Open MPI logo

Open MPI User's Mailing List Archives

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

Subject: [OMPI users] openMPI + Ubuntu 7.10 puzzling
From: Vincent Rotival (vincent.rotival_at_[hidden])
Date: 2008-04-21 18:12:00


Hello everyone

Sorry to bother you all about that but I am quite lost with a puzzling
problem concerning openMPI + Ubuntu 7.10. I could not find similar
threads on the archive

I have been designing complex openMPI program with Intel Fortran
compiler for quite some time, and everything worked perfectly until this
morning, when everything seemed to fall apart.

Basically, if I try now even the (very very) simple program below in
Fortran 90

************************************************** **
program test
use mpi

integer :: mpi_err

call mpi_init(mpi_err)

print *, "HELLO", mpi_err

call mpi_finalize(mpi_err)

end program test
************************************************** *

This code will compile (mpif90 -o test test_mpi.f90), but gives error
messages at the level of MPI_INIT, whether I call it with
> ./test
or
> mpirun -np N test

vrotival_at_vrotival-laptop:~/Work/workbench$ ./test
[vrotival-laptop:06375] *** Process received signal ***
[vrotival-laptop:06375] Signal: Segmentation fault (11)
[vrotival-laptop:06375] Signal code: Address not mapped (1)
[vrotival-laptop:06375] Failing at address: 0x8
[vrotival-laptop:06375] [ 0] [0xffffe440]
[vrotival-laptop:06375] [ 1] /usr/lib/libmpi.so.0(ompi_proc_init+0x13b)
[0xb7f0863b]
[vrotival-laptop:06375] [ 2] /usr/lib/libmpi.so.0(ompi_mpi_init+0x10c)
[0xb7f09fbc]
[vrotival-laptop:06375] [ 3] /usr/lib/libmpi.so.0(MPI_Init+0xab)
[0xb7f2b22b]
[vrotival-laptop:06375] [ 4]
/usr/local/lib/libmpi_f77.so.0(PMPI_INIT+0x32) [0xb7fd78d2]
[vrotival-laptop:06375] [ 5] ./test(MAIN__+0x24) [0x804dfe4]
[vrotival-laptop:06375] [ 6] ./test(main+0x3d) [0x804dfb1]
[vrotival-laptop:06375] [ 7]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe 0) [0xb7a39050]
[vrotival-laptop:06375] [ 8] ./test [0x804df11]
[vrotival-laptop:06375] *** End of error message ***

I am using openMPI version 1.2.6 (but same bug occured with 1.2.5),
which I compile directly from the source from www.openmpi.org
<http://www.openmpi.org>, with F90=ifort with Intel Fortran 10.1
I have not updated ifort since about 6 months, the only change between
last time I used MPI are small Ubuntu updates..... I can give you much
more complex codes which worked perfectly one week ago

As a small comment, the ouput from ompi_info is also weird

vrotival_at_vrotival-laptop:~/Work/workbench$ ompi_info --all
ompi_info: Symbol `mca_allocator_base_components' has different size in
shared object, consider re-linking
ompi_info: symbol lookup error: ompi_info: undefined symbol:
ompi_mtl_base_components_opened

but I hadn't looked at it before the bug

Has anyone else encountered this bug ? can it be relaed to recent
updates in Ubuntu (although I do not want to incriminate it at first)

I am upgrading my Ubuntu with the following repositories

deb http://archive.canonical.com/ubuntu gutsy partner
deb-src http://archive.canonical.com/ubuntu gutsy partner
deb http://archive.ubuntu.com/ubuntu/ gutsy main universe restricted
multiverse
deb http://security.ubuntu.com/ubuntu/ gutsy-security universe main
multiverse restricted
deb http://archive.ubuntu.com/ubuntu/ gutsy-updates universe main
multiverse restricted

If anyone has any advice it would be greatly appreciated (and would save
my PhD as a byproduct )

Cheers !