Fix ASSERTS in worker code in OSX....

Fix ASSERTS in worker code in OSX.
One issue was with using WTF::currentThread() from destructor of thread-specific data. In pthreads environment, this destructors are called after thread is detached and there is no ThreadIdentifier for it.
Also moved the destruction of dom objects from thread-specific destructor time to WorkerScriptController destruction time - to prevent similar issues with WTF ThreadIdentifier since WorkerScriptController is destroyed before thread proc actually exits.

[jianli, 2009-04-11 18:16:28]

http://codereview.chromium.org/67061/diff/1/5
File third_party/WebKit/WebCore/bindings/v8/V8DOMMap.cpp (right):

http://codereview.chromium.org/67061/diff/1/5#newcode109
Line 109: typedef WTF::Vector<ThreadSpecificDOMData*> DOMDataList;
Could you please add include for WTF::Vector?

http://codereview.chromium.org/67061/diff/1/4
File third_party/WebKit/WebCore/bindings/v8/WorkerScriptController.cpp (right):

http://codereview.chromium.org/67061/diff/1/4#newcode85
Line 85: m_domMapEraser.set(new DOMMapEraser(m_workerContext, this));
It seems that we still have some DOM objects in use, like WorkerContext object.
Not sure removing all DOM objects from the map will have any bad effect or not.

[Dmitry Titov, 2009-04-13 18:55:36]

> http://codereview.chromium.org/67061/diff/1/5#newcode109
> Line 109: typedef WTF::Vector<ThreadSpecificDOMData*> DOMDataList;
> Could you please add include for WTF::Vector?

Done.

> http://codereview.chromium.org/67061/diff/1/4#newcode85
> Line 85: m_domMapEraser.set(new DOMMapEraser(m_workerContext, this));
> It seems that we still have some DOM objects in use, like WorkerContext
object.
> Not sure removing all DOM objects from the map will have any bad effect or
not.

As discussed offline, I think it's allright since we just deref WorkerContext.
Tested on Win32 too - the worker tests pass.

[Dmitry Titov, 2009-04-13 21:59:06]

ready for another look. testing it in win32...

[jianli, 2009-04-13 22:02:31]

LGTM.