Reland again Android MediaCodecVideoDecoder: Manage lifetime of texture frames

talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java

Index: talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java
diff --git a/talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java b/talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java
index 86221c4cd9d7d4704841a844f59afa9b57352086..f60f3e56772f4a8cc887908e5841e5d910cc488b 100644
--- a/talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java
+++ b/talk/app/webrtc/java/src/org/webrtc/MediaCodecVideoDecoder.java
@@ -33,19 +33,19 @@ import android.media.MediaCodecInfo;
 import android.media.MediaCodecInfo.CodecCapabilities;
 import android.media.MediaCodecList;
 import android.media.MediaFormat;
-import android.opengl.GLES11Ext;
-import android.opengl.GLES20;
 import android.os.Build;
+import android.os.SystemClock;
 import android.view.Surface;
 
 import org.webrtc.Logging;
 
 import java.nio.ByteBuffer;
 import java.util.Arrays;
+import java.util.LinkedList;
 import java.util.List;
 import java.util.concurrent.CountDownLatch;
-
-import javax.microedition.khronos.egl.EGLContext;
+import java.util.Queue;
+import java.util.concurrent.TimeUnit;
 
 // Java-side of peerconnection_jni.cc:MediaCodecVideoDecoder.
 // This class is an implementation detail of the Java PeerConnection API.
@@ -104,14 +104,18 @@ public class MediaCodecVideoDecoder {
   private int height;
   private int stride;
   private int sliceHeight;
+  private boolean hasDecodedFirstFrame;
+  private final Queue<TimeStamps> decodeStartTimeMs = new LinkedList<TimeStamps>();
   private boolean useSurface;
-  private int textureID = 0;
-  private SurfaceTexture surfaceTexture = null;
-  private Surface surface = null;
-  private EglBase eglBase;
 
-  private MediaCodecVideoDecoder() {
-  }
+  // The below variables are only used when decoding to a Surface.
+  private TextureListener textureListener;
+  // Max number of output buffers queued before starting to drop decoded frames.
+  private static final int MAX_QUEUED_OUTPUTBUFFERS = 3;
+  private int droppedFrames;
+  private Surface surface = null;
+  private final Queue<DecodedOutputBuffer>
+      dequeuedSurfaceOutputBuffers = new LinkedList<DecodedOutputBuffer>();
 
   // MediaCodec error handler - invoked when critical error happens which may prevent
   // further use of media codec API. Now it means that one of media codec instances
@@ -223,12 +227,13 @@ public class MediaCodecVideoDecoder {
     }
   }
 
-  // Pass null in |sharedContext| to configure the codec for ByteBuffer output.
-  private boolean initDecode(VideoCodecType type, int width, int height, EGLContext sharedContext) {
+  // Pass null in |surfaceTextureHelper| to configure the codec for ByteBuffer output.
+  private boolean initDecode(
+      VideoCodecType type, int width, int height, SurfaceTextureHelper surfaceTextureHelper) {
     if (mediaCodecThread != null) {
       throw new RuntimeException("Forgot to release()?");
     }
-    useSurface = (sharedContext != null);
+    useSurface = (surfaceTextureHelper != null);
     String mime = null;
     String[] supportedCodecPrefixes = null;
     if (type == VideoCodecType.VIDEO_CODEC_VP8) {
@@ -250,9 +255,6 @@ public class MediaCodecVideoDecoder {
     Logging.d(TAG, "Java initDecode: " + type + " : "+ width + " x " + height +
         ". Color: 0x" + Integer.toHexString(properties.colorFormat) +
         ". Use Surface: " + useSurface);
-    if (sharedContext != null) {
-      Logging.d(TAG, "Decoder shared EGL Context: " + sharedContext);
-    }
     runningInstance = this; // Decoder is now running and can be queried for stack traces.
     mediaCodecThread = Thread.currentThread();
     try {
@@ -262,16 +264,8 @@ public class MediaCodecVideoDecoder {
       sliceHeight = height;
 
       if (useSurface) {
-        // Create shared EGL context.
-        eglBase = new EglBase(sharedContext, EglBase.ConfigType.PIXEL_BUFFER);
-        eglBase.createDummyPbufferSurface();
-        eglBase.makeCurrent();
-
-        // Create output surface
-        textureID = GlUtil.generateTexture(GLES11Ext.GL_TEXTURE_EXTERNAL_OES);
-        Logging.d(TAG, "Video decoder TextureID = " + textureID);
-        surfaceTexture = new SurfaceTexture(textureID);
-        surface = new Surface(surfaceTexture);
+        textureListener = new TextureListener(surfaceTextureHelper);
+        surface = new Surface(surfaceTextureHelper.getSurfaceTexture());
       }
 
       MediaFormat format = MediaFormat.createVideoFormat(mime, width, height);
@@ -290,6 +284,10 @@ public class MediaCodecVideoDecoder {
       colorFormat = properties.colorFormat;
       outputBuffers = mediaCodec.getOutputBuffers();
       inputBuffers = mediaCodec.getInputBuffers();
+      decodeStartTimeMs.clear();
+      hasDecodedFirstFrame = false;
+      dequeuedSurfaceOutputBuffers.clear();
+      droppedFrames = 0;
       Logging.d(TAG, "Input buffers: " + inputBuffers.length +
           ". Output buffers: " + outputBuffers.length);
       return true;
@@ -300,7 +298,7 @@ public class MediaCodecVideoDecoder {
   }
 
   private void release() {
-    Logging.d(TAG, "Java releaseDecoder");
+    Logging.d(TAG, "Java releaseDecoder. Total number of dropped frames: " + droppedFrames);
     checkOnMediaCodecThread();
 
     // Run Mediacodec stop() and release() on separate thread since sometime
@@ -338,11 +336,7 @@ public class MediaCodecVideoDecoder {
     if (useSurface) {
       surface.release();
       surface = null;
-      Logging.d(TAG, "Delete video decoder TextureID " + textureID);
-      GLES20.glDeleteTextures(1, new int[] {textureID}, 0);
-      textureID = 0;
-      eglBase.release();
-      eglBase = null;
+      textureListener.release();
     }
     Logging.d(TAG, "Java releaseDecoder done");
   }
@@ -359,13 +353,15 @@ public class MediaCodecVideoDecoder {
     }
   }
 
-  private boolean queueInputBuffer(
-      int inputBufferIndex, int size, long timestampUs) {
+  private boolean queueInputBuffer(int inputBufferIndex, int size, long presentationTimeStamUs,
+      long timeStampMs, long ntpTimeStamp) {
     checkOnMediaCodecThread();
     try {
       inputBuffers[inputBufferIndex].position(0);
       inputBuffers[inputBufferIndex].limit(size);
-      mediaCodec.queueInputBuffer(inputBufferIndex, 0, size, timestampUs, 0);
+      decodeStartTimeMs.add(new TimeStamps(SystemClock.elapsedRealtime(), timeStampMs,
+          ntpTimeStamp));
+      mediaCodec.queueInputBuffer(inputBufferIndex, 0, size, presentationTimeStamUs, 0);
       return true;
     }
     catch (IllegalStateException e) {
@@ -374,57 +370,183 @@ public class MediaCodecVideoDecoder {
     }
   }
 
-  // Helper structs for dequeueOutputBuffer() below.
-  private static class DecodedByteBuffer {
-    public DecodedByteBuffer(int index, int offset, int size, long presentationTimestampUs) {
+  private static class TimeStamps {
+    public TimeStamps(long decodeStartTimeMs, long timeStampMs, long ntpTimeStampMs) {
+      this.decodeStartTimeMs = decodeStartTimeMs;
+      this.timeStampMs = timeStampMs;
+      this.ntpTimeStampMs = ntpTimeStampMs;
+    }
+    private final long decodeStartTimeMs; // Time when this frame was queued for decoding.
+    private final long timeStampMs; // Only used for bookkeeping in Java. Used in C++;
+    private final long ntpTimeStampMs; // Only used for bookkeeping in Java. Used in C++;
+  }
+
+  // Helper struct for dequeueOutputBuffer() below.
+  private static class DecodedOutputBuffer {
+    public DecodedOutputBuffer(int index, int offset, int size, long timeStampMs,
+        long ntpTimeStampMs, long decodeTime, long endDecodeTime) {
       this.index = index;
       this.offset = offset;
       this.size = size;
-      this.presentationTimestampUs = presentationTimestampUs;
+      this.timeStampMs = timeStampMs;
+      this.ntpTimeStampMs = ntpTimeStampMs;
+      this.decodeTimeMs = decodeTime;
+      this.endDecodeTimeMs = endDecodeTime;
     }
 
     private final int index;
     private final int offset;
     private final int size;
-    private final long presentationTimestampUs;
+    private final long timeStampMs;
+    private final long ntpTimeStampMs;
+    // Number of ms it took to decode this frame.
+    private final long decodeTimeMs;
+    // System time when this frame finished decoding.
+    private final long endDecodeTimeMs;
   }
 
+  // Helper struct for dequeueTextureBuffer() below.
   private static class DecodedTextureBuffer {
     private final int textureID;
-    private final long presentationTimestampUs;
-
-    public DecodedTextureBuffer(int textureID, long presentationTimestampUs) {
+    private final float[] transformMatrix;
+    private final long timeStampMs;
+    private final long ntpTimeStampMs;
+    private final long decodeTimeMs;
+    // Interval from when the frame finished decoding until this buffer has been created.
+    // Since there is only one texture, this interval depend on the time from when
+    // a frame is decoded and provided to C++ and until that frame is returned to the MediaCodec
+    // so that the texture can be updated with the next decoded frame.
+    private final long frameDelayMs;
+
+    // A DecodedTextureBuffer with zero |textureID| has special meaning and represents a frame
+    // that was dropped.
+    public DecodedTextureBuffer(int textureID, float[] transformMatrix, long timeStampMs,
+        long ntpTimeStampMs, long decodeTimeMs, long frameDelay) {
       this.textureID = textureID;
-      this.presentationTimestampUs = presentationTimestampUs;
+      this.transformMatrix = transformMatrix;
+      this.timeStampMs = timeStampMs;
+      this.ntpTimeStampMs = ntpTimeStampMs;
+      this.decodeTimeMs = decodeTimeMs;
+      this.frameDelayMs = frameDelay;
+    }
+  }
+
+  // Poll based texture listener.
+  private static class TextureListener
+      implements SurfaceTextureHelper.OnTextureFrameAvailableListener {
+    private final SurfaceTextureHelper surfaceTextureHelper;
+    // |newFrameLock| is used to synchronize arrival of new frames with wait()/notifyAll().
+    private final Object newFrameLock = new Object();
+    // |bufferToRender| is non-null when waiting for transition between addBufferToRender() to
+    // onTextureFrameAvailable().
+    private DecodedOutputBuffer bufferToRender;
+    private DecodedTextureBuffer renderedBuffer;
+
+    public TextureListener(SurfaceTextureHelper surfaceTextureHelper) {
+      this.surfaceTextureHelper = surfaceTextureHelper;
+      surfaceTextureHelper.setListener(this);
+    }
+
+    public void addBufferToRender(DecodedOutputBuffer buffer) {
+      if (bufferToRender != null) {
+        Logging.e(TAG,
+            "Unexpected addBufferToRender() called while waiting for a texture.");
+        throw new IllegalStateException("Waiting for a texture.");
+      }
+      bufferToRender = buffer;
+    }
+
+    public boolean isWaitingForTexture() {
+      synchronized (newFrameLock) {
+        return bufferToRender != null;
+      }
+    }
+
+    // Callback from |surfaceTextureHelper|. May be called on an arbitrary thread.
+    @Override
+    public void onTextureFrameAvailable(
+        int oesTextureId, float[] transformMatrix, long timestampNs) {
+      synchronized (newFrameLock) {
+        if (renderedBuffer != null) {
+          Logging.e(TAG,
+              "Unexpected onTextureFrameAvailable() called while already holding a texture.");
+          throw new IllegalStateException("Already holding a texture.");
+        }
+        // |timestampNs| is always zero on some Android versions.
+        renderedBuffer = new DecodedTextureBuffer(oesTextureId, transformMatrix,
+            bufferToRender.timeStampMs, bufferToRender.ntpTimeStampMs, bufferToRender.decodeTimeMs,
+            SystemClock.elapsedRealtime() - bufferToRender.endDecodeTimeMs);
+        bufferToRender = null;
+        newFrameLock.notifyAll();
+      }
+    }
+
+    // Dequeues and returns a DecodedTextureBuffer if available, or null otherwise.
+    public DecodedTextureBuffer dequeueTextureBuffer(int timeoutMs) {
+      synchronized (newFrameLock) {
+        if (renderedBuffer == null && timeoutMs > 0 && isWaitingForTexture()) {
+          try {
+            newFrameLock.wait(timeoutMs);
+          } catch(InterruptedException e) {
+            // Restore the interrupted status by reinterrupting the thread.
+            Thread.currentThread().interrupt();
+          }
+        }
+        DecodedTextureBuffer returnedBuffer = renderedBuffer;
+        renderedBuffer = null;
+        return returnedBuffer;
+      }
+    }
+
+    public void release() {
+      // SurfaceTextureHelper.disconnect() will block until any onTextureFrameAvailable() in
+      // progress is done. Therefore, the call to disconnect() must be outside any synchronized
+      // statement that is also used in the onTextureFrameAvailable() above to avoid deadlocks.
+      surfaceTextureHelper.disconnect();
+      synchronized (newFrameLock) {
+        if (renderedBuffer != null) {
+          surfaceTextureHelper.returnTextureFrame();
+          renderedBuffer = null;
+        }
+      }
     }
   }
 
-  // Returns null if no decoded buffer is available, and otherwise either a DecodedByteBuffer or
-  // DecodedTexturebuffer depending on |useSurface| configuration.
+  // Returns null if no decoded buffer is available, and otherwise a DecodedByteBuffer.
   // Throws IllegalStateException if call is made on the wrong thread, if color format changes to an
   // unsupported format, or if |mediaCodec| is not in the Executing state. Throws CodecException
   // upon codec error.
-  private Object dequeueOutputBuffer(int dequeueTimeoutUs)
-      throws IllegalStateException, MediaCodec.CodecException {
+  private DecodedOutputBuffer dequeueOutputBuffer(int dequeueTimeoutMs) {
     checkOnMediaCodecThread();
-
+    if (decodeStartTimeMs.isEmpty()) {
+      return null;
+    }
     // Drain the decoder until receiving a decoded buffer or hitting
     // MediaCodec.INFO_TRY_AGAIN_LATER.
     final MediaCodec.BufferInfo info = new MediaCodec.BufferInfo();
     while (true) {
-      final int result = mediaCodec.dequeueOutputBuffer(info, dequeueTimeoutUs);
+      final int result = mediaCodec.dequeueOutputBuffer(
+          info, TimeUnit.MILLISECONDS.toMicros(dequeueTimeoutMs));
       switch (result) {
-        case MediaCodec.INFO_TRY_AGAIN_LATER:
-          return null;
         case MediaCodec.INFO_OUTPUT_BUFFERS_CHANGED:
           outputBuffers = mediaCodec.getOutputBuffers();
           Logging.d(TAG, "Decoder output buffers changed: " + outputBuffers.length);
+          if (hasDecodedFirstFrame) {
+            throw new RuntimeException("Unexpected output buffer change event.");
+          }
           break;
         case MediaCodec.INFO_OUTPUT_FORMAT_CHANGED:
           MediaFormat format = mediaCodec.getOutputFormat();
           Logging.d(TAG, "Decoder format changed: " + format.toString());
+          int new_width = format.getInteger(MediaFormat.KEY_WIDTH);
+          int new_height = format.getInteger(MediaFormat.KEY_HEIGHT);
+          if (hasDecodedFirstFrame && (new_width != width || new_height != height)) {
+            throw new RuntimeException("Unexpected size change. Configured " + width + "*" +
+                height + ". New " + new_width + "*" + new_height);
+          }
           width = format.getInteger(MediaFormat.KEY_WIDTH);
           height = format.getInteger(MediaFormat.KEY_HEIGHT);
+
           if (!useSurface && format.containsKey(MediaFormat.KEY_COLOR_FORMAT)) {
             colorFormat = format.getInteger(MediaFormat.KEY_COLOR_FORMAT);
             Logging.d(TAG, "Color: 0x" + Integer.toHexString(colorFormat));
@@ -442,18 +564,74 @@ public class MediaCodecVideoDecoder {
           stride = Math.max(width, stride);
           sliceHeight = Math.max(height, sliceHeight);
           break;
+        case MediaCodec.INFO_TRY_AGAIN_LATER:
+          return null;
         default:
-          // Output buffer decoded.
-          if (useSurface) {
-            mediaCodec.releaseOutputBuffer(result, true /* render */);
-            // TODO(magjed): Wait for SurfaceTexture.onFrameAvailable() before returning a texture
-            // frame.
-            return new DecodedTextureBuffer(textureID, info.presentationTimeUs);
-          } else {
-            return new DecodedByteBuffer(result, info.offset, info.size, info.presentationTimeUs);
-          }
+          hasDecodedFirstFrame = true;
+          TimeStamps timeStamps = decodeStartTimeMs.remove();
+          return new DecodedOutputBuffer(result, info.offset, info.size, timeStamps.timeStampMs,
+              timeStamps.ntpTimeStampMs,
+              SystemClock.elapsedRealtime() - timeStamps.decodeStartTimeMs,
+              SystemClock.elapsedRealtime());
+        }
+    }
+  }
+
+  // Returns null if no decoded buffer is available, and otherwise a DecodedTextureBuffer.
+  // Throws IllegalStateException if call is made on the wrong thread, if color format changes to an
+  // unsupported format, or if |mediaCodec| is not in the Executing state. Throws CodecException
+  // upon codec error. If |dequeueTimeoutMs| > 0, the oldest decoded frame will be dropped if
+  // a frame can't be returned.
+  private DecodedTextureBuffer dequeueTextureBuffer(int dequeueTimeoutMs) {
+    checkOnMediaCodecThread();
+    if (!useSurface) {
+      throw new IllegalStateException("dequeueTexture() called for byte buffer decoding.");
+    }
+    DecodedOutputBuffer outputBuffer = dequeueOutputBuffer(dequeueTimeoutMs);
+    if (outputBuffer != null) {
+      dequeuedSurfaceOutputBuffers.add(outputBuffer);
+    }
+
+    MaybeRenderDecodedTextureBuffer();
+    // Check if there is texture ready now by waiting max |dequeueTimeoutMs|.
+    DecodedTextureBuffer renderedBuffer = textureListener.dequeueTextureBuffer(dequeueTimeoutMs);
+    if (renderedBuffer != null) {
+      MaybeRenderDecodedTextureBuffer();
+      return renderedBuffer;
+    }
+
+    if ((dequeuedSurfaceOutputBuffers.size()
+         >= Math.min(MAX_QUEUED_OUTPUTBUFFERS, outputBuffers.length)
+         || (dequeueTimeoutMs > 0 && !dequeuedSurfaceOutputBuffers.isEmpty()))) {
+      ++droppedFrames;
+      // Drop the oldest frame still in dequeuedSurfaceOutputBuffers.
+      // The oldest frame is owned by |textureListener| and can't be dropped since
+      // mediaCodec.releaseOutputBuffer has already been called.
+      final DecodedOutputBuffer droppedFrame = dequeuedSurfaceOutputBuffers.remove();
+      if (dequeueTimeoutMs > 0) {
+        Logging.w(TAG, "Draining decoder. Dropping frame with TS: "
+            + droppedFrame.timeStampMs + ". Total number of dropped frames: " + droppedFrames);
+      } else {
+        Logging.w(TAG, "Too many output buffers. Dropping frame with TS: "
+            + droppedFrame.timeStampMs + ". Total number of dropped frames: " + droppedFrames);
       }
+
+      mediaCodec.releaseOutputBuffer(droppedFrame.index, false /* render */);
+      return new DecodedTextureBuffer(0, null, droppedFrame.timeStampMs,
+          droppedFrame.ntpTimeStampMs, droppedFrame.decodeTimeMs,
+          SystemClock.elapsedRealtime() - droppedFrame.endDecodeTimeMs);
+    }
+    return null;
+  }
+
+  private void MaybeRenderDecodedTextureBuffer() {
+    if (dequeuedSurfaceOutputBuffers.isEmpty() || textureListener.isWaitingForTexture()) {
+      return;
     }
+    // Get the first frame in the queue and render to the decoder output surface.
+    final DecodedOutputBuffer buffer = dequeuedSurfaceOutputBuffers.remove();
+    textureListener.addBufferToRender(buffer);
+    mediaCodec.releaseOutputBuffer(buffer.index, true /* render */);
   }
 
   // Release a dequeued output byte buffer back to the codec for re-use. Should only be called for
@@ -461,11 +639,11 @@ public class MediaCodecVideoDecoder {
   // Throws IllegalStateException if the call is made on the wrong thread, if codec is configured
   // for surface decoding, or if |mediaCodec| is not in the Executing state. Throws
   // MediaCodec.CodecException upon codec error.
-  private void returnDecodedByteBuffer(int index)
+  private void returnDecodedOutputBuffer(int index)
       throws IllegalStateException, MediaCodec.CodecException {
     checkOnMediaCodecThread();
     if (useSurface) {
-      throw new IllegalStateException("returnDecodedByteBuffer() called for surface decoding.");
+      throw new IllegalStateException("returnDecodedOutputBuffer() called for surface decoding.");
     }
     mediaCodec.releaseOutputBuffer(index, false /* render */);
   }