Support AudioContextOptions latencyHint as double.

third_party/WebKit/Source/platform/audio/AudioDestination.cpp

 /*
  * Copyright (C) 2010 Google Inc. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  *
  * 1.  Redistributions of source code must retain the above copyright
  *     notice, this list of conditions and the following disclaimer.
  * 2.  Redistributions in binary form must reproduce the above copyright
@@ skipping 55 matching lines @@
       m_isPlaying(false),
       m_callback(callback),
       m_outputBus(AudioBus::create(numberOfOutputChannels,
                                    AudioUtilities::kRenderQuantumFrames,
                                    false)),
       m_renderBus(AudioBus::create(numberOfOutputChannels,
                                    AudioUtilities::kRenderQuantumFrames)),
       m_fifo(
           WTF::wrapUnique(new PushPullFIFO(numberOfOutputChannels, kFIFOSize))),
       m_framesElapsed(0) {
-  m_callbackBufferSize = hardwareBufferSize();
-  if (!checkBufferSize()) {
-    NOTREACHED();
-  }
-
   // Create WebAudioDevice. blink::WebAudioDevice is designed to support the
   // local input (e.g. loopback from OS audio system), but Chromium's media
   // renderer does not support it currently. Thus, we use zero for the number
   // of input channels.
   m_webAudioDevice = WTF::wrapUnique(Platform::current()->createAudioDevice(
       0, numberOfOutputChannels, latencyHint, this, String(),
       std::move(securityOrigin)));
   DCHECK(m_webAudioDevice);
+
+  m_callbackBufferSize = m_webAudioDevice->framesPerBuffer();

[Andrew MacPherson, 2017/03/14 12:03:03]

This CL changed m_callbackBufferSize to hardwareBufferSize(), which breaks
latencyHint: https://codereview.chromium.org/2740103005 I've reverted this here
but we'll need to figure out why it was needed, and also why the unit tests
missed pointing out the breakage.

[hongchan, 2017/03/14 14:36:34]

If you rebase the patch, you'll see this is fixed. However, we're still missing
the test coverage as you mentioned.

[Andrew MacPherson, 2017/03/15 15:08:18]

Great, just rebased now and can see this, thanks! Regarding the unit tests it
looks like because the mock doesn't try to render anything the render() method
is never called and the crashing CHECK_EQ() is never hit. I tried to very
trivially force it to render a fake frame and ran into issues around thread
checking so left it for now. I assume it may require refactoring the render()
method a bit to make it more unit testable and wanted to check with you before
doing anything there.

+  if (!checkBufferSize()) {
+    NOTREACHED();
+  }
 }
 
 AudioDestination::~AudioDestination() {
   stop();
 }
 
 void AudioDestination::render(const WebVector<float*>& destinationData,
                               size_t numberOfFrames,
                               double delay,
                               double delayTimestamp,
@@ skipping 99 matching lines @@
   callbackBufferSizeHistogram.sample(m_callbackBufferSize);
 
   // Check if the requested buffer size is too large.
   bool isBufferSizeValid =
       m_callbackBufferSize + AudioUtilities::kRenderQuantumFrames <= kFIFOSize;
   DCHECK(isBufferSizeValid);
   return isBufferSizeValid;
 }
 
 }  // namespace blink