[Info-vax] VMS port to x86

[Ken Fairfield]

On Thursday, May 31, 2012 7:52:47 AM UTC-7, Paul Sture wrote:
> On Thu, 31 May 2012 09:27:39 -0500, Bob Koehler wrote:
> 
> >    IIRC, the earliest Fortran compiler for VMS was "Fortran IV Plus",
> >    and actually had CHARACTER and other Fortran 77 pieces, so there was
> >    no need to mix Fortran 66 and 77 calling conventions for character
> >    data.
> 
> Grr.  If I'd known that I could have saved myself much pain.  The manual 
> I was working from didn't mention those extensions.

You just would have put off the pain for later.  :-)

I never saw the pre-F77 VMS Fortran version.  What I do know 
is that "old" FORTRAN habits were hard to break and many "new"
programs were written in essentially F66 even after F77 (and 
Character variables) had been around for a while.  Not to
mention "established" libraries of code written in the older
style that were simply recompiled for the VAX with the F77 
compiler.

As a result, I saw more than my share of subprograms that
passed Character variables to numeric arguments (as was
done on the IBM FORTRAN H compiler...where many of these
codes originated).  VAX/VMS Fortran, if it did anything,
made it possible to recompile IBM-sourced programs with
few or no changes.  Thus this feature in the VAX linker to
manage the argument type mismatch across separately compiled
sub-programs.

I recall Steve Lionel discussing the port of Fortran to the
Alpha specifically noting that this was one of a handful of
incompatabilities between the VAX and the Alpha since they
were *not* going to put this "fix-up" into the Alpha linker.

    -Ken