Disable automatic token refreshing for tests

[marcoow]

As it causes problems (see #35), also it might actually randomly break tests.

[johnnyshields]

@marcoow your commit 75a73a6 affects our tests. can we make this behavior a config variable rather than hardcoding to if (!Ember.testing)?

[marcoow]

@johnnyshields: hm, why would you ever want that to run in tests?

[johnnyshields]

for long-running integration tests which test performance. I think the question is really "why would a library hardcode its behavior to not run in tests, rather than making it configurable?" Its fine if the default config is false for test env, true for non-test so as to affect existing users.

[marcoow]

I get your point - changing the behavior of sth. if it's running in test env isn't ideal but at the same time a pretty common pattern in Ember.js as you have to make sure tests are not leaking and run loop events.

I think if you're running long-running integration tests to test performance you shouldn't be running the app in test env anyway though as it will behave differently and might give you a wrong sense of its performance even (for example, ESA will always store the session in memory when running in test mode which of course is much faster than using localStorage or a cookie).

[johnnyshields]

@marcoow can we make this a config?

[marcoow]

@johnnyshields: I honestly don't see the need for doing that

[johnnyshields]

could we make a more general config test_mode which replaces all usages of Ember.testing within the library, which by default is set to the value of Ember.testing but can be overridden, so that we are not tightly coupled to Ember itself?

[marcoow]

@johnnyshields: I honestly don't see the need for doing that. As I said above - when you're executing long-running integration tests, specifically for performance testing you shouldn't be running the app in test mode anyway.

[johnnyshields]

I pray for the day when the Ember community will learn the merits of loose coupling.