[NumPy-Tickets] [NumPy] #1646: numpy crash inside anki (and mathplotlib)

NumPy Trac numpy-tickets@scipy....
Sun Oct 24 10:14:12 CDT 2010


#1646: numpy crash inside anki (and mathplotlib)
-------------------------------+--------------------------------------------
 Reporter:  batusek            |       Owner:  somebody  
     Type:  defect             |      Status:  needs_info
 Priority:  low                |   Milestone:  2.0.0     
Component:  Official binaries  |     Version:  1.5.0     
 Keywords:                     |  
-------------------------------+--------------------------------------------
Changes (by pv):

  * status:  new => needs_info
  * component:  numpy.core => Official binaries


Comment:

 Pinpointing the source of this crash requires more information. It seems
 it is also likely that this is not a bug in Numpy itself, but appears
 because of some interactions between the components in the binaries you
 use.

 Anyway, the crash points to a crash inside Lapack. As a workaround, you
 can probably just delete _dotblas.pyd.

 First,

   - Which versions of Numpy have you tried? And from which binaries? Which
 version of Python?
   - How to reproduce this issue? What files would I need to download?
   - Is it so that only binary releases of Numpy cause a crash? I assume by
 "development version" you mean a version of Numpy you compiled yourself?
   - What do you mean by "I mean .py instead of the .pyd library"?

 On debugging:

   - I would first try to use e.g. Winpdb (winpdb.org) to locate the point
 in Python code inside matplotlib at which the crash occurs.
   - It's not `setSizePolicy` and `updateGeometry` calls per se -- these
 routines probably call some resizing callbacks inside `matplotlib`. What
 are those?

 Building binaries with debug symbols on Windows is somewhat tricky, IIRC
 you'd need to build also Python with debug symbols. Ask on the Numpy
 mailing list if someone knows how to do this. Or, asking on a suitable
 Python mailing list on how to debug on Windows might also give clues.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1646#comment:1>
NumPy <http://projects.scipy.org/numpy>
My example project


More information about the NumPy-Tickets mailing list