[patch,gfortran] Fix PR17917 - module equivalences (redux)

[Tobias.Schlueter@Physik.Uni-Muenchen.DE]

Quoting Paul Thomas <paulthomas2@wanadoo.fr>:
> >I couldn't see how uniqueness is important, the code only depends on == / !=
> >NULL, so I think you should just use
> >+ 	    e->module = gfc_get_string(module_name);
> >
> >
> Not so, e->module is used to name the union and is mangled to provide
> the external symbol for the loader.

Uh, well, yes of course.  After I had read through all the code up to the point
where it calls build_decl and had understood why a name is needed, I had of
course forgotten where the name came from :-)

> Fine.  I have just spent more time than I should on responding to FX's
> excellent remark; turns out that I had the solution in minutes but that
> the test case was wrong.  It's just now regtesting.

Great.

> >(who promises to dedicate every moment of boredom at the office to gfortran)
> >
> >
> Reading Dilbert cartoons is pretty good too.

But it's not of equal benefit for mankind.  (BTW I tried to convince my
officemates to learn Fortran instead of C++ for their numerical applications,
but somehow they thought this was a reason to point out to me that their hair
has not yet turned gray.)

Cheers,
- Tobi