taking this up a level, what is the rationale for having MPICH abort (without any override mechanism) the configuration process because of the definition of ANY environment variable that the configuration process does not reference anywhere?
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.
the reason that i suggested "F90FLAGS" for google is that "F90" will return a number of false positives that are not environment variables, not because i consider F90FLAGS to be a global setting.
no one is asking anyone to set F90FLAGS globally. that's ridiculous.
"export F90=ifort" is absolutely reasonable as a global setting since there is essentially no legitimate reason to use gfortran if one has a late-model ifort available.
One can easily scope the environment variables to the invocation of configure:
/home/robl/src/mpich/configure CFLAGS="-DJEFF_IS_A_HOTTHEAD" F90FLAGS="-Wall -Something_only_mpich_needs" --enable-robl-is-awesome
i understand how to do that because i am not just a Fortran application programmer. in fact, this is how i build every package. do i need to ask a friendly sysadmin to run a global search through ~/.bashrc files for "export F90" to make my point? jeff -- Jeff Hammond [email protected] http://jeffhammond.github.io/