Fortran 2008 feature of pointer rank remapping not properly diagnosed

[Harald]

Hello,

the Fortran 2008 feature of pointer rank remapping with a target that
has rank greater than 1 is not diagnosed when -stand f03 is specified:

[fortran]program ifort_pointer_assign
  implicit none
  integer, parameter :: n1 = 2, n2 = 3
  integer, target    :: base(n1,n2)
  integer, pointer   :: j(:)

  base = 0
  j(1:n1*n2) => base(:,:)       ! F2008 feature!
  print *, j(n1*n2)

end program ifort_pointer_assign
[/fortran]

% ifort ifort_pointer_assign.f90 -stand f03 -standard-semantics -warn all

(No warning produced)

% ifort -V
Intel Fortran Compiler XE for applications running on IA-32, Version 12.1.1.256 Build 20111011

[Steven_L_Intel1]

Nice catch. Thanks for pointing this out to us. I have let the developers know - the issue ID is DPD200176779.

For those who care, the relevant constraints are F2003 C720 and F2008 C718.

[Steven_L_Intel1]

This has been fixed for a future product release.

[jimdempseyatthecove]

Steve,

>>j(1:n1*n2)=>base(:,:)!F2008feature!

This looks like a great feature and like you can eliminate convoluted TRANSFER statements.

What happens when the target is a DUMMY argument from a call using assumed shape with a stride other than 1 (of an array that had stride 1). IOW Dummy represents a logically contiguous array that isnon-contiguous in physical memory.

Jim

[Steven_L_Intel1]

Jim, it should work just fine. The compiler will need to generate code to compute the new strides, that's all.