I was not aware of that. But code is now fixed. Check the latest version from svn.

_________________
http://3das.noeska.com - create adventure games without programming

At the moment i work on a real vektor based implementation of FreeType Fonts and svg.
http://www.delphigl.com/forum/viewtopic.php?t=8534
FreeType use quadric bezier curves which are really easy compared to cubic.

_________________
"Wer die Freiheit aufgibt um Sicherheit zu gewinnen, der wird am Ende beides verlieren"
Benjamin Franklin

Projekte: https://github.com/tak2004

@TAK2004: I had a look on you shaders but at first looks i dont understand it yet. But you feed a triangle with texture coords and the fragment shader then knows how to draw the spline?

That would be much faster then my current proces of triangulation. And not to mention the high triangle count i get with complex shapes. But you mention a precalcuation step?
The article states the triangulation is still needed as pre process step (but at low detail only?)).

Could the texture coords be used as an paramter to the quadratic bezier line from, to and control point (relative coords) or am i going haywire on that theory.

_________________
http://3das.noeska.com - create adventure games without programming

The first step is to get the type of curve(line,quadratic bezier, serpetine,loop or cusp).
This need some math and knowledge(first half of paper) and the second part is drawing and calculation of uv coords.
The second step is to react on the type, like convert data to a normal glLine, work with a quadratic bezier, work with cubic bezier.
Cubic bezier give u 3 different version, serpetine,loop and cusp, which you can find on the begin of the paper with the lines l k m n s.
Serpetine is easy, calculate the stuff, loop you have to find the double point, break down the beziercurve into two and calculate the stuff and third i didn't implement yet(i think there are 2 double points=4 bezier curves but not sure).
You have to calculate k and check the sign of the determinant D2 to switch it.
After that you calculate s,m,n and get the new Texcoords.

The shader calculate the with the short formular the distance to the nearest point of the cubic bezier curve.
That means, you get the lines k(u) l(v) and m(w) (n was eliminated through the conditions explained in the paper).
This and the sign can be used to decide which side is to draw and which texel can discard.

At the moment i have 300lines of code for the stuff but its not complete.
Some points of the paper i don't unterstand too.

_________________
"Wer die Freiheit aufgibt um Sicherheit zu gewinnen, der wird am Ende beides verlieren"
Benjamin Franklin

Projekte: https://github.com/tak2004

I am back on my old way and even got circular gradients working. Alpha with an circular gradient is a work in progress.

_________________
http://3das.noeska.com - create adventure games without programming