Number of test failures is calculated incorrectly

[Roman Podoliaka]

Consider the following test suite:

import unittest

class Test(unittest.TestCase):
    def test_case1(self):
        self.assertTrue(3 > 1) # success

    def test_case2(self):
        self.assertTrue(5 > 1) # success

    def test_case3(self):
        self.assertTrue(2 < 1) # fail

    def test_case4(self):
        self.assertTrue(1 > 2) # fail

    def test_case5(self):
        self.assertTrue(0 > 2) # fail

.testr.conf:

[DEFAULT]
test_command=python -m subunit.run discover . $LISTOPT $IDOPTION
test_id_option=--load-list $IDFILE
test_list_option=--list

Running of "testr run" shows 4 failures instead of 3.
Running of "testr run --parallel" (with 4 CPU cores) shows either 5 or 6 failures instead of 3.
Running of "testr run --parallel --concurrency 1" shows 4 failures instead of 3.
Running of "testr run --parallel --concurrency 2" shows 5 failures instead of 3.
 ...

It seems, that the number of failures reported is equal to the actual number of test failures + the number of testr subprocesses running at least one failing test case.