Hello,
I am in the process of building parallel HDF5 1.8.8 in an HPC environment
running Red Hat Enterprise Linux 5, kernel 2.6.18-238 with Intel Xeon X5660
cpus. It is being built with mpicc set to icc version 12.1 and mpif90 set
to ifort 12.1. Here are the configuration parameters:
export CC=mpicc
export FC=mpif90
./configure \
--prefix=$hdf51.8.8/ \
--enable-parallel \
--enable-fortran
"make" completes successfully, but when I run "make check" I get the
following error and segfault during execution of "testh5diff.sh -p":
Testing ph5diff --count=200 h5diff_basic1.h5 h5diff_basic2.h5
g1/dset3*FAILED*
====Expected result (expect_sorted) differs from actual result
(actual_sorted)
*** expect_sorted 2012-01-20 17:15:15.038781109 -0700
--- actual_sorted 2012-01-20 17:15:15.037781100 -0700
ยทยทยท
***************
*** 1,2 ****
! 6 differences found
! dataset: </g1/dset3> and </g1/dset4>
--- 1,3 ----
!
--------------------------------------------------------------------------
!
--------------------------------------------------------------------------
! mpiexec noticed that process rank 2 with PID 16751 on node node0001
exited on signal 11 (Segmentation fault).
====The actual output (./testfiles/h5diff_627.out-sav)
====The actual stderr (./testfiles/h5diff_627.err-sav)
--------------------------------------------------------------------------
mpiexec noticed that process rank 2 with PID 16751 on node node0001 exited
on signal 11 (Segmentation fault).
--------------------------------------------------------------------------
====End of actual stderr (./testfiles/h5diff_627.err-sav)
I ran an strace on ph5diff --count=200 h5diff_basic1.h5 h5diff_basic2.h5
g1/dset3 g1/dset4, but I cannot reproduce the segfault. I am unsure
whether there is a serious error in the build process, or whether there is
a bug in testh5diff.sh. Any help would be greatly appreciated.
Regards,
Dan Milroy