avoids a zkrb_iterate_event_loop which can be relatively expensive
Make sure that if a fork has happened, closing the child's connection won't cause a problem in the parent, even if the user doesnt' close the connection on purpose.
suggested by Martin Kou on the zookeeper mailing list http://is.gd/5JxVnQ Before the child calls zookeeper_close, close the socket behind its back, I'm not fully certain why this makes a difference, but i assume it's that in the child, some of the "shutdown chat" succeeds with the server, and causes funky state with the parent connection. "What good is a phone call if you can't...speak?"
…when attention is needed
provide a sync api for the async backend blacklist this branch travis
forgot to *use* the attr_accessor after I changed things...
…s branch Looks like a rewrite with the 'st' library is the only way to ensure fork safety