Kill CleanUpAfterMessageLoopDestruction.

chrome/browser/io_thread.cc

Index: chrome/browser/io_thread.cc
diff --git a/chrome/browser/io_thread.cc b/chrome/browser/io_thread.cc
index 7839925f46b4d5a39b40a0d656aabd021a290b4c..a8cd62ecdfb3030a0c3dec5abbde990498d855a6 100644
--- a/chrome/browser/io_thread.cc
+++ b/chrome/browser/io_thread.cc
@@ -439,20 +439,12 @@ void IOThread::CleanUp() {
   delete globals_;
   globals_ = NULL;
 
-  BrowserProcessSubThread::CleanUp();
-}
+  // net::URLRequest instances must NOT outlive the IO thread.
+  base::debug::LeakTracker<net::URLRequest>::CheckForLeaks();

[eroman, 2011/03/02 02:23:31]

What has changed to make this safe?

If I remember correctly, this had to be done in
CleanupAfterMessageLoopDestruction because URLFetcher was releasing its
URLRequest from a destruction observer on the IO thread (and the destruction
observers don't run until AFTER CleanUp().

[willchan no longer on Chromium, 2011/03/02 05:31:16]

http://code.google.com/p/chromium/issues/detail?id=59630 documents how I fixed
them all. I think you reviewed most of those changes :)

If you search for DestructionObserver, you'll see that all the URLRequest
related ones are gone now.

 
-void IOThread::CleanUpAfterMessageLoopDestruction() {
   // This will delete the |notification_service_|.  Make sure it's done after
   // anything else can reference it.
-  BrowserProcessSubThread::CleanUpAfterMessageLoopDestruction();
-
-  // net::URLRequest instances must NOT outlive the IO thread.
-  //
-  // To allow for URLRequests to be deleted from
-  // MessageLoop::DestructionObserver this check has to happen after CleanUp
-  // (which runs before DestructionObservers).
-  base::debug::LeakTracker<net::URLRequest>::CheckForLeaks();
+  BrowserProcessSubThread::CleanUp();
 }
 
 // static