calculated model / view matrices up to that point. The IntersectionVisitor would instead keep the
view matrices calculated up to that point even though the Transform class will throw out the
calculated model matrix via “computeLocalToWorldMatrix.”
The change I made will push an identity matrix as the view matrix when running into a transform
with an absolute reference frame and will pop the matrix off after the traverse.
To test this, I created a camera with a perspective view and added a transform with some geometry
in it. Afterwards, I set the transform’s reference frame to ABSOLUTE_RF and spun the camera around
using the trackball manipulator. When trying to pick with a LineSegmentIntersector, it would not
pick the geometry in the transform with the reference frame set to ABSOLUTE_RF."
submission email from Nathan: "I discovered a problem with POINT_ROT_EYE billboards in IntersectionVisitor: because we pass in just the model matrix to Billboard::computeBillboardMatrix, the billboard gets the wrong up vector. It really needs to take the view matrix into account to get the correct up vector.
This version of IntersectionVisitor.cpp is made against today's SVN. It corrects the problem by computing the billboard matrix using the complete modelview, and then multiplies by the inverse of the view matrix before pushing onto IntersectionVisitor's model stack. The only code I changed is in apply(Billboard&)."
notes from Robert, refactored the matrix multiplication code and the use of RefMatrix to make Nathan's changes more efficient.
1) The "highest res" child is assumed to be the child with index "getNumFileNames()-1" or "getNumChildren()-1". As a result, PagedLODs that do not sort children from furthest to nearest will intersect with the wrong child. (see attached "case1.osg" to reproduce this problem.)
2) The code assumes there is only one highest res child. As a result. PagedLODs with multiple children at the same highest res range can only intersect one of those children. ("case2.osg" demonstrates this issue; you can only pick the quad on the right.)
I've attached a modified IntersectionVisitor.cpp that attempts to resolve these issues. It identifies a highest res range based on the range mode, then continues traversal on all valid children corresponding to that range description. Only in the case of a malformed PagedLOD does the code fall back to getting the last child in the list.
"
This was easy to implement simply by overriding IntersectionVisitor::apply(PagedLOD). My question is: Are there any opinions on whether this should be the default behavior? If it makes sense, I will submit the change; if not, no worries."
Performance tests on big models did not indicate any performance penalty in using doubles over floats,
so the move to doubles should mainly impact precision improvements for whole earth databases.
Also made improvements to osgUtil::PlaneIntersector and osgSim::ElevationSlice classes
