Fix compiler warnings and fail on warnings in future

[rhaschke]

No description provided.

[v4hn]

Wouldn't we want -pedantic -std=c++14 as well?
Also I would have expected issues with upstream library headers...
Are all potentially-problematic ones included as -isystem now?

[rhaschke]

We never used -pedantic in the past and as our target compilers are gcc + clang, I personally don't mind allowing non-standard compiler extensions.

[codecov]

Codecov Report

Merging #2687 (7e90e6d) into master (722b97f) will increase coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #2687      +/-   ##
==========================================
+ Coverage   60.55%   60.56%   +0.01%     
==========================================
  Files         402      402              
  Lines       29601    29601              
==========================================
+ Hits        17922    17924       +2     
+ Misses      11679    11677       -2     

Impacted Files	Coverage Δ
...anning/planning_pipeline/src/planning_pipeline.cpp	63.64% <ø> (ø)
...meterization/work_space/pose_model_state_space.cpp	83.34% <0.00%> (+1.29%)	⬆️

---

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 722b97f...7e90e6d. Read the comment docs.

[v4hn]

We never used `-pedantic` in the past

I believe so, but we also never compiled with `-Werror`. I see `-pedantic` simply as an extension from "warning-free" to "warning-free according to specified standard" here.

and as our target compilers are gcc + clang,

I agree, but some users provide fixes for MS Visual Studio as well (see upcoming maintainer meeting). (Not that MSVS would be standard conform...)

I personally don't mind allowing non-standard compiler extensions.

We [explicitly disable extension support in our compile instructions](https://github.com/ros-planning/moveit/blob/master/moveit_core/CMakeLists.txt#L7-L8) and I see no reason not to actually enforce this on CI (unless huge chunks would start to fail).

[rhaschke]

We also never compiled with -Werror.

We didn't do this explicitly but implicitly: Travis CI had a WARNINGS_OK=false variable, which made the build fail on any warnings (not only compiler warnings).

I see no reason not to actually enforce this on CI

We can give it a try, yes.