Wrong outcome when test fails

[DavideCanton]

It seems that the plugin doesn't log the test outcome but the teardown one.

For example, this file:

import pytest


@pytest.mark.parametrize("n", [1, 2, 3])
def test_name(n):
    assert n > 1

when run with pytest --replay-record-dir ., generates a file containing

{"nodeid": "tests/test_foo.py::test_name[1]", "start": 0.03358029099990745}
{"nodeid": "tests/test_foo.py::test_name[1]", "start": 0.03358029099990745, "finish": 0.03893194400006905, "outcome": "passed"}
{"nodeid": "tests/test_foo.py::test_name[2]", "start": 0.039133442000093055}
{"nodeid": "tests/test_foo.py::test_name[2]", "start": 0.039133442000093055, "finish": 0.039566410000134056, "outcome": "passed"}
{"nodeid": "tests/test_foo.py::test_name[3]", "start": 0.039726459000121395}
{"nodeid": "tests/test_foo.py::test_name[3]", "start": 0.039726459000121395, "finish": 0.04012836599986258, "outcome": "passed"}

while the test output is:

$ pytest --replay-record-dir .
# other stuff
FAILED tests/test_foo.py::test_name[1] - assert 1 > 1
============================================ 1 failed, 2 passed in 0.06s =============================================

The problem seems to be in the pytest_runtest_makereport hook, that logs the outcome of the "teardown" phase, that is success, while the test outcome is in the "call" phase.

I'm opening a PR for this soon.