TaskDescription.waitForAllTaskTermination() on waitOnCommand()

[norio-nomura]

This is rebased version of #496.

Fix segmentation fault on termination of Carthage.

Segmentation fault on termination of Carthage 0.7.4 beta-2 happens following scenario:

1. launchTask's thread call sendCompleted() and kick main thread
   2a. main thread wake up and call exit()
   2b. launchTask's thread continue disposing
2. main thread cause segmentation fault.

Waiting completed from launchTask's thread is not enough for fixing the segmentation fault.
It needs waiting actual termination of launchTask's thread

This PR depends on Carthage/ReactiveTask#24

#474 (comment)

[robrix]

Should this be finer-grained, e.g. waiting per-task when they’re executed?

[norio-nomura]

I can't determine where I should place dispatch_group_wait() on another place.
RAC's completed event timing is not fit for place it.

[norio-nomura]

The segmentation fault happens while disposing after completed event from launch task's thread and main thread calling exit() concurrently.
Waiting per-task will need waiting some RAC event. That will be same what just current implementation is doing.
I think fixing the segmentation fault needs doing outside of RAC event, and waiting per-task without RAC event may be too complicated unnecessarily.

[robrix]

Okay, thanks for looking into this!

[robrix]

I’ve merged the dependency and have a beta tag of ReactiveTask to include it.

Could you please update the Cartfile & submodules to pin to that tag?

[robrix]

Once that’s done I think we can merge and release a new beta of Carthage. I may not be aesthetically content with waiting for all child processes like this, but I can’t argue with its results.

[norio-nomura]

I've pushed the commit updating the Cartfile & submodules to pin to 0.7-beta1

[robrix]

Thanks very much!

[norio-nomura]

You're welcome. 🙇