Hello,
I think you have a point, I've been bitten by this sort
of thing myself.
It's too late for any sort of more-than-editorial change
to f2k, but it's a good time to propose you wishes for the wish list
for the next standard.
I agree with James Giles that the import statement is a more likely
place for this sort of thing.
In some weeks, the latest draft of f2k should be on the web site,
it will be called 03-007, and probably be available as PDF and PS.
If you have a specific proposal, it will be more likely to receive
the sort of response you want than a ill-defined "fix it" sort
of thing (written without regard to the merits of the case).
Note that whatever is proposed must be compatible with existing
practice (as is James Giles' proposal) and will probably
be modified as it goes through the process (which is a long process).
Experienced members of J3 have difficulty having a paper
approved without modifications.
The best you can do now is to have an item on the wish list
for next time. Since you're posting from a ".UK" domain,
contact someone in the British standards organization
for more help.
HTH
--
Cheers!
Dan Nagle
Purple Sage Computing Solutions, Inc.
On Fri, 11 Apr 2003 11:30:24 +0100, P Suckling
<[log in to unmask]> wrote:
>Dear Dan Nagle and others.
>
>I was programing away just now, and felt unsatisfied by a habit I have got into
>to try and keep track of variables in module procedures obtained by host
>association.
>
>My habit is to keep a commented list in each module procedure of all variables I
>think are being obtained by host association and either `used and changed' or
>`used and unchanged'.
>
>A problem with host association as I see it, is that if my commented list is
>wrong and that I have accidentally changed or used a host associated variable
>there is no way of a compiler picking this up and generating an error. (An
>exception is the case where the module procedure is PURE, and you change a host
>associated variable.)
>
>I don't know if any move has been made to address this issue with Fortran 2000,
>but if not, can I suggest something like the following for some future version
>of fortran:
>
>USE HOST, NONE
>to say you don't want to obtain any variables by host association
>
>USE HOST, ONLY : [local_name =>] host_name, ...
>as with modules, but for host associated variables, and
>
>USE HOST
>to do the default which is to leave all variables available for host
>association.
>
>Even better would be the inclusion of an additional attribute that could be
>added to the above syntax would serve a similar purpose to that of the INTENT
>attribute. It could indicate whether you want to allow this variable to be
>changed by the module procedure or not.
>
>This would make debugging host association problems far simpler.
>
>Cheers
>
>Paul Suckling
|