On 17/04/14 10:45, Florian Rathgeber wrote:
> On 17/04/14 10:10, Garth N. Wells wrote:
>> Could you test with the latest SWIG?
> 
> I'm assuming you mean SWIG 2.0.12 rather than 3.0.0? I'll try that.

That doesn't help unfortunately, the error persists. I've updated the
gist: https://gist.github.com/kynan/e13e8a14c9ef2413330d

Any other clues? Seeing that it fails casting the return type of
PyArray_DATA, could it have anything to do with the NumPy version?

Florian

> Florian
> 
>> Garth
>>
>> On 17 Apr 2014, at 08:57, Florian Rathgeber <[email protected]> 
>> wrote:
>>
>>> I'm trying to build current DOLFIN master with the Intel 2013 compilers,
>>> however it fails to build the SWIG wrappers as follows:
>>>
>>> Building CXX object
>>> dolfin/swig/modules/mesh/CMakeFiles/_mesh.dir/modulePYTHON_wrap.cxx.o
>>> dolfin/swig/modules/mesh/modulePYTHON_wrap.cxx(4873): error: invalid
>>> type conversion: "char *" to "double *"
>>>        _array  = static_cast<double*>(PyArray_DATA(xa));
>>>                  ^
>>> [and many more type conversion errors]
>>>
>>> The platform is the cx1 compute cluster at Imperial College with the
>>> following software/package versions:
>>> * Intel compilers 14.0.1 20131008
>>> * Intel MPI Version 3.1.038
>>> * SWIG 2.0.11
>>> * Python 2.7.3
>>> * NumPy 1.7.0b2
>>>
>>> CMake is called as follows:
>>>
>>> CC=mpicc CXX=mpicxx LD_PRELOAD= cmake -DCMAKE_CXX_FLAGS=-std=c++11 \
>>>  -DCMAKE_INSTALL_PREFIX=$DOLFIN_DIR \
>>>  -DUFC_DIR=$UFC_DIR -DBOOST_ROOT=$BOOST_HOME \
>>>  -DARMADILLO_DIR=$ARMADILLO_HOME -DSCOTCH_DIR=$SCOTCH_HOME \
>>>  -DEIGEN3_INCLUDE_DIR=$EIGEN_HOME/include/eigen3 \
>>>  -DDOLFIN_ENABLE_UNIT_TESTS=OFF ..
>>>
>>> The full log from CMake configure and make and the CMakeCache.txt is at
>>> https://gist.github.com/kynan/e13e8a14c9ef2413330d
>>>
>>> I'm happy to provide more information to help tracking down the issue.
>>>
>>> Any hints are much appreciated,
>>> Florian

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

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

Reply via email to