Sub-trees not triggering end task edge case

[ashblue]

Possible edge case discovered with sub-trees. Maybe connected with async tasks? Needs further investigation as this edge case was not triggering on a recent project I worked on.

---

I will say, the cases where I had issues were Tasks that would run actions on a seperate system for async work in the game, e.g. 'Go over there and attack that character'.

I found that when the sub-tree holding that Task (and therefore Actions) exited, it didn't End the Task, so the Actions didn't end either, leading to characters continuing to move/animate/etc. while other new behaviours were also kicking off.

The ending of the Actions was called from the OnExit of the Task, and that seemed to work fine once I added the code in this PR.

Originally posted by @achynes in #113 (comment)

[achynes]

Responding to: #113 (comment)

@achynes thanks for the follow up response. I've broken this off into a separate ticket to research and figure out what might be going on here as the edge case will need a patch.

I'm assuming you noticed this behavior with an async task maybe? Or was it specifically just any sub-tree. Just curious as I'll have to tinker to figure out what loses the end event. Sub-trees shouldn't cause issues as they're copied into the main tree normally at runtime. I'm wondering if you found some sorta async task bug.

I'm off that codebase now due to a change of job, but here are the details from memory.
Aside: I recall now, I had an intermediate base type for decorators, just to add OnStart and OnExit virtuals. I wonder if I mucked something up when I added that. :/ Hopefully I've not reported a red herring!

• The BT branches affected were part of a spliced sub-tree; the model had a base tree for all characters, and then a sub-tree for individual behaviours
• Both the base tree and sub-tree made extensive use of ticked Decorators that performed either data updates or self-cancellations based on same data
• Every task was a very simple async task of the same type for running an ability (which would take over the linear flow in a separate system)
• The Tasks that ran abilities were definitely staying alive when the sub-tree was exited

[ashblue]

@achynes super duper thx for the context in which you saw this happen. I'll definitely see if I can't reproduce this when I dig into updating the library again. The splice shouldn't have any issues since everything is deep copied into the target tree (but worth taking a look). The issue may have been in complex decorator logic, as the decorators by design are meant to be super duper simple. They don't run a lot of hooks seen in the more complete actions and conditions by design.