https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124070
--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Actually, I think it is a bug in convert_nontype_argument.
For reflections it does
else if (REFLECTION_TYPE_P (type))
{
if (!REFLECTION_TYPE_P (TREE_TYPE (expr)))
{
if (complain & tf_error)
error ("%qE is not a valid template argument for type %qT "
"because it is of type %qT", expr, type, TREE_TYPE (expr));
return NULL_TREE;
}
return expr;
}
while e.g. for scalars it does:
/* Notice that there are constant expressions like '4 % 0' which
do not fold into integer constants. */
if (!CONSTANT_CLASS_P (expr) && !val_dep_p)
{
if (complain & tf_error)
{
int errs = errorcount, warns = warningcount + werrorcount;
if (!require_potential_constant_expression (expr))
expr = error_mark_node;
else
expr = cxx_constant_value (expr);
...
}
else
return NULL_TREE;
}
So bet for reflections it should also do something similar, if !val_dep_p and
expr is not REFLECT_EXPR, return NULL_TREE or use cxx_constant_value.
And I wonder if the same problem isn't in the nullptr_t case, there we have
else if (NULLPTR_TYPE_P (type))
{
if (!NULLPTR_TYPE_P (TREE_TYPE (expr)))
{
if (complain & tf_error)
error ("%qE is not a valid template argument for type %qT "
"because it is of type %qT", expr, type, TREE_TYPE (expr));
return NULL_TREE;
}
return expr;
}
and nothing that would similarly check for integer_zerop (expr).
