Hi there!

I was looking at the packages and decided to run the tests (in race mode) in each. Unfortunately I didn't make it past the router tests before hitting a race condition and a separate deadlock.

The race occurred between:

and

I didn't look into the deadlock, but it occurred during TestLoad, I had to kill the test after 7 minutes.

=== RUN TestLoad
2019/02/27 11:00:31 Testing load
2019/02/27 11:00:31 Publisher: Starting publisher
2019/02/27 11:00:31 Publisher: Publisher ready.
2019/02/27 11:00:31 ERROR starting publisher listener: listen tcp :60010: bind: address already in use
2019/02/27 11:00:31 Starting distribution Channel
2019/02/27 11:00:31 Router: starting router
2019/02/27 11:00:31 Reciever: staring reciever
2019/02/27 11:00:31 Starting distribution Channel
2019/02/27 11:00:31 ERROR starting publisher listener: listen tcp :60011: bind: address already in use
2019/02/27 11:00:31 Starting subscription Manager.
2019/02/27 11:00:31 Reciever: reciever ready
2019/02/27 11:00:31 Reciever: Starting connection with localhost:60010
2019/02/27 11:00:31 Starting subscription Manager.
2019/02/27 11:00:31 Subscribed to localhost:60011.
2019/02/27 11:00:31 Starting listener
2019/02/27 11:00:31 Receiver: subscription to localhost:60010 added
2019/02/27 11:00:31 Subscribed to localhost:60010.
2019/02/27 11:00:31 Starting listener
^Csignal: interrupt
FAIL github.com/xuther/go-message-router/router 436.926s

This may or may not be the cause of this deadlock, but I suspect passing the sync.WaitGroup by value is causing the issue. As the vet tool points out, passing sync types by value results in misuse as the internal data structure is copied and the atomic operations for synchronization no longer work (each goroutine essentially has it's own copy of the WaitGroup, defeating the purpose of the WaitGroup).