Hi,
we have found that packaging evaluates some comparisons unreasonably / illogically.

I'll use this format:

(version, operator, version, comparison_evaluation),

For these comparisons, packaging ignores equality and evaluates as if the operator was simply <:

('2.4.8', '<=', '2.4.8.*', False),
('2.4.8.0', '<=', '2.4.8.*', False),
('2.4.8.1', '<=', '2.4.8.*', False),
('2.4.8.post1', '<=', '2.4.8.*', False),

And in this set, also the > operator is evaluated wrong:

('2.4.8', '>', '2.4.8.*', True),
('2.4.8.0', '>', '2.4.8.*', True),
('2.4.8.1', '>', '2.4.8.*', True),
('2.4.8b5', '>', '2.4.8.*', True),
('2.4.8.post1', '>', '2.4.8.*', True),

Both these comparisons are undefined according to PEP 440. However, as someone will surely be using them, tests have been added for them in the Python-to-RPM-version project pyreq2rpm, see PR gordonmessmer/pyreq2rpm#7.

The tests currently verify that the comparisons are consistent with this illlogical upstream behaviour. But of course, an even better solution would be to fix the behaviour.