Fix crawlComplete exception when running tests

[marcelgerber]

Fixes #12658. (cc @ficristo)

Turns out the fix is quite straightforward.

[ficristo]

The change itself LGTM, but I'm not sure if instead we should always have the project name at this point in the code.
Maybe @abose knows.

[petetnt]

Small nit / bikeshedding: instead of assigning the return value getProjectRoot() to projectName, it could be a bit cleaner if it was something like:

var projectRoot = ProjectManager.getProjectRoot();
var projectName = projectRoot ? projectRoot.name : null;

if (!projectName) {
  projectName = "noName00";
}

In general I think that there should never be an situation where projectRoot.name is empty anyways until Brackets supports opening up with no projects at all.

[petetnt]

I agree with @ficristo, there really shouldn't be any situation where there wasn't a projectRoot but the search was running, should there?

[marcelgerber]

To me, it looks like the Node instance sends the event both to the test window and the Jasmine window, where obviously, no project is loaded. So it runs fine in the test window, but throws this exception in the parent Test Runner window.

[ficristo]

@marcelgerber to me you are the expert so if you think is fine it's ok for me.
But would be nice, if possible, to add at least a log or something in the unexpected case:

if no project name then
  if not test window then
    log "unexpected"
  end
  projectname = NoName00
end

[marcelgerber]

So, I just made sure my assumption is correct: Both the Test Runner and the test window receive an event in this case, and the Test Runner cannot handle it properly. I'm not sure whether it's the same event though (I think it is).

One way would be (assuming it is one event) for the Test Runner to ignore all Node events it is being sent, which is definitely the more general solution and may prevent other issues as well. On the other hand, I'm not sure if some Node events are expected for tests to pass.
Also, I'm not quite sure how to know whether the current window is the Test Runner or not.

I'll look into it.

[marcelgerber]

So, it turns out the SpecRunner definitely needs to receive Node events.
I updated the PR to have it ignore just this event.

[ficristo]

Would be nice to check for the /brackets/i regex too. But I don't know the location on the other OS.

[marcelgerber]

We cannot be certain what the git clone is named on the system. Also, it's very likely that a Brackets Shell window that has a SpecRunner.html loaded, which in turn loaded the rest of Brackets, actually belongs to Brackets ;)

[ficristo]

I left a couple of comments, but otherwise LGTM.
@petetnt needs to check too.

[petetnt]

LGTM too 👍