This code in the tessellation evaluation shader determines the area over which a texture is applied to a surface:
float p = length(tess_vertexWorldPosition[0].xyz - tess_vertexWorldPosition[1].xyz);
float t = length(tess_texCoords[0].xy - tess_texCoords[1].xy);
if (p > 0.0001f && t > 0.0001f) uvscale = p / t; else uvscale = 0.0f;
The code results in a different value for two adjacent triangles with two shared vertices:
Here are the edges highlighted in RGB for edges 0, 1, and 2:
The texture mapping scale across the two triangles is constant and equal on the U/V axes. Even if a different length edge is being tested, the length of the position and texcoord difference should be scaled exactly the same, so it should return the same value.
I applied a grid texture to the material and it shows there is nothing unexpected going on with the UV coords:
I tried replacing it with this more convoluted code that finds the longest edge and uses that for the calculation:
float GetUVScale(in vec3 a, in vec3 b, in vec3 c, in vec2 ta, in vec2 tb, in vec2 tc)
{
float ab = length(a - b);
float bc = length(c - b);
float ac = length(a - c);
float tab = length(ta - tb);
float tbc = length(tc - tb);
float tac = length(ta - tc);
if (length(ab) > length(bc))
{
if (ab > ac)
{
return ab / tab;
}
else
{
return ac / tac;
}
}
else
{
if (bc > ac)
{
return bc / tac;
}
else
{
return ac / tac;
}
}
}
This code returns the same result for both triangles, eliminating the crack, but I do not understand why testing any of the edges would not return the same value.
Any ideas what the problem might be? Am I overlooking something really obvious?
