Better error handling when server fails to start during testing (see issue #7)

[aquamoth]

[See issue #7]

This fixes the bad error message when the server port is already occupied, but it doesn't address the real problem with firefox and selenium.

The firefox/selenium-bug seems to be hard to get by. Selenium crashes with a SIGSEGV and gives mocha no chance to clean up. I can't see a simple way to try/catch that in nodejs either. https://github.com/ddopson/node-segfault-handler created a module that writes a stack-trace when nodejs dies, but that module doesn't make any attempt to stop the actual core-dump.

[jamesshore]

Thanks. The crash is a rough one. I suppose you could work around it by spawning the smoke tests in a new process, similar to the way the PhantomJS tests worked, but I'm not sure the trouble would be worth it.

[jamesshore]

This is a good change, but I've decided not to integrate it at this time. See issue #7 for discussion.