I filed an upstream bug about 'No such file or directory' message leaking out of configure output. Upstream proposes to set F77 and F90 vars in my case to 'gfortran'. See https://trac.mcs.anl.gov/projects/mpich2/ticket/619#comment:3 for details.
Look like something for fortran eclass.
Sorry for the noise, messed up that reassignment...
I don't think F90 is actually a proper make environment variable though sometimes used, mpich2 upstream should may be consider FC instead. My understanding was that F77 is being slowly deprecated in favor of FC. On our side, the fortran.eclass needs either a complete rework or simply a dump. Lately I've been bumping packages without the fortran eclass and used the toolchain-funcs, but the toolchain also needs a bit of work (e.g. bug #278772). We should try to stick to FC and FFLAGS only so it makes it easier for makefiles.
(In reply to comment #3) > I don't think F90 is actually a proper make environment variable though > sometimes used, mpich2 upstream should may be consider FC instead. > My understanding was that F77 is being slowly deprecated in favor of FC. My impression is that F77 and F90 are preferred because it allows one specify for example F77=icc F90=gfortran. Imagine somebody has a parallelized and licensed F77-capable compiler but has to stick to F90 with gfortan from gcc. But, personally am not a programmer in fortran so I don't care, it just seems a bad decision. ;-)
fortran.eclass not used anymore in the tree