4.1.1 laggy pinch detection

[iljaburzev]

Consistently running into an issue where pinch detection on one of the hands stops working completely, until you bring the hands out of fov and in again to initiate, after which the pinch detection works again. Funny thing is that when I bring the hand out and in again - pinch detection is off again, and this happens consistently - looks like a broken toggle somewhere.

Might be helpful to locate the problem:
In another experiment I am just scaling up the default capsule hands, and I observe exactly the same toggle behavior - I see capsule hands either scaled as I want them to, or initiating with default cyllinder thickness, as move one of the hands in and out of leap fov.

Please let us know which part of the code this could be related to.

[iljaburzev]

I've made a video for you, I got a custom hand models there but it behaves the same way with default ones.
The way to reproduce I think is to hold pinch with one hand and simultaneously exit and enter the fov with the other.

[Joe_Ward]

Does this occur in the PinchMovementDemo scene? I'm not able to reproduce the problem.

[iljaburzev]

Joe, yes, this happens in PinchMovementDemo, I've just tried on an empty project.

I also don't manage to figure out a precise recipe to reporduce, but it happens after some combination of bringing hands in and out of fov, while pinching and not.

[Joe_Ward]

There might be a bug in the way we handle duplicate hands. The LeapRTS script used to manipulate the Menger sponge uses explicit, editor-time references to the PinchDetector instances. Conceivably, if a hand is duplicated for some reason, it's pinch detector instance is also duplicated and since it isn't the instance referenced by the LeapRTS script, pinching has no affect.

To verify this, you could look at the hierarchy when the problem occurs and see if there are multiple copies of the same hand and if, so check whether the original of the duplicate is active.

[iljaburzev]

Joe, is this related to the propery Can Duplicate? Would turning this off theoretically help?

[Joe_Ward]

Yes, that's the property. If this is really the cause, then, yes, turing it off should correct the problem. There could be some downsides, though -- if, for example, it recognizes the left hand as a right hand, the true right hand won't be displayed until the false right hand corrects itself (on the service side). A more flexible approach would be to dynamically pick up the active pinch detectors rather than assigning them at editor-time.