[mpich-devel] overzealous F90/F90FLAGS error

[Jeff Hammond]

It is obnoxious that M(VA)PICH assumes that it is the only Fortran
code I will ever build and asserts that I do not have the right to set
the F90 environment variable in my shell, when this variable is not
used at all by M(VA)PICH but might be used by other packages.  I can
understand prohibiting the use of --with-f90, but prohibiting "export
F90" has far-reaching side-effects.

I understand that configure is trying to help the user do the right
thing (set FC and FCFLAGS), but this is the wrong way to do it.

This is the code that I would like to see removed:

# Since F90/F90FLAGS are replaced by FC/FCFLAGS, rather than silently
# substituting them, i.e. FC=$F90 and FCFLAGS=$F90FLAGS, we choose to emit
# an error message and abort to avoid any ambiguous/hidden bug in choosing
# Fortran90 compilers.
if test -n "$F90" -o -n "$F90FLAGS" ; then
    AC_MSG_ERROR([F90 and F90FLAGS are replaced by FC and FCFLAGS
respectively in this configure, please unset F90/F90FLAGS and set
FC/FCFLAGS instead and rerun configure again.])
fi

Googling "F90FLAGS -mpich" reveals a litany of scientific applications
that rely upon the F90 environment variable to be configured.  Some
users don't want to set F90 manually for every build if they know they
always want to use F90=ifort, hence the shell export.

Thanks,

Jeff

-- 
Jeff Hammond
jeff.science at gmail.com
http://jeffhammond.github.io/