(j3.2006) resolving grammar ambiguities

[Van Snyder]

On Sat, 2009-10-17 at 00:56 -0700, Robert Corbett wrote:
> One of the members of Sun's Fortran team has been reviewing
> the Fortran 2008 draft.  I have been able to answer most of
> the questions he has raised, but tonight he raised one for
> which I could not find an answer.
> 
> Consider syntax rule R737 on page 158.  It is
> 
>      data-target    is  variable
>                     or  expr
> 
> The syntax rule for variable is rule R602 on page 117.
> It is
> 
>      variable       is  designator
>                     or  expr
> 
> Thus, expr can be generated from data-target in two ways
> 
>      data-target  ->  expr
> 
> and
> 
>      data-target  ->  variable  ->  expr
> 
> If the second derivation is used, data-target is subject
> to constraint C724, but if the first definition is used,
> it is not.

I don't think there is a problem with the constraints, since C602
requires <expr> to be a reference to a function that returns a pointer.

The "or expr" part of R737 is unnecessary, but harmless, since the
second derivation arrives at the same syntax, with the constraint that
<expr> is a reference to a function that returns a pointer.  Applying
C724 to the second derivation just says "<variable> has to have the
POINTER attribute" and then C602 says "and by the way, <expr> has to be
a reference to a function that returns a pointer."  It doesn't hurt
anything to say it twice.

> I could find no explanation of how such ambiguities should
> be resolved.
> 
> Bob Corbett
> _______________________________________________
> J3 mailing list
> J3 at j3-fortran.org
> http://j3-fortran.org/mailman/listinfo/j3