Open levonavagyanagility opened 2 years ago
If compiled in debug, instead the assert here is hit: https://github.com/flexible-collision-library/fcl/blob/master/include/fcl/narrowphase/distance-inl.h#L163
It looks like in certain cases the distance measurements can come up as negative, but when the collision call returns no contact points it overruns bounds.
It's assumed that there must have been contact points if the distance was negative, but the two functions have different tolerances, so it ends up reporting as not in collision but it ends up segfaulting when accessing a contact point at size_t::max here: https://github.com/flexible-collision-library/fcl/blob/master/include/fcl/narrowphase/distance-inl.h#L181
size_t::max
If compiled in debug, instead the assert here is hit: https://github.com/flexible-collision-library/fcl/blob/master/include/fcl/narrowphase/distance-inl.h#L163
It looks like in certain cases the distance measurements can come up as negative, but when the collision call returns no contact points it overruns bounds.
It's assumed that there must have been contact points if the distance was negative, but the two functions have different tolerances, so it ends up reporting as not in collision but it ends up segfaulting when accessing a contact point at
size_t::max
here: https://github.com/flexible-collision-library/fcl/blob/master/include/fcl/narrowphase/distance-inl.h#L181