(j3.2006) Questions about 29113

[Van Snyder]

On Mon, 2011-10-24 at 05:35 -0700, Bill Long wrote:
> 
> On 10/21/11 10:11 PM, Van Snyder wrote:
> > I have a few questions about 29113.
> >
> > What should the value of the "attribute" member of CFI_cdesc_t be for
> > the case of a nonpointer nonallocatable scalar, which I assume is only
> > meaningful for an assumed-length character?  None of the values in Table
> > 8.1 seem appropriate.
> 
> If you  have a nonallocatable nonpointer  assumed-rank dummy in a 
> Fortran interface,  the attribute in the corresponding C descriptor will 
> be CFI_attribute_assumed.  It is  possible for the actual argument to be 
> a scalar in that case, so CFI_attribute_assumed already covers at least 
> one case of scalar.  CFI_attribute_assumed would seem like the natural 
> option for nonallocatable nonpointer assumed-length character as well.

That's about what I expected.  29113 ought to make that explicit, and
have have a few examples.  N1885 has no discussion of or examples
dealing with character strings.

> Cheers,
> Bill
> 
> >
> > Do we need CFI_attribute_scalar, or is CFI_attribute_unknown_size OK?
> > If the latter, the final sentence of 8.3.4p6 ought to be changed to
> > explain this usage.  Maybe something like "... an object that is, or is
> > argument associated with, a scalar or an assumed-size dummy argument."
> >
> > Or, should we say that the "attribute" member is not used, and can have
> > any value, if the "rank" member is zero?
> >
> > CFI_cdesc_t seems to be a rather large hammer for nonpointer
> > nonallocatable assumed-length scalar character arguments.
> >
> > _______________________________________________
> > J3 mailing list
> > J3 at j3-fortran.org
> > http://j3-fortran.org/mailman/listinfo/j3
>