New jinja2 pinning makes it impossible to build our codebase

[kramer65]

This isn't a bug in the codebase of locust per se, but rather a new jinja2 pinning here which makes it impossible to use locust in our codebase in which a different library requires a version of jinja2 higher than 3.1.0.

Is there a way that we can get rid of this pinning in locust? Or is there any known workaround?

[cyberw]

The reason I pinned it was because it is not backwards compatible and broke Locust. (I couldn't bother myself to figure out exactly how :)

If you can make Locust compatible with 3.1 then we can remove the pinning. This is probably a good starting point https://jinja.palletsprojects.com/en/3.1.x/changes/#version-3-1-0

https://github.com/locustio/locust/runs/5693857807?check_suite_focus=true

[Apteryks]

For what it's worth, unpinning Jinja2 to use version 3.1.1 in GNU Guix on Python 3.9.9 and Locust 2.8.6, I had only one new test failure, which looks benign and unrelated:

=================================== FAILURES ===================================
________________ TestMasterWorkerRunners.test_distributed_shape ________________

self = <locust.test.test_runners.TestMasterWorkerRunners testMethod=test_distributed_shape>

    def test_distributed_shape(self):
        """
        Full integration test that starts both a MasterRunner and three WorkerRunner instances
        and tests a basic LoadTestShape with scaling up and down users
        """
    
        class TestUser(User):
            @task
            def my_task(self):
                pass
    
        class TestShape(LoadTestShape):
            def tick(self):
                run_time = self.get_run_time()
                if run_time < 2:
                    return 9, 9
                elif run_time < 4:
                    return 21, 21
                elif run_time < 6:
                    return 3, 21
                else:
                    return None
    
        with mock.patch("locust.runners.WORKER_REPORT_INTERVAL", new=0.3):
            test_shape = TestShape()
            master_env = Environment(user_classes=[TestUser], shape_class=test_shape)
            master_env.shape_class.reset_time()
            master = master_env.create_master_runner("*", 0)
    
            workers = []
            for i in range(3):
                worker_env = Environment(user_classes=[TestUser])
                worker = worker_env.create_worker_runner("127.0.0.1", master.server.port)
                workers.append(worker)
    
            # Give workers time to connect
            sleep(0.1)
    
            # Start a shape test
            master.start_shape()
            sleep(1)
    
            # Ensure workers have connected and started the correct amount of users
            for worker in workers:
                self.assertEqual(3, worker.user_count, "Shape test has not reached stage 1")
                self.assertEqual(
                    9, test_shape.get_current_user_count(), "Shape is not seeing stage 1 runner user count correctly"
                )
            self.assertDictEqual(master.reported_user_classes_count, {"TestUser": 9})
    
            # Ensure new stage with more users has been reached
            sleep(2)
            for worker in workers:
                self.assertEqual(7, worker.user_count, "Shape test has not reached stage 2")
                self.assertEqual(
                    21, test_shape.get_current_user_count(), "Shape is not seeing stage 2 runner user count correctly"
                )
            self.assertDictEqual(master.reported_user_classes_count, {"TestUser": 21})
    
            # Ensure new stage with less users has been reached
            sleep(2)
            for worker in workers:
                self.assertEqual(1, worker.user_count, "Shape test has not reached stage 3")
                self.assertEqual(
                    3, test_shape.get_current_user_count(), "Shape is not seeing stage 3 runner user count correctly"
                )
            self.assertDictEqual(master.reported_user_classes_count, {"TestUser": 3})
    
            # Ensure test stops at the end
            sleep(2)
            for worker in workers:
                self.assertEqual(0, worker.user_count, "Shape test has not stopped")
                self.assertEqual(
                    0, test_shape.get_current_user_count(), "Shape is not seeing stopped runner user count correctly"
                )
            self.assertDictEqual(master.reported_user_classes_count, {"TestUser": 0})
    
>           self.assertEqual("stopped", master.state)
E           AssertionError: 'stopped' != 'stopping'
E           - stopped
E           + stopping

locust/test/test_runners.py:1121: AssertionError
[...]
=============================== warnings summary ===============================
../../../gnu/store/l4nsxlz4hlxabj97vaylj1mwlvcwm4sy-python-pyquery-1.2.17/lib/python3.9/site-packages/pyquery/pyquery.py:50: 15 warnings
  /gnu/store/l4nsxlz4hlxabj97vaylj1mwlvcwm4sy-python-pyquery-1.2.17/lib/python3.9/site-packages/pyquery/pyquery.py:50: DeprecationWarning: inspect.getargspec() is deprecated since Python 3.0, use inspect.signature() or inspect.getfullargspec()
    name, inspect.getargspec(func).defaults)

-- Docs: https://docs.pytest.org/en/stable/warnings.html
=========================== short test summary info ============================
FAILED locust/test/test_runners.py::TestMasterWorkerRunners::test_distributed_shape
= 1 failed, 434 passed, 1 skipped, 8 deselected, 15 warnings in 349.99s (0:05:49) =
error: in phase 'check': uncaught exception:
%exception #<&invoke-error program: "python" arguments: ("-m" "pytest" "locust" "-k" "not test_default_headless_spawn_options and not test_default_headless_spawn_options_with_shape and not test_headless_spawn_options_wo_run_time and not test_html_report_option and not test_web_options and not test_skip_logging and not test_custom_exit_code and not test_webserver") exit-status: 1 term-signal: #f stop-signal: #f>

[cyberw]

Yes, that issue doesnt look very serious. But it doesnt appear that jinja2 has had any relevant changes since 3.1.0, so its unlikely to be fixed. If somene makes a PR unpinning it and it works I'll be happy to look at it though.