[Numpy-tickets] [NumPy] #378: NumPy or Python crashing with segfault
NumPy
numpy-tickets at scipy.net
Thu Nov 16 18:28:37 CST 2006
#378: NumPy or Python crashing with segfault
--------------------------------+-------------------------------------------
Reporter: tecki | Owner: somebody
Type: defect | Status: closed
Priority: normal | Milestone:
Component: numpy.core | Version:
Severity: normal | Resolution: fixed
Keywords: segfault vectorize |
--------------------------------+-------------------------------------------
Changes (by oliphant):
* status: reopened => closed
* resolution: => fixed
Comment:
Thanks for this detective work. Getting the reference counting of OBJECT
arrays right in the buffered section of the ufunc loop is certainly
challenging. I thought I had accounted for the cases. Look in
numpy/numpy/doc/ufunc.txt for some of the explanation.
At any rate. Using Py_CLEAR certaintly doesn't hurt as the object in the
casting buffer should not
be re-used anyway and it seems to help. Perhaps what is happening, is
that when the buffer is re-used it is not filling up and a new decref is
being performed on an "un-copied" object.
So, I've accepted and applied the changes. I changed both XDECREF's in
that section of code to CLEAR.
Thanks very much,
Travis
--
Ticket URL: <http://projects.scipy.org/scipy/numpy/ticket/378#comment:13>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.
More information about the Numpy-tickets
mailing list