fix segfault at robot-model update (#255 revisited)

[corot]

Resurrecting #255:

We found exactly the same issue #245, reproduced in the same way, and I can confirm that @dorezyuk's PR #255 fixes it

I open a new PR because:

• Need to solve some conflicts
• I also want to patch noetic and ros2
• I'm not sure about the need for commit f4c43757fc59f647ae7f5c305cb7036056764934

About @amakarow's comment, I think the code becomes cleaner now (if nothing else, we have almost 4 to 1 deletions!)

@dorezyuk, can u explain why the lock is needed? Thanks, and thanks for the fix!

[siferati]

We need the lock because otherwise you're modifying a field of the config object outside the protection of the mutex. The other fields are protected by the mutex inside TebConfig::reconfigure, but robot_model is not since you set it outside in TebLocalPlannerROS::reconfigureCB.

It's the same problem we had with DWA a while ago

[corot]

We need the lock because otherwise you're modifying a field of the config object outside the protection of the mutex. The other fields are protected by the mutex inside TebConfig::reconfigure, but robot_model is not since you set it outside in TebLocalPlannerROS::reconfigureCB.

It's the same problem we had with DWA a while ago

added

[zst1406217]

This is the exact problem I have in my research. When I reconfigure the parameters, the segfault occurs occasionally, but not from the beginning, this has confused me a long time. I prove this is helpful!

[corot]

This is the exact problem I have in my research. When I reconfigure the parameters, the segfault occurs occasionally, but not from the beginning, this has confused me a long time. I prove this is helpful!

Thanks for reporting. Merged