Closed issue 1924 as duplicate of this one, but I'm copying here the
text from David, as it's very explanative:

"""
I ran across this bug in some legacy production code when numbers got high:

>>> '%i' % 2e9
'2000000000'
>>> '%i' % 3e9
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: int argument required

It looks like the float is being automatically converted to an int, but
floats > sys.maxint cause an error.  However,

>>> int(3e9)
3000000000L

So the implicit float-to-int conversion is imperfect; large floats are
not being converted to long ints.

Same error in Python 2.3 through 2.6a0 (as of 2007-12-28).

In Python 2.1.3 & 2.2.3 the error is "OverflowError: float too large to
convert".  The same error is triggered by int(3e9) though.

While it's arguably not-quite-sane to have code that triggers this
error, the inconsistency is what concerns me.
"""