[Scipy-tickets] [SciPy] #1414: Seg. fault from scipy.interpolate or numpy

SciPy Trac scipy-tickets@scipy....
Fri Mar 25 17:19:55 CDT 2011

#1414: Seg. fault from scipy.interpolate or numpy
 Reporter:  jturner            |       Owner:  somebody
     Type:  defect             |      Status:  new     
 Priority:  normal             |   Milestone:          
Component:  scipy.interpolate  |     Version:  0.9.0   
 Keywords:  segmentation       |  
 (As per discussion with Pauli Virtanen on scipy-user:)

 I'm getting a segmentation fault with no traceback from the attached
 Python test script (a highly cut-down version of my real script), even
 after upgrading to the latest stable NumPy 1.5.1 and SciPy 0.9.0. My
 minimal script depends on NumPy, SciPy and PyFITS. The crash is happening
 with Python 2.5.1 on an Intel Xeon running RedHat Enterprise 5.4 (x86_64).
 My PyFITS version (though I doubt it matters) is 2.3.2. I don't see the
 problem on my laptop, where I am running Python 2.6.4 under Ubuntu 10.04
 (x86_64) on an Intel Core i7.

 You can get a 1.1Mb input file for the script, scipybug2.fits, with sftp
 from ftp.gemini.edu until one month from now, by logging in as "scipy1"
 with the password "QHDGX08r" (sorry I don't have an HTTP URL where I can
 easily put it). This only works with secure ftp, not normal ftp. I'll also
 try attaching the file here if possible.

 As you can see from attached gdb output, the crash occurs in numpy, but I
 suspect the cause of the problem may be the array that is produced at the
 previous line by interp.splev(). You can see some funny-looking characters
 and and an apparently-huge reference count from "print **des" in gdb,
 though I don't really know what I'm doing / looking for here. My
 familiarity with gdb (and Python C extensions) is very limited, but I'm
 happy to type some more stuff if you let me know what.

 Thanks a lot!


Ticket URL: <http://projects.scipy.org/scipy/ticket/1414>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.

More information about the Scipy-tickets mailing list