Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] Linux opteron infiniband sunstudio configure, problem
From: Kevin McManus (k.mcmanus_at_[hidden])
Date: 2009-03-31 17:25:35


On Tue, Mar 31, 2009 at 04:59:00PM -0400, Jeff Squyres wrote:
> My goal in having you try that statement in a standalone shell script
> wasn't the success or failure of the uname command -- but rather to
> figure out if something in that statement itself was causing the
> syntax error.
>
> Apparently it is not. There's an errant character elsewhere that is
> causing the problem. FWIW, you should be able to run ./config/
> config.guess by itself and get a single line of output with no errors
> -- so you don't even have to test this in the larger context of Open
> MPI's configure process. As you showed in an earlier mail, even that
> doesn't work.
>
> We get config.guess from git.savannah.gnu.org when we make OMPI
> tarballs. FWIW, here's the md5 and sha1sums from the config.guess
> that I got from a 1.3.1 tarball:
>
> 13:57] svbu-mpi:/home/jsquyres/openmpi-1.3.1 % md5sum config/
> config.guess
> 6384bbc496ef6e69328a9a428c623cc5 config/config.guess
> [13:58] svbu-mpi:/home/jsquyres/openmpi-1.3.1 % sha1sum config/
> config.guess
> 199adb16fc94e6cf62912be5fbb69a77214b33e5 config/config.guess
>
> I would be pretty surprised if yours don't match. If they don't
> there's some larger issue going on (e.g., filesystem corruption?) that
> would also be pretty surprising.

BINGO!!

mine don't match but I have openmpi-1.3 not openmpi-1.3.1
just downloaded a fresh copy
checked md5 and sha1 match
run my build.sh without x86_64 - hurrah

#!/bin/sh

../configure \
        CC=cc CXX=CC F77=f77 FC=f90 \
        CFLAGS=-m64 CXXFLAGS=-m64 FFLAGS=-m64 FCFLAGS=-m64 \
        --enable-static \
        --enable-orterun-prefix-by-default \
        --enable-mpi-f90 \
        --disable-mpi-threads \
        --with-psm \
        --with-psm-libdir=/usr/lib64

This runs until I get to...

--- MCA component mtl:psm (m4 configuration macro)
checking for MCA component mtl:psm compile mode... static
checking --with-psm value... simple ok (unspecified)
checking --with-psm-libdir value... sanity check ok (/usr/lib64)
checking psm.h usability... no
checking psm.h presence... no
checking for psm.h... no
configure: error: PSM support requested but not found. Aborting

Which is an entirely different problem

The lesson here seems to be that when things just don't make sense
there has to be a corruption someplace.
This is not the first time this week that I have seen undetected
bit level errors.
Now kicking myself for not checking the md5s earlier.

Time for me to pursue the psm installs
Start by trawling the archives and FAQs :-)

I think we should consider this thread closed
Many thanks to all who contributed.

k.mcmanus_at_[hidden] - http://staffweb.cms.gre.ac.uk/~k.mcmanus
--------------------------------------------------------------
Dr Kevin McManus Queen Mary 413
School of Computing & Mathematical Sciences,
University of Greenwich,
Old Royal Naval College,
Park Row, Greenwich, Tel +44 (0)208 331 8719
London, SE10 9LS Fax +44 (0)208 331 8665
--------------------------------------------------------------

University of Greenwich, a charity and company limited by
guarantee, registered in England (reg no. 986729)
Registered Office: Old Royal Naval College