the i at the second for loop should be reinit to be zero

[dahuacai]

Issue Place:
navigation2/nav2_theta_star_planner/src/ThetaStar::initializePosn(int size_inc)
raw code:
`void ThetaStar::initializePosn(int size_inc)
{
int i = 0;

if (!node_position_.empty()) {
for (; i < size_x_ * size_y_; i++) {
node_position_[i] = nullptr;
}
}

for (; i < size_inc; i++) {
node_position_.push_back(nullptr);
}
}`

Question:
the i at the second for loop if need reinit to be 0 ? otherwise the second loop may not work some time and access transgressions can also occur at other place

[SteveMacenski]

Incorrect - look at the algorithm again. Any element within node_position_ is set to nullptr if it exists. For the case that there are more requested (e.g. size_inc > size_x_ * size_y_) then we need to add in new elements to node_position_, which is what's happening in the second loop.

[dahuacai]

but size_inc = curr_size_y * curr_size_x - size_x * size_y; meanwhile the first for loop size_x,size_y is old . we should reset old size_x * size_y size data to null and push back new size size_inc, so the i at second for loop should be re init to be zero ,can we push back the newly added size_inc num data .if not set to be zero i test happened memory leak. so the second loop should be "for (i = 0; i < size_inc; i++) {"

[dahuacai]

from line 240 initializePosn(curr_size_y * curr_size_x - last_size_y * last_size_x); we known that input param size_inc should be incremental data and it may below to size_x_ * size_y_ and it is diff to curr_size_y * curr_size_x

[dahuacai]

@SteveMacenski

[SteveMacenski]

Agreed, I did not see that. This looks like it would work for the original initialization (e.g. map size is zero and then becomes non-zero) & at steady state while navigating with localization. I think the issue you bring up would only appear if navigating while performing SLAM so the map is increasing in size over time.

I read this that it should cause issues when the map is expanding (contracting looks OK). Do you see issues in that case which is why you're filing the ticket? That was missing context that there was a detected crash or issue at runtime.

A PR to resolve with a unit test would be appreciated @dahuacai !

[dahuacai]

@SteveMacenski yes we use the theta star with the map is always expanding(clean robot use online slam),so i find the problem.

[SteveMacenski]

Can you open the PR for a fix?

[SteveMacenski]

Just following up here! I can too, but its always nice to have folks contribute their own work :-)

[SteveMacenski]

#3597 resolves