A reduction capture must actually have reduction type

[nelsc]

The grammar for a reduction capture allows an arbitrary declarator. This makes it possible to write a reduction capture which looks like it declares a pointer-to-reduction, or array-of-reduction, or even function-returning-reduction.

[nelsc]

These possibilities were never intended; consequently, the existing text doesn't actually say what happens if any kind of declarator derivation is attempted.

[nelsc]

The simple solution would be to change the grammar to allow only an identifier instead of a declarator. This is not quite consistent with the way a bit-field declaration is handled.

[nelsc]

I decided to go a different direction, less explicit, but consistent with the way member and bit-field declarations are described. (Members are described without reference to the specifiers in their declarations. Basically, they get their type and such just by virtue of the fact that a member declaration works like a declaration because it looks like a declaration.)