dorsal team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On 01/09/10 12:30, Anders Logg wrote:
> On Wed, Sep 01, 2010 at 12:22:27PM +0100, Garth N. Wells wrote:
>>
>>
>> On 01/09/10 12:17, Anders Logg wrote:
>>> On Wed, Sep 01, 2010 at 01:13:30PM +0200, Anders Logg wrote:
>>>> On Wed, Sep 01, 2010 at 12:58:36PM +0200, Johannes Ring wrote:
>>>>> On Wed, Sep 1, 2010 at 12:50 PM, Anders Logg <logg@xxxxxxxxx> wrote:
>>>>>> On Wed, Sep 01, 2010 at 12:43:12PM +0200, Anders Logg wrote:
>>>>>>> On Wed, Sep 01, 2010 at 12:33:40PM +0200, Johannes Ring wrote:
>>>>>>>> On Wed, Sep 1, 2010 at 12:12 PM, Anders Logg <logg@xxxxxxxxx> wrote:
>>>>>>>>> On Wed, Sep 01, 2010 at 12:09:20PM +0200, Johannes Ring wrote:
>>>>>>>>>> On Wed, Sep 1, 2010 at 12:03 PM, Anders Logg <logg@xxxxxxxxx> wrote:
>>>>>>>>>>> On Wed, Sep 01, 2010 at 11:36:11AM +0200, Johannes Ring wrote:
>>>>>>>>>>>> On Wed, Sep 1, 2010 at 11:13 AM, Harish Narayanan
>>>>>>>>>>>> <harish.mlists@xxxxxxxxx> wrote:
>>>>>>>>>>>>> On 9/1/10 2:35 PM, Harish Narayanan wrote:
>>>>>>>>>>>>>> On 9/1/10 2:09 PM, Johannes Ring wrote:
>>>>>>>>>>>>>>> On Wed, Sep 1, 2010 at 9:17 AM, Johannes Ring <johannr@xxxxxxxxx> wrote:
>>>>>>>>>>>>>>>> On Wed, Sep 1, 2010 at 8:39 AM, Harish Narayanan
>>>>>>>>>>>>>>>> <harish.mlists@xxxxxxxxx> wrote:
>>>>>>>>>>>>>>>>> On 9/1/10 12:02 PM, Harish Narayanan wrote:
>>>>>>>>>>>>>>>>>> On 9/1/10 2:44 AM, Marie Rognes wrote:
>>>>>>>>>>>>>>>>>>> On 31. aug. 2010 22:56, Anders Logg wrote:
>>>>>>>>>>>>>>>>>>>> On Tue, Aug 31, 2010 at 04:28:32PM +0200, Marie Rognes wrote:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> On 31. aug. 2010 16:22, Johannes Ring wrote:
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> On Tue, Aug 31, 2010 at 4:13 PM, Marie Rognes <meg@xxxxxxxxx> wrote:
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> I having some issues with PETSc (surprise) as of today.
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Dorsal tells me that (full log attached):
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> [\snip]
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> -- The following optional packages were found:
>>>>>>>>>>>>>>>>>>>>>>> -- (OK) MPI
>>>>>>>>>>>>>>>>>>>>>>> -- (OK) PETSC
>>>>>>>>>>>>>>>>>>>>>>> -- (OK) SLEPC
>>>>>>>>>>>>>>>>>>>>>>> -- (OK) UMFPACK
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> [\snap]
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> But:
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>> from dolfin import *
>>>>>>>>>>>>>>>>>>>>>>>>>> has_la_backend("PETSc")
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> False
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>> A = PETScMatrix()
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Traceback (most recent call last):
>>>>>>>>>>>>>>>>>>>>>>>   File "<stdin>", line 1, in <module>
>>>>>>>>>>>>>>>>>>>>>>> NameError: name 'PETScMatrix' is not defined
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Help?
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> There is no -DHAS_PETSC in your log. Try to clean up/remove the
>>>>>>>>>>>>>>>>>>>>>> dorsal_build_dir inside the DOLFIN source directory.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> Removed. Check.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> Then rerun Dorsal
>>>>>>>>>>>>>>>>>>>>>> and see if the PETSc test passes.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> -- The following optional packages were found:
>>>>>>>>>>>>>>>>>>>>> -- (OK) MPI
>>>>>>>>>>>>>>>>>>>>> -- (OK) PETSC
>>>>>>>>>>>>>>>>>>>>> -- (OK) SLEPC
>>>>>>>>>>>>>>>>>>>>> -- (OK) UMFPACK
>>>>>>>>>>>>>>>>>>>>> -- (OK) PARMETIS
>>>>>>>>>>>>>>>>>>>>> -- (OK) GMP
>>>>>>>>>>>>>>>>>>>>> -- (OK) CGAL
>>>>>>>>>>>>>>>>>>>>> -- (OK) ZLIB
>>>>>>>>>>>>>>>>>>>>> -- (OK) PYTHON
>>>>>>>>>>>>>>>>>>>>> -- The following optional packages could not be found:
>>>>>>>>>>>>>>>>>>>>> -- (**) TRILINOS
>>>>>>>>>>>>>>>>>>>>> -- (**) MTL4
>>>>>>>>>>>>>>>>>>>>> -- (**) CHOLMOD
>>>>>>>>>>>>>>>>>>>>> -- (**) SCOTCH
>>>>>>>>>>>>>>>>>>>>> -- Configuring done
>>>>>>>>>>>>>>>>>>>>> -- Generating done
>>>>>>>>>>>>>>>>>>>>> -- Build files have been written to: /home/meg/local/dorsal_downloads/dolfin
>>>>>>>>>>>>>>>>>>>>> make: *** No rule to make target `install'.  Stop.
>>>>>>>>>>>>>>>>>>>>> Failure with exit status: 2
>>>>>>>>>>>>>>>>>>>>> Exit message: There was a problem building dolfin.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> I had the same problem (I think). It turned out to be a problem with my
>>>>>>>>>>>>>>>>>>>> PETSC_ARCH variable which was set to linux-gnu-c-opt instead of
>>>>>>>>>>>>>>>>>>>> linux-gnu-cxx-opt.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Doesn't solve the problem for me :(
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> - My PETSC_ARCH is set to linux-gnu-cxx-opt (cf gentoo.platform)
>>>>>>>>>>>>>>>>>>> - cmake reports: -- PETSC_ARCH is linux-gnu-cxx-opt
>>>>>>>>>>>>>>>>>>> - Still get same problem.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> I think this changed in Dorsal at some point recently.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Everything updated as of 20 seconds ago ...
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> I too am now seeing strange things such as this:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> -- PETSC_DIR is /Users/harish/Work/FEniCS/dev/src/petsc-3.1-p4
>>>>>>>>>>>>>>>>>> -- PETSC_ARCH is darwin10.0.0-cxx-opt
>>>>>>>>>>>>>>>>>> -- Found petscconf.h
>>>>>>>>>>>>>>>>>> -- PETSc test failed
>>>>>>>>>>>>>>>>>> -- PETSc could not be found. Be sure to set PETSC_DIR and PETSC_ARCH.
>>>>>>>>>>>>>>>>>> (missing:  PETSC_TEST_RUNS)
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>> .
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> -- (OK) MPI
>>>>>>>>>>>>>>>>>> -- (OK) MTL4
>>>>>>>>>>>>>>>>>> -- (OK) UMFPACK
>>>>>>>>>>>>>>>>>> -- (OK) CHOLMOD
>>>>>>>>>>>>>>>>>> -- (OK) SCOTCH
>>>>>>>>>>>>>>>>>> -- (OK) PARMETIS
>>>>>>>>>>>>>>>>>> -- (OK) GMP
>>>>>>>>>>>>>>>>>> -- (OK) CGAL
>>>>>>>>>>>>>>>>>> -- (OK) ZLIB
>>>>>>>>>>>>>>>>>> -- (OK) PYTHON
>>>>>>>>>>>>>>>>>> -- The following optional packages could not be found:
>>>>>>>>>>>>>>>>>> -- (**) PETSC
>>>>>>>>>>>>>>>>>> -- (**) SLEPC
>>>>>>>>>>>>>>>>>> -- (**) TRILINOS
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Following which (and independently), CMake fails catastrophically:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Building CXX object dolfin/CMakeFiles/dolfin.dir/ale/HarmonicSmoothing.cpp.o
>>>>>>>>>>>>>>>>> [  1%] cd /foo/dolfin/dorsal_build_dir/dolfin && /opt/local/bin/c++
>>>>>>>>>>>>>>>>> -Ddolfin_EXPORTS -DPACKAGE_VERSION=\"0.9.8\" -DBOOST_UBLAS_NDEBUG
>>>>>>>>>>>>>>>>> -DHAS_MTL4 -DMTL_HAS_BLAS -DHAS_UMFPACK -DHAS_CHOLMOD -DHAS_SCOTCH
>>>>>>>>>>>>>>>>> -DHAS_PARMETIS -DHAS_CGAL -DHAS_GMP -DHAS_ZLIB -DHAS_CPPUNIT -DHAS_MPI
>>>>>>>>>>>>>>>>> -DMPICH_IGNORE_CXX_SEEK -frounding-math;-D_REENTRANT -O2 -g -fPIC
>>>>>>>>>>>>>>>>> -I/foo/dolfin -I/opt/local/include/libxml2
>>>>>>>>>>>>>>>>> -I/Users/harish/Work/FEniCS/dev/include -I/opt/local/include
>>>>>>>>>>>>>>>>> -I/Users/harish/Work/FEniCS/dev/src/mtl4 -I/opt/local/include/ufsparse
>>>>>>>>>>>>>>>>> -I/opt/local/include/cppunit   -o
>>>>>>>>>>>>>>>>> CMakeFiles/dolfin.dir/ale/HarmonicSmoothing.cpp.o -c
>>>>>>>>>>>>>>>>> /foo/dolfin/dolfin/ale/HarmonicSmoothing.cpp
>>>>>>>>>>>>>>>>> i686-apple-darwin10-g++-4.2.1: no input files
>>>>>>>>>>>>>>>>> i686-apple-darwin10-g++-4.2.1: no input files
>>>>>>>>>>>>>>>>> Building CXX object
>>>>>>>>>>>>>>>>> dolfin/CMakeFiles/dolfin.dir/ale/TransfiniteInterpolation.cpp.o
>>>>>>>>>>>>>>>>> cd /foo/dolfin/dorsal_build_dir/dolfin && /opt/local/bin/c++
>>>>>>>>>>>>>>>>> -Ddolfin_EXPORTS -DPACKAGE_VERSION=\"0.9.8\" -DBOOST_UBLAS_NDEBUG
>>>>>>>>>>>>>>>>> -DHAS_MTL4 -DMTL_HAS_BLAS -DHAS_UMFPACK -DHAS_CHOLMOD -DHAS_SCOTCH
>>>>>>>>>>>>>>>>> -DHAS_PARMETIS -DHAS_CGAL -DHAS_GMP -DHAS_ZLIB -DHAS_CPPUNIT -DHAS_MPI
>>>>>>>>>>>>>>>>> -DMPICH_IGNORE_CXX_SEEK -frounding-math;-D_REENTRANT -O2 -g -fPIC
>>>>>>>>>>>>>>>>> -I/foo/dolfin -I/opt/local/include/libxml2
>>>>>>>>>>>>>>>>> -I/Users/harish/Work/FEniCS/dev/include -I/opt/local/include
>>>>>>>>>>>>>>>>> -I/Users/harish/Work/FEniCS/dev/src/mtl4 -I/opt/local/include/ufsparse
>>>>>>>>>>>>>>>>> -I/opt/local/include/cppunit   -o
>>>>>>>>>>>>>>>>> CMakeFiles/dolfin.dir/ale/TransfiniteInterpolation.cpp.o -c
>>>>>>>>>>>>>>>>> /foo/dolfin/dolfin/ale/TransfiniteInterpolation.cpp
>>>>>>>>>>>>>>>>> Building CXX object dolfin/CMakeFiles/dolfin.dir/common/real.cpp.o
>>>>>>>>>>>>>>>>> /bin/sh: -D_REENTRANT: command not found
>>>>>>>>>>>>>>>>> /bin/sh: -D_REENTRANT: command not found
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I made a fix for this yesterday. Maybe you just need to clean up the
>>>>>>>>>>>>>>>> CMake cache (remove the dorsal_build_dir folder)?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> My fix was reverted but it should be fixed again now.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> This problem is fixed, leaving only the petsc/slepc error.
>>>>>>>>>>>>>
>>>>>>>>>>>>> I spoke too soon. I now receive the following during the linking stage
>>>>>>>>>>>>> (too).
>>>>>>>>>>>>>
>>>>>>>>>>>>> Undefined symbols:
>>>>>>>>>>>>>  "_METIS_mCPartGraphRecursive2", referenced from:
>>>>>>>>>>>>>      _ParMETIS_V3_PartKway in libparmetis.a(kmetis.o)
>>>>>>>>>>>>>      _PartitionSmallGraph__ in libparmetis.a(weird.o)
>>>>>>>>>>>>>      _Mc_InitPartition_RB__ in libparmetis.a(initpart.o)
>>>>>>>>>>>>>      _Mc_InitPartition_RB__ in libparmetis.a(initpart.o)
>>>>>>>>>>>>>      _Balance_Partition__ in libparmetis.a(initbalance.o)
>>>>>>>>>>>>>      _Balance_Partition__ in libparmetis.a(initbalance.o)
>>>>>>>>>>>>> ld: symbol(s) not found
>>>>>>>>>>>>> collect2: ld returned 1 exit status
>>>>>>>>>>>>> make[2]: *** [dolfin/libdolfin.0.9.8.dylib] Error 1
>>>>>>>>>>>>> make[1]: *** [dolfin/CMakeFiles/dolfin.dir/all] Error 2
>>>>>>>>>>>>> make: *** [all] Error 2
>>>>>>>>>>>>
>>>>>>>>>>>> I also got this on the Mac buildbot. The problem is that there is a
>>>>>>>>>>>> file /opt/local/lib/libmetis.dylib that is used instead of the one
>>>>>>>>>>>> that comes with ParMETIS (that Dorsal builds). I worked around this by
>>>>>>>>>>>> configuring DOLFIN with
>>>>>>>>>>>> -DMETIS_LIBRARY=/Users/fenicsslave2/Work/FEniCS/lib/libmetis.a (the
>>>>>>>>>>>> one that Dorsal installed with ParMETIS).
>>>>>>>>>>>>
>>>>>>>>>>>> This problem only shows up when PETSc is not found because
>>>>>>>>>>>> PETSC_LIBRARIES contains the correct link flags (-L/path/to/libmetis.a
>>>>>>>>>>>> -lmetis).
>>>>>>>>>>>>
>>>>>>>>>>>> Johannes
>>>>>>>>>>>
>>>>>>>>>>> Everything seems to work fine except BOOST_DIR is missing when I run
>>>>>>>>>>> the demos.
>>>>>>>>>>
>>>>>>>>>> /sw (Fink) was the default BOOST_DIR. I have changed this to
>>>>>>>>>> /opt/local (MacPorts) now.
>>>>>>>>>
>>>>>>>>> It works fine when I set BOOST_DIR to /opt/local
>>>>>>>>>
>>>>>>>>> Should we write BOOST_DIR to dolfin.conf?
>>>>>>>>
>>>>>>>> Yes, that might be a good idea, however, it might not always be defined.
>>>>>>>
>>>>>>> Your latest fix seems to have done the trick. It's working fine now.
>>>>>>>
>>>>>>> So, clean install from Dorsal on Mac OS X without any extra tweaks
>>>>>>> seems to work perfectly now, at least for me.
>>>>>>>
>>>>>>> Anyone else still having trouble?
>>>>>>>
>>>>>>> I'll try removing some PETSc files next to see if it breaks cleanly.
>>>>>>
>>>>>> One thing that's still missing on OS X is SLEPc. It has worked before
>>>>>> but the test program fails during configuration with the following message:
>>>>>>
>>>>>> Undefined symbols:
>>>>>>  "_ompi_mpi_comm_self"
>>>>>
>>>>> I'm having a similar problem with PETSc on the Mac buildbot:
>>>>>
>>>>>   http://fenics.org:8080/builders/dolfin-mac-osx/builds/721/steps/dolfin%20build/logs/CMakeError.log
>>>>>
>>>>>> I think I remember seeing this in a discussion recently. Is there a
>>>>>> solution for this? It has worked before.
>>>>>
>>>>> One solution will be to set CXX=mpic++ before running cmake. I guess
>>>>> the reason it worked before might be that the test was not required to
>>>>> work in order for PETSc and SLEPc to be found. This was changed a few
>>>>> days back.
>>>>
>>>> Did we have CXX=mpic++ or similar in the Dorsal platform file for Snow Leopard?
>>>
>>> We used to have this in dolfin.package:
>>>
>>> package_specific_setup () {
>>>     export CC=mpicc
>>>     export CXX=mpic++
>>> }
>>>
>>> Can we add it back? SLEPc works again for me on OS X when I add it.
>>>
>>
>> Best would be to get the MPI flags right in dolfin/CMakeLists.txt. There
>> are a few FIXME related the link flags.
>>
>> If we use the MPI wrappers, we should somehow test for this and then not
>> add the MPI compiler flags in dolfin/CMakeLists.txt.
> 
> I don't understand. That seems like a secondary issue, that we would
> add flags we don't really need if we are using the MPI wrappers.
>

CMake might pick up a different version of MPI to what mpicxx points to.
It's not very tidy to mix things ups.

> Does anyone know which extra flag is needed to get things to run?
> Otherwise, I suggest adding the above to dolfin.platform and also a
> FIXME to dolfin/CMakeLists.txt about needing to figure out how to get
> those libraries on Mac. At least that would get things working now.
> 

Read the FIXME:

    # FIXME: Is this linker flag required?

Garth

> --
> Anders