OrbitControls throws on pointerUp event about three.js

[sciecode]

I was able to reproduce now, the issue can occur if both left/right buttons are pressed simultaneously. The controls becomes stuck as explained by this issue. I'll investigate ways to solve this behaviour, thanks for reporting.

[wnr]

I think this issue should be treated separately as it is much easier to fix than #24566. I believe that reverting https://github.com/mrdoob/three.js/pull/27425/files would fix this particular issue, but of course it would be preferable to keep it while fixing this regression somehow.

I too use a magic mouse. I notice that it is really easy to reproduce like 75% of the times, but sometimes I can't reproduce it a few times in a row. Very strange. Does this video help?

[wnr]

Can you make a test if Firefox and/or Safari shows the same behavior?

I tried both of them and I'm unable to get the controls stuck (the root issue) so I think this is a Chrome-only issue.

[Mugen87]

Can we merge this issue into #24566? The root cause of both issues seems to be identical.

Sidenote: I still can't reproduce any pointerup related issues on macOS with latest Chrome and a magic mouse. I'll try to reproduce the issue with a different mouse later.

[Mugen87]

/cc @sciecode

[Mugen87]

Can you make a test if Firefox and/or Safari shows the same behavior?

[sciecode]

Unable to reproduce on Windows - FF/Chrome, and Ubuntu - FF/Chrome.

[sciecode]

Although I can't reproduce, I can definitely say there's something inconsistent happening either with user agent or magic mouse hardware.

There are two things that shouldn't occur, judging from the reported error. The block that triggers the error is on OnPointerUp.

three.js/examples/jsm/controls/OrbitControls.js

Lines 1065 to 1071 in a984c6f

	case 1:
	const pointerId = pointers[ 0 ];
	const position = pointerPositions[ pointerId ];
	// minimal placeholder event - allows state correction on pointer-up
	onTouchStart( { pointerId: pointerId, pageX: position.x, pageY: position.y } );

• Specifically the case where one pointer is lifted and one pointer remains active. However, I'm unaware of anyway this would happen on a mouse interface. So, this is problematic already.

• Furthermore, the error implies the saved pointer-position, that is still in contact with the canvas, has undefined coordinates, which should also never happen.

I suspect this isn't an issue on our end, but would need more information of the events and pointerIds/positions to confirm.