Implement the ServiceWorker "activate" event during activation

content/browser/service_worker/embedded_worker_registry.cc

Index: content/browser/service_worker/embedded_worker_registry.cc
diff --git a/content/browser/service_worker/embedded_worker_registry.cc b/content/browser/service_worker/embedded_worker_registry.cc
index 308108e60515ef00dfb1ca3be171b22a3e3c1049..f07e943a438fdab7775a8b867fd61388c66ed9ea 100644
--- a/content/browser/service_worker/embedded_worker_registry.cc
+++ b/content/browser/service_worker/embedded_worker_registry.cc
@@ -78,7 +78,8 @@ void EmbeddedWorkerRegistry::OnSendMessageToBrowser(
   }
   // Perform security check to filter out any unexpected (and non-test)
   // messages. This must list up all message types that can go through here.
-  if (message.type() == ServiceWorkerHostMsg_InstallEventFinished::ID ||
+  if (message.type() == ServiceWorkerHostMsg_ActivateEventFinished::ID ||

[jschuh, 2014/03/28 21:27:29]

Why are we filtering here? Shouldn't found->second->OnMessageReceived() have an
explicit message map?

[dominicc (has gone to gerrit), 2014/03/31 03:59:33]

This class acts as a proxy routing messages to a worker; the point of checking
here is to limit the kinds of messages sent via this  payload in the event the
renderer is compromised. The kinds of methods that should go via this route may
be fewer than all the kinds of messages the worker responds to.

This is based on advice from Tom Sepez here:
<https://codereview.chromium.org/140893002#msg7>.

[jschuh, 2014/03/31 13:36:15]

Okay, I guess @tsepez and I disagree on the value f this, but it certainly
doesn't hurt anything.

+      message.type() == ServiceWorkerHostMsg_InstallEventFinished::ID ||
       message.type() == ServiceWorkerHostMsg_FetchEventFinished::ID ||
       IPC_MESSAGE_CLASS(message) == TestMsgStart) {
     found->second->OnMessageReceived(request_id, message);