[Plplot-devel] Re: What is your opinion of a bug-fix 5.6.1 release in (say) two weeks?

[Alan W. I. <ir...@be...>]

On 2006-05-08 10:07+0200 Arjen Markus wrote:

>> [Alan said]
>> BTW, you may have noticed in one of my commit messages that I had to 
>> change
>> the name of that example file from x01f.f95 to x01f.f.  The autotools
>> weren't properly understanding that f95 suffix despite documentation 
>> saying
>> that they should.  I think they are stuck part way through the fortran 77 
>> to
>> fortran 95 transition so funny things are going on that we have to work
>> around (as in renaming the examples file with a ".f" suffix.)
>
> This is bad news: it means that most compilers will assume the files to be
> in fixed-form. You can usually add an option to force them to read the
> file as free-form, but that is not the solution I would prefer.

OK, I took the hint, and tried harder, and we now have good news again.  I
was just making some dumb mistake with the suffix rule so once I figured
that out I switched back to using x01f.f95 (and changed it to use
free-form).  When I made the appropriate adjustment to
examples/f95/Makefile.am to use x01f.f95 rather than x01f.f, automake did
the right thing and started using $FC to build and $FCFLAGS and $AM_FCFLAGS
for the flags.  Therefore, I decided to make a consistent change for *.f to
*.f95 (using free form) in bindings/f95 as well. Despite these rather large
changes everything (including make check now) seems to continue to work
fine.  Also, during this latest test, I cvs updated to get Andrew's
bounding-box change, and that works fine on my system as well.

Arjen, please be careful.  All the f95-related fortran source files in
bindings/f95 and examples/f95 now have the f95 suffix (and it should remain
that way).  So cvs update will remove any *.f form from those directories.
So back those files up if you want to preserve your work before executing
cvs update, and use the *.f95 form from now on for those particular
directories.

My previous conclusion that FC, FCFLAGS, and AM_FCFLAGS was unused was only
correct for the *.f case.  These flags are the ones used (and F77, FFLAGS,
and AM_FFLAGS ignored) for the present case where bindings/f95 and
examples/f95 only have *.f95 files.  For the case of bindings/f77 and
examples/f77 where all source files are in the *.f form (and should remain
that way), the reverse is true.  So autotools is doing exactly the right
thing and using the correct set of variables when *.f files are encountered
and similarly when *.f95 files are encountered.  I now output FC and FCFLAGS
in the configure summary since those flags are important in the .f95 suffix
case.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the Yorick front-end to PLplot (yplot.sf.net); the
Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________