Message11172

I would argue:

1. If we don't expect the test to pass under standard conditions as written, then we should fix the test.

2. If we believe the test demands no more than users can legitimately expect from Jython, then we should fix the behaviour (on all platforms).

I tend to think 2 applies in this case. A particular expectation (2000) is acknowledged in _marshal.java, and I think the test should match it (roughly as the comment says, but not marshal.c).

This looks relevant: http://bugs.python.org/issue22734, but I suspect the motivation is specific to C-on-Windows, not Windows per se. With Java I think we can expect uniformity cross-platform.