On May 20, 2005, at 7:37 PM, Shishir Shroff wrote:
>> You used "--without-FC", which is not the same thing as "without-fc".
>>
>> You might want to try again with --without-fc.
>>
>> Also, ensure that you have one consistent installation of LAM/MPI.
>> It's a little confusing from the info you provided, but laminfo says
>> that there is not Fortran compiler, but yet mpicc and mpif77 are
>> both trying to link in the Fortran MPI library, which shouldn't
>> happen if your LAM doesn't support Fortran. I'm guessing that
>> laminfo is coming from one installation and mpicc/mpif77 are coming
>> from a different LAM installation.
>
> I configured, built and installed lam-7.1.1 with the "--without-fc"
> option. That worked fine.
> Then, I untarred the lamtests suite source in a diff. location than
> my lam installation.
> The output of the configure script and the config.log file that I
> sent earlier was the result of trying to configure lamtests and not
> lam. (The --without-FC was just a an experiment I tried for the
> lamtests configuration when --without-fc did not work). I still have
> the same problem during lamtests configuration. How do I tell the
> configure script for lamtests to avoid checks for Fortran bindings?
You can't. The lamtests configure checks to see if the LAM
installation is supposed to have Fortran bindings (that is, it checks a
#define in the LAM header files). If LAM is supposed to have Fortran
bindings, it then tries to compile and link a simple Fortran MPI
application.
Please see my previous text about possible problems with your install
-- the information that you sent contradicted itself, implying that you
have either a broken LAM install, or multiple LAM installs on that
machine (perhaps one was installed via RPM on your base system?).
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
