On 27/09/2016 02:53, Ryan Gonzalez wrote:

I'm not saying Boost.Build is bad. I love it. It's just that it doesn't leave the best first impression, and lots of people give up after the first try.

I didn't, I gave up after many month of pain an suffering, debugging and a fix that ended up in the trashcan. My two last attempts ended up in a hanging b2 (the answer I got the first time went along the lines of "well, Intel should fix it's mpi engine to fit our completely unknown tool.", is not that useful in the world I live in. Second time, having a background script sending appropriate signal to the build process ended up qualifying as a convenient solution, didn't bother reporting the issue).

CMake might be (well, is) a piece of junk, but it's a documented , supported and widespread piece of junk. And I'd probably use AutoTool over b2 anyway. At least I have a failsafe on scripting when everything else fails.
Not to mention I do not work in a C++ only universe, and support for Fortran and most third party tools seems to be kind of limited.

But yes, documentation would be a nice improvement to consider if world domination is under consideration.

Performances is not such an issue when you have parallel build, although CMake seems to have a faster startup time for me.