numpy would not load: libgcc_s vs. libgfortran

[Olivier Duchateau]

On Mon, 04 Jan 2016 16:29:56 -0500
"Mikhail T." <mi+thun at aldan.algebra.com> wrote:

> On 04.01.2016 16:31, Olivier Duchateau wrote:
> > After all it's perhaps Numpy's problem, but nobody reported this issue.
> Well, I decided to try figuring this out myself and got the following
> error from a very simply one line script consisting of "import numpy;":
> 
>     Traceback (most recent call last):
>       File "tn.py", line 1, in <module>
>         import numpy;
>       File "/opt/lib/python2.7/site-packages/numpy/__init__.py", line
>     180, in <module>
>         from . import add_newdocs
>       File "/opt/lib/python2.7/site-packages/numpy/add_newdocs.py", line
>     13, in <module>
>         from numpy.lib import add_newdoc
>       File "/opt/lib/python2.7/site-packages/numpy/lib/__init__.py",
>     line 8, in <module>
>         from .type_check import *
>       File "/opt/lib/python2.7/site-packages/numpy/lib/type_check.py",
>     line 11, in <module>
>         import numpy.core.numeric as _nx
>       File "/opt/lib/python2.7/site-packages/numpy/core/__init__.py",
>     line 14, in <module>
>         from . import multiarray
>     ImportError:
>     /opt/lib/python2.7/site-packages/numpy/core/multiarray.so: Undefined
>     symbol "cblas_cdotc_sub"
> 
> Any ideas? Thanks!
> 
>     -mi
> 

It's strange. Unfortunately I can't reproduce this behaviour.

Under Python session (simply run python):

% python
>>> import sys
>>> import numpy

if no errors

>>> sys.modules # display loaded modules (numpy should be here)

>>> keys = sys.modules.keys()
>>> for k in keys:
...  if 'numpy' in k:
...   print k
...

On my system I get this:

numpy.core.info
numpy.lib.numpy
numpy.core.numerictypes
numpy.polynomial.legendre
numpy.random.mtrand
numpy.matrixlib.defmatrix
numpy.random.info
numpy.linalg
numpy.polynomial.laguerre
numpy.testing
numpy.core.umath
numpy.lib._version
numpy.linalg.info
numpy.polynomial._polybase
numpy.lib.arraysetops
numpy.version
numpy.core.machar
numpy.lib.type_check
numpy.core._internal
numpy.random.operator
numpy.lib._datasource
numpy.fft.fftpack_lite
numpy.core.multiarray
numpy.linalg._umath_linalg
numpy.polynomial
numpy.add_newdocs
numpy.core.getlimits
numpy.matrixlib
numpy.ma.extras
numpy.testing.decorators
numpy.lib.npyio
numpy.core.shape_base
numpy.random.threading
numpy.fft.helper
numpy.lib.info
numpy.core.memmap
numpy.fft.fftpack
numpy.linalg.lapack_lite
numpy.ma
numpy.random
numpy.lib.twodim_base
numpy.ma.core
numpy.lib.arrayterator
numpy.lib.financial
numpy.core.arrayprint
numpy.core._methods
numpy.core.function_base
numpy.lib.stride_tricks
numpy
numpy.polynomial.chebyshev
numpy.polynomial.hermite_e
numpy.__config__
numpy.testing.utils
numpy.core.fromnumeric
numpy.ctypeslib
numpy.lib.scimath
numpy.fft
numpy.lib
numpy.lib.nanfunctions
numpy.random.numpy
numpy.lib.function_base
numpy.random.warnings
numpy.compat.py3k
numpy.lib.polynomial
numpy.compat
numpy._import_tools
numpy.polynomial.hermite
numpy.fft.info
numpy.core.records
numpy.polynomial.polynomial
numpy.core.numeric
numpy.core
numpy.lib.ufunclike
numpy.lib._iotools
numpy.lib.index_tricks
numpy.lib.utils
numpy.core.defchararray
numpy.polynomial.polyutils
numpy.lib.shape_base
numpy.compat._inspect
numpy.lib.format
numpy.testing.nosetester
numpy.lib.arraypad
numpy.linalg.linalg

'multiarray' library is even available:

>>> from numpy.core import multiarray
>>> multiarray.__file__
'/usr/local/lib/python2.7/site-packages/numpy/core/multiarray.so'
>>> multiarray.__version__
'3.1'
>>>

Tested on 9.3 (i386) and 10.2 (amd64) machines. 

-- 
olivier