scipy.test segfaulting in Windows
![](https://secure.gravatar.com/avatar/93476f0f6e1767a3de6043fa02e267bc.jpg?s=120&d=mm&r=g)
Hello, First off, let me thank all the developers of scipy for creating such an excellent tool for scientific computation. After a week of struggling with some C++ packages and their python bindings, I found that the scipy.sandbox.delaunay package is just what I needed. It was a cinch to compile and install in Windows under Cygwin and natively. I got greedy and wanted to make sure that the installation tests worked and got the segfault on both the native Python and the Cygwin Python sides that others have mentioned on the web. I've tried building it with: python setup.py build config_fc --noopt but got the same segfault. Any idea of how this can be fixed? What needs to be tweaked? p.s. Using UMFPACK in Cygwin requires that UMFPACK be linked to UFconfig/xerbla/libcerbla.a to avoid linker errors while building scipy with umfpack support -- Andres Corrada-Emmanuel Research Fellow Aerial Imaging and Remote Sensing Lab Computer Science Department University of Massachusetts at Amherst Blog: www.corrada.com/blog
![](https://secure.gravatar.com/avatar/93476f0f6e1767a3de6043fa02e267bc.jpg?s=120&d=mm&r=g)
The problem is related to the test_histogram tests in test_ndimage.py. Commenting out the tests results in the test suite completing but two failures occur later on (with no segfault) also related to nd_image: generic filter 1 generic 1d filter 1 Andres Corrada-Emmanuel wrote:
Hello,
First off, let me thank all the developers of scipy for creating such an excellent tool for scientific computation. After a week of struggling with some C++ packages and their python bindings, I found that the scipy.sandbox.delaunay package is just what I needed. It was a cinch to compile and install in Windows under Cygwin and natively.
I got greedy and wanted to make sure that the installation tests worked and got the segfault on both the native Python and the Cygwin Python sides that others have mentioned on the web.
I've tried building it with:
python setup.py build config_fc --noopt
but got the same segfault.
Any idea of how this can be fixed? What needs to be tweaked?
p.s.
Using UMFPACK in Cygwin requires that UMFPACK be linked to UFconfig/xerbla/libcerbla.a to avoid linker errors while building scipy with umfpack support
-- Andres Corrada-Emmanuel Research Fellow Aerial Imaging and Remote Sensing Lab Computer Science Department University of Massachusetts at Amherst Blog: www.corrada.com/blog
![](https://secure.gravatar.com/avatar/af6c39d6943bd4b0e1fde23161e7bb8c.jpg?s=120&d=mm&r=g)
Hi Andres On Sat, Dec 29, 2007 at 08:26:14PM -0500, Andres Corrada-Emmanuel wrote:
The problem is related to the test_histogram tests in test_ndimage.py. Commenting out the tests results in the test suite completing but two failures occur later on (with no segfault) also related to nd_image:
generic filter 1 generic 1d filter 1
This has been fixed in SVN for some time, so try that version instead. Regards Stéfan
![](https://secure.gravatar.com/avatar/93476f0f6e1767a3de6043fa02e267bc.jpg?s=120&d=mm&r=g)
Thanks Stefan, using the svn versions of nd_image.[ch] fixed the problem. Stefan van der Walt wrote:
Hi Andres
On Sat, Dec 29, 2007 at 08:26:14PM -0500, Andres Corrada-Emmanuel wrote:
The problem is related to the test_histogram tests in test_ndimage.py. Commenting out the tests results in the test suite completing but two failures occur later on (with no segfault) also related to nd_image:
generic filter 1 generic 1d filter 1
This has been fixed in SVN for some time, so try that version instead.
Regards Stéfan _______________________________________________ SciPy-user mailing list SciPy-user@scipy.org http://projects.scipy.org/mailman/listinfo/scipy-user
-- Andres Corrada-Emmanuel Research Fellow Aerial Imaging and Remote Sensing Lab Computer Science Department University of Massachusetts at Amherst Blog: www.corrada.com/blog
participants (2)
-
Andres Corrada-Emmanuel
-
Stefan van der Walt