From: Thomas Koenig on
Has anybody been thinking about C++ / Fortran interoperability?

Obviously, there are a lot of things in C++ that Fortran cannot do,
such as multiple inheritance, templates and constructors. Still,
a common subset larger than the C interoperability probably exists,
and might well be useful.

For practical purposes, such an interoperability might be restricted
to a compiler family.

Comments?
From: Richard Maine on
Thomas Koenig <tkoenig(a)netcologne.de> wrote:

> Has anybody been thinking about C++ / Fortran interoperability?

Not me, partly because I don't know enough C++ to do even a half-decent
job. But...

> For practical purposes, such an interoperability might be restricted
> to a compiler family.

I'd think that a really bad idea. Sounds like something that would be
literally worse than useless in that it could do harm. Unless...

Perhaps I misunderstand what you are saying. There does seem a good
chance of that. If you want to think about doing an experimental
implementation that works with only some compilers, that seems sensible
and natural. But the specification of an interoperability feature should
not be specific to a compiler; it should be in terms of the relevant
standards. To me, there is a huge difference between the specification
and an implementation. The above-quoted words sound to me like they are
talking about a specification that only applied to a compiler family,
which I would find objectionable. I suspect instead that you might be
talking about implementation for a compiler family.

Otherwise, it reminds me a bit of a NASA "interoperability" panel that I
was once asked to be our center's representative on, as people knew I
had an interest in interoperability. In this case, it had to do with
interoperability of things like software applications such as word
processors and others widely used throughout the agency. Then I found
out the ground rules the panel was to operate under. It was decreed by
the high-level bureaucrat who commissioned the panel that only Microsoft
products were to be used. It was made clear to me that even mentioning
other possibilities was likely to draw ire from above. It was one of
those orders that you were not supposed to question. As far as I could
tell, the panel was not about actually facilitating interoperability,
but instead about how to work around an arbitrary and absolute mandate
to use non-interoperable software. We could only look at things like how
to set up Windows server systems that Mac, Linux, Sun, and other users
could access. I declined to serve on the panel under those ground rules,
as it sounded to me more like a panel on non-interoperability.

--
Richard Maine | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle | -- Mark Twain
From: Nick Maclaren on
In article <1jmrdhz.4760cd1mbcaxeN%nospam(a)see.signature>,
Richard Maine <nospam(a)see.signature> wrote:
>Thomas Koenig <tkoenig(a)netcologne.de> wrote:
>
>> Has anybody been thinking about C++ / Fortran interoperability?
>
>Not me, partly because I don't know enough C++ to do even a half-decent
>job. But...

Nor do I, though I am horribly afraid that I may be approaching the
state of being one of the more knowledgeable people - and my ignorance
of both is considerable!

>> For practical purposes, such an interoperability might be restricted
>> to a compiler family.
>
>I'd think that a really bad idea. Sounds like something that would be
>literally worse than useless in that it could do harm. Unless...

Yes. And to finish your dots ... unless pigs start to fly.

The killer isn't the general facilities, but the details and, far,
far worse, the semantic model. I don't know either at all well,
but even the existing C interoperability is a mess in those
respects. The chances of there being a close enough match for
the advanced features are infinitesimal.

I was involved in IBM's CEE project, and even a single company had
major problems getting PL/I and Cobol to interoperate at that level.
And, yes, that was differences in the semantic model, though I can
no longer remember the details.

However, that doesn't mean to say that NOTHING can be done. I have
a back-burner project to design and implement some interoperability
for data alone. If I can get enough going in a test version of
gfortran, it would justify proposing a TR. Another great IF ....

It is possible that there is enough commonality to do something,
but the first step would be to do the investigation of the details
and semantics.



Regards,
Nick Maclaren.