Race condition in example?

[hammingweight]

I think that there might be a (low probability) race condition in your example in the interaction between handle_info/3 and the worker/3 function. If the commands execute in the order indicated by the numbers below, the wrong process owns the socket after sending the response data.

handle_info({tcp, Sock, Data}, State) ->

1. Me = self(),
2. P = spawn(fun() -> worker(Me, Sock, Data) end),
3. gen_tcp:controlling_process(Sock, P),
4. {noreply, State};

worker(Owner, Sock, Data) ->
3 . gen_tcp:send(Sock, Data),
4 . inet:setopts(Sock, [{active, once}]),
5. gen_tcp:controlling_process(Sock, Owner).

I simulated the issue by inserting a timer:sleep(1000) between instructions 2 and 6.