eastl::shared_ptr<>::reset(): function not thread safe

[Wawha]

reset() function of shared_ptr is not thread safe.
It takes so many times to understand/find where come from that memory leak I have in my code. The problem was so difficult to reproduce and isolate!

So here the code:

	inline void ref_count_sp::release()
	{
		EASTL_ASSERT((mRefCount > 0) && (mWeakRefCount > 0));
		if(Internal::atomic_decrement(&mRefCount) > 0)
			Internal::atomic_decrement(&mWeakRefCount);
		else
		{
			free_value();

			if(Internal::atomic_decrement(&mWeakRefCount) == 0)
				free_ref_count_sp();
		}
	}

If 2 threads, with 2 shared_ptr<> pointed the same value, and are calling the release() function at the same time, the free_ref_count_sp(); could be not called.
It happens, when thread 1 is calling Internal::atomic_decrement(&mWeakRefCount); (due m_RefCount > 0) after the thread 2 call the code if(Internal::atomic_decrement(&mWeakRefCount) == 0). In that case free_ref_count_sp(); is not called, because on the thread 2, mWeakRefCount will have value 1.

The fix consist to remove that else condition and process each counter independently :

	inline void ref_count_sp::release()
	{
		EASTL_ASSERT((mRefCount > 0) && (mWeakRefCount > 0));
		if(Internal::atomic_decrement(&mRefCount) == 0)
			free_value();

		if(Internal::atomic_decrement(&mWeakRefCount) == 0)
			free_ref_count_sp();
	}

I try to write UTest for that case. For the moment, I used std::future and std::async because it's easier for me. If I have to used EAThread, could you tell me the best eastl function to used instead.

[MaxEWinkler]

Thank you so much for the fix.
This is a really good find and a definite issue.
Thanks again for taking the time to figure this out and sending us the patch :).