Bug 120799 – Make Vector::uncheckedAppend work with move-only types

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Anders Carlsson]

Make Vector::uncheckedAppend work with move-only types

[Anders Carlsson]

Created attachment 210667 [details]
Patch

[Andreas Kling]

Comment on attachment 210667 [details]
Patch

Awesome! Can't wait for append()!

[Anders Carlsson]

Committed r155152: <http://trac.webkit.org/changeset/155152>

[Darin Adler]

Comment on attachment 210667 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=210667&action=review

> Source/WTF/wtf/Vector.h:646
> +    template<typename U> void uncheckedAppend(U&& val);

Does removing the old version cause any problems if we want to append something given a const&?

Why name this argument, val, here?

> Source/WTF/wtf/Vector.h:1071
> +inline void Vector<T, inlineCapacity, OverflowHandler>::uncheckedAppend(U&& val)

Why val? How about a word instead of this abbreviation.

[Anders Carlsson]

(In reply to comment #4)
> (From update of attachment 210667 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=210667&action=review
> 
> > Source/WTF/wtf/Vector.h:646
> > +    template<typename U> void uncheckedAppend(U&& val);
> 
> Does removing the old version cause any problems if we want to append something given a const&?

Nope, what happens if you pass a const T& to it is that U&& will become “const T&” && and reference collapsing will turn that into const T&:

struct A { };

template<typename T> void f(T&&)
{
    static_assert(std::is_same<T, const A&>::value, "");
}

void g(const A& a) { f(a); }

> 
> Why name this argument, val, here?

Good point, I’ll fix.

> 
> > Source/WTF/wtf/Vector.h:1071
> > +inline void Vector<T, inlineCapacity, OverflowHandler>::uncheckedAppend(U&& val)
> 
> Why val? How about a word instead of this abbreviation.

Yeah, will fix this too.