Compiling needs -fpermissive on gcc 7.3.0

[gassmoeller]

Hi Natsuki, I have started looking into the code, and will collect problems that I find in these github issues.

First I ran into a compiling problem with gcc 7.3.0. I get the following error message:

init/GI.h: At global scope:
init/GI.h:8:22: error: ‘constexpr’ needed for in-class initialization of static data member ‘const double GI<Ptcl>::END_TIME’ of non-integral type [-fpermissive]
  static const double END_TIME = 1.0e+4;
                      ^~~~~~~~
In file included from main.cpp:15:0:
force.h:82:23: error: ‘constexpr’ needed for in-class initialization of static data member ‘const double STD::CalcGravityForce<TPtclJ>::G’ of non-integral type [-fpermissive]
   static const double G = 6.67e-11;
                       ^
In file included from main.cpp:14:0:
init/GI.h: In instantiation of ‘class GI<STD::RealPtcl>’:
main.cpp:37:10:   required from here
init/GI.h:8:22: error: ‘constexpr’ needed for in-class initialization of static data member ‘const double GI<STD::RealPtcl>::END_TIME’ of non-integral type [-fpermissive]
  static const double END_TIME = 1.0e+4;
                      ^~~~~~~~
In file included from main.cpp:15:0:
force.h: In instantiation of ‘class STD::CalcGravityForce<STD::EPJ::Grav>’:
main.cpp:77:77:   required from here
force.h:82:23: error: ‘constexpr’ needed for in-class initialization of static data member ‘const double STD::CalcGravityForce<STD::EPJ::Grav>::G’ of non-integral type [-fpermissive]
   static const double G = 6.67e-11;
                       ^
force.h: In instantiation of ‘class STD::CalcGravityForce<ParticleSimulator::SPJMonopole>’:
main.cpp:77:120:   required from here
force.h:82:23: error: ‘constexpr’ needed for in-class initialization of static data member ‘const double STD::CalcGravityForce<ParticleSimulator::SPJMonopole>::G’ of non-integral type [-fpermissive]
In file included from main.cpp:16:0:

Using the -fpermissive flags reduces these errors to warnings (then it compiles for me), but the more correct fix would of course be to use constexpr in front of these initializations.

[gassmoeller]

This should have been fixed by #4.