[CIG-SHORT] pylith1.8.0-0/openmpi build failing on OS X

[Matthew Knepley]

On Tue, Jan 29, 2013 at 2:00 PM, Brad Aagaard <baagaard at usgs.gov> wrote:

> Richard,
>
> We have not encountered this problem you are describing with regards to
> openmpi and darwin. I am still using 10.6 on my Mac; I am not sure if we
> have tested building from source under 10.8.
>
> We didn't setup the pylith-installer configure to have openmpi options
> because we didn't anticipate the user needing to customize the openmpi
> options. For clusters and other computers where custom MPI settings are
> used, we expect a sysadmin to install MPI in a system wide location.
>
> There are two easy workarounds to this problem:
> (1) Install mpi yourself in a custom location that is in the path and
> then don't install it using the pylith installer. You can customize the
> mpi configure arguments to make them work on your machine.
> (2) Manually adjust the openmpi configure options used by the pylith
> installer by editing the install_openmpi target in the Makefile.
>

(3) Have it install MPICH instead.

  Thanks,

    Matt


> Regards,
> Brad
>
>
> On 1/29/13 7:34 AM, Richard Styron wrote:
> > Hi all,
> >
> > I am trying to build Pylith from source on a mac (OS X 10.8) following
> > the install instructions for Darwin/CUDA:
> >
> >    mkdir -p $HOME/build/pylith
> >    $HOME/src/pylith/pylith-installer-1.8.0-0/configure \
> >      --enable-gcc=fortran \
> >      --enable-python \
> >      --enable-mpi=openmpi \
> >      --with-fetch=curl \
> >      --with-make-threads=2 \
> >      --prefix=$HOME/pylith \
> >      --with-cuda (<-- also tried without this.  CUDA is installed and
> > functional on my machine.)
> >
> >
> >
> >
> > However, OpenMPI fails when during the build:
> >
> > libtool: link: g++ -fopenmp -DVT_OMP -g -O2 -DMPICH_IGNORE_CXX_SEEK
> > -fopenmp -o vtfilter vtfilter-vt_filter.o vtfilter-vt_filthandler.o
> > vtfilter-vt_otfhandler.o vtfilter-vt_tracefilter.o ../../util/util.o
> > -Wl,-bind_at_load
> >
> -L/Users/itchy/src/pylith/openmpi-build/ompi/contrib/vt/vt/extlib/otf/otflib/.libs
> -L/Users/itchy/pylith/lib
> -L/Users/itchy/src/pylith/openmpi-build/ompi/contrib/vt/vt/extlib/otf/otflib
> /Users/itchy/src/pylith/openmpi-build/ompi/contrib/vt/vt/extlib/otf/otflib/.libs/libotf.a
> -lz -lutil
> > Undefined symbols for architecture x86_64:
> >    "___builtin_expect", referenced from:
> >        _main.omp_fn.0 in vtfilter-vt_tracefilter.o
> >        _main.omp_fn.1 in vtfilter-vt_tracefilter.o
> > ld: symbol(s) not found for architecture x86_64
> > collect2: ld returned 1 exit status
> > make[11]: *** [vtfilter] Error 1
> > make[10]: *** [all-recursive] Error 1
> > make[9]: *** [all-recursive] Error 1
> > make[8]: *** [all] Error 2
> > make[7]: *** [all-recursive] Error 1
> > make[6]: *** [all-recursive] Error 1
> > make[5]: *** [all-recursive] Error 1
> > make[4]: *** [openmpi] Error 2
> > make[3]: *** [installed_mpi] Error 2
> > make[2]: *** [installed_hdf5] Error 2
> > make[1]: *** [installed_netcdf] Error 2
> > make: *** [installed_pylith] Error 2
> >
> > The error appears to be related to building VampirTrace, which may not
> > be critical (?), and following this
> > http://stackoverflow.com/questions/7785464/open-mpi-compilation-errorI'd like
> > to disable VampirTrace during the build, but I don't know how to pass
> > the 'disable-vt' flag during the pylith configuration process.  I tried
> > building openmpi from the terminal using the configuration generated by
> > autoconf during the Pylith install ($ ../openmpi-1.4.5/configure
> > --prefix=/Users/itchy/pylith CPPFLAGS=-I/Users/itchy/pylith/include
> >   LDFLAGS=-L/Users/itchy/pylith/lib  CFLAGS=-g -O2 CXXFLAGS=-g -O2
> > -DMPICH_IGNORE_CXX_SEEK FCFLAGS= CC=gcc CXX=g++ FC=gfortran) plus
> > '--disable-vt' but that doesn't work because many of the flags there
> > aren't recognized as options from the command line.  I also tried adding
> > '--mpi_options="--disable-vt" ' and '--openmpi_options="--disable-vt" '
> > during pylith configuration (similar to what can be done with petsc) but
> > those don't fly.
> >
> > The config.log files for both pylith and openmpi are attached.
> >
> > Has anyone dealt with this or a similar issue?
> >
> > Thanks,
> > Richard
> >
> >
> >
> > _______________________________________________
> > 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
>



-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://geodynamics.org/pipermail/cig-short/attachments/20130129/3317276b/attachment.htm