[fixed in trunk] getCollisonNode() fail in IMetaTriangleSel

[Strong99]

Here is a patch which fixes bug1 and 2. But it will still occur for many other collisionmethods
Patch

For example there is still a nested problem with the getCollisionResultPosition. If you look at the stack:
1. getCollisionResultPosition
2. collideEllipsoidWithWorld
3. collideWithWorld

Where 2 uses all the triangles to search for the triangle index returned from 3, which uses a box as filter. I can add the box filter for the getSceneNodeFromTriangle but the collision manager itself would still have the problem.

The filter used in the collisionmanager isn't only the ray or bounding box but it also requires the correct transformation for the box. That transforation is applied after setting the node's transformation in the TriangleSelectors. So to be sure to you get the correct triangle for the given index we should pass the exact same parameters like filter and transformation to the getSceneNodeFromTriangle methods.

Thus we should somehow pass these from 3. collideWithWorld to the 2. collideEllipsoidWithWorld and I'm not sure if thats the correct thing. This could be done through extending the SCollisionData structure with the boundingbox and the used transformation,

@Hybrid, any suggestions?

[hybrid]

I'm not really into this collision system, maybe CuteAlien can help better!?

[CuteAlien]

Sorry, that hope's in vain currently.

[CuteAlien]

Stuff got rewritten now (svn trunk r5350) and I hope it also fixed all those bugs. Basically in the past the indices of triangles returned got mixed up with indices used inside triangle-selectors - which couldn't work as those are not identical. But this couldn't easily be fixed with old interface, so won't be fixed for Irrlicht 1.8, but only Irrlicht 1.9
On the up-side a new feature got added while fixing this (meshbuffer selection).

edit: Fix had a bug. Fixed again in r5352 and another bug with box selectors in r5353.