automated test coverage

[andrewrk]

need coverage for:

• hot code swap
  • make very quick successive requests and ensure that none of them fail
  • test with 1 worker, test with N workers
  • when a deploy is already in progress
• ability to change environment variables of workers with zero downtime
• starting a server twice doesn't do it
• stopping a server twice doesn't do it
• manually killing a server and then checking the status reports correctly
• make sure there is zero downtime when a worker dies and the worker is restarted
• logging
  • stderr/stdout routed to correct file
  • naught log has events
  • appending to an existing log file
  • starting a new one
  • log rotation
  • mkdirp
  • gzipping
• stop server with timeout
• sending a process.message other than 'online' first thing
• deploying code in between a worker dying and coming back online
• deploy abort
  • when waiting for old workers
  • when waiting for new workers
  • when no deploy is in progress
• sending command line args to script
• multiple naught instances connecting to daemon simultaneously