Possible regression in Intel 17?

[eric_n_]

Code below compiles and functions as expected with Intel 2013, 15, and 16.  Intel 17 gives ICE.  Other compilers seem to handle it as well.  We have a workaround in the "big" code, but would be curious to know if this is valid code or not.

ifort (IFORT) 17.0.1 20161005
Copyright (C) 1985-2016 Intel Corporation.  All rights reserved.

Thanks!

Eric

module prob

   type ddt1
     real              :: f ! function
     real              :: d ! the derivatives of f
   end type ddt1

   interface assignment(=)
     module procedure ddt1_assign_rd
   end interface

contains

   function fun_rd(v1)  result(f1)
     real, intent(in) :: v1
     type(ddt1) :: f1

     f1%f = v1
     f1%d = 0.0

   end function fun_rd


   subroutine ddt1_assign_rd(r, d)
     real,    intent(out)  :: r
     type(ddt1),  intent(in)   :: d

      r = d%f

   end subroutine ddt1_assign_rd

end module


program problem

   use prob

   implicit none
   real       ::  var1

   var1 = 1.0
   var1 = fun_rd(var1)

   print *,  var1

end program problem

[Kevin_D_Intel]

The code looks conformant. I note the internal error occurs where var1 appears as the argument to fun_rd. The internal error should not occur whether the code is conformant or not. Thank you for reporting this and for the convenient reproducer. I reported this to our Developers.

(Internal tracking id: DPD200417568)