Takatoshi Kondo

Thank you for creating the PR. If the PR resolves the issues raised by both @otegami and @saper, I will merge it. @otegami, could you please check it?

Thank you for reporting the issue. When you pass the flag `-DMSGPACK_USE_BOOST=OFF` to **cmake** then C++ preprocessor macro `MSGPACK_NO_BOOST` should be defined. https://github.com/msgpack/msgpack-c/blob/fb64ea0b68cda185f1e45b87713ba9cd7be3e394/CMakeLists.txt#L68-L82 And if MSGPACK_NO_BOOST is defined, chrono adaptor...

Could you try #1038 ?

No problem. I will release fixed version soon.

@flo-carrere , I'm not sure what you mean. Could you elaborate? MSGPACK_NO_BOOST is publish flag for users. See https://github.com/msgpack/msgpack-c/wiki/v2_0_cpp_configure#msgpack_no_boost--since-410 User can set the flag on their project. I think that...

I think I just understood what is the biggest difference of our concept. > The whole point here, is I think it is not acceptable to force the user to...

Another important point. > you could maybe rely on a templated version include/msgpack/sysdep.hpp.in that could be "patched" at build time to the correct output. This is opposite direction of C++...

Sorry for the late reply. I'd like to know more about the usecase. Could you write the minimal code example that demonstrate your usecase ? In the code, please mention...

Old A and new A cannot exist in the same program even if compiler doesn't report error. It is ODR violation.

Ok. Try this: ```cpp #include #include struct old_a { int version = 1; struct nest1 { struct nest2 { int oldx; int oldy; }; nest2 n2; }; nest1 n1; };...