Express audio delay more precisely in frames rather than milliseconds.

media/base/audio_renderer_mixer_input.cc

Index: media/base/audio_renderer_mixer_input.cc
diff --git a/media/base/audio_renderer_mixer_input.cc b/media/base/audio_renderer_mixer_input.cc
index d0d38d3b8eef902f69b9a342cdaffcbe0109235b..bd69f44b318e12ca4c94c0c93324689a0825d7fa 100644
--- a/media/base/audio_renderer_mixer_input.cc
+++ b/media/base/audio_renderer_mixer_input.cc
@@ -165,8 +165,13 @@ OutputDeviceStatus AudioRendererMixerInput::GetDeviceStatus() {
 
 double AudioRendererMixerInput::ProvideInput(AudioBus* audio_bus,
                                              base::TimeDelta buffer_delay) {
-  int frames_filled = callback_->Render(
-      audio_bus, static_cast<int>(buffer_delay.InMillisecondsF() + 0.5), 0);
+  double microseconds_per_frame =

[DaleCurtis, 2016/02/11 01:57:16]

Ditto on the TODO.

[chcunningham, 2016/02/11 21:02:52]

Done.

+      static_cast<double>(base::Time::kMicrosecondsPerSecond) /
+      params_.sample_rate();
+  uint32_t frames_delayed =
+      std::round(buffer_delay.InMicroseconds() / microseconds_per_frame);
+
+  int frames_filled = callback_->Render(audio_bus, frames_delayed, 0);
 
   // AudioConverter expects unfilled frames to be zeroed.
   if (frames_filled < audio_bus->frames()) {