Workaround for building on clang and libstdc++ (part 2)

[czangela]

Enables compilation with clang 11.

[msmk0]

Hi @czangela, could you let us know in which environment you encounter this error (operating system, how are the external dependencies installed) and the exact compile error you encounter?

[paulgessinger]

Hi @czangela, could you let us know in which environment you encounter this error (operating system, how are the external dependencies installed) and the exact compile error you encounter?

Most likely LCG clang + libstdc++, as we've had before. We really need a build of that going again...

[czangela]

Hi @czangela, could you let us know in which environment you encounter this error (operating system, how are the external dependencies installed) and the exact compile error you encounter?

Hi @msmk0,
I forgot to mention that it is a follow-up.

There was a patch about it a few months ago, which content was then moved to a separate header file.

As far as we know the error occurs because clang and gcc standard header files differ a little bit, and mixing one compiler with the other's header files might introduce errors. Usually, it doesn't mean any problem. However, putting a (const) std::any (GeometryContext) type in a std::reference_wrapper (and moving that around) seems to trigger an error. compile_error.txt

(So clang and libstdc++ are not 100% compatible. It has to do something with the order of template instantiation. Or something like that.)

This is just a workaround for this. It is definitely not a good solution but it works.

[krasznaa]

Hi Paul,

We see this when using Intel's compiler (https://github.com/intel/llvm) on top of GCC 8's libstdc++ version. But @cgleggett was also reporting about some strange issues with a little different LLVM + libstdc++ combination. So yeah, it's very much a moving target. 😦

Also, don't miss that the code adds a missing include and a missing #pragma once statement as well. 😉

Cheers,
Attila

[codecov]

Codecov Report

Merging #343 into master will increase coverage by 0.03%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #343      +/-   ##
==========================================
+ Coverage   48.63%   48.66%   +0.03%     
==========================================
  Files         325      325              
  Lines       16356    16356              
  Branches     7588     7588              
==========================================
+ Hits         7954     7959       +5     
+ Misses       3093     3084       -9     
- Partials     5309     5313       +4     

Impacted Files	Coverage Δ
Core/include/Acts/Fitter/KalmanFitter.hpp	36.43% <ø> (ø)
...re/include/Acts/Material/SurfaceMaterialMapper.hpp	50.00% <ø> (ø)
...ore/include/Acts/Material/VolumeMaterialMapper.hpp	60.00% <ø> (ø)
Core/include/Acts/Propagator/AtlasStepper.hpp	71.42% <ø> (ø)
...lude/Acts/Propagator/DenseEnvironmentExtension.hpp	59.64% <ø> (ø)
Core/include/Acts/Propagator/EigenStepper.hpp	93.10% <ø> (ø)
...re/include/Acts/Propagator/StraightLineStepper.hpp	68.88% <ø> (ø)
...ude/Acts/TrackFinder/CombinatorialKalmanFilter.hpp	27.69% <ø> (ø)
Core/include/Acts/Vertexing/VertexingOptions.hpp	100.00% <ø> (ø)
...include/Acts/Vertexing/FullBilloirVertexFitter.ipp	98.69% <0.00%> (+0.65%)	⬆️
... and 2 more

---

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 35bcdc4...0f5e54c. Read the comment docs.

[paulgessinger]

Right, makes sense and LGTM