Mock's assertion failures can be extremely hard to read for a few reasons -- mostly the noisy default repr that mock objects have, but also because they don't give you the hints that `unittest.TestCase`'s `assert*` methods give you (things like diffing two lists for example).

unittest.mock.Mock's `assert*` methods could hook into the TestCase's assertion methods if Mock either gained a MockInTestCase subclass or started taking an arg to `__init__` that was an instance of `TestCase`, so that `assert*` could then use the assertion methods on the instance.

#11664 could (should) then obviously use this argument by default when patching.

Another added advantage would be that the raised exception could then be `TestCase.failureException`, whatever that might be, rather than `AssertionError`, though I doubt that's that big a deal since that's usually a subclass of `AssertionError` I bet.

Yep, interesting idea. In Python 3 you can't import unittest.mock without importing unittest, so there could even be a "default testcase" to fall back on. For the external release I'd rather not have unittest as a dependency, but allowing a testcase to be provided at instantiation would be fine.

The only issue is that the testcase would have to be propagated to child mocks too.

See also https://bugs.python.org/issue28054