Implement Automated Testing

[simonwex]

• Implement testing framework
• Pick two recent or open bugs
  • Bug: ____
  • Bug: ____
• implement tests that validate fixes for two open bugs
• Document how to add a new test
• Set up project on TravisCI
• Add Travis status Image to README
• Add mofo-style linter
• Ensure all tests are passing

MozillaFoundation/plan#351

[Pomax]

There is an interesting problem with newrelic here: the testing with mocha works quite well provided either newrellic is available, or a newrelic shim is used in event-controller.js. However, I cannot seem to find any particular way to reliably detect whether newrelic will work and then fallback to the shim. Mocha does throw-interceptions that cause the following code to just crash the mocha run:

var newrelic;
try {
  newrelic = require("newrelic");
} catch (err) {
  if (process.env.NODE_ENV === "development") {
    newrelic = {
      addCustomParameter: function() {
        // no need to do anything
      }
    };
  } else {
    throw err;
  }
}

Mind you, this is the same trick we already use in server.js, but because things are wrapped by Mocha, this trick no longer works. The throw from the require call outright terminates the mocha run.

[Pomax]

as it turns out, newrelic does not throw errors on require() (any throws it generates it self-catches and simply logs instead), so the try/catch we're using in server.js doesn't even do anything (tested that hypothesis, confirmed it by the catch block never actually running).

So, I'm going to solve this with a new, backward-compatible .env variable called WITHOUT_NEW_RELIC which, when set, effectis loading of a newrelic shim instead of the actual newrelic module.