With the changes introduced for bpo-4591, test_grp and test_pwd now fail on OS X systems when Python is run in 32-bit mode.

Previously, using a 64-bit/32-bit universal build of 2.7.3:

$ arch -i386 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nobody").gr_gid; print(g,type(g))'
(-2, <type 'int'>)
$ arch -x86_64 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nobody").gr_gid; print(g,type(g))'
(4294967294, <type 'int'>)

Now, with 2.7.4rc1:

$ arch -i386 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nobody").gr_gid; print(g,type(g))'
(4294967294L, <type 'long'>)
$ arch -x86_64 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nobody").gr_gid; print(g,type(g))'
(4294967294, <type 'int'>)

This results in test failures in both test_grp and test_pwd:

======================================================================
FAIL: test_values (test.test_grp.GroupDatabaseTestCase)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/test/test_grp.py", line 27, in test_values
    self.check_value(e)
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/test/test_grp.py", line 19, in check_value
    self.assertIsInstance(value.gr_gid, int)
AssertionError: 4294967294L is not an instance of <type 'int'>

----------------------------------------------------------------------

======================================================================
FAIL: test_values (test.test_pwd.PwdTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/test/test_pwd.py", line 23, in test_values
    self.assertIsInstance(e.pw_gid, int)
AssertionError: 4294967294L is not an instance of <type 'int'>

Failure also seen on 32-bit buildbot:
http://buildbot.python.org/all/builders/x86%20Tiger%202.7

New changeset 2aa817e0a645 by Benjamin Peterson in branch '2.7':
return int instead long when possible (bpo-17531)
http://hg.python.org/cpython/rev/2aa817e0a645

I hope that does it?

No, that was wrong, sorry.

I think the tests are wrong.

New changeset c982393bea4e by Benjamin Peterson in branch '2.7':
group ids and user ids can be longs now (bpo-17531)
http://hg.python.org/cpython/rev/c982393bea4e

That fixes the test for nobody (-2). Now the next part of the test fails more seriously for nogroup (-1). Simplifying:

2.7.3
$ arch -i386 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nogroup").gr_gid; print(grp.getgrgid(g),type(g))'
(grp.struct_group(gr_name='nogroup', gr_passwd='*', gr_gid=-1, gr_mem=[]), <type 'int'>)
$ arch -x86_64 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nogroup").gr_gid; print(grp.getgrgid(g),type(g))'
(grp.struct_group(gr_name='nogroup', gr_passwd='*', gr_gid=4294967295, gr_mem=[]), <type 'int'>)

2.7.4rc1:
$ arch -i386 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nogroup").gr_gid; print(grp.getgrgid(g),type(g))'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
OverflowError: group id is greater than maximum
$ arch -x86_64 /usr/local/bin/python2.7 -c 'import grp; g=grp.getgrnam("nogroup").gr_gid; print(grp.getgrgid(g),type(g))'
(grp.struct_group(gr_name='nogroup', gr_passwd='*', gr_gid=4294967295, gr_mem=[]), <type 'int'>)

Whack, whack. a4dbe53577cb

a4dbe53577cb appears to fix the latter problem. Thanks!

New changeset 87d266988905 by Benjamin Peterson in branch '2.7':
update NEWS for bpo-17531
http://hg.python.org/cpython/rev/87d266988905