[SciPy-user] [OpenOpt] Segmentation fault

Emanuele Olivetti emanuele@relativita....
Mon Jul 14 12:06:47 CDT 2008


Emanuele Olivetti wrote:
> Dear all,
>
> Whenever I run example 'nlp_1.py' of OpenOpt (both tarball 0.18
> and svn) on both my linux boxes I get "Segmentation fault"!
> Or better, I get it 9 times out of ten (due to random
> initialization without fixed seed, I guess).
>
> My configuration:
> box1: Ubuntu Gutsy, amd64 (Intel C2D)
> box2: Ubuntu Hardy, amd64 (Intel QuadCore)
>
> Python, Numpy, Scipy, Matplotlib etc. are standard versions
> from Ubuntu repositories.
>
> Can anyone confirm/reproduce this behavior?
>
> I've a report that in i386 debian it seems to work flawlessly.
>
>   

Here is the a full log on box1:
----
/usr/lib/python2.5/site-packages/scikits/openopt/examples# python nlp_1.py
OpenOpt checks user-supplied gradient df (shape: (150,) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(df_user - df_numerical)) = 1.72111294887e-05
(is registered in df number 67)
========================
OpenOpt checks user-supplied gradient dc (shape: (2, 150) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(dc_user - dc_numerical)) = 8.42460444801e-05
(is registered in dc number 0)
========================
OpenOpt checks user-supplied gradient dh (shape: (2, 150) )
according to prob.diffInt = [  1.00000000e-07]
lines with 1 - info_user/info_numerical greater than maxViolation = 0.01
will be shown
max(abs(dh_user - dh_numerical)) = 0.00043945014113
(is registered in dh number 149)
========================
-----------------------------------------------------
solver: ralg   problem: unnamed   goal: minimum
 iter    objFunVal    log10(maxResidual)  
    0  8.596e+03               3.91
Segmentation fault (core dumped)
----

The core dump is available here:
http://sra.fbk.eu/people/olivetti/tmp/core.bz2


HTH,

Emanuele



More information about the SciPy-user mailing list