AABB bounds == integer max

I think I have found a rather dubious bug.

The bug shows its face as a large memory leak that consumes all the memory on my computer and freezes in just a few seconds. The bug was difficult to reproduce because it only occurs after about an hour of simulation. After some searching the issue was being caused by a sphere I created in the simulation that I (probably should not have) left fall forever. After a large amount of time the position of the ball becomes ~ (10, -536870949.07..., 10). When the y value is divided by cellsize in dxHashSpace::collide it returns integer_max. Then when the loop is run over the dbounds in the y dimension, yi= integer_max can never be greater than dbounds[3] which is also integer_max. It loops very quickly allocating Nodes and my computer runs out of memory in a few seconds.

I am not sure I would consider this a bug because it is an obvious misuse of the software but I think a least a warning might be nice.

Comments (2)