forums.silverfrost.com

[dk]

Hi,

I also came across the following spurious error when overloading procedures in FTN95.

Line 44: error 283 - DOT_PRODUCT must appear in a type declaration because IMPLICIT NONE has been used

this is despite the fact that in this example DOT_PRODUCT invokes the intrinsic procedure, not the overloaded extension. Adding DOT_PRODUCT to the use statement removes the error yet, according to the standard, should not be necessary. Not the worst bug possible but a bug nonetheless.

!*****************
module mod
implicit none
private
public::func,dot_product
! this overloads the new routine onto the
! same name as the intrinsic dot_product,
! which is legal F-95.
interface dot_product
module procedure dot_product2
endinterface dot_product
contains
!------
function func(x)
implicit none
real,intent(in)::x
real::func
func=x**2
endfunction func
!------
function dot_product2(x)
implicit none
real,intent(in)::x(:)
real::dot_product2
dot_product2=dot_product(x,x)
endfunction dot_product2
!------
endmodule mod
!*****************
module mod2
use mod,only:func
implicit none
contains
subroutine test(x)
real,intent(inout)::x(:)
x(1)=func(x(1))
! the next call uses the intrinsic dot_product, not the
! overloaded extension. My reading of the F-95 standard
! suggests it is not necessary to declare
! 'dot_product' in the USE statement.
! However when compiling with FTN95 the compiler
! reports an error and asks that dot_product be
! declared. This should not be necessary.
x=dot_product(x,(/3.0,6.0/))
endsubroutine test
endmodule mod2
!*****************

[PaulLaidler]

I have logged this as something to investigate.