fix cluster test for disconnect with open connections

[sam-github]

fix#1933
replaces #1934

passed all tests except arm7 and raspberry pi 1 and 2, they are still ongoing

@Olegas@trevnorris PTAL

note that I think there are ALSO bugs in cluster message passing... this test attempts to evade them by waiting for the socket to be definitively open... by waiting for data from the worker/server

[sam-github]

last test: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/814/

re-test after rebase: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/815/

[jbergstroem]

LGTM. Verified on a machine I could reproduce on prior to patch.

[shigeki]

Timeout error of test-cluster-worker-wait-server-close.js on my FreeBSD machine was resolved with this patch.

[rvagg]

LGTM

[sam-github]

Landed in a5987d4d5e21be04a6239579634fa8b19446d5ef

[jbergstroem]

@sam-github I can't find this commit in the master branch. Github doesn't either. Confuserating.

[jbergstroem]

$ git fetch -q iojs && git branch -a --contains a5987d4d5e21be04a6239579634fa8b19446d5ef error: no such commit a5987d4d5e21be04a6239579634fa8b19446d5ef

[jbergstroem]

I think I figured out what's going on. You pushed it to your own master branch, but for some reason the generated github link thinks the commit also lives in the nodejs/io.js repo (which it doesn't).

[Olegas]

@sam-github
Test is fine. Passing on my environment. Thanks!
But can't see your commit on master also (
Reopened #1933 just to not get things lost.

[sam-github]

Landed in 03ce84d

Sorry for the confusion

[jbergstroem]

Probably not related to this specific test, but part of the same problem: Two freebsd bots saw this issue in a recent run: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/842/

I couldn't find any trailing processes on either machine. Could it possibly be some other test opening it and now being closed by this fix?