Running test_errno on my 32-bit Ubuntu Gutsy machine gives me this:

$ ./python Lib/test/test_errno.py
test_for_improper_attributes (__main__.ErrnoAttributeTests) ... FAIL
test_using_errorcode (__main__.ErrnoAttributeTests) ... ok
test_attributes_in_errorcode (__main__.ErrorcodeTests) ... ok

======================================================================
FAIL: test_for_improper_attributes (main.ErrnoAttributeTests)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "Lib/test/test_errno.py", line 46, in test_for_improper_attributes
    "%s is an unexpected error value" % attribute)
AssertionError: EREMOTEIO is an unexpected error value

Ran 3 tests in 0.001s

FAILED (failures=1)
Traceback (most recent call last):
  File "Lib/test/test_errno.py", line 68, in <module>
    test_main()
  File "Lib/test/test_errno.py", line 64, in test_main
    test_support.run_unittest(ErrnoAttributeTests, ErrorcodeTests)
  File "/home/andy/cvs/python/Lib/test/test_support.py", line 573, in
run_unittest
    _run_suite(suite)
  File "/home/andy/cvs/python/Lib/test/test_support.py", line 556, in
_run_suite
    raise TestFailed(err)
test.test_support.TestFailed: Traceback (most recent call last):
  File "Lib/test/test_errno.py", line 46, in test_for_improper_attributes
    "%s is an unexpected error value" % attribute)
AssertionError: EREMOTEIO is an unexpected error value

I've attached a patch which changes test_errno.py so that its list of
expected errors exactly matches the possible errors listed in
Modules/errnomodule.c in the latest SVN trunk.

I don't know whether this is the right solution, but the patch is there
if it is :)

Apologies if I've misunderstood something, or formatted the patch wrong
etc. This is my first Python patch.

Thanks for the patch, Andy, but I went ahead and fixed test_errno to
only explicitly test errno values from Standard C. That way we are not
constantly chasing our tail to support every errno value on every
platform that Python runs on.

Adding Brett Cannon since it looks like his checkin created the test
which fails on my machine. Apologies if this is very bad etiquette. I
couldn't find any guidelines about this in the developers' docs, but
probably that's because I am incompetent.

Woah! fast response, and what looks like a much more sensible fix.
Thanks Brett.