Explicit notification and removal of lifecycle observers.

Source/core/workers/WorkerThread.cpp

 /*
  * Copyright (C) 2008 Apple Inc. All Rights Reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
@@ skipping 117 matching lines @@
     m_workerReportingProxy.workerGlobalScopeStarted(m_workerGlobalScope.get());
 
     WorkerScriptController* script = m_workerGlobalScope->script();
     InspectorInstrumentation::willEvaluateWorkerScript(workerGlobalScope(), startMode);
     script->evaluate(ScriptSourceCode(sourceCode, scriptURL));
 
     runEventLoop();
 
     ThreadIdentifier threadID = m_threadID;
 
+#if !ENABLE(OILPAN)
+    ASSERT(m_workerGlobalScope->hasOneRef());
+#endif
     // The below assignment will destroy the context, which will in turn notify messaging proxy.
     // We cannot let any objects survive past thread exit, because no other thread will run GC or otherwise destroy them.
     // If Oilpan is enabled, we detach of the context/global scope, with the final heap cleanup below sweeping it out.
-#if ENABLE(OILPAN)
     m_workerGlobalScope->dispose();
-#else
-    ASSERT(m_workerGlobalScope->hasOneRef());
-#endif
     m_workerGlobalScope = nullptr;
 
     ThreadState::detach();

[haraken, 2014/03/17 01:22:56]

Shall we move ThreadState::detach() at the very end of this method? A thread
should be detached at the very end of the thread execution.

[haraken, 2014/03/17 01:25:45]

As far as I read your comment in
https://codereview.chromium.org/196033004/#msg6, probably do you need to call
ThreadState::detach() before
workerReportingProxy().workerGlobalScopeDestroyed()? Then I might want to have a
comment about that.

[sof, 2014/03/17 07:46:41]

Added a comment, which might address this? (not sure..) Notification has to come
after the (ThreadState) detach.

 
+    // Notify proxy that WorkerGlobalScope is now gone. This can free the WorkerThread object, so do not access it after this.
+    workerReportingProxy().workerGlobalScopeDestroyed();
+
     // Clean up PlatformThreadData before WTF::WTFThreadData goes away!
     PlatformThreadData::current().destroy();
 
     // The thread object may be already destroyed from notification now, don't try to access "this".
     detachThread(threadID);
 }
 
 void WorkerThread::runEventLoop()
 {
     // Does not return until terminated.
@@ skipping 77 matching lines @@
 void WorkerThread::releaseFastMallocFreeMemoryInAllThreads()
 {
     MutexLocker lock(threadSetMutex());
     HashSet<WorkerThread*>& threads = workerThreads();
     HashSet<WorkerThread*>::iterator end = threads.end();
     for (HashSet<WorkerThread*>::iterator it = threads.begin(); it != end; ++it)
         (*it)->runLoop().postTask(adoptPtr(new ReleaseFastMallocFreeMemoryTask));
 }
 
 } // namespace WebCore