Fixes memory leak in utils/queue.js.

[iamolivinius]

Also fixes #60

[frederikbosch]

@iamolivinius One of the tests fail with this PR. Do you know why?

[iamolivinius]

Unfortunately there is no failing test on my local machine. Using node 5.7.0, phantomjs 2.1, and chrome 48

[iamolivinius]

I merged the latest master into my local copy today and rerun the tests in phantomjs 2.1.1 and chrome 53.0. In both environments all tests passed

------ RESULTS ------

default: PASS ember-release: PASS ember-beta: PASS ember-canary: PASS

[frederikbosch]

@iamolivinius I restarted the Travis build, but it fails. Do you know why?

[iamolivinius]

Not exactly yet but it's related to phantomjs. Afer I replaced phantomjs@2.1 with phantomjs@1.9 I got the same failures as in the travis environment.

[iamolivinius]

I run the tests in several saucelab environments and all passed. At this point in time I'd say the tests and the code under test is correct but the two year old versions of PhantomJS is somehow flawed :wink:

[iamolivinius]

Wait, I cloned the wrong repo. :fearful: I'll rerun the tests.

[iamolivinius]

These are the actual results! It's always the same test that fails: "perform multiple changes in bulk => 2nd assertion". But I don't have any clue why the query doesn't return the previously saved result of the "Rambo" list.

[iamolivinius]

Looks like the first build was OK and the rest had network issues.

First I rewrote the attach function to this:

import Ember from 'ember';
export default Ember.Object.extend({
  queue: [Ember.RSVP.resolve()],
  attach(callback) {
    const newPromise = new Ember.RSVP.Promise((resolve, reject) => {
      const previousPromise = this.queue.pop();
      previousPromise.then(() => {
        callback(resolve, reject);
      });
    });
    this.queue.push(newPromise);
    return newPromise;
  }
});

Then I realized that I spliced the wrong Promise from the queue array. I adapted the splice index et voilà! the tests pass continuously.