Reconnection fails if server holds connection

[3rd-Eden]

There's a missing feature in the reconnection implementation which is the ability to specify a timeout for the reconnect attempt. Currently it's possible that you attempt to connect to a server which never answers the requests causing the request to hang indefinitely.

The easiest way to reproduce this is to trigger a reconnection from the server side and remove all the upgrade listeners from the server and add an empty function which will handle the request. For example:

primus.on('connection', function (spark) {
  setTimeout(function () {
    primus.server.removeAllListeners('upgrade');
    primus.server.on('upgrade', function () {});
    spark.end(null, { reconnect: true });
  }, 200);
});

While nobody in their sane minds would do something like the code above it's certainly possible that load balancers or proxies buffer the connection while an application is dead or restarting.

Originally reported by @developerdizzle which ran into these issues at Cloud9

[lpinca]

I think that this is an issue of the proxy / load balancer.
That said, adding a timeout to our reconnection attempt is not a bad idea.

[3rd-Eden]

@lpinca Yes, it's definitely an edge case, but still something we should support.

[3rd-Eden]

FYI: Wrote a new module to fix this behavior: http://npmjs.org/package/recovery . So this issue will be fixed once the browserify branch is completed and merged in to master (as the browserify branch implements this new reconnect module)

[3rd-Eden]

Resolved in master