CompositionPlayer: skip decode-only frames upstream of the VideoGraph

This is necessary for prewarming. With prewarming, in a sequence of 2
videos, the second renderer is enabled before the first one is disabled,
and decode-only frames should be skipped before the second renderer is
started. The problem is that the second renderer will forward frames to
a BufferingVideoSink before it is started, which  will delay the frame
handling and therefore not skip the frame before the renderer is
started.

PiperOrigin-RevId: 721032049

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/DefaultVideoSink.java

@@ -173,8 +173,7 @@ public void setVideoEffects(List<Effect> videoEffects) {
   }
 
   @Override
-  public void setStreamTimestampInfo(
-      long streamStartPositionUs, long bufferTimestampAdjustmentUs, long lastResetPositionUs) {
+  public void setStreamTimestampInfo(long streamStartPositionUs, long bufferTimestampAdjustmentUs) {
     if (streamStartPositionUs != this.streamStartPositionUs) {
       videoFrameRenderControl.onStreamStartPositionChanged(streamStartPositionUs);
       this.streamStartPositionUs = streamStartPositionUs;

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/MediaCodecVideoRenderer.java

@@ -814,9 +814,7 @@ protected void onPositionReset(long positionUs, boolean joining) throws ExoPlayb
         videoSink.flush(/* resetPosition= */ true);
       }
       videoSink.setStreamTimestampInfo(
-          getOutputStreamStartPositionUs(),
-          getBufferTimestampAdjustmentUs(),
-          getLastResetPositionUs());
+          getOutputStreamStartPositionUs(), getBufferTimestampAdjustmentUs());
       pendingVideoSinkInputStreamChange = true;
     }
     super.onPositionReset(positionUs, joining);
@@ -1480,6 +1478,11 @@ protected boolean processOutputBuffer(
     long presentationTimeUs = bufferPresentationTimeUs - outputStreamOffsetUs;
 
     if (videoSink != null) {
+      // Skip decode-only buffers, e.g. after seeking, immediately.
+      if (isDecodeOnlyBuffer && !isLastBuffer) {
+        skipOutputBuffer(codec, bufferIndex, presentationTimeUs);
+        return true;
+      }
       long framePresentationTimeUs = bufferPresentationTimeUs + getBufferTimestampAdjustmentUs();
       return videoSink.handleInputFrame(
           framePresentationTimeUs,
@@ -1631,9 +1634,7 @@ protected void onProcessedStreamChange() {
       // Signaling end of the previous stream.
       videoSink.signalEndOfCurrentInputStream();
       videoSink.setStreamTimestampInfo(
-          getOutputStreamStartPositionUs(),
-          getBufferTimestampAdjustmentUs(),
-          getLastResetPositionUs());
+          getOutputStreamStartPositionUs(), getBufferTimestampAdjustmentUs());
     } else {
       videoFrameReleaseControl.onProcessedStreamChange();
     }

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/PlaybackVideoGraphWrapper.java

@@ -441,7 +441,7 @@ public void onOutputFrameAvailableForRendering(long framePresentationTimeUs) {
     if (newOutputStreamStartPositionUs != null
         && newOutputStreamStartPositionUs != outputStreamStartPositionUs) {
       defaultVideoSink.setStreamTimestampInfo(
-          newOutputStreamStartPositionUs, bufferTimestampAdjustmentUs, /* unused */ C.TIME_UNSET);
+          newOutputStreamStartPositionUs, bufferTimestampAdjustmentUs);
       outputStreamStartPositionUs = newOutputStreamStartPositionUs;
     }
     boolean isLastFrame =
@@ -588,8 +588,7 @@ private void flush(boolean resetPosition) {
     if (streamStartPositionsUs.size() == 1) {
       long lastStartPositionUs = checkNotNull(streamStartPositionsUs.pollFirst());
       // defaultVideoSink should use the latest startPositionUs if none is passed after flushing.
-      defaultVideoSink.setStreamTimestampInfo(
-          lastStartPositionUs, bufferTimestampAdjustmentUs, /* unused */ C.TIME_UNSET);
+      defaultVideoSink.setStreamTimestampInfo(lastStartPositionUs, bufferTimestampAdjustmentUs);
     }
     lastOutputBufferPresentationTimeUs = C.TIME_UNSET;
     finalBufferPresentationTimeUs = C.TIME_UNSET;
@@ -611,7 +610,7 @@ private void setPlaybackSpeed(float speed) {
   private void setBufferTimestampAdjustment(long bufferTimestampAdjustmentUs) {
     this.bufferTimestampAdjustmentUs = bufferTimestampAdjustmentUs;
     defaultVideoSink.setStreamTimestampInfo(
-        outputStreamStartPositionUs, bufferTimestampAdjustmentUs, /* unused */ C.TIME_UNSET);
+        outputStreamStartPositionUs, bufferTimestampAdjustmentUs);
   }
 
   private boolean shouldRenderToInputVideoSink() {
@@ -638,7 +637,6 @@ private final class InputVideoSink implements VideoSink, PlaybackVideoGraphWrapp
     @Nullable private Format inputFormat;
     private @InputType int inputType;
     private long inputBufferTimestampAdjustmentUs;
-    private long lastResetPositionUs;
 
     /**
      * The buffer presentation timestamp, in microseconds, of the most recently registered frame.
@@ -796,7 +794,7 @@ public void setVideoEffects(List<Effect> videoEffects) {
 
     @Override
     public void setStreamTimestampInfo(
-        long streamStartPositionUs, long bufferTimestampAdjustmentUs, long lastResetPositionUs) {
+        long streamStartPositionUs, long bufferTimestampAdjustmentUs) {
       // Input timestamps should always be positive because they are offset by ExoPlayer. Adding a
       // position to the queue with timestamp 0 should therefore always apply it as long as it is
       // the only position in the queue.
@@ -809,7 +807,6 @@ public void setStreamTimestampInfo(
       // directly at the output of the VideoGraph because no frame has been input yet following the
       // flush.
       PlaybackVideoGraphWrapper.this.setBufferTimestampAdjustment(inputBufferTimestampAdjustmentUs);
-      this.lastResetPositionUs = lastResetPositionUs;
     }
 
     @Override
@@ -837,26 +834,10 @@ public void enableMayRenderStartOfStream() {
     public boolean handleInputFrame(
         long framePresentationTimeUs, boolean isLastFrame, VideoFrameHandler videoFrameHandler) {
       checkState(isInitialized());
+
       if (!shouldRenderToInputVideoSink()) {
         return false;
       }
-
-      // The sink takes in frames with monotonically increasing, non-offset frame
-      // timestamps. That is, with two ten-second long videos, the first frame of the second video
-      // should bear a timestamp of 10s seen from VideoFrameProcessor; while in ExoPlayer, the
-      // timestamp of the said frame would be 0s, but the streamOffset is incremented by 10s to
-      // include the duration of the first video. Thus this correction is needed to account for the
-      // different handling of presentation timestamps in ExoPlayer and VideoFrameProcessor.
-      //
-      // inputBufferTimestampAdjustmentUs adjusts the frame presentation time (which is relative to
-      // the start of a composition) to the buffer timestamp (that corresponds to the player
-      // position).
-      long bufferPresentationTimeUs = framePresentationTimeUs - inputBufferTimestampAdjustmentUs;
-      if (bufferPresentationTimeUs < lastResetPositionUs && !isLastFrame) {
-        videoFrameHandler.skip();
-        return true;
-      }
-
       if (checkStateNotNull(videoFrameProcessor).getPendingInputFrameCount()
           >= videoFrameProcessorMaxPendingFrameCount) {
         return false;
@@ -865,7 +846,17 @@ public boolean handleInputFrame(
         return false;
       }
 
-      lastBufferPresentationTimeUs = bufferPresentationTimeUs;
+      // The sink takes in frames with monotonically increasing, non-offset frame
+      // timestamps. That is, with two 10s long videos, the first frame of the second video should
+      // bear a timestamp of 10s seen from VideoFrameProcessor; while in ExoPlayer, the timestamp of
+      // the said frame would be 0s, but the streamOffset is incremented by 10s to include the
+      // duration of the first video. Thus this correction is needed to account for the different
+      // handling of presentation timestamps in ExoPlayer and VideoFrameProcessor.
+      //
+      // inputBufferTimestampAdjustmentUs adjusts the frame presentation time (which is relative to
+      // the start of a composition) to the buffer timestamp (that corresponds to the player
+      // position).
+      lastBufferPresentationTimeUs = framePresentationTimeUs - inputBufferTimestampAdjustmentUs;
       // Use the frame presentation time as render time so that the SurfaceTexture is accompanied
       // by this timestamp. Setting a realtime based release time is only relevant when rendering to
       // a SurfaceView, but we render to a surface in this case.

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/VideoSink.java

@@ -217,10 +217,8 @@ interface VideoFrameHandler {
    *     current stream, in microseconds.
    * @param bufferTimestampAdjustmentUs The timestamp adjustment to add to the buffer presentation
    *     timestamps to convert them to frame presentation timestamps, in microseconds.
-   * @param lastResetPositionUs The renderer last reset position, in microseconds.
    */
-  void setStreamTimestampInfo(
-      long streamStartPositionUs, long bufferTimestampAdjustmentUs, long lastResetPositionUs);
+  void setStreamTimestampInfo(long streamStartPositionUs, long bufferTimestampAdjustmentUs);
 
   /** Sets the output surface info. */
   void setOutputSurfaceInfo(Surface outputSurface, Size outputResolution);

libraries/transformer/src/main/java/androidx/media3/transformer/BufferingVideoSink.java

@@ -193,12 +193,10 @@ public void setVideoEffects(List<Effect> videoEffects) {
   }
 
   @Override
-  public void setStreamTimestampInfo(
-      long streamStartPositionUs, long bufferTimestampAdjustmentUs, long lastResetPositionUs) {
+  public void setStreamTimestampInfo(long streamStartPositionUs, long bufferTimestampAdjustmentUs) {
     executeOrDelay(
         videoSink ->
-            videoSink.setStreamTimestampInfo(
-                streamStartPositionUs, bufferTimestampAdjustmentUs, lastResetPositionUs));
+            videoSink.setStreamTimestampInfo(streamStartPositionUs, bufferTimestampAdjustmentUs));
   }
 
   @Override

libraries/transformer/src/main/java/androidx/media3/transformer/SequenceRenderersFactory.java

@@ -505,9 +505,7 @@ protected boolean processOutputBuffer(
       if (inputStreamPending) {
         checkState(streamStartPositionUs != C.TIME_UNSET);
         videoSink.setStreamTimestampInfo(
-            streamStartPositionUs,
-            /* bufferTimestampAdjustmentUs= */ offsetToCompositionTimeUs,
-            getLastResetPositionUs());
+            streamStartPositionUs, /* bufferTimestampAdjustmentUs= */ offsetToCompositionTimeUs);
         videoSink.onInputStreamChanged(
             VideoSink.INPUT_TYPE_BITMAP,
             new Format.Builder()