When I first saw this, I thought it was a bug in the 
PEP 264 patch.  Then I thought it was because 
PyEval_MergeCompilerFlags hadn't been taught about 
CO_FUTURE_DIVISION.  But I just repaired (and checked 
in) the latter, so that's not it either (although that 
was a cause of related bugs).  If it's not more 
obvious to you than it is to me, assign it back to me 
and I'll get back to it later.

C:\Code\python\PCbuild>python
Python 2.2a1+ (#21, Aug 17 2001, 17:16:14) [MSC 32 bit 
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for 
more information.
>>> from __future__ import division
>>> 1/2
0.5
>>> eval("1/2")
0
>>>

The same thing happens if you run it from file:

from __future__ import division
print 1/2
print eval("1/2")

That prints "0.5" and "0".

Logged In: YES 
user_id=6380

Looks to me like builtin_eval() doesn't do the necessary
dance to pass in any compiler flags -- it uses the old API,
PyRun_String().

Logged In: YES 
user_id=31435

Fixed and Closed (Guido's observation was spot on):

Lib/test/test_binop.py; new revision: 1.2
Python/bltinmodule.c; new revision: 2.229