Is it intended that an aggregate with a deleted or nonexistent default constructor satisfy DefaultConstructible?

[timsong-cpp]

For example, given

struct A{
    A(const A&) = default;
};

The following expressions are all valid:

A{};
new A{};
new A[5]{};

which would seem to mean that A meets DefaultConstructible as defined in [concepts.lib.object.defaultconstructible], unless I missed something.

If this is intended, then the note in [concepts.lib.object.defaultconstructible] is incorrect.

Proposed Resolution

See P0547R1.

[CaseyCarter]

Close enough to intended, though I'm interested to get feedback from users on our choice of list-initialization for Constructible et al.

Proposed Resolution:

Modify the note in [concepts.lib.object.defaultconstructible]/1 as follows:
1 [ Note: The array allocation expression new T[n]{} implicitly requires that T
- has a non-explicit
+ does not have an explicit
default constructor. --end note ]

[CaseyCarter]

Committee discussion of this issue needs to consider the fact that this implies the existence of DefaultConstructible types T for which the declaration T foo; is ill-formed. This is enough of a freakish corner case that I'm happy to leave it ill-formed and let the fix be "don't do that."

[ericniebler]

As with std::is_constructible, we can require DefaultConstructible to check the well-formed-ness of the statement T tmp;. That will require a compiler intrinsic, which makes it harder to implement the TS, but that may be a price worth paying. From googling around, I'll note the __is_constructible intrinsic already exists on MSVC 2013 and clang.