associate construct and targets

[IanH]

Sometimes the "is this thing a target" logic for the associate name of an associate construct goes awry.

PROGRAM p
  IMPLICIT NONE
  
  TYPE ta
    INTEGER :: i
  END TYPE ta
  
  TYPE tb
    TYPE(ta), POINTER :: p
  END TYPE tb
  
  TYPE(tb) :: obj
  INTEGER, POINTER :: pi
  
  !****
  
  ! 6.7.1.4p1 Allocation of a pointer implicitly creates an object that 
  ! has the target attribute.  So the thing pointed at by `obj%p` has 
  ! the TARGET attribute, though `obj` itself does not.
  ALLOCATE(obj%p)
  
  ! The component `i` is a subobject of the thing pointed at by `obj%p`.  
  ! As the thing pointed at by `obj%p` is a variable, `obj%p%i` is a 
  ! variable.
  !
  ! 5.3.17p2 If an object has the TARGET attribute, then all of its 
  ! nonpointer subobjects also have the TARGET attribute.
  !
  ! 8.1.3.3p1 The associating entity has the TARGET attribute if an 
  ! only if the selector is a variable (which it is) and has either 
  ! the TARGET (which it does) or POINTER attribute.
  ASSOCIATE(x => obj%p%i)
    pi => x           ! ifort says "error #6976: the variable must 
                      ! have the target attribute...".  Bad ifort!
  END ASSOCIATE
END PROGRAM p

>ifort /check:all /warn:all /standard-semantics "2014-10-09 associate-ptr.f90"
Intel(R) Visual Fortran Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 15.0.0.108 Build 201407
26
Copyright (C) 1985-2014 Intel Corporation.  All rights reserved.

2014-10-09 associate-ptr.f90(32): error #6796: The variable must have the TARGET attribute or be a subobject of an objec
t with the TARGET attribute, or it must have the POINTER attribute.   
    pi => x           ! ifort says "error #6976: the variable must
----------^
compilation aborted for 2014-10-09 associate-ptr.f90 (code 1)

[Kevin_D_Intel]

Thank you for the compelling bug report Ian. I reported this to Development (see internal tracking id below) for repair.

(Internal tracking id: DPD200361928)

(Resolution Update on 02/06/2015): This defect is fixed in the Intel® Parallel Studio XE 2015 Update 2 release (pkg ids: 2015.2.179 - Windows -and- 2015.2.164 - Linux)

[Kevin_D_Intel]

I confirmed this issue is resolved in the Intel® Parallel Studio XE 2015 Update 2 now available from the Intel Registration Center.

[JVanB]

About that Update 2 situation...

I went to Premier support. Wouldn't let me file a bug report as usual. Clicked on the button for Intel Registration Center. Logged in again (this is on a different account BTW). It said my expiration date is 03 Nov 2015. Tried to download the update. The UPDATE button was dead. Logged out (the web page makes me log in again to log out and brings up an error message, perhaps in Japanese. Finally got out of there. Went to Intel Software Manager from the Start menu. Said there were two updates available, ifort and MKL. Tried to update ifort but after I accepted the EULA and rejected the spyware, was left at a page that said there was nothing to download and left no path to advance. Tried Customize but nothing useful happened. Then I got out of that and went back to software manager and tried to update MKL, and this time everything went through.

Accordingly I tried to update ifort again, and this time I got through the point where I was blocked before, but at the end of the process the installer said in one place that my installation was complete and in another that removal of ifort (or whatever) was complete. In the event, my command prompt shortcuts have been deleted from the Start menu and from where they were under Intel's submenus, so my installation is now totally borked and I can't install anything or even determine what version of ifort I have installed on my machine. Thank you for this.

Maybe someday Intel will get a machine running MSIE 11 on Windows 8.1, but until then the untested installers and bug reporting page are somewhere between useless and counterproductive.

[Steven_L_Intel1]

RO, please send me an email (steve dot lionel at intel dot com) and I'll try to help.