[SciPy-User] Install Scipy Errors: ImportError: /path_to/liblapack.so: undefined symbol: ztbsv_

Bruce Southey bsouthey@gmail....
Sun Aug 28 09:40:15 CDT 2011


On Sun, Aug 28, 2011 at 1:16 AM, Johnny <dbigbear@gmail.com> wrote:
> Hi all,
>
> I am installing lapack, atlas, numpy, scipy on my LINUX for TEN times,
> but always encountering the problem:
>
> [work <at> XXX]$ python -c 'import scipy.optimize;
> scipy.optimize.test()'
> Traceback (most recent call last):
>  File "<string>", line 1, in <module>
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> scipy/optimize/__init__.py", line 11, in <module>
>    from lbfgsb import fmin_l_bfgs_b
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> scipy/optimize/lbfgsb.py", line 28, in <module>
>    import _lbfgsb
> ImportError: /home/work/local/lib/liblapack.so: undefined symbol:
> ztbsv_
>
> I can pass some other tests like:
>
>
> [work <at> XXX:~/local]$ python -c 'import scipy.ndimage;
> scipy.ndimage.test()'
> Running unit tests for scipy.ndimage
> NumPy version 1.6.1
> NumPy is installed in /home/work/local/python-2.7.1/lib/python2.7/site-
> packages/numpy
> SciPy version 0.9.0
> SciPy is installed in /home/work/local/python-2.7.1/lib/python2.7/site-
> packages/scipy
> Python version 2.7.1 (r271:86832, Jan 13 2011, 22:17:56) [GCC 3.4.5
> 20051201 (Red Hat 3.4.5-2)]
> nose version 1.1.2
> .........S.................................................................................................................................................................................................................................................................................................................................................................................................................
> ----------------------------------------------------------------------
> Ran 411 tests in 1.247s
>
> OK (SKIP=1)
>
> The problem seems due to the lib of Lapack. So I tried the solutions
> posted on the internet before.
>
> 1) The liblapack.so may be not complete...SO I tried this:
>    # integrate lapack with atlas:
>    cd lib/
>    mkdir tmp
>    cd tmp/
>    ar x ../liblapack.a
>    cp ~/path_to/lapack-3.1.1/lapack_LINUX.a ../liblapack.a
>    ar r ../liblapack.a *.o
>    cd ../..
>    make check
>    make ptcheck
>    cp include/* ~/include/
>    cp lib/*.a ~/lib/
>
> That is, after installing atlas, there is another liblapack.a (in
> addition to the lapack_LINUX.a after Lapack) in its lib, but it is
> about 500k, so I integrate it with the lapack_LINUX.a from installing
> Lapack. The final liblapack.a is about 9.3m, The liblapack.so is about
> 5m
>
> 2) re-install Lapack and atlas many times....No use
>
> 3) I found there is a lapack.so under scipy/lib, and it is about 500K,
> but I think it may be not the problem, becaues the failure is
> "ImportError: /home/work/local/lib/liblapack.so: undefined symbol:
> ztbsv_". Scipy seemed to import liblapack.so in my general lib
> directory...
>
> 4) One thing  I am not sure is that I used gcc 4.7 and gfortran to
> compile lapack and atlas, but my python 2.7 was built using gcc
> 3.4.5.....Is this a problem?
>
>
> Anyone can help?
> _______________________________________________________________
> My configuration of the installation:
>
> * ATLAS 3.8.4
> * lapack 3.3.1
> * numpy 1.6.1
> * SciPy version 0.9.0
> * dateutil 1.5
> * Python version 2.7.1 (r271:86832, Jan 13 2011, 22:17:56) [GCC 3.4.5
> 20051201 (Red Hat 3.4.5-2)]
> * nose version 1.1.2
> * gcc (GCC) 4.7.0 20110820 (experimental)
> * LINUX: Linux XXX 2.6.9_5-9-0-0 #1 SMP Wed Jun 23 14:03:19 CST 2010
> x86_64 x86_64 x86_64 GNU/Linux
>
> site.cfg of Scipy:
>
> [DEFAULT]
> library_dirs = /home/work/local/lib
> include_dirs = /home/work/local/include
> [lapack_opt]
> libraries = lapack, f77blas, cblas, atlas
>
> site.cfg of Numpy:
>
> [DEFAULT]
> library_dirs = /home/work/local/lib
> include_dirs = /home/work/local/include
> [lapack_opt]
> libraries = lapack, f77blas, cblas, atlas
>
>
> In addition, there are failures as well when test Numpy:
>
>>>> import numpy
>>>> numpy.test('1')
> Running unit tests for numpy
> NumPy version 1.6.1
> NumPy is installed in /home/work/local/python-2.7.1/lib/python2.7/site-
> packages/numpy
> Python version 2.7.1 (r271:86832, Jan 13 2011, 22:17:56) [GCC 3.4.5
> 20051201 (Red Hat 3.4.5-2)]
> nose version 1.1.2
> ======================================================================
> FAIL: Test basic arithmetic function errors
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/testing/decorators.py", line 215, in knownfailer
>    return f(*args, **kwargs)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_numeric.py", line 367, in
> test_floating_exceptions_power
>    np.power, ftype(2), ftype(2**fi.nexp))
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_numeric.py", line 271, in assert_raises_fpe
>    "Type %s did not raise fpe error '%s'." % (ftype, fpeerr))
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/testing/utils.py", line 34, in assert_
>    raise AssertionError(msg)
> AssertionError: Type <type 'numpy.float64'> did not raise fpe error
> 'overflow'.
>
> ======================================================================
> FAIL: Test generic loops.
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_ufunc.py", line 86, in test_generic_loops
>    assert_almost_equal(fone(x), fone_val, err_msg=msg)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/testing/utils.py", line 448, in assert_almost_equal
>    raise AssertionError(msg)
> AssertionError:
> Arrays are not almost equal to 7 decimals PyUFunc_F_F
>  ACTUAL: array([ 0.+0.j,  0.+0.j,  0.+0.j,  0.+0.j,  0.+0.j],
> dtype=complex64)
>  DESIRED: 1
>
> ======================================================================
> FAIL: test_umath.TestComplexFunctions.test_loss_of_precision(<type
> 'numpy.complex64'>,)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/nose/
> case.py", line 197, in runTest
>    self.test(*self.arg)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_umath.py", line 931, in check_loss_of_precision
>    check(x_basic, 2*eps/1e-3)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_umath.py", line 901, in check
>    'arcsinh')
> AssertionError: (0, 0.0010023052, 0.9987238, 'arcsinh')
>
> ======================================================================
> FAIL: test_umath.TestComplexFunctions.test_precisions_consistent
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/nose/
> case.py", line 197, in runTest
>    self.test(*self.arg)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/core/tests/test_umath.py", line 812, in
> test_precisions_consistent
>    assert_almost_equal(fcf, fcd, decimal=6, err_msg='fch-fcd %s'%f)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/testing/utils.py", line 448, in assert_almost_equal
>    raise AssertionError(msg)
> AssertionError:
> Arrays are not almost equal to 6 decimals fch-fcd <ufunc 'arcsin'>
>  ACTUAL: 2.3561945j
>  DESIRED: (0.66623943249251527+1.0612750619050355j)
>
> ======================================================================
> FAIL: test_kind.TestKind.test_all
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/nose/
> case.py", line 197, in runTest
>    self.test(*self.arg)
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/f2py/tests/test_kind.py", line 30, in test_all
>    'selectedrealkind(%s): expected %r but got %r' %  (i,
> selected_real_kind(i), selectedrealkind(i)))
>  File "/home/work/local/python-2.7.1/lib/python2.7/site-packages/
> numpy/testing/utils.py", line 34, in assert_
>    raise AssertionError(msg)
> AssertionError: selectedrealkind(19): expected -1 but got 16
>
> ----------------------------------------------------------------------
> Ran 3552 tests in 29.977s
>
> FAILED (KNOWNFAIL=3, failures=5)
> <nose.result.TextTestResult run=3552 errors=0 failures=5>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
Hi,
What Linux distro are you actually using?
Unless you have some issue, I would install the atlas version provided
by the distro as I have long term success with Fedora's packages
across multiple versions.

If you still want to build it yourself, then you need to be using the
same compiler version everywhere.

The ztbsv_ error suggests that you have not build blas, lapack and
atlas correctly. It is hard to get those right so very carefully check
the build logs and run the associated tests.

Bruce


More information about the SciPy-User mailing list