Return a new can.compute every time (new can.view.Scope()).compute() is called

[simpleTechs]

You need to actually return the new compute, so https://github.com/bitovi/canjs/blob/master/view/bindings/bindings.js#L62 won't get undefined back, resulting in an error here https://github.com/bitovi/canjs/blob/master/view/bindings/bindings.js#L187

[simpleTechs]

Here's a JSFiddle to show the issue: http://jsfiddle.net/simpleFabian/TWHFs/2/
Comment out the first few lines of js to see the problem.
Also, livebinding will only work on properties that were present when binding was initialized - can this be circumvented somehow? (See in this example only "Middle West"'s checkbox will update correctly, as it has had the property selected before being bound.)

[daffl]

This does seem to make sense but a lot of the tests are failing now.

[simpleTechs]

Well, yes, many tests are failing, but this is due to the fact I submitted the pull against current head, not against, say, tag 2.0.0. Just tried the latter locally and every single test is fine.
Should I re-submit?

(In fact, 7d38e29 seems to have solved the test-failures)

[daffl]

Travis should be running against the head. I will merge it in locally and see what is happening. Would it be possible to scale the Fiddle down a bit so that I can turn it into an easy test to add for the issue that your change is fixing?

Thanks

[justinbmeyer]

@simpleTechs / @daffl

Anyway you can provide a simple explanation of what's going on? Reading this issue on its own gives no insight to the problem.

[imjoshdean]

The issue is that calling can.view.Scope.prototype.compute doesn't consistently return a compute. It will only return a compute if the requested attribute's parent is observable, otherwise it creates a can.compute and does nothing with it. I'm to assume this boils down to something as simple as "this was probably a typo" or "this was an unintentional slip."

I agree with @daffl, this makes sense. And merging against master it works just fine. My theory as to what cause the test to break was that the pull request came right before cfb6bd0 which fixed some breaking tests.