TCP replication fails with mysterious exceptions if worker_replication_port is unset

[richvdh]

If you fail to configure worker_replication_port in a worker which is configured to use old-style TCP replication, you get an unhelpful stack trace:

2020-10-07 09:23:51,954 - synapse.replication.tcp.client - 71 - INFO - - Connecting to replication: IPv4Address(type='TCP', host='127.0.0.1', port=None)

... 

2020-10-07 09:23:51,959 - twisted - 231 - CRITICAL - - Unhandled Error
Traceback (most recent call last):
  File "/usr/lib/python3.8/site-packages/synapse/app/_base.py", line 116, in run
    run_command()
  File "/usr/lib/python3.8/site-packages/twisted/internet/base.py", line 1283, in run
    self.mainLoop()
  File "/usr/lib/python3.8/site-packages/twisted/internet/base.py", line 1292, in mainLoop
    self.runUntilCurrent()
  File "/usr/lib/python3.8/site-packages/synapse/metrics/__init__.py", line 517, in f
    ret = func(*args, **kwargs)
--- <exception caught here> ---
  File "/usr/lib/python3.8/site-packages/twisted/internet/base.py", line 913, in runUntilCurrent
    call.func(*call.args, **call.kw)
  File "/usr/lib/python3.8/site-packages/twisted/internet/tcp.py", line 453, in resolveAddress
    self._setRealAddress(self.addr)
  File "/usr/lib/python3.8/site-packages/twisted/internet/tcp.py", line 474, in _setRealAddress
    self.doConnect()
  File "/usr/lib/python3.8/site-packages/twisted/internet/tcp.py", line 603, in doConnect
    connectResult = self.socket.connect_ex(self.realAddress)
builtins.TypeError: an integer is required (got type NoneType)

(it seems there's also a bug in twisted in that the error is not passed into factory.clientConnectionFailed, which is why the stacktrace is unclear.)

[richvdh]

we should just get rid of this code: #11728

[dklimpel]

This is not supported anymore and can be closed?