Conveyor belts and combined drift speed: wrong speed result

[pasqualemilano]

Discussed in #4577

^{Originally posted by pasqualemilano February 19, 2024}
Hi, this is the first time I post here:
I've been working with bullet for some time, I use it for testing some parcel behavior running on Conveyor belts:
I already read and follow suggestions from forum (something like: https://pybullet.org/Bullet/phpBB3/viewtopic.php?t=6730) and engine work well.
Recently I found out a really weird case, which seems more like a bug than a my code error.
In this explanation I'll try to simplify as much as possible:
As I told before, standard belts with bullets work fine.
Then I try to use a belt applying a drift angle, and again everything works fine.
In the following image there are two belts with drift angle and no drift angle: the behavior is exactly what I expect:

Then I try to simulate TWO drift belt that works in opposite drift, at the same speed and a parcel on it exactly on the middle:

The combined speed result faster than the speed applied.
I also check if negative drift can mess up something, but working exactly like positive drift.
Then I Ask to a friend to recreate the same case on a commercial software that use the same bullet engine: "Emulate3D" (actually the screenshot are from that software, is very simple generate a similar case): again the same problem.
This commercial software handle different engines, so we switch to another one and... the problem disappeared!

I search into open issues if there is something related but I didn't find anything.
This seems happen when the set of btManifoldPoint of the same rigidbody has different and partially conflictual direction (m_lateralFrictionDir1)
I can show you my code (that is inspired to the code suggested in the link that I wrote at the beginning of the post) but after the last "external" test I start to exclude the coding error and think about the engine bug.
Anyway I am available for any clarification.

thanks