MGI::setStartState: Only fetch current state when new state is diff

[rhaschke]

Fixes #2774.

[codecov]

Codecov Report

Merging #2775 (f2c0ce9) into master (8ff9cdf) will decrease coverage by 0.01%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master    #2775      +/-   ##
==========================================
- Coverage   60.16%   60.16%   -0.00%     
==========================================
  Files         356      356              
  Lines       31321    31323       +2     
==========================================
  Hits        18841    18841              
- Misses      12480    12482       +2     

Impacted Files	Coverage Δ
.../move_group_interface/src/move_group_interface.cpp	46.94% <0.00%> (-0.08%)	⬇️

---

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 8ff9cdf...f2c0ce9. Read the comment docs.

[v4hn]

Obvious fix. Thanks for pointing it out @werner291
Your use-case is quite non-standard though and I would not be surprised if you found other similar issues.

[werner291]

Your use-case is quite non-standard though and I would not be surprised if you found other similar issues.

It is? I was informed that MoveIt was the de-facto standard in terms of motion planning and benchmarking a new planner, and that I would probably get some questions at review time (when submitting something to a conference) if I didn't use it.

[v4hn]

It is? I was informed that MoveIt was the de-facto standard in terms of motion planning and benchmarking a new planner, and that I would probably get some questions at review time (when submitting something to a conference) if I didn't use it.

You would get some from me I'm sure 😄
It's very useful for users to have an open source MoveIt implementation of your planner available!

What I meant is that the full MoveIt framework (which you can control through the MoveGroupInterface) usually connects to a simulated or real robot.
For pure benchmarking, moveit_ros_benchmarks, robowflex and others already provide wrappers so that you do not have to call these user-level control methods.

[werner291]

Oooh, shiny! I'll be sure to go take a look. Especially seeing that robowflex is from Kavrakilab.

Gosh, there sure are a lot of layers to all of these systems... Of course, I'm speaking with the benefit of hindsight, obviously, before we actually figured out what we want to do with all of this.

[v4hn]

Especially seeing that robowflex is from Kavrakilab.

That being said, I'll point out that the only maintainer is a PhD student there. He does a great job though. :-)

Gosh, there sure are a *lot* of layers to all of these systems... Of course, I'm speaking with the benefit of hindsight, obviously.

Tell me about it. You don't know half of it :-)