Make the ChromeNetworkDelegate use the ExtensionEventRouterForwarder

chrome/browser/profiles/profile_io_data.h

Index: chrome/browser/profiles/profile_io_data.h
diff --git a/chrome/browser/profiles/profile_io_data.h b/chrome/browser/profiles/profile_io_data.h
index ce651912ffc616d43f2f85009eecce07e4494477..6c37ecc411c949d28a0090b70563074200566caa 100644
--- a/chrome/browser/profiles/profile_io_data.h
+++ b/chrome/browser/profiles/profile_io_data.h
@@ -19,7 +19,6 @@ class ChromeBlobStorageContext;
 class ChromeURLRequestContext;
 class ChromeURLRequestContextGetter;
 class ExtensionInfoMap;
-class ExtensionIOEventRouter;
 namespace fileapi {
 class FileSystemContext;
 }
@@ -103,13 +102,16 @@ class ProfileIOData : public base::RefCountedThreadSafe<ProfileIOData> {
     scoped_refptr<ChromeBlobStorageContext> blob_storage_context;
     scoped_refptr<fileapi::FileSystemContext> file_system_context;
     scoped_refptr<ExtensionInfoMap> extension_info_map;
-    scoped_refptr<ExtensionIOEventRouter> extension_io_event_router;
     scoped_refptr<prerender::PrerenderManager> prerender_manager;
     scoped_refptr<ProtocolHandlerRegistry> protocol_handler_registry;
     // We need to initialize the ProxyConfigService from the UI thread
     // because on linux it relies on initializing things through gconf,
     // and needs to be on the main thread.
     scoped_ptr<net::ProxyConfigService> proxy_config_service;
+    // The profile this struct was populated from. We must not dereference
+    // this pointer on any but the UI thread with prior checking that it is
+    // still valid, e.g. using ProfileManager::IsValidProfile.
+    Profile* profile;

[willchan no longer on Chromium, 2011/02/26 02:11:35]

jochen, I think I told you this was ok before, but I am concerned about this
Profile*, especially since it's stored in ChromeNetworkDelegate where you don't
have a comment like this. Also, people never read comments. I think there's
probably something we could do where the global event forwarder could maintain a
mapping of Profile*<=>Foo, where Foo can be passed in ProfileParams and used by
the ChromeNetworkDelegate when calling back into the event forwarder.

I'm curious what ericroman thinks here. I suspect he shares my concern with
exposing the Profile* on the IO thread. 

   };
 
   explicit ProfileIOData(bool is_off_the_record);