[Scipy-tickets] [SciPy] #1717: QZ segfault on Windows for 0.11.0rc1

SciPy Trac scipy-tickets@scipy....
Thu Aug 9 14:04:25 CDT 2012

#1717: QZ segfault on Windows for 0.11.0rc1
 Reporter:  rgommers      |       Owner:  somebody
     Type:  defect        |      Status:  new     
 Priority:  highest       |   Milestone:  0.11.0  
Component:  scipy.linalg  |     Version:  devel   
 Keywords:                |  

Comment(by jseabold):

 Sure you can't copyright a matrix of random numbers?

 I'm starting to think it's a LAPACK bug now, but I'm way out of my depth
 and at the end of my rope. If you change the matrices slightly it doesn't
 crash, which makes me thing they're just ill-conditioned and it leads to
 that reported bug that was fixed in LAPACK 3.2.x or so. Unfortunately, we
 can't upgrade ATLAS/LAPACK AFAICT because it requires gcc >= 4.2 once you
 get beyond the 3.8.4 or so, though I'm sure there's a way to disable to
 incompatible flags I don't know how and am tired of reading. ATLAS 3.9.79
 builds fine on cygwin, but again requires gcc >= 4.2. I don't understand
 the details of why we can't just do this. I could try again and use the MS
 compilers included with the SDK to see if it's compiler or LAPACK, but
 I've spent too much time on this already. I'm just going to submit a patch
 with slightly different matrices.

Ticket URL: <http://projects.scipy.org/scipy/ticket/1717#comment:20>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.

More information about the Scipy-tickets mailing list