Ciro Santilli

> @royjacobson do you have build time comparisons for a couple big projects, with and without this PR? > > I forgot where I wrote this before, but splitting up...

> > but from my perspective, a serious cleanup of the file organization has to come first. > > Do you feel you could help with that? I'm afraid I...

> I think .inc is used sometimes, but may not get nice highlighting in some editors. fmtlib uses -inl.h, which I think is my favorite. /src then has cpp files...

> I'm happy to help with the CMake setup if this continues. :) Thanks, Henry, I might need it later on! Does the offer extends to quantum field theory help...

OK, the cmake help might be needed now XD The new failures are for `make test_cmake_build` which appears to test other projects using pybind11 via cmake. I tried to add...

OK, I'm going to do another round of coding now :-) > I don't like the idea of having an #include file, or *.cpp files being distributed as headers. Personally...

OK, I've converted all files to .cpp now, and previous testing state is unchanged. If after my last comment you still feel it is better to split in -inl.h +...

> I'm still in favor of `-inl.h` because: a) having a .cpp file in `/include` and shipped with your library in packaging is odd, b) compilers sometimes complain if you...

Thanks so much for the cmake update @henryiii !!! I squashed with my commit, and I also had to add a new commit to make things work with gem5 with...

OK, I split the second unrelated commit to https://github.com/pybind/pybind11/pull/2476 now.