http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41951
--- Comment #17 from janus at gcc dot gnu.org 2012-11-16 12:33:00 UTC ---
(In reply to comment #16)
> One way to reject the abstract case in comment 11, would be to just do the
> checking, but not add the procedure to the operator list:
The only problem with the patch in comment 16 is that the error message it
gives on comment 11 ...
generic :: assignment(=) => assign
1
Error: Entity 'assign_it' at (1) is already present in the interface
... is technically not correct. 'assign_it' is not added twice, but in one case
is just the interface of the deferred type-boud procedure 'assign'. Instead, it
might be better to complain about 'Ambiguous interfaces' (since any
non-abstract extension of the type 'sort_t' will have to override/implement the
type-bound 'assign' with a procedure which is ambiguous to 'assign_it').
However, such an error about 'Ambiguous interfaces' is also missing for the
following case (which is still accepted):
module m_sort
implicit none
type, abstract :: sort_t
contains
generic :: assignment(=) => assign
procedure(assign_ifc), deferred :: assign
end type
abstract interface
subroutine assign_ifc (lhs, rhs)
import
class(sort_t), intent(out) :: lhs
class(sort_t), intent(in) :: rhs
end subroutine
end interface
interface assignment(=)
subroutine assign_it (lhs, rhs)
import
class(sort_t), intent(out) :: lhs
class(sort_t), intent(in) :: rhs
end subroutine
end interface
end module
