[Scipy-tickets] [SciPy] #1265: ARPACK seg fault on 64bit Linux

SciPy Trac scipy-tickets@scipy....
Fri Aug 13 19:49:01 CDT 2010

#1265: ARPACK seg fault on 64bit Linux
 Reporter:  mbudisic               |       Owner:  wnbell
     Type:  defect                 |      Status:  new   
 Priority:  normal                 |   Milestone:  0.9.0 
Component:  scipy.sparse.linalg    |     Version:  devel 
 Keywords:  arpack, 64 bit, linux  |  
 I have encountered a seg fault in test_arpack at the following point:

 test_nonsymmetric_modes (test_arpack.TestEigenNonSymmetric) ... ERROR

 I'm running a 64 bit CentOS machine, gcc 4.5.0, custom-compiled ATLAS
 3.8.3 with full LAPACK, both Python 2.6 (main) and Python 2.7 (alt-
 installed), Numpy 1.4.1 and 1.5.0b1 (respectively) and the Scipy
 0.9.0.dev6646. I've noticed the problem first in 0.8.0 release, and have
 since upgraded to try to avoid the problem, but without luck. The rest of
 the Scipy seems to be working fine.

 This problem might be related to
 [http://projects.scipy.org/scipy/ticket/1220 #1220]

 I'm attaching the full gdb crash log to this ticket.

 In trying to remedy this, I have swapped out scipy-distributed ARPACK
 source with the one distributed by Andreas Kloeckner, in which he added a
 few newer patches [http://mathema.tician.de/software/arpack link]. By
 this, I mean that I replaced scipy's source with Kloeckner's source, and
 let scipy installer compile them as they install. Note that in that
 distribution the (in)famous "second.f" edit that changes ETIME from
 external to internal function happens in "arscnd.f", in case you need it.
 In any case, the error happens in the same fashion, in the same spot, so I
 figure this is not an issue with an internal ARPACK distribution, at least
 not one that is fixed by Kloeckner's distributions.

Ticket URL: <http://projects.scipy.org/scipy/ticket/1265>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.

More information about the Scipy-tickets mailing list