On Wed, 29 Apr 2015 13:06:35 +0000
"Fux  Samuel (ID SIS)" <[email protected]> wrote:

> Dear Jan,
> 
> thank you for your reply. I followed your suggestion and tried to use
> FFC instead of UFC and disabled petsc4py and slepc4py support for
> Dolfin.
> 
> Now I get some different errors, but this most likely means that the
> original error was caused either by petsc4py or slepc4py.
> 
> Is it possible to get a list of dependencies (version numbers, for
> instance for petsc4py or slepc4py) that have been tested with fenics
> 1.5.0 and are known to result in a stable build ?

I have a system where the following combination seems to work:

Python 2.7.6
NumPy 1.8.2
FEniCS 1.5.0
SWIG 3.0.5
PETSc 3.5.2
petsc4py 3.5
SLEPc 3.5.2
slepc4py 3.5.1

You can also try to look how Debian/Ubuntu package has been built. It's
somewhere on Launchpad.

Jan

> 
> Best regards
> 
> Sam
> 
> 
> --
> ETH Zurich, Dr. Samuel Fux
> High Performance Computing, Scientific IT Services
> WEC D 13, Weinbergstrasse 11, 8092 Zurich, Switzerland
> Phone +41 44 632 57 63, http://www.id.ethz.ch/
> [email protected]
> 
> ________________________________________
> Von: Jan Blechta [[email protected]]
> Gesendet: Donnerstag, 23. April 2015 15:22
> An: Fux  Samuel (ID SIS)
> Cc: [email protected]
> Betreff: Re: [FEniCS-support] FEniCS 1.5.0 installation fails with
> strange error when building Dolfin 1.5.0
> 
> On Thu, 23 Apr 2015 12:44:56 +0000
> "Fux  Samuel (ID SIS)" <[email protected]> wrote:
> 
> > Dear Jan,
> >
> > thank you very much for your reply. I tried to follow your first
> > suggestion to remove UFC 2.3.0 as you told me that it is no longer
> > used because it is now part of ffc. But then, when I try to compile
> > Dolfin, it is complaining about not finding UFC:
> >
> >  CMake Error at CMakeLists.txt:322 (message):
> >    Could not find a configuration file for package UFC that is
> > compatible with requested version 1.5.0.
> >
> >    Set UFC_DIR to the directory containing a CMake configuration
> > file for UFC.
> 
> Install FFC 1.5.0 to <prefix> and then configure DOLFIN 1.5.0 by
> 
>   cmake -DUFC_DIR:PATH=<prefix>/share/ufc <other args> <src dir>
> 
> Jan
> 
> >
> > Best regards
> >
> > Sam
> >
> > --
> > ETH Zurich, Dr. Samuel Fux
> > High Performance Computing, Scientific IT Services
> > WEC D 13, Weinbergstrasse 11, 8092 Zurich, Switzerland
> > Phone +41 44 632 57 63, http://www.id.ethz.ch/
> > [email protected]
> >
> > ________________________________________
> > Von: Jan Blechta [[email protected]]
> > Gesendet: Montag, 20. April 2015 12:15
> > An: Fux  Samuel (ID SIS)
> > Cc: [email protected]
> > Betreff: Re: [FEniCS-support] FEniCS 1.5.0 installation fails with
> > strange error when building Dolfin 1.5.0
> >
> > On Thu, 16 Apr 2015 13:56:46 +0000
> > "Fux  Samuel (ID SIS)" <[email protected]> wrote:
> >
> > > Dear FEniCS community,
> > >
> > > since a few days I am trying to install FEniCS 1.5.0 from source
> > > on a HPC cluster (CentOS 6.5, Intel Xeon CPU's).
> > >
> > > The dependencies were all installed from source:
> > >
> > > - GCC 4.8.2
> > > - CMake 2.8.12
> > > - Python 2.7.6
> > > - Swig 3.0.5
> > > - QT 4.8.4
> > > - VTK 6.1.0
> > > - szip 2.1
> > > - hdf5 1.8.12
> > > - BOOST 1.55.0
> > > - Eigen 3.2.1
> > > - ScientificPython 2.9.4
> > > - numpy 1.9.2
> > > - fiat 1.5.0
> > > - instant 1.5.0
> > > - ufl 1.5.0
> > > - ffc 1.5.0
> > > - ufc 2.3.0
> >
> > Don't install this! UFC 2.3.0 is much more older, from the times
> > when version numbering scheme was different. Now, UFC is part of
> > FFC.
> >
> > > - ply 3.4
> > > - petsc 3.5.3 + petsc4py 3.5.1
> > > - slepc 3.5.3 + slepc4py 3.5.1
> > >
> > > On the FEniCS web page, it was indicated which dependencies FEniCS
> > > has, but not which versions of the dependencies work with FEniCS
> > > 1.5.0.
> > >
> > > When trying to finally install Dolfin, I disabled the parallel
> > > features and made sure that CMake finds all the dependencies. When
> > > starting the compilation everything looked good until I reached 96
> > > %, then I got this strange error:
> > >
> > > [ 96%] Building CXX object
> > > dolfin/swig/modules/mesh/CMakeFiles/_mesh.dir/modulePYTHON_wrap.cxx.o
> > > cd 
> > > /cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build/dolfin/swig/modules/mesh
> > > && /cluster/apps/gcc/4.8.2/bin/c++   -DBOOST_UBLAS_NDEBUG
> > > -DDOLFIN_GIT_COMMIT_HASH=\"\" -DDOLFIN_LA_INDEX_SIZE=4
> > > -DDOLFIN_SIZE_T=8 -DDOLFIN_VERSION=\"1.5.0\" -DENABLE_PETSC_SNES
> > > -DENABLE_PETSC_TAO -DHAS_OPENMP -DHAS_PETSC -DHAS_PETSC4PY
> > > -DHAS_QT4 -DHAS_QVTK -DHAS_SLEPC -DHAS_SLEPC4PY -DHAS_VTK
> > > -DHAS_ZLIB -DNPY_NO_DEPRECATED_API=NPY_1_9_API_VERSION
> > > -DNUMPY_VERSION_MAJOR=1 -DNUMPY_VERSION_MICRO=2
> > > -DNUMPY_VERSION_MINOR=9 -DQT_CORE_LIB -DQT_GUI_LIB -DQT_NO_DEBUG
> > > -D_mesh_EXPORTS
> > > -DvtkRenderingCore_AUTOINIT="4(vtkInteractionStyle,vtkRenderingFreeType,vtkRenderingFreeTypeOpenGL,vtkRenderingOpenGL)"
> > > -DvtkRenderingVolume_AUTOINIT="1(vtkRenderingVolumeOpenGL)"
> > > -std=c++11  -Wno-deprecated  -fopenmp -O2 -g -DNDEBUG -fPIC
> > > -I/cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build
> > > -isystem /cluster/apps/qt/4.8.4/x86_64/lib64/qt/include
> > > -isystem /cluster/apps/qt/4.8.4/x86_64/lib64/qt/include/QtGui
> > > -isystem /cluster/apps/qt/4.8.4/x86_64/lib64/qt/include/QtCore
> > > -isystem /cluster/apps/vtk/6.1.0/x86_64/include/vtk-6.1
> > > -isystem /usr/include/freetype2
> > > -isystem /cluster/apps/python/2.7.6/x86_64/include/python2.7
> > > -isystem /cluster/apps/netcdf/4.3.2/x86_64/gcc_4.8.2/serial/include
> > > -isystem /cluster/apps/hdf5/1.8.13/x86_64/gcc_4.8.2/serial/include
> > > -isystem /usr/include/libxml2
> > > -I/cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0
> > > -I/cluster/home/muelfabi/python/include
> > > -I/cluster/home/muelfabi/fenics/SLEPC/include
> > > -isystem /cluster/home/muelfabi/fenics/EIGEN3/include/eigen3
> > > -isystem /cluster/apps/boost/1.55.0/x86_64/serial/gcc_4.8.2/include
> > > -isystem /cluster/home/muelfabi/fenics/PETSC/include
> > > -isystem /cluster/home/muelfabi/fenics/PETSC/include/mpiuni
> > > -I/cluster/home/muelfabi/python/lib64/python2.7/site-packages/numpy/core/include
> > > -I/cluster/home/muelfabi/python/lib64/python2.7/site-packages/petsc4py/include
> > > -I/cluster/home/muelfabi/python/lib64/python2.7/site-packages/slepc4py/include
> > > -o CMakeFiles/_mesh.dir/modulePYTHON_wrap.cxx.o
> > > -c 
> > > /cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build/dolfin/swig/modules/mesh/modulePYTHON_wrap.cxx
> > >  
> > > /cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build/dolfin/swig/modules/mesh/modulePYTHON_wrap.cxx:
> > > In function ât SWIG_AsVal_MPI_Comm(PyObject*,
> > > MPI_Comm*)â/cluster/home/muelfabfenics/SOURCES/dolfin-1.5.0/build/dolfin/swig/modules/mesh/modulePYTHON_wrap.cxx:4780:43:
> > > error: âSWIGTYPE_p_MPI_Commâas not declared in this scope int
> > > res = SWIG_ConvertPtr(input,&argp,SWIGTYPE_p_MPI_Comm, 0);
> > > ^ 
> > > /cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build/dolfin/swig/modules/mesh/modulePYTHON_wrap.cxx:2959:37:
> > > note: in definition of macro âIG_ConvertPtrâ
> > > SWIG_convert_ptr(o, pptr, type, flags) ^ make[2]: ***
> > > [dolfin/swig/modules/mesh/CMakeFiles/_mesh.dir/modulePYTHON_wrap.cxx.o]
> > > Error 1 make[2]: Leaving directory
> > > `/cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build'
> > > make[1]: *** [dolfin/swig/modules/mesh/CMakeFiles/_mesh.dir/all]
> > > Error 2 make[1]: Leaving directory
> > > `/cluster/home/muelfabi/fenics/SOURCES/dolfin-1.5.0/build' make:
> > > *** [all] Error 2
> > >
> > > There was a similar post in the FEniCS mailing list
> > > (http://fenicsproject.org/pipermail/fenics-support/2014-December/001026.html),
> > > but there the problem happened for Dolfin 1.4.0, and I could not
> > > really understand if the problem was solved or if the author of
> > > the thread gave up.
> >
> > You could start bisecting the problem by switching off petsc4py and
> > slepc4py support in DOLFIN. Configure with
> >
> > cmake -DDOLFIN_ENABLE_PETSC4PY:BOOL=OFF
> >  -DDOLFIN_ENABLE_SLEPC4PY:BOOL=OFF <other args> dir
> >
> > Jan
> >
> > >
> > > Any help is appreciated.
> > >
> > > Best regards
> > >
> > > Samuel Fux
> > >
> >
> 
> _______________________________________________
> fenics-support mailing list
> [email protected]
> http://fenicsproject.org/mailman/listinfo/fenics-support

_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support

Reply via email to