Switch to DiscoverRunner from DjangoTestSuiteRunner

(imported from commit 77dd694b20509bfdc7327186e0524d686f0415f7)

tools/lint-all

@@ -145,8 +145,6 @@ def check_pyflakes():
                      'redefinition of unused' in ln or
                      ("zerver/models.py" in ln and
                       " undefined name 'bugdown'" in ln) or
-                     ("zerver/tests.py" in ln and
-                      " unable to detect undefined names" in ln) or
                      ("zerver/lib/tornado_ioloop_logging.py" in ln and
                       "redefinition of function 'instrument_tornado_ioloop'" in ln) or
                      ("zephyr_mirror_backend.py:" in ln and

zerver/lib/test_runner.py

@@ -1,4 +1,4 @@
-from django.test.simple import DjangoTestSuiteRunner
+from django.test.runner import DiscoverRunner
 
 from zerver.lib.cache import bounce_key_prefix_for_testing
 
@@ -72,11 +72,11 @@ def run_test(test):
 
     test._post_teardown()
 
-class Runner(DjangoTestSuiteRunner):
+class Runner(DiscoverRunner):
     option_list = ()
 
     def __init__(self, *args, **kwargs):
-        DjangoTestSuiteRunner.__init__(self, *args, **kwargs)
+        DiscoverRunner.__init__(self, *args, **kwargs)
 
     def run_suite(self, suite):
         for test in suite:

zerver/tests.py

@@ -1,11 +1,6 @@
 # -*- coding: utf-8 -*-
 from __future__ import absolute_import
 
-# This is less than ideal, but I couldn't figure out Django's extremely
-# opaque test runner.  We want to split out tests into different files,
-# but it would be even better if we could run the files individually.
-from zerver.test_hooks import * # noqa
-
 from django.test import TestCase
 from django.utils.timezone import now
 from django.core.exceptions import ValidationError