Possibly spurious error with deferred type bound procedures

[MR]

Hi,
compiling the attached code, ifort complains that

ifort -c test-add.f90 test-add.f90(36): error #8314: If the rightmost part-name is of abstract type, data-ref shall be polymorphic [ADD]
call a%assign( a%add(b) )
--------------------^
compilation aborted for test-add.f90 (code 1)

My question is whether this error is correct; in particular, all my
variables are of abstract type c_ode, so I don't see the problem. As a
note, the code compiles with gfortran.

Thank you,
Marco

[bash]module mod_m
  
 implicit none  
  
 type, abstract :: c_ode
 contains
  procedure(i_add),    deferred, pass(x) :: add
  procedure(i_assign), deferred, pass(y) :: assign
 end type c_ode
  
 abstract interface
  pure function i_add(x,y) result(z)
   import :: c_ode
   implicit none
   class(c_ode), intent(in) :: x, y
   class(c_ode), allocatable :: z
  end function i_add
 end interface

 abstract interface
  pure subroutine i_assign(y,x)
   import :: c_ode
   implicit none
   class(c_ode), intent(in)  :: x
   class(c_ode), intent(out) :: y
  end subroutine i_assign
 end interface

contains

 pure subroutine sub(a,b)
  class(c_ode), intent(in)  :: b
  class(c_ode), intent(out) :: a
 
   ! a = a+b
   call a%assign( a%add(b) )
 
 end subroutine sub
  
end module mod_m
[/bash]

[Steven_L_Intel1]

Yes, this is our bug. It's an issue we've been wrestling with for a while now as the wording in the standard seemed ambiguous. However, we've resolved that and fixed the bug - the fix should appear in an update later this year. The issue ID is DPD200170220.

[MR]

OK, thank you, then I'm looking forward to seing the new version.

Marco