, videoStartPosition=6377, videoSize=4686]
+tracksEnded = true

testdata/src/test/assets/extractordumps/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg.3.dump

+seekMap:
+  isSeekable = true
+  duration = 867000
+  getPosition(0) = [[timeUs=0, position=6425]]
+  getPosition(1) = [[timeUs=0, position=6425]]
+  getPosition(433500) = [[timeUs=0, position=6425]]
+  getPosition(867000) = [[timeUs=0, position=6425]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 3865
+  sample count = 1
+  format 0:
+    id = 1
+    sampleMimeType = video/avc
+    codecs = avc1.64000A
+    maxInputSize = 3895
+    width = 180
+    height = 120
+    pixelWidthHeightRatio = 0.5
+    initializationData:
+      data = length 32, hash 1F3D6E87
+      data = length 10, hash 7A0D0F2B
+  sample 0:
+    time = 0
+    flags = 536870913
+    data = length 3865, hash 5B0DEEC7
+track 1024:
+  total output bytes = 0
+  sample count = 0
+  format 0:
+    metadata = entries=[Motion photo metadata: photoStartPosition=0, photoSize=6377, photoPresentationTimestampUs=1232840, videoStartPosition=6377, videoSize=4686]
+tracksEnded = true

testdata/src/test/assets/extractordumps/jpeg/pixel-motion-photo-jfif-segment-shortened.jpg.unknown_length.dump

+seekMap:
+  isSeekable = false
+  duration = UNSET TIME
+  getPosition(0) = [[timeUs=0, position=0]]
+numberOfTracks = 1
+track 1024:
+  total output bytes = 0
+  sample count = 0
+  format 0:
+    metadata = entries=[]
+tracksEnded = true