I wish I could be more active like I used to, but I've got to spend some time coding if we're going to make any progress! After v3 is released I expect things to be more exciting around here as we address some smaller issues and add smaller new features.
I'm glad to say that last night I've finished the new IK algorithm. It would be great if you guys would like to test it. Our test harness is here:
http://esotericsoftware.com/temp/iktest.jar
You can click and drag the buttons at the top to change their values. Right click drag to change scale uniformly.
All aspects of IK need testing, both with uniform and nonuniform scale on any or all of the bones. Also negative scale should always work. Other features are important to: bend direction, mixing, etc. All the things should work!
Please remember that if the child Y position is not zero and the parent is scaled nonuniformly then the IK solution will be incorrect. This is an expected drawback of the nonuniform scale IK solution and is not a bug. You can either have a child Y position != 0 and use only uniform scaling of the parent bone, or you can have child Y position = 0 and use uniform or nonuniform scaling of the parent bone. Note that nonuniform scale inherited from the parent bone's parents doesn't affect using a child Y position != 0.
For other cases, keep in mind that the correct solution may not be what you expect at first when using nonuniform scaling. The goal is for the tip of the yellow child bone to be as close to the target as possible. With nonuniform scaling, rotating the parent changes the length of the child bone. This means that when the target is too far (or too close), the best IK solution may not result in the bones being in a straight line toward the target. The reason is if the parent were rotated to point at the target, it would make the child bone shorter or longer which would make the tip of the child further from the target.
I hope that is entertaining for you! 🙂
