(j3.2006) Did we intend to prohibit this?

[Van Snyder]

On Fri, 2017-03-10 at 21:47 +0000, Clune, Thomas L. (GSFC-6101) wrote:
> > On Mar 10, 2017, at 2:09 PM, Van Snyder <Van.Snyder at jpl.nasa.gov> wrote:
> > 
> > On Fri, 2017-03-10 at 16:13 +0900, Cohen Malcolm wrote:
> >> 
> >>> So how does it get deallocated?
> >> 
> >> "it" is not a variable, so does not get deallocated (in the Fortran
> >> language sense of "deallocated").
> >> 
> >> You know what the function result variable is, and it is not the value
> >> that is returned by the function, nor has it ever been.  The result
> >> variable is what gets deallocated.
> > 
> > So what does 9.7.3.2p4 mean?
> > 
> >        "If an executable construct references a function whose result
> >        is allocatable or has an allocatable subobject, and the function
> >        reference is executed, an allocatable result and any allocated
> >        allocatable subobject of the result is deallocated after
> >        execution of the innermost executable construct containing the
> >        reference."
> > 
> > If the result variable is in fact what gets deallocated, even though the
> > word "variable" does not appear in 9.7.3.2p4, then the result variable
> > of F(X) in
> > 
> >  call move_alloc ( f(x), v )
> > 
> > IS a variable, right?  So what's the problem with executing it?
> 
> The result variable of F(X) is not in the expression you provide.
> Outside of F all you can ?see" is the result _value_.

According to 9.7.3.2p4, whatever the function result is, it must be
allocatable, or it couldn't be deallocated.  So what's the problem with

  call move_alloc ( f(x), v )

> I think that what you probably want in this situation is:
> 
> allocate(v, source=f(x))

If it's not known not to be allocated, one would need the somewhat more
verbose

  if ( allocated(v) ) deallocate ( v )
  allocate(v, source=f(x))

As of F08, this is no different from

  v = f(x)

This brings me back to the other question in my original message, which
Malcolm erroneously interpreted as a question about the standard, rather
than a question about what processors actually do:

In

  v = f(x)

where v and the result of F() are allocatable, do processors typically
reallocate v with the correct size if necessary, copy the value of f(x)
into it, and deallocate the result of f(x), or do they do what

  call move_alloc ( f(x), v )

would obviously have done if it were allowed?