Adding policy support to the new network configuration stack.

chromeos/dbus/shill_profile_client.cc

Index: chromeos/dbus/shill_profile_client.cc
diff --git a/chromeos/dbus/shill_profile_client.cc b/chromeos/dbus/shill_profile_client.cc
index a86ff785cce6526e8b947cf323de98af35445efe..511a22e88bb782bdf46beb06f2e70816e0538884 100644
--- a/chromeos/dbus/shill_profile_client.cc
+++ b/chromeos/dbus/shill_profile_client.cc
@@ -20,12 +20,10 @@ namespace chromeos {
 
 namespace {
 
-// The ShillProfileClient implementation.
 class ShillProfileClientImpl : public ShillProfileClient {
  public:
   explicit ShillProfileClientImpl(dbus::Bus* bus);
 
-  /////////////////////////////////////
   // ShillProfileClient overrides.
   virtual void AddPropertyChangedObserver(
       const dbus::ObjectPath& profile_path,
@@ -51,6 +49,10 @@ class ShillProfileClientImpl : public ShillProfileClient {
                            const base::Closure& callback,
                            const ErrorCallback& error_callback) OVERRIDE;
 
+  virtual TestInterface* GetTestInterface() OVERRIDE {
+    return NULL;

[pastarmovj, 2013/04/11 14:46:29]

Maybe put a NOTREACHED() here too to make sure this code is not called at all.

[stevenjb, 2013/04/11 18:20:08]

Actually, I think it's better (easier to debug) if the code  crashes where the
test interface is accessed.

[pneubeck (no reviews), 2013/04/15 12:16:24]

NOTREACHED will still need the NULL return. CHECK(false) would be the
alternative.

I prefer it as it is.

[pneubeck (no reviews), 2013/04/15 12:16:24]

Done.

[pastarmovj, 2013/04/16 11:31:17]

I am not sure if it will be easier to debug since the crash might occur very far
away from the point that calls GetTestInterface. I suggest to keep the return
NULL but add a NOTREACHED for speeding up the crash in debug builds. It will not
add any extra code to the release build.

+  }
+
  private:
   typedef std::map<std::string, ShillClientHelper*> HelperMap;