Cleanup WebState and WebStateUserData API

ios/public/provider/web/web_state_user_data.h

Index: ios/public/provider/web/web_state_user_data.h
diff --git a/ios/public/provider/web/web_state_user_data.h b/ios/public/provider/web/web_state_user_data.h
index 17d2fab6f90b744fdbe00133444d12e8da0f8cb9..a326d08210919fd2b76bedaeb185dcef80182a96 100644
--- a/ios/public/provider/web/web_state_user_data.h
+++ b/ios/public/provider/web/web_state_user_data.h
@@ -5,7 +5,7 @@
 #ifndef IOS_PUBLIC_PROVIDER_WEB_WEB_STATE_USER_DATA_H_
 #define IOS_PUBLIC_PROVIDER_WEB_WEB_STATE_USER_DATA_H_
 
-#include "ios/public/consumer/base/supports_user_data.h"
+#include "base/supports_user_data.h"
 #include "ios/public/provider/web/web_state.h"
 
 namespace ios {
@@ -27,7 +27,7 @@ namespace ios {
 // DEFINE_WEB_CONTENTS_USER_DATA_KEY(Foo);
 //
 template <typename T>
-class WebStateUserData : public ios::SupportsUserData::Data {
+class WebStateUserData : public base::SupportsUserData::Data {

[droger, 2014/12/04 09:39:47]

Important: This class needs a virtual destructor.

[sdefresne, 2014/12/05 14:32:05]

Done. But base::SupportsUserData::Data already has a virtual destructor and
since this class does not change anything, so AFAICT there is no need to
explicitly add a destructor there, right?

[droger, 2014/12/05 14:44:33]

I think there is a problem in this case, am I wrong?

class MyUserData : public WebStateUserData<MyUserData> {
  ...
};

WebStateUserData<MyUserData>* a = new MyUserData;
delete a;

Since the destructor is not virtual, it will not call the destructor of
MyUserData.

[droger, 2014/12/05 14:57:41]

Ok, it turns out I was wrong.
As long as the base class has a virtual destructor, all the subclasses are
implicitly virtual too.
Sorry about that.

  public:
   // Creates an object of type T, and attaches it to the specified WebState.
   // If an instance is already attached, does nothing.