I also installed nose and tried running the numpy test suite. This seems to show multiple issues in numpy.linalg and actually crashes while running these tests.

The relevant output:

test_cdouble (test_linalg.TestCond2) ... ERROR
test_csingle (test_linalg.TestCond2) ... ERROR
test_double (test_linalg.TestCond2) ... ERROR
test_empty (test_linalg.TestCond2) ... ok
Check that matrix type is preserved. ... ERROR
Check that matrix type is preserved. ... ERROR
test_nonarray (test_linalg.TestCond2) ... ERROR
test_single (test_linalg.TestCond2) ... ERROR
error: [/home/moloney/Software/slicer/Slicer-SuperBuild-Debug/Slicer-build/bin/./SlicerApp-real] exit abnormally - Report the problem.

After investigating, it resumes it is pointless to compile CLAPACK and build numpy against it. Indeed, the modules "blas_lite.c" and "dlapack_lite.c" already bundle part of CLAPACK that are required by "linalg" numpy module.

While looking for the root cause of the problem, I also realized that CLAPACK can't be linked against numpy. The "f2c.h" bundled within clapack-3.2.1-CMAKE.tgz is different from the "f2c.h" available in "numpy/linalg" folder. For example, the typedefs used to specify "integer" or "uinteger" are different.

In clapack/INCLUDE/f2c.h => "integer" is defined as "long int"
In numpy/linalg/f2c.h => "integer" is defined as "int"

Within the python module "lapack_litemodule.c", we can read that the function signature used to specify the function to link against is of the form:

extern int foo_(char arg1, char arg2, int *n)

whereas in CLapack, the same function is defined as:

int foo_(char arg1, char arg2, integer *n) // integer being "long int"

This difference in type was causing the error reported in the bug.
While working on that issue, I also created a small project illustrating the problem. See https://github.com/jcfr/CFunclash#readme

This apply at least from numpy 1.4.1 to 1.6.2

The following topic resolved the issue on linux: https://github.com/jcfr/Slicer/tree/2249-fix-numpy-runtime-issue

Note: Will have to test it on both windows and macosx.

Associated commit: https://github.com/jcfr/Slicer/commit/fd21352c9212f64c5a1b40b2f3356e3a0f5dc17f

On linux 64bit (Ubuntu 10.04), the following numpy test suite run without
any errors:

=> nose version 1.1.2
=> gcc 4.4.3

numpy.test()
[...]
Ran 2483 tests in 10.194s

OK (KNOWNFAIL=4, SKIP=12)

Details (package by package)

np.linalg.test()
Running unit tests for numpy.linalg
NumPy version 1.4.1
[...]
Ran 116 tests in 0.203s

OK
<nose.result.TextTestResult run=116 errors=0 failures=0>

np.ma.test()
Running unit tests for numpy.ma
NumPy version 1.4.1
[...]
Ran 282 tests in 1.011s

OK
<nose.result.TextTestResult run=282 errors=0 failures=0>

np.polynomial.test()
Running unit tests for numpy.polynomial
NumPy version 1.4.1
[...]
Ran 86 tests in 0.324s

OK
<nose.result.TextTestResult run=86 errors=0 failures=0>

np.random.test()
Running unit tests for numpy.random
NumPy version 1.4.1
[...]
Ran 11 tests in 0.020s

OK
<nose.result.TextTestResult run=11 errors=0 failures=0>

np.fft.test()
Running unit tests for numpy.fft
NumPy version 1.4.1
[...]
Ran 5 tests in 0.023s

OK
<nose.result.TextTestResult run=5 errors=0 failures=0>

np.core.test()
Running unit tests for numpy.core
NumPy version 1.4.1
[...]
Ran 861 tests in 2.181s

OK (KNOWNFAIL=4, SKIP=11)
<nose.result.TextTestResult run=861 errors=0 failures=0>

np.linalg.test()
Running unit tests for numpy.linalg
NumPy version 1.4.1
[...]
Ran 116 tests in 0.184s

OK
<nose.result.TextTestResult run=116 errors=0 failures=0>

np.matrixlib.test()
Running unit tests for numpy.matrixlib
NumPy version 1.4.1
[...]
Ran 38 tests in 0.077s

OK
<nose.result.TextTestResult run=38 errors=0 failures=0>

np.lib.test()
Running unit tests for numpy.lib
NumPy version 1.4.1
[...]
Ran 994 tests in 6.313s

OK (SKIP=1)

numpy.oldnumeric.test()
[...]
Ran 2 tests in 0.020s

OK

numpy.testing.test()
[...]
Ran 69 tests in 0.128s

OK

bmoloney> Would be great if you could try the topic I referenced in the previous note. Thanks

Tested with https://github.com/jcfr/Slicer/tree/2249-fix-numpy-runtime-issue.

I followed the steps in the original report, and there was no error.

import numpy as np
test=np.eye(3)
np.linalg.det(test)
1.0

I also tried to do the nose tests, but that did not work:

[fedorov@bear nose-1.1.2] pwd
/Users/fedorov/github/Slicer-Superbuild/python-build/nose-1.1.2
[fedorov@bear nose-1.1.2] ../bin/python

import numpy as np
np.test()
Running unit tests for numpy
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy/testing/nosetester.py", line 335, in test
t = NumpyTestProgram(argv=argv, exit=False, plugins=plugins)
File "nose/core.py", line 118, in init
**extra_args)
File "/Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/unittest.py", line 816, in init
self.parseArgs(argv)
File "nose/core.py", line 135, in parseArgs
self.config.configure(argv, doc=self.usage())
File "nose/config.py", line 276, in configure
options, args = self._parseArgs(argv, cfg_files)
File "nose/config.py", line 264, in _parseArgs
return parser.parseArgsAndConfigFiles(argv[1:], cfg_files)
File "nose/config.py", line 132, in parseArgsAndConfigFiles
self._applyConfigurationToValues(self._parser, config, values)
File "nose/config.py", line 118, in _applyConfigurationToValues
name=name, filename=filename)
File "nose/config.py", line 261, in warn_sometimes
raise ConfigError(msg)
nose.config.ConfigError: Error reading config file 'setup.cfg': no such option 'with-doctest'

Andriy> Edit ~/nose.cfg and remove the like having "with doc test". That should do it.

The file I had to edit was nose-1.1.2/setup.cfg

[fedorov@bear Slicer-Superbuild] ./python-build/bin/python
Python 2.6.6 (r266:84292, Jul 1 2012, 17:09:31)
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin
Type "help", "copyright", "credits" or "license" for more information.

import numpy as np
import nose
np.test()
Running unit tests for numpy
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
...

Ran 2483 tests in 4.686s

OK (KNOWNFAIL=4, SKIP=1)
<nose.result.TextTestResult run=2483 errors=0 failures=0>

np.linalg.test()
Running unit tests for numpy.linalg
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
S...................................................................................................................

Ran 116 tests in 0.109s

OK (SKIP=1)
<nose.result.TextTestResult run=116 errors=0 failures=0>

np.ma.test()
Running unit tests for numpy.ma
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
..........................................................................................................................................................................................................................................................................................

Ran 282 tests in 0.775s

OK
<nose.result.TextTestResult run=282 errors=0 failures=0>

np.polynomial.test()
Running unit tests for numpy.polynomial
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
......................................................................................

Ran 86 tests in 0.272s

OK
<nose.result.TextTestResult run=86 errors=0 failures=0>

np.random.test()
Running unit tests for numpy.random
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
...........

Ran 11 tests in 0.015s

OK
<nose.result.TextTestResult run=11 errors=0 failures=0>

np.core.test()
Running unit tests for numpy.core
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
...

Ran 861 tests in 1.247s

OK (KNOWNFAIL=4)
<nose.result.TextTestResult run=861 errors=0 failures=0>

np.matrixlib.test()
Running unit tests for numpy.matrixlib
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
......................................

Ran 38 tests in 0.062s

OK
<nose.result.TextTestResult run=38 errors=0 failures=0>

np.lib.test()
Running unit tests for numpy.lib
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
...

Ran 994 tests in 1.130s

OK
<nose.result.TextTestResult run=994 errors=0 failures=0>

np.oldnumeric.test()
Running unit tests for numpy.oldnumeric
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
..

Ran 2 tests in 0.017s

OK
<nose.result.TextTestResult run=2 errors=0 failures=0>

np.testing.test()
Running unit tests for numpy.testing
NumPy version 1.4.1
NumPy is installed in /Users/fedorov/github/Slicer-Superbuild/python-build/lib/python2.6/site-packages/numpy
Python version 2.6.6 (r266:84292, Jul 1 2012, 17:09:31) [GCC 4.2.1 (Apple Inc. build 5666) (dot 3)]
nose version 1.1.2
.....................................................................

Ran 69 tests in 0.103s

OK
<nose.result.TextTestResult run=69 errors=0 failures=0>

Fixed in r20539
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&amp;revision=20539

Dear issue reporter,

Good news :) Slicer developers SOLVED the problem you reported - YOU now need to VERIFY and CLOSE this issue.

Sorry for my delayed response. I can confirm the issue is resolved now.

Thanks so much for the hard work!