Hello,

The tests in test_lowlevel.py use the testserver.server.Server thread object.

When using the object as a context manager, we wait until the server thread is ready with a timeout.

Unfortunately, if the timeout is reached, we would not propagate the error. Instead, we would return from the __enter__ method and access the with body with an uninitialized value for port (i.e., port = 0).

Therefore, the tests would fail with:

E           urllib3.exceptions.NewConnectionError: <urllib3.connection.HTTPConnection object at 0x7fc3a3e30a90>: Failed to establish a new connection: [Errno 111] Connection refused

/usr/lib/python3/dist-packages/urllib3/connection.py:181: NewConnectionError

Reading this error does not make it obvious that this is the result of a timeout.

Fixed by raising an error in case of timeout, which makes it more obvious:

        if not self.ready_event.wait(self.WAIT_EVENT_TIMEOUT):
>           raise RuntimeError("Timeout waiting for server to be ready.")
E           RuntimeError: Timeout waiting for server to be ready.

tests/testserver/server.py:119: RuntimeError

The reason why I had a timeout on my end was because my environment was not resolving "localhost" as it should have. Requests were going through the DNS and therefore ended-up hitting the timeout

The name resolution taking forever was happening during this instruction (where self.host is localhost):

sock.bind((self.host, self.port))

Thanks,
Olivier