Issue1628469

PyString
  __int__: don't overflow, return a long instead

PyFloat
  _pow: ValueError when raising number to fractional power
  float___int__: don't overflow, return a long instead

__builtin__
  ord: TypeError if not a string or a string longer than 1
  unichr: ValueError when not 0 < len < 65536
  chr: ValueError when not 0 < len < 256
  hex: TypeError when no conversion possible
  min: TypeError when zero args
  vars: TypeError when arg has no __dict__

Nice. The attached patch also includes binascii updates but will skip that part assuming its  the same as (or superceded by) patch 1626844. I'll look at applying the rest to 2.3.

Some of these could probably also be applied to 2.2/trunk, but lets return to that after.

Thanks!
- kz

About the chr/unichr changes, as far as i know the jython strategy has been for all strings to be unicode, so this difference from cpython is intentional. There is even a jython bugtest (test073) that checks that chr() will accept larger values. Also mentioned here:

  http://jython.org/docs/differences.html

Unless i've misunderstood something, i'm inclinded to stick with historical jython behaviour in this case. The rest of the patch is looking okay to me so far.

ah, didn't know about the different chr semantics.  sorry about that.

and, yeah, the binascii stuff shouldn't have been part of this patch, my bad, again.

Commited to 2.3 in r3051, less the chr/unichr & binascii changes as mentioned. Didn't notice any regressions in the test suites, and several additional tests from test_builtin now pass.

Will drop a few of these into 2.2. I believe int->long was a change of behaviour in cpython from 2.2 to 2.3, but the others are probably good, have to check...

- kz (of the agonizingly-slow-patch-review-society ;-)

applied float _pow, __builtin__ hex and vars fixes to 2.2 in r3203.

- kz