[CIG-SHORT] problem with installation of pylith 64bit from source code

Wed Oct 8 06:28:13 PDT 2008

Makoto-

Checking to make sure the correct NetCDF library was being used was going to 
be my next suggestion. Glad to hear that you have things working!

Brad

On Wednesday 08 October 2008 3:28:02 am Makoto Hoshino wrote:
> Brad,
>
> I'm happy to tell you that I have solved by myself the problem I
> consulted you in the previous email.
>
> To be specific, when compiling Pylith, I should have set
> CPPFLAGS=/usr/local2/include
> LDFLAGS=/usr/local2/lib
> but in fact, I did set only
> LDFLAGS=/usr/local2/lib
>
> The problem arose because different versions of Netcdf packages have
> been installed in the directories /usr/local and /usr/local2. The former
> is exported from a 32-bit machine, while the latter from a 64-bit
> machine. I didn't set CPPFLAGS in the previous compilation of Pylith, as
> a result, the header file of the netcdf package in the directory
> /usr/local/include was used. This leaded to the runtime error that
> reports the inconsistency of data size at the time to read a datafile
> created by Cubit.
>
> Thank you for the kind supports.
>
> Sincerely,
> Makoto
>
> Brad Aagaard  wrote:
> > Makoto-
> >
> > The assertion that was triggered is a consistency check to make sure that
> > the size of the coordinates array is the number of vertices times the
> > spatial dimension. This error seems to indicate that the
> > box_hex8_1000m.exo file is corrupt. Did you run CUBIT and generate the
> > file or are you using the file from the source code? I would extract a
> > clean version of this file from the source tarball (or repository if you
> > used the SVN repository) and then try again.
> >
> > Brad
> >
> > On Monday 06 October 2008 5:28:19 pm Leif Strand wrote:
> >> Sorry, I just noticed that your last e-mail didn't go to Matt or the
> >> cig-short list.
> >>
> >> I don't know what that assertion means... I've CC'd Brad, who should be
> >> able to help.
> >>
> >> --Leif
> >>
> >> Makoto Hoshino wrote:
> >>> Matt and Leif,
> >>>
> >>> Thanks for your prompt reply.
> >>>
> >>> As you pointed out,  I actually built Nemesis by running 'setup.py'. 
> >>> So I deleted the directory
> >>> /usr/local2/lib/python2.5/site-packages/nemesis-1.0-py2.5.egg
> >>> and installed Nemesis and Pylith again. After all, Pylith started to
> >>> run correctly, but only when a meshfile made by Lagrit was used. For
> >>> example, a test run
> >>> cd src/pylith/examples/3d/tet4; pylith dislocation.cfg
> >>> seemed to go well, but another run “cd src/pylith/examples/3d/hex8;
> >>> pylith dislocation.cfg”
> >>> ended with error messages:
> >>>
> >>> [dry at n102 hex8]$ pylith dislocation.cfg
> >>>
> >>> /usr/local2/lib/python2.5/site-packages/pylith/utils/PetscManager.py:47
> >>>:i nitialize
> >>>
> >>> -- petsc(info)
> >>> -- Initializing PETSc.
> >>>
> >>>  >> /usr/local2/lib/python2.5/site-packages/pylith/meshio/MeshIO.py:45:
> >>>  >>re ad
> >>>
> >>> -- meshiocubit(info)
> >>> -- Reading finite-element mesh
> >>>
> >>>  >> meshio/MeshIOCubit.cc:138:<unknown>
> >>>
> >>> -- meshiocubit(info)
> >>> -- Reading 245 vertices.
> >>> mpinemesis: meshio/MeshIOCubit.cc:154: void
> >>> pylith::meshio::MeshIOCubit::_readVertices(NcFile&,
> >>> pylith::double_array*, int*, int*) const: Assertion
> >>> `(*numVertices)*(*numDims) == size' failed.
> >>> rank 0 in job 5  n102_53796   caused collective abort of all ranks
> >>>  exit status of rank 0: killed by signal 6
> >>> --pyre-start: mpirun: exit 6
> >>> /usr/local2/bin/pylith: /usr/local2/bin/nemesis: exit 1
> >>>
> >>> Should I install any library related to Cubit?
> >>>
> >>>
> >>> On Thu, 02 Oct 2008 11:57:11 -0700
> >>> Leif Strand <leif at geodynamics.org> wrote:
> >>>
> >>> As Matt said, something probably went wrong when you installed
> >>>
> >>> 'nemesis'. I notice that in your 'sys.path', you have:
> >>>> /usr/local2/lib/python2.5/site-packages/nemesis-1.0-py2.5.egg
> >>>>
> >>>> This directory should not exist. If it does exist, it suggests you ran
> >>>> the Nemesis 'setup.py' script. But the nemesis 'setup.py' script
> >>>> should not be run directly. Instead, install 'nemesis' as follows:
> >>>>
> >>>> ./configure --prefix=/usr/local2
> >>>> make
> >>>> make install
> >>>>
> >>>> Make sure the correct MPI tools are on your path when you do this.
> >>>> Also make sure the correct 'python' is on your path.
> >>>>
> >>>> Afterwards, you should have a 'nemesis' executable
> >>>> (/usr/local2/bin/nemesis) which behaves just like Python, except that
> >>>> it knows about '_mpi':
> >>>>
> >>>> [dry at n100 ~]$ nemesis
> >>>> Python 2.5.2 (r252:60911, Sep 25 2008, 15:56:45)
> >>>> [GCC 3.4.6 20060404 (Red Hat 3.4.6-3)] on linux2
> >>>> Type "help", "copyright", "credits" or "license" for more information.
> >>>>
> >>>>>>> import _mpi
> >>>>
> >>>> I must confess that I'm baffled by the error message you sent:
> >>>>
> >>>> ImportError: No module named _mpi
> >>>>
> >>>> If you installed 'nemesis' incorrectly, then you wouldn't have a
> >>>> 'nemesis' program, and you wouldn't be able to run 'pylith' at all.
> >>>> Note that the very first line of the 'pylith' script is:
> >>>>
> >>>> #!/usr/bin/env nemesis
> >>>>
> >>>> On the other hand, if 'nemesis' does exist, then it must have been
> >>>> successfully installed... and it should be able to "import _mpi".
> >>>>
> >>>> --Leif
> >>>>
> >>>> Matthew Knepley wrote:
> >>>>> This is a problem with the nemesis installation, since this is what
> >>>>> loads the
> >>>>> _mpi module for Python. It is common (if you have multiple MPI
> >>>>> installations)
> >>>>> to select the wrong one at nemesis installation time. You muse have
> >>>>> the correct 'mpirun' or 'mpiexec' in your path when installing
> >>>>> nemesis.
> >>>>>
> >>>>> Thanks,
> >>>>>
> >>>>> Matt
> >>>>>
> >>>>> 2008/10/2 Makoto Hoshino <m-hoshino at aist.go.jp>:
> >>>>>> Dear pylith developers,
> >>>>>>
> >>>>>> As written in the previous email from Ikuo, I have installed Pylith
> >>>>>> from
> >>>>>> the source code into a 64bit computer machine with specification:
> >>>>>>
> >>>>>> OS: CentOS release 4.4 (Final)
> >>>>>> linux: 2.6.9-42.0.10.ELsmp x86_64
> >>>>>> CPU(64bit): Intel(R) Xeon(R) X5355 @ 2.66GHz
> >>>>>>
> >>>>>> As a note, I installed Python2.5 to /usr/local2/ and export
> >>>>>> PYTHONPATH=/usr/local2/lib/python2.5/site-packages,
> >>>>>> and installed pylith so that it does NOT depend on MPICH but MPICH2.
> >>>>>>
> >>>>>> No noticeable warning was found in the installation process of
> >>>>>> Pylith, but I was not able to run it correctly. To be specific, I
> >>>>>> tried to run Pylith in the directory src/pylith/examples/3d/hex8 by
> >>>>>> typing " pylith dislocation.cfg". However, I only obtained the
> >>>>>> following messages.
> >>>>>>
> >>>>>> Traceback (most recent call last):
> >>>>>> File "/usr/local2/bin/pylith", line 31, in <module>
> >>>>>> from pylith.PyLithApp import PyLithApp
> >>>>>> File "/usr/local2/lib/python2.5/site-packages/pylith/PyLithApp.py",
> >>>>>> line
> >>>>>> 17, in <module>
> >>>>>> from mpi import Application
> >>>>>> File
> >>>>>> "/usr/local2/lib/python2.5/site-packages/pythia-0.8.1.7-py2.5.egg/mp
> >>>>>>i/ __init__.py",
> >>>>>>
> >>>>>> line 14, in <module>
> >>>>>> from _mpi import *
> >>>>>> ImportError: No module named _mpi
> >>>>>>
> >>>>>> Could you tell me what happened? I think the following information
> >>>>>> obtained from the python shell is useful to make a diagnosis.
> >>>>>>
> >>>>>> [dry at n100 ~]$ python
> >>>>>> Python 2.5.2 (r252:60911, Sep 25 2008, 15:56:45)
> >>>>>> [GCC 3.4.6 20060404 (Red Hat 3.4.6-3)] on linux2
> >>>>>> Type "help", "copyright", "credits" or "license" for more
> >>>>>> information.
> >>>>>>
> >>>>>>>>>>>> import sys
> >>>>>>>>>>>> sys.path
> >>>>>>
> >>>>>> ['', '/usr/local2/lib/python2.5/site-packages/merlin-1.4.egg',
> >>>>>> '/usr/local2/lib/python2.5/site-packages/Cheetah-2.0.1-py2.5-linux-x
> >>>>>>86 _64.egg',
> >>>>>>
> >>>>>> '/usr/local2/lib/python2.5/site-packages/pythia-0.8.1.7-py2.5.egg',
> >>>>>> '/usr/local2/lib/python2.5/site-packages/nemesis-1.0-py2.5.egg',
> >>>>>> '/usr/local2/lib/python2.5/site-packages',
> >>>>>> '/usr/local2/lib/python25.zip', '/usr/local2/lib/python2.5',
> >>>>>> '/usr/local2/lib/python2.5/plat-linux2',
> >>>>>> '/usr/local2/lib/python2.5/lib-tk',
> >>>>>> '/usr/local2/lib/python2.5/lib-dynload']
> >>>>>>
> >>>>>>>>>>>> import mpi
> >>>>>>
> >>>>>> Traceback (most recent call last):
> >>>>>> File "<stdin>", line 1, in <module>
> >>>>>> File
> >>>>>> "/usr/local2/lib/python2.5/site-packages/pythia-0.8.1.7-py2.5.egg/mp
> >>>>>>i/ __init__.py",
> >>>>>>
> >>>>>> line 14, in <module>
> >>>>>> from _mpi import *
> >>>>>> ImportError: No module named _mpi
> >>>>>>
> >>>>>> ------------------------------------------------------
> >>>>>> Makoto HOSHINO
> >>>>>> Seismogenic Process Research Group
> >>>>>> Institute of Geology and Geoinformation
> >>>>>> National Institute of Advanced Industrial Science and Technology
> >>>>>> tel; 029-861-2460, fax; 029-861-3682
> >>>>>> email; m-hoshino at aist.go.jp
> >>>>>> -------------------------------------------------------
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> _______________________________________________
> >>>>>> CIG-SHORT mailing list
> >>>>>> CIG-SHORT at geodynamics.org
> >>>>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short