MeshBender.cs floating point precision problem with Repeat mode

[brightest1102]

Hi,
I've encounter some problem these days when I adjust the scale X in repeat mode.

When using curve spcce, If I adjust the (scale X) that
(size of mesh) x (scale X ) x (repetitionCount) ≒ spline.Length
Some meshes would fail with following error message.
Error message like

"Mesh.uv is out of bounds. The supplied array needs to be the same size as the Mesh.vertices array."
"Failed setting triangles. Some indices are referencing out of bounds vertices. IndexCount: X, VertexCount: Y"

It seems that the number of vertices does not match to (source.Mesh.vertices.Count() x repetitionCount)

After some test, I found that the problem is caused by floating point precision problem
The value "distOnSpline" of "distance" will be slight greater than slpine.Length (about 10^06).
And it will cause different problem in non-curve space and curve space setting.

In Non-curve space:

} else {
	float distOnSpline = intervalStart + distance;
	//if (true) { //spline.isLoop) {
		while (distOnSpline > spline.Length) {
			distOnSpline -= spline.Length;
		}
	//} else if (distOnSpline > spline.Length) {
	//    continue;
	//}
	sample = spline.GetSampleAtDistance(distOnSpline);
}

This will make the end of the mesh incorrectly links to the begin of the mesh.

In Curve space:

if (!useSpline) {
	if (distance > curve.Length) 
		continue;
	sample = curve.GetSampleAtDistance(distance);
}

And it will make these vertices skips when adding them into 'bentVertices' after iterations.

bentVertices.Add(sample.GetBent(vert));

I think that's how the vertices number do not meet uv and triangle numbers.

Here's my workaround now, but I think there's some better way.

if (!useSpline)
{
	if (distance > curve.Length)
	{ 
		distance = curve.Length;
	}
	sample = curve.GetSampleAtDistance(distance);
} else {
	float distOnSpline = intervalStart + distance;
	if (distOnSpline > spline.Length + 0.0001f)
	{
		while (distOnSpline > spline.Length)
		{
			distOnSpline -= spline.Length;
		}
	}
	else if (distOnSpline > spline.Length)
		distOnSpline = spline.Length;
	sample = spline.GetSampleAtDistance(distOnSpline);        
}
sampleCache[distance] = sample;

I just brutally add some offset (0.0001f) to avoid floating point precision.
It works for me but there should be some better solutions.

Thanks.

[methusalah]

I understand you issue, thank you for your report. In the repeat mode, there is indeed a chance that the repetition count Mathf.FloorToInt(intervalLength / source.Length) is smaller than the number of cumulated interval, due to float precision.

In loop spline mode, it is ok to go back to the start of the spline. It may cause the last mesh to overlap the first one, but the user can just change the mesh scale according to the current spline length to avoid that.

In curve and non-loop spline, though, you can't go past the spline length in any way. My startegy would be slightly different than yours. Instead of computing a possibly wrong repetition count before iterating, I would iterate until there is no room left and consider the number of iterations as the actual repetition count. The assignation of UVs arrays should be moved down in the method and use this resulting repetition count.

Would you like to test it and show some more code?