Closed
Description
I wrote a test class for node, using yield-base tests, and it does not behave as expected with py-test 2.9.2.
Consider the following class:
class TestYield(object):
SETUP = 0
def setup(self):
print "setup %r" % self
self.SETUP += 1
assert self.SETUP == 1
def teardown(self):
print "teardown %r" % self
assert 0 <= self.SETUP <= 1 # setup has not been used more than once
self.SETUP = 0
def test_a(self):
print "test_a %r" % self
assert self.SETUP == 1 #check that class is setup
self.SETUP -= 1 # "use" the setup
def test_b(self):
for i in range(3):
print "test_b %r yields %r" % (self, i)
yield self._do_test_b, i
def _do_test_b(self, val):
print "_do_test_b %r %r" % (self, val)
assert self.SETUP == 1 # check that class is setup
self.SETUP -= 1 # "use" the setupWhen I run it with nosetests -qs, all four tests (1test_a and 3 test_b's) pass. For each of them, setup and teardown are called, as can be seen from the output:
setup <test_yield.TestYield object at 0x7f08de452110>
test_a <test_yield.TestYield object at 0x7f08de452110>
teardown <test_yield.TestYield object at 0x7f08de452110>
test_b <test_yield.TestYield object at 0x7f08de4520d0> yields 0
setup <test_yield.TestYield object at 0x7f08de452a50>
_do_test_b <test_yield.TestYield object at 0x7f08de452a50> 0
teardown <test_yield.TestYield object at 0x7f08de452a50>
test_b <test_yield.TestYield object at 0x7f08de4520d0> yields 1
setup <test_yield.TestYield object at 0x7f08de452bd0>
_do_test_b <test_yield.TestYield object at 0x7f08de452bd0> 1
teardown <test_yield.TestYield object at 0x7f08de452bd0>
test_b <test_yield.TestYield object at 0x7f08de4520d0> yields 2
setup <test_yield.TestYield object at 0x7f08de452d10>
_do_test_b <test_yield.TestYield object at 0x7f08de452d10> 2
teardown <test_yield.TestYield object at 0x7f08de452d10>
----------------------------------------------------------------------
Ran 4 tests in 0.001s
When I run the tests with py.test -qs, only test_a passes, all 3 test_b's fail on the assert in _do_test_b. The output is:
test_b <test_yield.TestYield object at 0x7f6df825b410> yields 0
test_b <test_yield.TestYield object at 0x7f6df825b410> yields 1
test_b <test_yield.TestYield object at 0x7f6df825b410> yields 2
setup <test_yield.TestYield object at 0x7f6df825b8d0>
test_a <test_yield.TestYield object at 0x7f6df825b8d0>
.teardown <test_yield.TestYield object at 0x7f6df825b8d0>
setup <test_yield.TestYield object at 0x7f6df825b8d0>
_do_test_b <test_yield.TestYield object at 0x7f6df825b410> 0
F_do_test_b <test_yield.TestYield object at 0x7f6df825b410> 1
F_do_test_b <test_yield.TestYield object at 0x7f6df825b410> 2
F
(...)
We can see that the three test_b's are run on the same instance, which is not consistent with Nose. Worse, setup is not even called on that instance, instead it is called on the instance that was previously used to run test_a, which does not make much sense... Finally, teardown is never called for test_b.