WTF::Function adoption should be explicit instead of implicit

[Geoffrey Garen]

WTF::Function adoption should be explicit instead of implicit

[Geoffrey Garen]

Created attachment 404961 [details]
Patch

[Geoffrey Garen]

If someone can figure out the template fu to automatically deduce the function type in adoptImpl, that would be great!

[Darin Adler]

Comment on attachment 404961 [details]
Patch

Does it really need to be name adoptImpl and not just adopt? Are there name conflicts?

[Geoffrey Garen]

I don't think there are name conflicts.

Would you rename leakImpl() to leak() too?

[Darin Adler]

I am increasingly unsure about this.

Why should WTF::Function adoption be explicit?

In other cases, we do that because we are concerned about object lifetime. Like a raw pointer being adopted by a std::unique_ptr. Does that apply here.

I don’t think "impl" is the correct term for a C++ function.

[Geoffrey Garen]

Yes, this is all about the lifetime of the unique_ptr that holds the functor. leakImpl() leaks the unique_ptr and adopt takes ownership of the pointer.

[Geoffrey Garen]

Created attachment 404971 [details]
Patch

[Darin Adler]

Comment on attachment 404971 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=404971&action=review

> Source/WTF/wtf/Function.h:125
> +template <typename T> inline Function<T> adopt(typename Function<T>::Impl* impl)

I think you want to write something more like this:

    template<typename Out, typename... In> Function<Out(In...)> adopt(Detail::CallableWrapperBase<Out, In...>* pointer)

That should allow you to use adopt() without template arguments.

[Geoffrey Garen]

Created attachment 404992 [details]
Patch for landing

[EWS]

Committed r264743: <https://trac.webkit.org/changeset/264743>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 404992 [details].

[Radar WebKit Bug Importer]

<rdar://problem/65970781>