Hi Matt,

I agree with everything you say.  Thanks for taking the time to explain.  As someone who uses fortran all the time, it makes sense for me to own Ifort, but since it's only free for academics on Linux, that's not true for most people.

Actually, for FEFF our philosophy is also that it should compile with any compiler on any platform.  We respond to EVERY compilation issue and never tell people to use another compiler.  However -- most of our users don't want to compile code at all: they like a nice binary that I've compiled for them.  (I doubt that the ifeffit/demeter users would object to this option ;).)  For that, I like to use Ifort - for my purpose it's the better compiler.  But we make sure that FEFF compiles well using gfortran etc. also.

If I were to successfully built the codes on this mailing list, I'll definitely share the recipe, as well as any kind of binary other people can use without having to build/compile/... .   It won't be for today, unfortunately :).

Cheers,

Kevin




On Fri, Jan 10, 2014 at 7:28 AM, Matt Newville <newville@cars.uchicago.edu> wrote:
Hi Kevin,

On Thu, Jan 9, 2014 at 11:37 PM, Kevin Jorissen
<kevinjorissenpdx@gmail.com> wrote:
> I can concur with others here that making something work on Mac can be a
> miserable experience.  Sometimes you have to admit defeat and change the
> approach.
>
> Anyway, I tried "sudo port install p5.16-ifeffit" and failed:
>
> ~/Downloads% sudo port install p5.16-ifeffit
> --->  Computing dependencies for p5.16-ifeffit
>
> --->  Dependencies to be installed: ifeffit pgplot perl5.16 gdbm
>
> --->  Extracting pgplot
>
> Error: org.macports.extract for port pgplot returned: command execution
> failed
>
> Error: Failed to install pgplot
>
> Please see the log file for port pgplot for details:
>
>
> /opt/local/var/macports/logs/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_graphics_pgplot/pgplot/main.log
>
> Error: The following dependencies were not installed: ifeffit pgplot
> perl5.16 gdbm
>
> To report a bug, follow the instructions in the guide:
>
>     http://guide.macports.org/#project.tickets
>
> Error: Processing of port p5.16-ifeffit failed
>
> ~/Downloads%
>
>
> (This is upon rerunning the command - the first time, it installed many more
> dependencies -- does anyone else suspect "port" of being designed to fill
> hard disks as fast as possible?)  System is 10.9 generally up-to-date with
> most things.  pgplot seems to fail during extraction.
>
> I doubt that I'd be able to contribute a lot to this project, but I can test
> things.
>
> Also I'm surprised to read so much about gfortran about this thread, but
> maybe I'm missing some common knowledge on the project ...  I am generally
> much happier with Ifort.

If you or anyone else wants to use ifort or any other compiler, that
would be fine.  For many years, I did keep a license for
Digital/Compaq/Intel Fortran on Windows in order to make the Windows
dlls for distribution of the Ifeffit package.  Even on Windows, this
stopped being necessary, as the MinGW suite began including gfortran.

Unlike Feff, these are open source projects, and we absolutely insist
that anyone interested can use, (try to) build, and distribute these
programs -- both Bruce and I are begging for such help, actually.
Requiring  non-free tools simply to build these programs would
definitely put a burden on potential developers.   I don't see any
reason for why that would be necessary,  especially on Linux and Mac,
for which the system compilers are gcc (or gcc/llvm) which includes a
working fortran compiler.

Many free software projects (notably for me, scipy, but also the
macports that Frank mentioned) use gfortran on Macs.  And gfortran
works fine for me for many things on 10.6 and 10.8. I can compile the
basic ifeffit library and other fortran libraries used in Larch
(though I did recently notice some dependency issues between 10.6 and
10.8, as the compiler version changed, and haven't tried 10.9 yet).
 I have not been able to build PGPLOT (perhaps not crucial for
Demeter) or the Ifeffit-Perl bridge successfully on 10.8.  I believe
this is probably a matter of getting the compiler / linker arguments
correct....

If you or anyone else can get these codes to build using ifort, more
power to you.  If so, can you post your solution?

--Matt
_______________________________________________
Ifeffit mailing list
Ifeffit@millenia.cars.aps.anl.gov
http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit