[J3] [EXTERNAL] Re: Allocatation upon assignment of a dummy argument in an interoperable procedure

[Van Snyder]

On Mon, 2019-10-07 at 21:02 +0000, Bill Long via J3 wrote:
> The statement 
> 
> foo = 42
> 
> and the statements 
> 
> allocate( foo ); foo = 42 
> 
> should do the same thing, since the INTENT(OUT) will (should) ensure
> that the argument is deallocated on entry.   So, while it is possible
> that the descriptor is messed up somehow, that would not be a
> sufficient explanation of why one works and the other does not.   If
> the descriptor is bad, both versions should fail in the Fortran
> allocate routine. 
> 
> It would be useful to see the C code that sets up the descriptor. 

It would also be interesting to know which Fortran processor "fails,"
and whether others "fail" as well. I.e., is there a problem with a
(specific) Fortran processor, or is the descriptor damaged?

> Also, as an aside, it is usually considered poor programming style to
> use the same name for a dummy argument and the binding label for the
> routine. 
> 
> Cheers,
> Bill
> 
> 
> 
> > On Oct 7, 2019, at 3:53 PM, Robert Corbett via J3 <j3 at mailman.j3-fortran.org> wrote:
> > 
> > How does it fail?
> > 
> > Was the C descriptor initialized by CFI_establish and then passed changed to the interoperable subroutine, or was the C descriptor changed after it was established?
> > 
> > Bob Corbett
> > 
> >> On Oct 7, 2019, at 11:09 AM, Vipul Parekh via J3 <j3 at mailman.j3-fortran.org> wrote:
> >> 
> >> Is the following code conforming?
> >> 
> >>  subroutine sub(foo) bind(C, name="foo")
> >>     !.. Argument list
> >>     integer(c_int), allocatable, intent(out) :: foo
> >>     foo = 42  !<-- fails
> >>     !allocate( foo ); foo = 42  !<-- works
> >>  end subroutine
> >> 
> >> When foo is invoked from a companion C processor with a successfully
> >> established C descriptor with the right attributes, the use of
> >> ALLOCATE statement followed by the assignment works as I expect.
> >> However allocation upon assignment, as shown above, fails.
> >> 
> >> Thank you,
> >> Vipul Parekh
> > 
> 
> Bill Long                                                                       longb at cray.com
> Principal Engineer, Fortran Technical Support &   voice:  651-605-9024
> Bioinformatics Software Development                      fax:  651-605-9143
> Cray, a Hewlett Packard Enterprise company/ 2131 Lindau Lane/  Suite 1000/  Bloomington, MN  55425
> 
> 
>