Scott Snyder recently reported a bug in the pmath
module for psyco. Since that code appears to be derived
from the pfloatobject code, I checked that and sure
enought the bugs exists there as well.
The bug can be tickled by running the following little
snippet with Psyco compiled with DEBUG=1 (shamelessly
stolen from Scott's test case):
from math import pow
#~ def foo(etasize):
#~ pow (1.0, etasize)
return 5.0 + etasize
I'm attaching a patch that appears to fix this.