DR: Resolutions to CA 008 and CA 016 do not mix

[hubert-reinterpretcast]

CA 016 indicates that applying reflexpr on a constant id-expression naming a non-volatile const (but not constexpr) variable does not produce a meta-object type that satisfies Constant, but CA 008 indicates that the meta-object type so produced should satisfy Constant.

I think we can't fix this without CWG or EWG input. I'll send a note to CWG if we can confirm that we can't fix this during the current review process.

[camio]

HT: Please address as an issue introduced during NB comment resolution process
(would like to hear from Axel)

AN: Good catch. The intent is that the type specified by a reflex operand of
id-expression kind, where the id-expression is usable in constant expressions,
satisfies Constant. That will include const variables. I.e.

const int s = 12;
int arr[s];
static_assert(Constant<reflexpr(s)>)

AN: should not assert. Hubert, does "id-expression usable in constant
expressions" sound like the right magic spell here? It's what we use in
[expr.const]p3.

[camio]

Left open as an issue to handle post-publication.