Add a virtual beamforming audio device on ChromeOS.

chromeos/audio/cras_audio_handler.cc

Index: chromeos/audio/cras_audio_handler.cc
diff --git a/chromeos/audio/cras_audio_handler.cc b/chromeos/audio/cras_audio_handler.cc
index 8ee6009f35c69e6ee1ec3070716b5bf75654fd30..3ce04ad7bea07b74e9fa9eef28d436a7227dcf28 100644
--- a/chromeos/audio/cras_audio_handler.cc
+++ b/chromeos/audio/cras_audio_handler.cc
@@ -10,6 +10,7 @@
 #include "base/bind.h"
 #include "base/bind_helpers.h"
 #include "base/logging.h"
+#include "base/sys_info.h"
 #include "chromeos/audio/audio_devices_pref_handler.h"
 #include "chromeos/audio/audio_devices_pref_handler_stub.h"
 #include "chromeos/dbus/dbus_thread_manager.h"
@@ -864,6 +865,22 @@ void CrasAudioHandler::UpdateDevicesAndSwitchActive(
   size_t new_input_device_size = 0;
   for (size_t i = 0; i < nodes.size(); ++i) {
     AudioDevice device(nodes[i]);
+
+    // TODO(ajm): This is a temporary proof-of-concept hack. These values should
+    // come from the cras configs for each board.

[tommi (sloooow) - chröme, 2015/09/02 07:43:07]

will this be removed before checkin?

[ajm, 2015/09/02 17:00:55]

hychao@ is doing work to plumb this through on the ChromeOS side. Looks like
that will be finished before this lands, but if not, I intend to keep this. (It
doesn't make anything worse; these values were previously hard-coded in
MediaStreamAudioProcessor.)

[ajm, 2015/09/09 01:01:29]

Now planning to remove this before landing. hychao's work is almost finished.

+    if (device.type == AUDIO_TYPE_INTERNAL_MIC) {
+      const std::string& board = base::SysInfo::GetLsbReleaseBoard();

[tommi (sloooow) - chröme, 2015/09/02 07:43:07]

GetLsbReleaseBoard returns std::string (non-ref), so using ref here is a bit
misleading (although compilers will do the right thing and give you reference to
a temporary)

[ajm, 2015/09/02 17:00:55]

Right; I had a discussion with emerican@ about this in an earlier CL:
https://codereview.chromium.org/1224623014/#msg17

Given GetLsbReleaseBoard's non-ref return, there is no performance advantage of
making the receiver a ref and it's a bit misleading, as you mention. The one
benefit I can think of is if the return type were to change to a const ref; we'd
then be making a copy unnecessarily if we changed the receiver to non-ref.

What's your opinion? There are other examples of this pattern in the CL.

+      if (board.find("nyan_kitty") != std::string::npos) {
+        device.mic_positions = "-0.03 0 0 0.03 0 0";
+      } else if (board.find("peach_pi") != std::string::npos) {
+        device.mic_positions = "-0.025 0 0 0.025 0 0";
+      } else if (board.find("samus") != std::string::npos) {
+        device.mic_positions = "-0.032 0 0 0.032 0 0";
+      } else if (board.find("swanky") != std::string::npos) {
+        device.mic_positions = "-0.026 0 0 0.026 0 0";
+      }
+    }
+
     audio_devices_[device.id] = device;
 
     if (!has_alternative_input_ &&