RotatePointAroundVector Broken?

[jitspoe]

I wanted to tweak the angle of the stars a bit for the skybox I'm working on, so RotatePointAroundVector seemed to be the appropriate function to use, however, instead of just rotating, it seems I've managed to collapse the universe!

I'm rotating around 0,1,0 (y axis), and here are the results from 0 to -100 degrees:








Does anybody have a fixed version of the function, or is this how it's expected to work? Doesn't seem right...

[Spike]

AngleVectors('0 1 0' * time, forward,left,up);
neworg_x = dot(forward,origorg);
neworg_y = dot(left,origorg);
neworg_z = dot(up,origorg);

or in other words, use a matrix.
and rotate the matrix a little so that you're not standing at the north pole.

if you're just making a skybox image, you could just swizzle+negate accordingly without bothering with complex rotations.

[jitspoe]

I'm just doing the rotation on load.

I ended up rewriting the function like this, and it appears to work:

Code: Select all

void RotatePointAroundVector (vec3_t dst, const vec3_t dir, const vec3_t point, float degrees)
{
   float u = dir[0], v = dir[1], w = dir[2];
   float x = point[0], y = point[1], z = point[2];
   float ux = u * x, uy = u * y, uz = u * z, vx = v * x, vy = v * y, vz = v * z, wx = w * x, wy = w * y, wz = w * z;
   float uu = u * u, ww = w * w, vv = v * v;
   float s = (float)sin(DEG2RAD(degrees));
   float c = (float)cos(DEG2RAD(degrees));

   dst[0] = u * (ux + vy + wz) + (x * (vv + ww) - u * (vy + wz)) * c + (vz - wy) * s;
   dst[1] = v * (ux + vy + wz) + (y * (uu + ww) - v * (ux + wz)) * c + (wx - uz) * s;
   dst[2] = w * (ux + vy + wz) + (z * (uu + vv) - w * (ux + vy)) * c + (uy - vx) * s;
}

Edit: Wonder if model culling was buggy because of this? Or maybe this fix will create new bugs?

[Jay Dolan]

You might want to look at LordHavoc's matrix lib to replace or at least augment the sparse / incomplete / seemingly arbitrary / questionably buggy transform functions included in Quake2:

https://github.com/jdolan/quake2world/b ... r_matrix.c

It works very well; I've never had to touch it.