registering a context twice results in keyerror

[tacaswell]

See this with threading client, but looks like it's root is in _hub and _commands

In [7]: ct = Context()

In [8]: ct.register()
DEBUG:caproto.Broadcaster:Serializing 1 commands into one datagram
DEBUG:caproto.Broadcaster:1 of 1 RepeaterRegisterRequest(client_address='127.0.0.1')
sending to ('172.17.255.255', 5065) b'\x00\x18\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x7f\x00\x00\x01'
DEBUG:caproto.Broadcaster:Parsed 1 commands from first datagram in queue.
DEBUG:caproto.Broadcaster:0 of 1: RepeaterConfirmResponse(repeater_address='172.17.0.1')
DEBUG:caproto.Broadcaster:Processing 1/1 commands left in datagram. 0 more datagrams are cached.
Registered with repeater

In [11]: ct.register()
DEBUG:caproto.Broadcaster:Serializing 1 commands into one datagram
DEBUG:caproto.Broadcaster:1 of 1 RepeaterRegisterRequest(client_address='127.0.0.1')
sending to ('172.17.255.255', 5065) b'\x00\x18\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x7f\x00\x00\x01'
Registered with repeater

In [12]: Exception in thread Thread-3:
Traceback (most recent call last):
  File "/home/tcaswell/.virtualenvs/dd36/lib/python3.6/threading.py", line 916, in _bootstrap_inner
    self.run()
  File "/home/tcaswell/.virtualenvs/dd36/lib/python3.6/threading.py", line 864, in run
    self._target(*self._args, **self._kwargs)
  File "/home/tcaswell/source/bnl_source/caproto/caproto/threading/client.py", line 41, in __call__
    self.target_obj.next_command(bytes_recv, address)
  File "/home/tcaswell/source/bnl_source/caproto/caproto/threading/client.py", line 158, in next_command
    command = self.broadcaster.next_command()
  File "/home/tcaswell/source/bnl_source/caproto/caproto/_hub.py", line 445, in next_command
    commands = read_datagram(byteslike, address, self.their_role)
  File "/home/tcaswell/source/bnl_source/caproto/caproto/_commands.py", line 262, in read_datagram
    _class = get_command_class(role, header)
  File "/home/tcaswell/source/bnl_source/caproto/caproto/_commands.py", line 248, in get_command_class
    _class = Commands[role][header.command]
KeyError: 24

[danielballan]

The RepeaterRegisterRequest is making its way back to the client itself, and the client doesn't know what to do with a RepeaterRegisterRequest (or a *Request in general). The correct behavior is to ignore. I'll fix this in read_datagram.

[tacaswell]

Interesting! Multicast / udp has some weird side effects.

Why do not we see this problem if you register two broadcasters with the same repeater?

[danielballan]

Maybe you do. Have we ever checked that case? Not as far as I'm aware. Unfortunately, tolerating this at the protocol level really dilutes the usefulness of the Broadcaster. Good thing that the interesting action happens over TCP.

[tacaswell]

I did check and it worked :-p

[tacaswell]

it may also be that our reapater is bouncing stuff it should not.

[danielballan]

We now understand that the repeater was occasionally receiving a registration request twice (because UDP) and "forwarding" the second request back to the requester. This was fixed by #69.