[J3] USE association of namelist groups?

Van Snyder van.snyder at jpl.nasa.gov
Sat Nov 24 13:35:01 EST 2018


On Sat, 2018-11-24 at 10:10 -0800, Steven G. Kargl via J3 wrote:
> On Sat, Nov 24, 2018 at 04:31:37PM +0900, Malcolm Cohen via J3 wrote:
> > 
> > >Well, it seems to me that C8102 prohibits access to a name of
> > >a namelist-group-name by use association.
> > 
> > And yet literally no one else thinks this is true.  Because it is NOT TRUE. 
> > It only prohibits use of a namelist-group-name by use association IN THAT 
> > BNF RULE.  Which happens to be the NAMELIST statement.
> > 
> 
> Apparently the person who wrote gfortran's support for
> NAMELIST has interpreted the constraint in the Fortran
> 95 final committee draft (X3J3/96-007, page 66, line 8)
> the same way as I have explained.  That support was
> written 13 years ago, and only today has a bug report
> surface.   So, you're use of "literally, no one else"
> seems somewhat condescending.
> 
> But, I have the information I need.  c8102 is sufficiently
> poorly worded that someone may misintrepret its meaning
> and thus someone might write a buggy Fortran processor.

The purpose of C8103 is to prevent this:

  module M1
    real :: A, B
    namelist /foo/ A
    namelist /foo/ B
  end module M1

  module M2
    use M1
    real :: X
    namelist /foo/ X
  end module M2

A NAMELIST can be constructed by several statements because of the
statement-length restrictions.

The NAMELIST statement in M2 would extend the namelist group.  While
technically feasible, it's prohibited because it would be a mess for
implementors.




More information about the J3 mailing list