You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have found an issue with the GImpact contact test.
It reports false contacts if the triangles lay behind each other (and do not touch each other) but are close enough to be within their aabbs.
This can be the case if you have "small" objects or meshes where both sides of an triangle are visible and relevant for the collision.
GImpact mesh should work double sided.
I investigated the problem further and build an example consisting of only two objects which have only one triangle each (see end of this post), that shows the issue.
I looked in the code and found the issues. I fix it so that only triangles get reported as contact that really intersect each other and no longer all triangles that lay behind each other and within the aabb of a triangle.
I am happily providing a pull request for this, if there is any chance to get it merged. Is there any thing I need to pay special attention to?
Hi,
I have found an issue with the GImpact contact test.
It reports false contacts if the triangles lay behind each other (and do not touch each other) but are close enough to be within their aabbs.
This can be the case if you have "small" objects or meshes where both sides of an triangle are visible and relevant for the collision.
GImpact mesh should work double sided.
I investigated the problem further and build an example consisting of only two objects which have only one triangle each (see end of this post), that shows the issue.
I looked in the code and found the issues. I fix it so that only triangles get reported as contact that really intersect each other and no longer all triangles that lay behind each other and within the aabb of a triangle.
I am happily providing a pull request for this, if there is any chance to get it merged. Is there any thing I need to pay special attention to?
The text was updated successfully, but these errors were encountered: