[SciPy-user] problem with UMFPACK in scipy.test

[Johann Cohen-Tanugi]

hi again,
ok there was an uncommented CFLAG line at the bottom of the UFConfig.mk, 
my bad.... Commenting that line does bring the -fPIC arg back.
As far as I can tell, there is nothing wrong occurring during the build :
[cohen at jarrett scipy-svn]$ grep -i umfpack build.log
umfpack_info:
    libraries = ['umfpack', 'amd']
    library_dirs = ['/data1/sources/MATHSTUFF/UMFPACK/Lib/', '/usr/lib']
    swig_opts = ['-I/data1/sources/MATHSTUFF/UMFPACK/Include']
    define_macros = [('SCIPY_UMFPACK_H', None)]
    include_dirs = ['/data1/sources/MATHSTUFF/UMFPACK/Include']
building extension "scipy.sparse.linalg.dsolve.umfpack.__umfpack" sources
creating build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
  adding 'scipy/sparse/linalg/dsolve/umfpack/umfpack.i' to sources.
swig: scipy/sparse/linalg/dsolve/umfpack/umfpack.i
swig -python -I/data1/sources/MATHSTUFF/UMFPACK/Include 
-I/usr/local/atlas/include -o 
build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack/_umfpack_wrap.c 
-outdir build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack 
scipy/sparse/linalg/dsolve/umfpack/umfpack.i
creating build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying scipy/sparse/linalg/dsolve/umfpack/info.py -> 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying scipy/sparse/linalg/dsolve/umfpack/umfpack.py -> 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying scipy/sparse/linalg/dsolve/umfpack/setup.py -> 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying scipy/sparse/linalg/dsolve/umfpack/setupscons.py -> 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying scipy/sparse/linalg/dsolve/umfpack/__init__.py -> 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
copying 
build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack/_umfpack.py 
-> build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
building 'scipy.sparse.linalg.dsolve.umfpack.__umfpack' extension
creating 
build/temp.linux-i686-2.5/build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack
compile options: '-DSCIPY_UMFPACK_H -DATLAS_INFO="\"3.8.0\"" 
-I/data1/sources/MATHSTUFF/UMFPACK/Include -I/usr/local/atlas/include 
-I/usr/lib/python2.5/site-packages/numpy/core/include 
-I/usr/include/python2.5 -c'
gcc: 
build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack/_umfpack_wrap.c
gcc -pthread -shared 
build/temp.linux-i686-2.5/build/src.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack/_umfpack_wrap.o 
-L/data1/sources/MATHSTUFF/UMFPACK/Lib/ -L/usr/lib 
-L/usr/local/atlas/lib -L/usr/lib -Lbuild/temp.linux-i686-2.5 -lumfpack 
-lamd -lptf77blas -lptcblas -latlas -llapack -lptf77blas -lptcblas 
-latlas -lpython2.5 -o 
build/lib.linux-i686-2.5/scipy/sparse/linalg/dsolve/umfpack/__umfpack.so

and the test now reaches :
Prefactorize (with UMFPACK) matrix for solving with multiple rhs ... ok
Prefactorize matrix for solving with multiple rhs ... ok
Solve with UMFPACK: double precision complex ... ok
Solve: single precision complex ... ok
Solve with UMFPACK: double precision, sparse rhs ... ok
Solve with UMFPACK: double precision ... ok

and then continue until it seems to hang at:
test_gammaincinv (test_basic.TestGamma) ...

I guess that is for another thread! and another day....
thanks a lot for the hints,
Johann

Robert Kern wrote:
> On Thu, Apr 24, 2008 at 7:28 AM, Johann Cohen-Tanugi
> <cohen at slac.stanford.edu> wrote:
>   
>> hello,
>>  I have compiled and built UMFPACK, following the wiki, and when
>>  configuring scipy I read :
>>  umfpack_info:
>>  amd_info:
>>   libraries amd not found in /usr/local/lib
>>   FOUND:
>>     libraries = ['amd']
>>     library_dirs = ['/usr/lib']
>>
>>   FOUND:
>>     libraries = ['umfpack', 'amd']
>>     library_dirs = ['/data1/sources/MATHSTUFF/UMFPACK/Lib/', '/usr/lib']
>>     swig_opts = ['-I/data1/sources/MATHSTUFF/UMFPACK/Include']
>>     define_macros = [('SCIPY_UMFPACK_H', None)]
>>     include_dirs = ['/data1/sources/MATHSTUFF/UMFPACK/Include']
>>
>>  I would infer from that that UMFPACK was correctly found, but after
>>  building scipy from svn, I get :
>>     
>
> You should actually look later in the build log to see if the .so
> files got linked correctly.
>
>   
>>  In [4]: scipy.test(verbose=2)
>>  <lots of stuff>
>>  ======================================================================
>>  SKIP: Getting factors of complex matrix
>>  ----------------------------------------------------------------------
>>  Traceback (most recent call last):
>>   File "/usr/lib/python2.5/site-packages/scipy/testing/decorators.py",
>>  line 81, in skipper
>>     raise nose.SkipTest, msg
>>  SkipTest: UMFPACK appears not to be compiled
>>
>>  <several other outputs of the same kind>
>>
>>  ======================================================================
>>  SKIP: Solve: single precision
>>  ----------------------------------------------------------------------
>>  Traceback (most recent call last):
>>   File "/usr/lib/python2.5/site-packages/scipy/testing/decorators.py",
>>  line 81, in skipper
>>     raise nose.SkipTest, msg
>>  SkipTest: UMFPACK appears not to be compiled
>>
>>  ----------------------------------------------------------------------
>>  Ran 1524 tests in 39.269s
>>
>>  FAILED (failures=5, errors=12)
>>
>>
>>  What does that mean? I noticed that when building UMFPACK I put -fPIC in
>>  the UFConfig.mk but it does not seem to be honored. For instance :
>>  gcc -O3 -fexceptions -D_FILE_OFFSET_BITS=64 -D_LARGEFILE64_SOURCE
>>  -I../Include -I../Source -I../../AMD/Include -I../../UFconfig -DDINT -c
>>  ../Source/umfpack_qsymbolic.c -o umfpack_di_qsymbolic.o
>>
>>  Is that related?
>>     
>
> Quite possibly, yes.
>
>