Changed DefaultAdapter to RunCallbackOnAdapterReady function.

chrome/browser/chromeos/system/ash_system_tray_delegate.cc

Index: chrome/browser/chromeos/system/ash_system_tray_delegate.cc
diff --git a/chrome/browser/chromeos/system/ash_system_tray_delegate.cc b/chrome/browser/chromeos/system/ash_system_tray_delegate.cc
index 64ef92c61d8c01a46d949570c467563157749761..6172166b7814682f2e5eaaffcdc59b2aac68fe73 100644
--- a/chrome/browser/chromeos/system/ash_system_tray_delegate.cc
+++ b/chrome/browser/chromeos/system/ash_system_tray_delegate.cc
@@ -270,7 +270,14 @@ class SystemTrayDelegate : public ash::SystemTrayDelegate,
     network_icon_->SetResourceColorTheme(NetworkMenuIcon::COLOR_LIGHT);
     network_icon_dark_->SetResourceColorTheme(NetworkMenuIcon::COLOR_DARK);
 
-    bluetooth_adapter_ = device::BluetoothAdapterFactory::DefaultAdapter();
+    device::BluetoothAdapterFactory::RunCallbackOnAdapterReady(
+        base::Bind(&SystemTrayDelegate::InitializeOnAdapterReady,
+                   base::Unretained(this)));

[bryeung, 2013/01/08 22:24:27]

Remind me why using an Unretained here is safe...?

[youngki, 2013/01/09 15:12:03]

First of all my understanding is that AshSystemTrayDelegate is destroyed only
when the browser shuts down. That means that base::Unretained(this) is safe
unless it is a shutdown time. But during shutdown BluetoothAdapter will also be
destroyed and this callback won't be run then.

[bryeung, 2013/01/09 15:28:01]

And you're sure that:
- this is also true of the other places you've used this pattern?
- this doesn't introduce a race condition where shutdown has started and then
the callback is called

If so, then I agree it's okay.

[youngki, 2013/01/09 19:32:24]

I gave a second thought, and there might be race conditions if BluetoothAdapter
is destroyed later than the other modules during the shutdown. Also
BluetoothExtensionFunction is not live for the entire browser lifetime anyways.

So I changed the code to use weak pointer instead of base::Unretained(this).

+  }
+
+  bool InitializeOnAdapterReady(

[bryeung, 2013/01/08 22:24:27]

why does this need to return a value?

[youngki, 2013/01/09 15:12:03]

This is @asargent's suggestion to make BluetoothExtensionFunction::DoWork() to
return bool, not because the return data is meaningful, but so that we don't
have to manually put in the code to duplicate EXTENSION_FUNCTION_VALIDATE macro
in that function. EXTENSION_FUNCTION_VALIDATE returns false, and that requires
the signature of the function to return bool.

[bryeung, 2013/01/09 15:28:01]

Makes sense.  Okay.

[youngki, 2013/01/09 19:32:24]

I actually changed it to void because weak pointer does not allow to bind the
method that returns. Instead, I added a void RunOnAdapterCallback() which simply
calls DoWork() in BluetoothExtensionFunction and used that as a callback
instead.

+      scoped_refptr<device::BluetoothAdapter> adapter) {
+    bluetooth_adapter_ = adapter;
     bluetooth_adapter_->AddObserver(this);
 
     local_state_registrar_.Init(g_browser_process->local_state());
@@ -294,6 +301,8 @@ class SystemTrayDelegate : public ash::SystemTrayDelegate,
     if (policy_manager)
       policy_manager->core()->store()->AddObserver(this);
     UpdateEnterpriseDomain();
+
+    return true;
   }
 
   virtual ~SystemTrayDelegate() {