[J3] Type and kind conformance for VALUE dummy arguments

[Bill Long]

> On Nov 1, 2018, at 4:41 AM, Van Snyder via J3 <j3 at mailman.j3-fortran.org> wrote:
> 
> 15.5.2.3N1 in 18-007r1 suggests that if a dummy argument has the VALUE
> attribute, the effect is as if the actual argument were assigned to a
> temporary variable, and that variable were then associated with the
> dummy argument.

But you omitted the last sentence of the Note:

"The actual mechanism by which this happens is determined by the processor.”

Which is usually that the actual argument value is passed in a register rather than the address of that value. (As strongly hinted in the first sentence of the Note, also not presented above.)  In the case of a BIND(C) procedure, the passing scheme is determined by the C processor. 

> 
> It therefore does not appear to be necessary that the type and kind
> conform if the dummy argument is of numeric type.  The rules for
> intrinsic assignment could apply.

It is not clear how it can be concluded that something appears to be the case based on a non-normative Note.  But the suggested proposal seems to be promoting the idea of auto-conversion of arguments at call sites.  If that is want is actually needed, then vote for the Templates proposal.  That is the sensible way to accommodate multiple type/kind of actual argument for a particular dummy argument. Messing with the current argument association rules seems like a bad idea. 

Cheers,
Bill



> 
> Is this something we can relax?
> 
> 
> 

Bill Long                                                                       longb at cray.com
Principal Engineer, Fortran Technical Support &   voice:  651-605-9024
Bioinformatics Software Development                      fax:  651-605-9143
Cray Inc./ 2131 Lindau Lane/  Suite 1000/  Bloomington, MN  55425