fix emplace

[paulocoutinhox]

No description provided.

[paulocoutinhox]

Hi, can you merge? Thanks.

[pinguin999]

I don't understand the problem, but this code part blocks the build with the latest version of emscripten.

[hsdk123]

I don't understand the problem, but this code part blocks the build with the latest version of emscripten.

Just to give more context, the error #1611 here is shown with the latest version of emscripten.

This PR seems to fix it.

[cuavas]

I don’t think this implementation of emplace makes sense for optional<T&> as written.

Consider:

	template <class T>
	class optional<T&> {
...
		template <class... Args>
		T& emplace(Args&&... args) noexcept {
			static_assert(std::is_constructible<T, Args&&...>::value, "T must be constructible with Args");

			*this = nullopt;
			new (static_cast<void*>(this)) optional(std::in_place, std::forward<Args>(args)...);
			return **this;
		}
...
	};

Now look at the static_assert – it’s testing that T can be constructed using the argument types passed to emplace. However, the optional<T&> doesn’t hold T, it logically holds T& (although it’s implemented using T* to allow assignment). The only thing that T& can be constructed with is T&. It makes no sense to use variadic arguments.

When called with T&, the static_assert only succeeds by chance when T is copy constructible. It will fail if it isn’t, and it will succeed for things it shouldn’t.

Since you know T* is trivially destructible and copyable, you could implement emplace as:

		T& emplace(T& arg) noexcept {
			m_value = &arg;
			return **this;
		}