Support AudioContextOptions latencyHint as double.

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

Index: third_party/WebKit/Source/platform/audio/AudioDestination.cpp
diff --git a/third_party/WebKit/Source/platform/audio/AudioDestination.cpp b/third_party/WebKit/Source/platform/audio/AudioDestination.cpp
index 691ffb179e83c9f520c1d35e7ec0999ce6aac965..ff692918aa84800f9e5bb8ab952f7a81b3382a8c 100644
--- a/third_party/WebKit/Source/platform/audio/AudioDestination.cpp
+++ b/third_party/WebKit/Source/platform/audio/AudioDestination.cpp
@@ -73,11 +73,6 @@ AudioDestination::AudioDestination(AudioIOCallback& callback,
       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
@@ -86,6 +81,11 @@ AudioDestination::AudioDestination(AudioIOCallback& callback,
       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() {