Hi list,

the current master segfaults at import, which I find a very serious problem. It 
segfaulted the buildbot, and I can reproduce the problem under a computer with 
Python 2.7 (interstingly not Python 2.6). I spent 3 hours on this today but I 
must really switch to other urgent matters.

I could really use some help here: having the master is this state is very bad 
for the health of the project. Reverting the corresponding commits is not 
something easy, as we would have to revert a long list of commits in a pull 
request.

The current status of my investigation is:

* It's a sefault happening probably at garbage collection after importing 
sklearn.metrics.pairwise. Simply running "python -c 'import 
sklearn.metrics.pairwise'" is enough to trigger it sometimes.

* It's most probably related to the _fast_euclidean cython module

* valgrind on this doesn't give anything (but my knowledge of valgrind is small)

* the best way to reproduce is to run several times "nosetests -s sklearn"

* the gdb stack trace is hard to interpret, because it is a half-imported module

any help is more then welcome,

Gael

------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_sfd2d_oct
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to