test_repr of test.test_functools.TestPartialC fails sometimes because the test scenario depends on the order of items in a Python dict. The order is not stable and may change. Before PEP 456 was committed the order of items was stable most of the time. Nowadays the random seed has an impact on the order.

http://buildbot.python.org/all/builders/x86%20XP-4%203.x/builds/9609

======================================================================
FAIL: test_repr (test.test_functools.TestPartialC)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "D:\cygwin\home\db3l\buildarea\3.x.bolen-windows\build\lib\test\test_functools.py", line 174, in test_repr
    repr(f))
AssertionError: 'func[43 chars]D8>, b=<object object at 0x09D08B48>, a=<objec[20 chars]28>)' != 'func[43 chars]D8>, a=<object object at 0x09D08B28>, b=<objec[20 chars]48>)'
- functools.partial(<function capture at 0x0A2A72D8>, b=<object object at 0x09D08B48>, a=<object object at 0x09D08B28>)
+ functools.partial(<function capture at 0x0A2A72D8>, a=<object object at 0x09D08B28>, b=<object object at 0x09D08B48>)


======================================================================
FAIL: test_repr (test.test_functools.TestPartialCSubclass)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "D:\cygwin\home\db3l\buildarea\3.x.bolen-windows\build\lib\test\test_functools.py", line 174, in test_repr
    repr(f))
AssertionError: 'Part[41 chars]D8>, b=<object object at 0x09D08DC8>, a=<objec[20 chars]A8>)' != 'Part[41 chars]D8>, a=<object object at 0x09D08BA8>, b=<objec[20 chars]C8>)'
- PartialSubclass(<function capture at 0x0A2A72D8>, b=<object object at 0x09D08DC8>, a=<object object at 0x09D08BA8>)
+ PartialSubclass(<function capture at 0x0A2A72D8>, a=<object object at 0x09D08BA8>, b=<object object at 0x09D08DC8>)

The issue is causing flaky build bots and should be resolved soonish. I don't know what's the best way to fix the problem. Either the test needs to change or partial_repr() must print the keys in sorted order.

The repr test of userdict fails sometimes, too.

http://buildbot.python.org/all/builders/AMD64%20Ubuntu%20LTS%203.x/builds/3120/steps/test/logs/stdio

======================================================================
FAIL: test_all (test.test_userdict.UserDictTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/python/3.x.langa-ubuntu/build/Lib/test/test_userdict.py", line 48, in test_all
    self.assertEqual(repr(u2), repr(d2))
AssertionError: "{'one': 1, 'two': 2}" != "{'two': 2, 'one': 1}"
- {'one': 1, 'two': 2}
+ {'two': 2, 'one': 1}

See issue #19664 for userdict.

Here is a patch for test_functools.

permutation() looks overkill to me for such trivial test.

Here is a simpler approach: remove the second keyword :-) test_functools_repr.patch

I'm for Victors patch. Let's not get too fancy with tests.

I'm pretty sure the second keyword is there to ensure the repr includes the
appropriate commas. Can we just make a suitable dict and check the key
order?

And actually re-reading the existing code to see it already works that
way... how can two dicts with the same keys, without any extra keys being
added and deleted in either have different iteration orders? If it's just a
matter of copying the original dict changing the order, then we should be
able to restore consistency by copying it when building the expected repr
as well.

Ok, here is a different patch sorting keywords. So it still checks how the C code joins keyword parameters.

Both keys happen to have the same LSB bytes. I have added "print(hash('a') & 7, hash('b') & 7)" to the code and run test tests multiple times:

$ ./python -m test -m test_repr test_functools
[1/1] test_functools
3 6
3 6
1 test OK.
$ ./python -m test -m test_repr test_functools
[1/1] test_functools
2 2
2 2
test test_functools failed -- multiple errors occurred; run in verbose mode for details
1 test failed:
    test_functools

New changeset 18a44d65d34a by Christian Heimes in branch 'default':
Issue #19681: Apply a quick and minimal band-aid.
http://hg.python.org/cpython/rev/18a44d65d34a

What is the status of this issue?

It still needs a proper fix, though.

If you don't like permutation() (which already imported in this file), here is a patch which uses explicit hardcoded permutations.

The latest patch looks okay to my eye, can somebody do a formal commit review please.

Ping.

issue19681_2.patch LGTM.

New changeset 138c2afd9f1f by Serhiy Storchaka in branch '3.4':
Issue #19681: Test the repr of partial with more than one keyword argument.
https://hg.python.org/cpython/rev/138c2afd9f1f

New changeset 6f80e8bcb5ad by Serhiy Storchaka in branch 'default':
Issue #19681: Test the repr of partial with more than one keyword argument.
https://hg.python.org/cpython/rev/6f80e8bcb5ad