I installed HDF5 1.8.6 on Debian 5.0.8 with pvfs2 2.8.3 and mpich2 1.3.2p1.
I guess that I fully verified pvfs and mpich2 setup. My hdf5 setup passed
'make check-s', but 'make check-p' fails at 't_posix_compliant test'. The
followings show my configure and make options.
$ CC=/usr/local/bin/mpicc RUNPARALLEL="/usr/bin/srun -n \$\${NPROCS:=6}"
./configure --enable-fortran --enable-parallel --with-szlib=/usr/local
--with-mpe=/usr/local --prefix=/usr/local 2>&1 | tee c.txt
$ make 2>&1 | tee m.txt
$ make check-s 2>&1 | tee mcs.txt
$ make check-p 2>&1 | tee mcp.txt
I've attached the output of configure and make processes.
Thanks for the logs. I haven't run the HDF5 test suite against PVFS
in a few years but it did work at one time!
PVFS is (deliberately) not a posix-compliant file system, so the test
failures are in some way to be expected. However, it looks like the
PVFS servers, faced with the HDF5-generated workload, died a fiery
death.
Can you bring this up on the PVFS mailing list? I think we're dealing
with a PVFS issue here, and not an HDF5 defect.
==rob
···
On Fri, Feb 18, 2011 at 06:09:27PM +0900, 전경원 wrote:
Hi.
I installed HDF5 1.8.6 on Debian 5.0.8 with pvfs2 2.8.3 and mpich2 1.3.2p1.
I guess that I fully verified pvfs and mpich2 setup. My hdf5 setup passed
'make check-s', but 'make check-p' fails at 't_posix_compliant test'. The
followings show my configure and make options.
--
Rob Latham
Mathematics and Computer Science Division
Argonne National Lab, IL USA
