As discussed in http://bugs.python.org/issue13703, Python will eventually start
using a randomized hash seed, which affects the iteration order of things like
dicts and sets. Admittedly, this won't be enabled by default until Python 3.3,
but the functionality can be turned on selectively all the way back to Python
2.6. When it IS turned on, it breaks a number of unit tests in the transitfeed
library that make assumptions about the order of validation exceptions. The
order of the exceptions can be a function of underlying dictionary iteration
order, so the test end up being non-deterministic.
I happen to run the transitfeed library in a 2.7 environment where randomized
hash seeding has been enabled by default, so I'd like to get these tests fixed.
To do so, I will introduce a strict ordering of validation exceptions in unit
tests that will be deterministic even if the original order of the exceptions
is not.
Original issue reported on code.google.com by bdfer...@google.com on 16 Nov 2012 at 2:57
Original issue reported on code.google.com by
bdfer...@google.com
on 16 Nov 2012 at 2:57