Re: [boost] [move][unique_ptr] c++14 unique_ptr comes to town

Subject: Re: [boost] [move][unique_ptr] c++14 unique_ptr comes to town
From: Ion Gaztañaga (igaztanaga_at_[hidden])
Date: 2014-09-03 17:58:59

El 03/09/2014 22:33, Peter Dimov escribió:
> Ion Gaztañaga wrote:
>> One problem I've found for constructors taking the pointer is that the
>> pointer can be an incomplete type whereas has_trivial_destructor
>> requires a complete type.
>
> Are we in the default_delete case here, or in the arbitrary deleter case?

I was trying in the default_delete case, sfinaeing out this constructor:

template<class U, E>
unique_ptr(unique_ptr<U, E> &&u, enable_if<... E ....>... )

As E must be convertible to D, and to check this, it was checking if T 
was a base class of U and had no virtual destructor, it seems that the 
intrinsic complained about non-complete types. Maybe I was not properly 
filtering it first with is_convertible.

I've changed that approach to static_assert inside the function instead 
of sfinaeing the constructor. This seems to work. In any case, is 
expected is_convertible<T*, U*>::value to return false with pointers to 
incomplete types?

Best,

Ion