On Thu, Oct 11, 2018 at 11:35:23AM -0400, Jason Merrill wrote:
> > + /* [dcl.fct.spec]
> > + "the constant-expression, if supplied, shall be a contextually
> > + converted constant expression of type bool." */
> > + expr = build_explicit_specifier (expr, tf_warning_or_error);
> > + /* We could evaluate it -- mark the decl as appropriate. */
> > + if (expr == boolean_true_node)
> > + set_and_check_decl_spec_loc (decl_specs, ds_explicit, token);
> > + else if (explicit_specifier)
> > + /* The expression was value-dependent. Remember it so that we can
> > + substitute it later. */
> > + *explicit_specifier = expr;
>
> What if expr == boolean_false_node?
Then we proceed like no explicit was present and the decl isn't marked as
explicit/nonconverting. Perhaps I could have made this clearer with
else if (expr == boolean_true_node)
/* Don't mark the decl as explicit. */;
or somesuch.
> > + /* Handle explicit(dependent-expr). */
> > + if (DECL_HAS_DEPENDENT_EXPLICIT_SPEC_P (t))
> > + {
> > + tree spec = lookup_explicit_specifier (t);
> > + spec = tsubst_copy_and_build (spec, args, complain, in_decl,
> > + /*function_p=*/false,
> > + /*i_c_e_p=*/true);
> > + spec = build_explicit_specifier (spec, complain);
> > + DECL_NONCONVERTING_P (t) = (spec == boolean_true_node);
> > + }
>
> What if spec is still dependent, e.g. after partial substitution of a
> member template?
Something like this?
template<typename> struct A {
template<typename T, int N = 0>
explicit(N) operator T();
};
void
bar ()
{
A<int> a;
int i = a;
}
This also seemed to work: if spec is still dependent, the decl isn't marked as
DECL_NONCONVERTING_P, and we'll try again after deduction (fn_type_unification
in add_template_candidate).
Marek
