[Python-Dev] Nondeterministic long-to-float coercion

["Martin v. Löwis"]

Raymond Hettinger schrieb:
> My colleague got an odd result today that is reproducible on his build
> of Python (RedHat's distribution of Py2.4.2) but not any other builds
> I've checked (including an Ubuntu Py2.4.2 built with a later version of
> GCC).  I hypothesized that this was a bug in the underlying GCC
> libraries, but the magnitude of the error is so large that that seems
> implausible.  Does anyone have a clue what is going-on?  

I'd say it's memory corruption. Look:

r=array.array("d",[-194000000.0, -193995904.0, -193994880.0]).tostring()
print map(ord,r[0:8])
print map(ord,r[8:16])
print map(ord,r[16:24])

gives

[0, 0, 0, 0, 105, 32, 167, 193]
[0, 0, 0, 0, 73, 32, 167, 193]
[0, 0, 0, 0, 65, 32, 167, 193]

It's only one byte that changes, and then that in only two bits (2**3
and 2**5). Could be faulty hardware, too.

Regards,
Martin