Vague documentation with glm::refract

[tigert1998]

Documentation with glm::refract doesn't describe what will be returned when full reflection happened. Actually, it refers to GLSL refract man page while the official man page indicates that the result vector will be vec3(0).

I found a little bug with my project and debugged it for about an hour. At last, I found that the reason is that glm returned vec3(nan, nan, nan) which is different from GLSL.

This really confused me and I hope that we can make documentation more precise if possible.

[Groovounet]

Hi,
Your bug report is a little vague. :p What's missing in the documentation of refract?
What do you mean by "full reflection"? That 'eta' is 0 or I and N collinear? I guess I could add an assert for that to catch it.

[Groovounet]

I can't manage to reproduce the issue, would you share a code where the issue happen?

[tigert1998]

Sorry, my bad!

#include <iostream>
#include <string>
#include <glm/glm.hpp>
#include <glm/geometric.hpp>
#include <glm/gtc/constants.hpp>

using namespace glm;
using std::string;

void PrintVec3(string name, vec3 v, int tabs) {
    while (tabs--) printf("\t");
    printf("%s = (%.3f, %.3f, %.3f)\n", name.c_str(), v[0], v[1], v[2]);
}

int main() {
    float eta = 1.5;
    vec3 n(0, 1, 0);
    PrintVec3("N", n, 0);
    for (double outer_theta = 0; outer_theta < pi<double>() / 2; outer_theta += 0.1) {
        puts("{");
        vec3 i(sin(outer_theta), cos(outer_theta), 0);
        i = -normalize(i);
        PrintVec3("I", i, 1);
        auto o = refract(i, n, eta);
        PrintVec3("O", o, 1);
        puts("}");
    }
    return 0;
}

And that gives the following output:

N = (0.000, 1.000, 0.000)
{
	I = (-0.000, -1.000, -0.000)
	O = (0.000, -1.000, 0.000)
}
{
	I = (-0.100, -0.995, -0.000)
	O = (-0.150, -0.989, 0.000)
}
{
	I = (-0.199, -0.980, -0.000)
	O = (-0.298, -0.955, 0.000)
}
{
	I = (-0.296, -0.955, -0.000)
	O = (-0.443, -0.896, 0.000)
}
{
	I = (-0.389, -0.921, -0.000)
	O = (-0.584, -0.812, 0.000)
}
{
	I = (-0.479, -0.878, -0.000)
	O = (-0.719, -0.695, 0.000)
}
{
	I = (-0.565, -0.825, -0.000)
	O = (-0.847, -0.532, 0.000)
}
{
	I = (-0.644, -0.765, -0.000)
	O = (-0.966, -0.257, 0.000)
}
{
	I = (-0.717, -0.697, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.783, -0.622, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.841, -0.540, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.891, -0.454, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.932, -0.362, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.964, -0.267, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.985, -0.170, -0.000)
	O = (nan, nan, nan)
}
{
	I = (-0.997, -0.071, -0.000)
	O = (nan, nan, nan)
}

[Groovounet]

This issue is resolved in master branch for GLM 0.9.9.1.

Thanks for contributing!