Fix crash in the follow_actor example

[luca-della-vedova]

This PR fixes a sneaky segfault introduced in #858, while the fix works I'm open to better solutions if anything is recommended (it makes the function arguments a bit bulky).

The issue was that the previous version of the code was relying on an unfortunate coincidence related to the variable names for the entityPoses and trajectoryPoses data structures, specifically when updating them here the variables would refer to the local variables introduced in the beginning of the Update() function, here.
However, when during the refactor the code was moved into a new function, the Implementation class variables (i.e. this and this) with the same name are being silently used behind the scenes, the problem is that at the beginning of the Update() function there is a std::move call that moves from the member data structures into the local variables, hence it becomes a case of using a data structure after it has been moved, which causes a segfault for a null pointer exception in the data pointer.

To reproduce the bug / check the PR just run the follow_actor demo with and without the PR:
ign gazebo follow_actor.sdf

[codecov]

Codecov Report

Merging #958 (0e24087) into ign-gazebo4 (e95f17c) will increase coverage by 0.93%.
The diff coverage is 100.00%.

@@               Coverage Diff               @@
##           ign-gazebo4     #958      +/-   ##
===============================================
+ Coverage        66.13%   67.06%   +0.93%     
===============================================
  Files              243      243              
  Lines            18205    18207       +2     
===============================================
+ Hits             12039    12210     +171     
+ Misses            6166     5997     -169     

Impacted Files	Coverage Δ
src/rendering/RenderUtil.cc	48.67% <100.00%> (+6.59%)	⬆️
src/gui/plugins/entity_tree/EntityTree.cc	9.83% <0.00%> (-3.28%)	⬇️
src/SimulationRunner.cc	93.04% <0.00%> (-1.05%)	⬇️
src/rendering/SceneManager.cc	37.73% <0.00%> (+15.24%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 34eb437...0e24087. Read the comment docs.

[iche033]

nice, I can reproduce the crash and this fixes it for me too.

Adding a sensors system and a camera to the follow_actor integration test also causes the test to crash. I've added it in the fix/follow_actor_crash_test branch so we can test the rendering side as part of the integration test in the future. Can you take a look and merge if that looks good to you?

[adlarkin]

Thanks @luca-della-vedova for catching this! Sorry I didn't notice this when working on #858 😬

Adding a sensors system and a camera to the follow_actor integration test also causes the test to crash

Do we need to fix this as well?

[iche033]

Do we need to fix this as well?

oh I meant adding it to the test causes to crash before the changes in this PR, and confirmed that the test no longer crashes after this PR.

[luca-della-vedova]

nice, I can reproduce the crash and this fixes it for me too.

Adding a sensors system and a camera to the follow_actor integration test also causes the test to crash. I've added it in the fix/follow_actor_crash_test branch so we can test the rendering side as part of the integration test in the future. Can you take a look and merge if that looks good to you?

Done!

[iche033]

looks good to me!