(j3.2006) finalizer infelicity
Wed Nov 21 17:57:48 EST 2007
I think the question here is whether the standard requires that, for a
conforming program, the user is required to write the final routine such
that it conforms to the rules for argument association. I would say
that, yes, this is required. Just list the user is required to use a
name that has fewer than a prescribed number of characters.
The constraint I thought you wanted added by the interp process was your
suggestion that the final routine dummy argument had to be
assumed-shape. That's not acceptable.
Aleksandar Donev wrote:
> On Wednesday 21 November 2007 14:17, Bill Long wrote:
>> We should not, buy the interp process,
>> introduce new constraints that turn conforming programs into
>> non-conforming programs. Interps are for interpretations of what the
>> standard means.
> Maybe I am misunderstanding something here. Does the standard explicitly say
> this program is non-conforming or does that fall out of "oops, the standard
> does not provide an interpretation of this program"? If the latter, I find it
> much less convincing that we "changed our mind", rather, we lost our mind to
> begin with.
> The thing Mike was picking a fight with ws the sentence saying the finalizer
> will be invoked (even though the shapes do not agree). Why is fixing this
> statement to properly account for shape mismatch outside the interp process.
> Or do you honestly believe someone thought about this and decided only rank-1
> arrays of length 2 of type T are allowed, but any size rank-2 arrays are.
> J3 mailing list
> J3 at j3-fortran.org
Bill Long longb at cray.com
Fortran Technical Support & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120
More information about the J3