Problem running mpb-mpi

Hi Steven, I thought I would send these messages back to the mailing.

To my post, you answered:

>It might be hanging trying to output the epsilon.h5 file.  To check
>this, 
>you can disable epsilon output by adding the following line to your
>file:

>(define output-epsilon (lambda () (print "skipping output-epsilon\n"))

>The difficulty with HDF5 output may be that, last I checked, mpb-mpi
>works 
>best with the serial version of HDF5, and problems have been reported
>if 
>you compile HDF5 with MPI enabled.  (Conversely, meep-mpi works best
>with 
>the parallel version of HDF5.)

>Cordially,
>Steven G. Johnson

That got me passed the initial hang, but it stops very soon after with
this error in the output file:

Solving for bands 1 to 7...
    near maximum in trace
p0_3320:  p4_error: : 1
Broken pipe

and this in the mpi-output file:
+ logger -p daemon.err SGE job 14731 for rw231 on queue t4 starting
linmin: ERROR: INITIAL G .GE. ZERO
CHECK failure on line 91 of linmin.c: linmin failure
[0] MPI Abort by user Aborting program !
[0] Aborting program!
Killed by signal 2.
Killed by signal 2.
Killed by signal 2

Compiling mpb without hdf5 (by crippling the configure script since it
won't configure without, see other post "MPB Installation problem -
ctl.h not found"), I get this error in the mpi-output file:

CHECK failure on line 392 of matrixio.c: no matrixio implementation is
linked
[0] MPI Abort by user Aborting program !
[0] Aborting program!
Killed by signal 2.
CHECK failure on line 392 of matrixio.c: no matrixio implementation is
linked
CHECK failure on line 392 of matrixio.c: no matrixio implementation is
linked
Killed by signal 2.
Killed by signal 2.

Thanks for your help.
Nic