SmacHybrid planner can crash to segmentation fault about navigation2 HOT 3 CLOSED

Thanks for quick response and the feedback! I will do this once I am back from my trip on Monday.

from navigation2.

I have been investigating this now for a bit. According to StackOverflow comment dereferencing the iterator returned by std::max_element when the input array is empty, can lead into segmentation fault. Running std::max_element(EmptyArray) returns EmptyArray.end(). I believe that is what happens in this case.

The C++ docs that I usually reference have some remark if that would be a problem that I see aren't listed there. But just being a sane person knowing STL, yeah your expectation there would align with mine. I think the documentation just had that case missing.

Can you adjust that to:

auto max_cost_it = std::max_element(node_costs.begin(), node_costs.end());
if (max_cost_it != node_costs.end() && *node_costs > max_cost) {

Does it go away? If so, submit a PR to main and I'll merge + backport it to all the distributions. Thanks for the note, I agree with your analysis that this is the culprit - but as you point out - we shouldn't have been able to get here with that empty if all's well above.

For myself I fixed this by returning early in similar fashion as on line 186. Simple if (d < sqrt_2) return AnalyticExpansionNodes(); and that seems to be working fine for me.

You can add that too! Why not both, I practice defensive programming.

from navigation2.

Merging imminent

from navigation2.