since it's a simple one, I thought maybe a first step is to describe the bug
here. Then maybe someone will direct me to the right channel.
So, numpy appears not to correctly compute bitwise_and.reduce and
bitwise_or.reduce : instead of reducing over the complete axis, these methods
only take the extremities into account. Illustration :
from numpy import *
bitwise_or.reduce(array([8,256,32,8])) 8 import numpy
numpy.__version__ '0.9.8'
Platform : Win XP SP2, Python 2.4.2
HTH, bb