I believe this is may be because of a couple of problems:
The messages returned from the API are ordered by their 'created'
attribute, but the messages created from the fixtures had 'created'
set to None. This means they may have been returned in an arbitrary
order.
The list of public objects used to compare against was not
explicitly ordered, so the order these are returned would have been
arbitrary (well, database-specific anyway) as well.
This commit fixes both of those issues, so that hopefully these
transient test failures won't happen any more.
Coverage remained the same at 98.45% when pulling 6332936dc896abc986cd0126ae4b567fa12c0598 on fix-intermittent-test-failure into 1a0c81a4e78d8e9dfe84916dc0d850fa8264ec29 on master.
We would occasionally get failures from the test:
nuntium.tests.api.page_paginator_test:PagePaginationTestCase.test_get_paginated
I believe this is may be because of a couple of problems:
This commit fixes both of those issues, so that hopefully these transient test failures won't happen any more.