Expect PresentationTime Discontinuity During Stream Transitions

PiperOrigin-RevId: 440378974

libraries/common/src/main/java/androidx/media3/common/C.java

@@ -585,8 +585,9 @@ public final class C {
 
   /**
    * Flags which can apply to a buffer containing a media sample. Possible flag values are {@link
-   * #BUFFER_FLAG_KEY_FRAME}, {@link #BUFFER_FLAG_END_OF_STREAM}, {@link #BUFFER_FLAG_LAST_SAMPLE},
+   * #BUFFER_FLAG_KEY_FRAME}, {@link #BUFFER_FLAG_END_OF_STREAM}, {@link #BUFFER_FLAG_FIRST_SAMPLE},
-   * {@link #BUFFER_FLAG_ENCRYPTED} and {@link #BUFFER_FLAG_DECODE_ONLY}.
+   * {@link #BUFFER_FLAG_LAST_SAMPLE}, {@link #BUFFER_FLAG_ENCRYPTED} and {@link
+   * #BUFFER_FLAG_DECODE_ONLY}.
    */
   @UnstableApi
   @Documented
@@ -597,6 +598,7 @@ public final class C {
       value = {
         BUFFER_FLAG_KEY_FRAME,
         BUFFER_FLAG_END_OF_STREAM,
+        BUFFER_FLAG_FIRST_SAMPLE,
         BUFFER_FLAG_HAS_SUPPLEMENTAL_DATA,
         BUFFER_FLAG_LAST_SAMPLE,
         BUFFER_FLAG_ENCRYPTED,
@@ -608,6 +610,8 @@ public final class C {
   /** Flag for empty buffers that signal that the end of the stream was reached. */
   @UnstableApi
   public static final int BUFFER_FLAG_END_OF_STREAM = MediaCodec.BUFFER_FLAG_END_OF_STREAM;
+  /** Indicates that a buffer is known to contain the first media sample of the stream. */
+  @UnstableApi public static final int BUFFER_FLAG_FIRST_SAMPLE = 1 << 27; // 0x08000000
   /** Indicates that a buffer has supplemental data. */
   @UnstableApi public static final int BUFFER_FLAG_HAS_SUPPLEMENTAL_DATA = 1 << 28; // 0x10000000
   /** Indicates that a buffer is known to contain the last media sample of the stream. */

libraries/decoder/src/main/java/androidx/media3/decoder/Buffer.java

@@ -34,6 +34,11 @@ public abstract class Buffer {
     return getFlag(C.BUFFER_FLAG_DECODE_ONLY);
   }
 
+  /** Returns whether the {@link C#BUFFER_FLAG_FIRST_SAMPLE} flag is set. */
+  public final boolean isFirstSample() {
+    return getFlag(C.BUFFER_FLAG_FIRST_SAMPLE);
+  }
+
   /** Returns whether the {@link C#BUFFER_FLAG_END_OF_STREAM} flag is set. */
   public final boolean isEndOfStream() {
     return getFlag(C.BUFFER_FLAG_END_OF_STREAM);

libraries/decoder/src/main/java/androidx/media3/decoder/SimpleDecoder.java

@@ -235,6 +235,9 @@ public abstract class SimpleDecoder<
       if (inputBuffer.isDecodeOnly()) {
         outputBuffer.addFlag(C.BUFFER_FLAG_DECODE_ONLY);
       }
+      if (inputBuffer.isFirstSample()) {
+        outputBuffer.addFlag(C.BUFFER_FLAG_FIRST_SAMPLE);
+      }
       @Nullable E exception;
       try {
         exception = decode(inputBuffer, outputBuffer, resetDecoder);

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/audio/DecoderAudioRenderer.java

@@ -130,6 +130,7 @@ public abstract class DecoderAudioRenderer<
   private int encoderPadding;
 
   private boolean experimentalKeepAudioTrackOnSeek;
+  private boolean firstStreamSampleRead;
 
   @Nullable private T decoder;
 
@@ -389,6 +390,9 @@ public abstract class DecoderAudioRenderer<
         decoderCounters.skippedOutputBufferCount += outputBuffer.skippedOutputBufferCount;
         audioSink.handleDiscontinuity();
       }
+      if (outputBuffer.isFirstSample()) {
+        audioSink.handleDiscontinuity();
+      }
     }
 
     if (outputBuffer.isEndOfStream()) {
@@ -470,6 +474,10 @@ public abstract class DecoderAudioRenderer<
           inputBuffer = null;
           return false;
         }
+        if (!firstStreamSampleRead) {
+          firstStreamSampleRead = true;
+          inputBuffer.addFlag(C.BUFFER_FLAG_FIRST_SAMPLE);
+        }
         inputBuffer.flip();
         inputBuffer.format = inputFormat;
         onQueueInputBuffer(inputBuffer);
@@ -587,6 +595,13 @@ public abstract class DecoderAudioRenderer<
     }
   }
 
+  @Override
+  protected void onStreamChanged(Format[] formats, long startPositionUs, long offsetUs)
+      throws ExoPlaybackException {
+    super.onStreamChanged(formats, startPositionUs, offsetUs);
+    firstStreamSampleRead = false;
+  }
+
   @Override
   public void handleMessage(@MessageType int messageType, @Nullable Object message)
       throws ExoPlaybackException {

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/audio/DefaultAudioSink.java

@@ -887,7 +887,6 @@ public final class DefaultAudioSink implements AudioSink {
 
   @Override
   public void handleDiscontinuity() {
-    // Force resynchronization after a skipped buffer.
     startMediaTimeUsNeedsSync = true;
   }
 

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/audio/DecoderAudioRendererTest.java

@@ -21,7 +21,12 @@ import static androidx.media3.exoplayer.RendererCapabilities.DECODER_SUPPORT_PRI
 import static androidx.media3.exoplayer.RendererCapabilities.TUNNELING_NOT_SUPPORTED;
 import static androidx.media3.exoplayer.RendererCapabilities.TUNNELING_SUPPORTED;
 import static androidx.media3.test.utils.FakeSampleStream.FakeSampleStreamItem.END_OF_STREAM_ITEM;
+import static androidx.media3.test.utils.FakeSampleStream.FakeSampleStreamItem.oneByteSample;
 import static com.google.common.truth.Truth.assertThat;
+import static org.mockito.ArgumentMatchers.any;
+import static org.mockito.ArgumentMatchers.anyInt;
+import static org.mockito.ArgumentMatchers.anyLong;
+import static org.mockito.Mockito.inOrder;
 import static org.mockito.Mockito.times;
 import static org.mockito.Mockito.verify;
 import static org.mockito.Mockito.when;
@@ -46,6 +51,7 @@ import com.google.common.collect.ImmutableList;
 import org.junit.Before;
 import org.junit.Test;
 import org.junit.runner.RunWith;
+import org.mockito.InOrder;
 import org.mockito.Mock;
 import org.mockito.MockitoAnnotations;
 import org.robolectric.annotation.Config;
@@ -139,6 +145,100 @@ public class DecoderAudioRendererTest {
     verify(mockAudioSink, times(1)).reset();
   }
 
+  @Test
+  public void firstSampleOfStreamSignalsDiscontinuityToAudioSink() throws Exception {
+    when(mockAudioSink.handleBuffer(any(), anyLong(), anyInt())).thenReturn(true);
+    when(mockAudioSink.isEnded()).thenReturn(true);
+    InOrder inOrderAudioSink = inOrder(mockAudioSink);
+    FakeSampleStream fakeSampleStream =
+        new FakeSampleStream(
+            new DefaultAllocator(/* trimOnReset= */ true, /* individualAllocationSize= */ 1024),
+            /* mediaSourceEventDispatcher= */ null,
+            DrmSessionManager.DRM_UNSUPPORTED,
+            new DrmSessionEventListener.EventDispatcher(),
+            FORMAT,
+            ImmutableList.of(
+                oneByteSample(/* timeUs= */ 0, C.BUFFER_FLAG_KEY_FRAME),
+                oneByteSample(/* timeUs= */ 1_000),
+                END_OF_STREAM_ITEM));
+    fakeSampleStream.writeData(/* startPositionUs= */ 0);
+    audioRenderer.enable(
+        RendererConfiguration.DEFAULT,
+        new Format[] {FORMAT},
+        fakeSampleStream,
+        /* positionUs= */ 0,
+        /* joining= */ false,
+        /* mayRenderStartOfStream= */ true,
+        /* startPositionUs= */ 0,
+        /* offsetUs= */ 0);
+
+    audioRenderer.setCurrentStreamFinal();
+    while (!audioRenderer.isEnded()) {
+      audioRenderer.render(/* positionUs= */ 0, /* elapsedRealtimeUs= */ 0);
+    }
+
+    inOrderAudioSink.verify(mockAudioSink, times(1)).handleDiscontinuity();
+    inOrderAudioSink.verify(mockAudioSink, times(2)).handleBuffer(any(), anyLong(), anyInt());
+  }
+
+  @Test
+  public void firstSampleOfReplacementStreamSignalsDiscontinuityToAudioSink() throws Exception {
+    when(mockAudioSink.handleBuffer(any(), anyLong(), anyInt())).thenReturn(true);
+    when(mockAudioSink.isEnded()).thenReturn(true);
+    InOrder inOrderAudioSink = inOrder(mockAudioSink);
+    FakeSampleStream fakeSampleStream1 =
+        new FakeSampleStream(
+            new DefaultAllocator(/* trimOnReset= */ true, /* individualAllocationSize= */ 1024),
+            /* mediaSourceEventDispatcher= */ null,
+            DrmSessionManager.DRM_UNSUPPORTED,
+            new DrmSessionEventListener.EventDispatcher(),
+            FORMAT,
+            ImmutableList.of(
+                oneByteSample(/* timeUs= */ 0, C.BUFFER_FLAG_KEY_FRAME),
+                oneByteSample(/* timeUs= */ 1_000),
+                END_OF_STREAM_ITEM));
+    fakeSampleStream1.writeData(/* startPositionUs= */ 0);
+    FakeSampleStream fakeSampleStream2 =
+        new FakeSampleStream(
+            new DefaultAllocator(/* trimOnReset= */ true, /* individualAllocationSize= */ 1024),
+            /* mediaSourceEventDispatcher= */ null,
+            DrmSessionManager.DRM_UNSUPPORTED,
+            new DrmSessionEventListener.EventDispatcher(),
+            FORMAT,
+            ImmutableList.of(
+                oneByteSample(/* timeUs= */ 1_000_000, C.BUFFER_FLAG_KEY_FRAME),
+                oneByteSample(/* timeUs= */ 1_001_000),
+                END_OF_STREAM_ITEM));
+    fakeSampleStream2.writeData(/* startPositionUs= */ 0);
+    audioRenderer.enable(
+        RendererConfiguration.DEFAULT,
+        new Format[] {FORMAT},
+        fakeSampleStream1,
+        /* positionUs= */ 0,
+        /* joining= */ false,
+        /* mayRenderStartOfStream= */ true,
+        /* startPositionUs= */ 0,
+        /* offsetUs= */ 0);
+
+    while (!audioRenderer.hasReadStreamToEnd()) {
+      audioRenderer.render(/* positionUs= */ 0, /* elapsedRealtimeUs= */ 0);
+    }
+    audioRenderer.replaceStream(
+        new Format[] {FORMAT},
+        fakeSampleStream2,
+        /* startPositionUs= */ 1_000_000,
+        /* offsetUs= */ 1_000_000);
+    audioRenderer.setCurrentStreamFinal();
+    while (!audioRenderer.isEnded()) {
+      audioRenderer.render(/* positionUs= */ 0, /* elapsedRealtimeUs= */ 0);
+    }
+
+    inOrderAudioSink.verify(mockAudioSink, times(1)).handleDiscontinuity();
+    inOrderAudioSink.verify(mockAudioSink, times(2)).handleBuffer(any(), anyLong(), anyInt());
+    inOrderAudioSink.verify(mockAudioSink, times(1)).handleDiscontinuity();
+    inOrderAudioSink.verify(mockAudioSink, times(2)).handleBuffer(any(), anyLong(), anyInt());
+  }
+
   private static final class FakeDecoder
       extends SimpleDecoder<DecoderInputBuffer, SimpleDecoderOutputBuffer, DecoderException> {
 

libraries/test_data/src/test/assets/audiosinkdumps/mka/bear-flac-16bit.mka.audiosink.dump

@@ -1,3 +1,4 @@
+discontinuity:
 config:
   pcmEncoding = 2
   channelCount = 2

libraries/test_data/src/test/assets/audiosinkdumps/mka/bear-flac-24bit.mka.audiosink.dump

@@ -1,3 +1,4 @@
+discontinuity:
 config:
   pcmEncoding = 536870912
   channelCount = 2