NatsConnection.close unnecessarily sleeps for one second

[brettkail-wk]

This call to NatsConnectionReader.stop simply sets a flag, but it does not actually interrupt the blocking read, which means that this call to await the stop always times out. Commit bb8968b recently removed the logging that happens when the await fails, but the underlying issue is still present. It seems like these calls to .get should be removed, and these these calls should be updated to use readStop.get and writeStop.get.

[sasbury]

Haven't dug into logic, but wanted to make a quick note, in case someone else digs in, but we shouldnt' use get(), there should always be a timeout on these. That said, if we are always reaching the timeout, that is bad.

[brettkail-wk]

@sasbury The code does use get-with-timeout (twice), but yes, the readStop.get(1, TimeUnit.SECONDS) always times out. Sorry for confusion; I've adjusted the description slightly.

[jglink]

Any news on this issue?
Is there an estimation on when this will be fixed?
Thanks

[scottf]

The read call is blocking. We can actually interrupt the read by telling the underlying socket to shutdown. Is this the desired behavior to basically forcibly stop the read? Here is a PR to review: #427

[scottf]

Closed in #427