William F Mitchell writes:
> Michael Metcalf wrote:
> >
> > Message text written by Lars Mossberg
> > >After all, Fortran DOES support zero-sized objects and they DO exist and
> > ARE defined and
> > "a" and "b" DO point to the same target, or?<
> >
> > Agreed. But the standard says, in the definition of ASSOCIATED, that for
> > zero-size arrays the result is false.
>
> I find this wording in the Fortran 95 standard, but not the Fortran 90 standard.
> Is it true that if Lars's compiler that returned T only claims to be Fortran 90,
> then it is not a bug in that compiler?
The words are in an interpretation (erratum) to f90. I think it was
interpretation 100, though I'm not looking at it right now, so I could
misremember the number. The number does sort of stick in my mind,
because it was a pretty controversial one. I recall disagreeing with
it, but I lost that one. (The interpretation did address a problem
that needed solving - I just thought this the wrong solution).
I'm afraid that my only personal suggestion here is to avoid using
the associated intrinsic with zero-sized arrays because it doesn't
work "right". ("Right" in this case being my personal definition of
what seems reasonable - not any objective measure).
You are almost certainly looking at the original f90 standard (that
being all that was ever printed or published in toto). To get those
words, you need to look at the original standard plus the 3 corrigenda
to it.
F95 is supposed to be compatable to f90. You shouldn't get changes in
behavior that was standard-conforming in f90.
--
Richard Maine
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
