Fix error: redefinition of ‘struct eigenpy::NumpyEquivalentType<...

[sloretz]

The ROS Noetic armhf build for eigenpy is failing - likely caused by #455, and it's causing regressions blocking the Noetic Sync. May I request a release to Noetic when this PR or another fix for the issue is merged?

This fixes the issue by using SFINAE and partial template specialization(as suggested here) to remove the specializations of NumpyEquivalentType<long long> and NumpyEquivalentType<unsigned long long> when they're the same as int64_t and uint64_t respectively.

https://build.ros.org/view/Nbin_ufhf_uFhf/job/Nbin_ufhf_uFhf__eigenpy__ubuntu_focal_armhf__binary/96

23:06:42 /usr/lib/ccache/c++  -Deigenpy_EXPORTS -I/tmp/binarydeb/ros-noetic-eigenpy-3.7.0/.obj-arm-linux-gnueabihf -I/tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include -isystem /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/.obj-arm-linux-gnueabihf/include -isystem /usr/include/eigen3 -isystem /usr/lib/python3/dist-packages/numpy/core/include -isystem /usr/include/python3.8  -Wno-long-long -Wall -Wextra -Wcast-align -Wcast-qual -Wformat -Wwrite-strings -Wconversion -g -O2 -fdebug-prefix-map=/tmp/binarydeb/ros-noetic-eigenpy-3.7.0=. -fstack-protector-strong -Wformat -Werror=format-security -DNDEBUG -Wdate-time -D_FORTIFY_SOURCE=2 -fPIC   -Wno-conversion -std=gnu++14 -o CMakeFiles/eigenpy.dir/src/solvers/preconditioners.cpp.o -c /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/src/solvers/preconditioners.cpp
23:06:50 In file included from /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/fwd.hpp:87,
23:06:50                  from /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/solvers/BasicPreconditioners.hpp:11,
23:06:50                  from /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/src/solvers/preconditioners.cpp:8:
23:06:50 /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/numpy.hpp:143:8: error: redefinition of ‘struct eigenpy::NumpyEquivalentType<long long int>’
23:06:50   143 | struct NumpyEquivalentType<long long> {
23:06:50       |        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
23:06:50 /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/numpy.hpp:115:8: note: previous definition of ‘struct eigenpy::NumpyEquivalentType<long long int>’
23:06:50   115 | struct NumpyEquivalentType<int64_t> {
23:06:50       |        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
23:06:50 /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/numpy.hpp:147:8: error: redefinition of ‘struct eigenpy::NumpyEquivalentType<long long unsigned int>’
23:06:50   147 | struct NumpyEquivalentType<unsigned long long> {
23:06:50       |        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
23:06:50 /tmp/binarydeb/ros-noetic-eigenpy-3.7.0/include/eigenpy/numpy.hpp:119:8: note: previous definition of ‘struct eigenpy::NumpyEquivalentType<long long unsigned int>’
23:06:50   119 | struct NumpyEquivalentType<uint64_t> {
23:06:50       |        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
23:06:50 make[4]: *** [CMakeFiles/eigenpy.dir/build.make:66: CMakeFiles/eigenpy.dir/src/solvers/preconditioners.cpp.o] Error 1

I tested that eigenpy builds with this PR on an amd64 machine, but that of course won't show the build failure. Instead I made sure the method for the fix works using godbolt.org with the 32bit ARM GCC 9.3.0 compiler option with the code below:

#include <cstdint>
#include <type_traits>

template<typename S, typename Enable = void>
struct FoobarType{};

template<>
struct FoobarType<int64_t>{};

template<typename S>
struct FoobarType<S, 
std::enable_if_t<
!std::is_same<int64_t, long long>::value &&
std::is_same<S, long long>::value>>{};

template<>
struct FoobarType<uint64_t>{};

template<typename S>
struct FoobarType<S, 
std::enable_if_t<
!std::is_same<uint64_t, unsigned long long>::value &&
std::is_same<S, unsigned long long>::value>>{};

int main() {
    FoobarType<long long> bar;
    FoobarType<int64_t> baz;

    FoobarType<unsigned long long> ubar;
    FoobarType<uint64_t> ubaz;
}

[sloretz]

@wxmerkt Mind reviewing this one? This bug is causing regressions in Noetic on armhf, and that's blocking the Noetic sync.

[wxmerkt]

Thank you for putting this together and the ping. Looks good - let's wait for CI to pass

[wxmerkt]

Good from my end. @jcarpent is this good to merge?

[jcarpent]

Thanks a lot, @sloretz, for providing this fix, and thanks a lot, @wxmerkt, for reviewing this PR carefully.
I will provide a new release soon.