CGAL users discussion list

[alexei <>]

Hi all, 
I've recently stumbled upon a bug in CGAL's implementation of AABB trees. 
The bug occurs when you construct an AABB tree using a single primitive as
input (I used triangles); it seems that there's a common assumption that
there would be at least two primitives, and this is reflected in the
following two places: 
 - AABB_tree.h:499 - if there's only one primitive, a 0-length block of
memory is allocated for the tree 
 - AABB_tree_node.h:133 - if there's only one primitive the recursion is
incorrect since the switch handles the cases for two, three and more
primitives 

The latter bug (or their combination) causes a segmentation fault (or some
other colorful panic) on my machine. 
I'm running Kubuntu 11.04 (Natty) in VM over Windows 7; CGAL was installed
via the package manager and there was no special configuration that needed
to be done. I'm using the latest Eclipse for editing but compile using the
CMake-generated Makefile. 

Attached you'll find some simple code example that recreates the bug: 
http://cgal-discuss.949826.n4.nabble.com/file/n3555851/bug.h bug.h 

Cheers, 
Alexei

--
View this message in context: 
http://cgal-discuss.949826.n4.nabble.com/Bug-in-AABB-tree-implementation-tp3555851p3555851.html
Sent from the cgal-discuss mailing list archive at Nabble.com.