Automated testing differences between browser test runner and CLI?

Question

I have an Ember CLI addon project -- ui-slider-input -- which has a VERY basic wrapper around a JS slider widget. This is implemented as a component and the out-of-the-box unit tests for whether a component does or does not render is run. Here's the strange part ...

1. There are no errors when running in the browser (UiSliderComponent: it renders)
2. When I run the same tests from CLI with npm test (could also have used ember test ... no difference) it fails with the following stack trace:

	not ok 7 PhantomJS 1.9 - UiSliderInputComponent: it renders
	    ---
	    actual: >
	        null
	    message: >
	        Died on test #2     at http://localhost:7357/assets/test-support.js:418
	            at test (http://localhost:7357/assets/test-support.js:284)
	            at http://localhost:7357/assets/dummy.js:273
	            at http://localhost:7357/assets/vendor.js:77
	            at http://localhost:7357/assets/test-loader.js:14: 'undefined' is not a function (evaluating 'this._applyPrecision.bind(this)')

Answer 1

It’s not a Ember CLI problem, it’s a PhantomJS problem.

PhantomJS lacks Function.prototype.bind, which you can see is being called in the stacktrace. This GitHub issue discusses the problem. The simplest solution would be to add one:

Function.prototype.bind = Function.prototype.bind || function (thisp) {
    var fn = this;
    return function () {
        return fn.apply(thisp, arguments);
    };
};

See the thread for more details.