I'm not a CMake user, and don't speak the language as it were, so while I'm not totally against it, such a submission would have to come with at least some minimal tests (hooked up to the CI system) to verify that it does actually work and pull in all the dependencies it needs.

That is great. I have just the stuff. Pull request is coming.

I believe it would be great to link this issue to PR #143, so it will close just as that PR will be merged.

@nemothenoone : Is there a chance that the CMake system can produce a C++-20 module for Boost.Multiprecision, without jeopardizing backwards compatibility for users who consume the header-only version?

@NAThompson Definitely. Kitware guys already has modules support implementation going, there is an example available: https://github.com/mathstuf/cxx-modules-sandbox. So, I'ill simply introduce same mechanism usage in here.

Or are you talking about generating the module definition itself?

Or are you talking about generating the module definition itself?

Sadly I haven't yet educated myself enough on how to do this to even understand this comment! What I had envisioned was that CMake would somehow make putting a module up on conan.io easier . . .