[SciPy-Dev] Two potential 64 bit issues

Christoph Gohlke cgohlke@uci....
Fri Jul 16 14:39:58 CDT 2010


I came across two potential 64 bit issues in scipy 0.8 while trying to 
track down an elusive scipy.test failure on win-amd64 
Unfortunately none of these two issues turns out to be related to any 
test failure so I report them in a new thread.

The first issue is described in 
In scipy/integrate/__quadpack.h a npy_intp variable is parsed as int in 
the PyArg_ParseTuple function. Shouldn't it be parsed as Py_ssize_t? 
Unfortunately that would work only on Python 2.5+. Are there any plans 
to drop support for Python 2.4?

Second, in scipy/sparse/linalg/dsolve/_superlu_utils.c a memory pointer 
is first cast to a long and then stored as PyInt, which value is later 
passed to free(). At least the cast to long will cause trouble on 
win-amd64 where long is 32 bit. I opened a ticket at 

Thank you.


More information about the SciPy-Dev mailing list