Support AudioContextOptions latencyHint as double.

content/renderer/media/audio_renderer_mixer_manager.cc

Index: content/renderer/media/audio_renderer_mixer_manager.cc
diff --git a/content/renderer/media/audio_renderer_mixer_manager.cc b/content/renderer/media/audio_renderer_mixer_manager.cc
index f22bdc3f6d7c94d38cff3ee3f6ac4e05466c47ac..5ca7bcfb848641f5fea356ac3470276c5fd6d21b 100644
--- a/content/renderer/media/audio_renderer_mixer_manager.cc
+++ b/content/renderer/media/audio_renderer_mixer_manager.cc
@@ -63,7 +63,13 @@ media::AudioParameters GetMixerOutputParams(
           output_sample_rate, preferred_high_latency_output_buffer_size);
       break;
     case media::AudioLatency::LATENCY_EXACT_MS:
-    // TODO(olka): add support when WebAudio requires it.
+      output_buffer_size =
+          std::max(std::min(input_params.frames_per_buffer(),
+                            media::AudioLatency::GetHighLatencyBufferSize(

[o1ka, 2017/03/14 15:55:43]

HighLatencyBufferSize is 20 ms. Is it the the maximum value WebAudio want to
have?

rtoy@: latencies more than 10 ms will expose
https://bugs.chromium.org/p/chromium/issues/detail?id=701000 - something to be
aware of. I marked it as blocking
https://bugs.chromium.org/p/chromium/issues/detail?id=564276 (which does not
have an owner?)

[Raymond Toy, 2017/03/14 16:14:35]

Is 20ms the value for "playback"?

In any case, there are Android devices with huge HW buffer sizes, much greater
than 20 ms long. We need to support these.

IIRC Andrew said in some cases Android "interactive" was greater than
"balanced".  We'll need to fix these issus (in a different CL).

+                                output_sample_rate,
+                                preferred_high_latency_output_buffer_size)),
+                   hardware_params.frames_per_buffer());

[Raymond Toy, 2017/03/14 15:19:39]

Blink has a clampTo() function that's easier to read than this.  Don't know if
there's a Chromium equivalent.

Also, you do very similar clamping in several places.  Perhaps refactoring this
clamping into its own function would be good.

[Andrew MacPherson, 2017/03/15 15:08:17]

Makes sense, I've refactored the AudioContextTest in Blink to use clampTo() and
I've added a media::AudioLatency::GetExactBufferSize() as suggested by olka for
the Chromium side.

+      break;
     default:
       NOTREACHED();
   }