`step` value prevents `moving` agents to wander

[RoiArthurB]

Describe the bug
The step value (seems to) un-schedule wandering agents on a radius starting from the middle of the map. They are not moving at all (displayed as not, but can also check that there position is not updated at all).

To Reproduce
Steps to reproduce the behavior:

model issue382

global {
	float step;
	init {
		create p number: 1000;
	}
}
species p skills: [moving]{
	reflex move {
		do wander;
	}
	aspect default {draw circle(1) color: #green;}
}
experiment e {
	parameter var: step min: 1#mn max: 3#mn init: 1.8#mn;
	output {display d {species p;}}
}

1. Open model above on Linux
2. Start simulation
3. Scroll step value
4. See error

Expected behavior
Execute do wander on any agent regardless of the step value.

Screenshots
If applicable, add screenshots to help explain your problem.

Screencast.From.2024-12-11.09-18-42.mp4

Desktop (please complete the following information):

• OS: Windows, MacOS, Linux
• PC Model: [if applicable, can be useful for displays, OpenGL problems)]
• GAMA version: git and alpha
• Java version: 21.0.5+11 and embedded
• Graphics cards / Display system: Both 2d and 3d

Additional context
The problem seems to be limited to Linux using Gnome desktop environment (the issue exists on Ubuntu and Arch Linux on Gnome, but not on LinuxMint on Cinnamon), and might be related to Wayland (even tho, I think it's not the related).
EDIT: This behavior happens on every OS

[AlexisDrogoul]

Particular combinations of the size of the environment (when it's not toroidal -- default is 1m x 1m), the speed of the agents (default is 1m/s) and the step (default is 1s) make this behavior appear. If one changes the speed to 1mm/s, for instance, it disappears. So I would bet that there is a problem of rounding somewhere that causes this behavior in the moving skill.

[ptaillandier]

A specific choice was made for the ‘wander’ action in a non-toroidal environment: when an agent’s movement would take it outside the environment, it doesn’t move. The higher the step value, the further an agent can move in a single time step using Wander, and as a result, fewer agents will actually move. We could make a different choice (such as placing the agent at the edge of the world), but that would change how previous models function.

[AlexisDrogoul]

Closing this issue as it is by design