Re: Uninclude perl script

[Matt Newville <newville@cars.uchicago.edu>]

Hi Jim,

I respectfully disagree with the sentiment expressed.  The issue is
not whether Farges now has a working code, but why the build process
is broken.

Does HP's f77 allow include files or not?  If so, the error is in
the makefile.  Reliance of a Makefile on perl is not a wise choice,
but neither is both insisting on using include files and on not
using them, which is what the intent of the perl script was.

Traditionally, there's been a very strange notion of adhering to
'The F77 Standard', with John claiming to want to be fully portable
while ignoring the standard right and left.  Include files and
end-of-line '!' comments are nonstandard.  There is not a standard
solution to having them.  No, not even a fortran program like
'noinc' can be guaranteed to work for recursive includes, characters
like '\' in the file, '!' in character strings, or even continuation
lines.  Thankfully, the carriage-control days are behind us, but if
you want the code to be ANSI standard, the *only* choice is to
write it that way in the first place.

On the other hand, if f2c and g77 compile a code, it could be said
to be more portable than having it be 'ANSI f77'.  To that end, I
would highly recommend using autoconf -- I don't have an HP machine,
but ifeffit uses include files, end-do's, long variable and routine
names, and mixed C/Fortran and compiles on every machine I've tried
(SunOS4, Solaris, IRIX, DEC OSF, cygwin, and linux), including those
without a fortran compiler!  I _would_ recommend it, but I won't
bother getting my hopes up.

You want true portability?  Use perl: there's only one compiler.

--Matt

|= Matthew Newville        email: newville@cars.uchicago.edu
|= GSECARS, Bldg 434A      voice: (630) 252-0431
|= Argonne Natl Lab        fax:   (630) 252-0443
|= 9700 South Cass Ave     http://cars.uchicago.edu/~newville/
|= Argonne, IL 60439 USA

On Tue, 6 Mar 2001, James S Sims wrote:

> Steve Zabinsky wrote:
> >
> > You'd think it would -- this is, after all, 2001, not 1980...
> >
> > -----Original Message-----
> > What is the the problem Francois is having?  Doesn't the HP fortran
> > compiler support 'include's?
> >
> > --Matt
>
> The problem is that the HP is a Unix box, and he only speaks Microsoft :=)
>
> Thanks for all of your help. I think Francois and I can take it from here.
> Francois had gotten the full source tree, with Uninclude files, which is for
> developers,and then ran into a problem. I have given him instructions to try
> to pin it down, but that's really mainly curiosity. He also has the source
> files that are the result of "Unincluding" on a machine at UWash, so the
> issue of HP fortran compiler support of 'include's is a moot point.
>
> Jim
>