Usage of sane build system

[scarabeus]

Hello, I am gentoo maintainer of boinc package currently and the new version fails me again in several ways (usage of internal libs, automagicness, etc.).

So I would like to ask you if you would accept cmake-based build system based on some branch you have or trunk, what ever would be best.

Why cmake? Because i am one of cmake maintainer too so i am quite solid there and in autotools i am not that good :] So i am definetly sure i could be able to write the cmake system, but I need to know if it is worth doing, you accept it, or i should keep whacking the autotools pretty badly with each release.

For starters i am capable of writing the linux build on it, where since cmake is multiplatform it should be quite easy to port it to osx and windows.

For simple example how cmake build system looks please point your eyes to ​http://repo.or.cz/w/tuxanci.git where i am trying to write simple and easy to understand build structure in cmake, so others can copy/learn from it.

Cheers

[Nicolas]

I have a CMake build system for the client here: ​http://synecdoche.googlecode.com/svn/branches/cmake/

Untested for a long while, and won't work directly for BOINC (because all server code is removed from there). And only seriously tested in Linux, did few tests on Windows, and didn't work on Mac. And it's based on old code (there may be new .cpp files to add by now).

I also wrote a CMakeLists.txt for the server code (also standalone, with client code removed). It compiles, but doesn't leave files where make_project expects them, so it doesn't really "work". And it's not publicly available at the moment.

Either way, I don't think it will be so easy for you to convince BOINC devs to completely switch build systems ;)

[scarabeus]

That is the reason why i am asking, i can write something, but definetly i wont test all features, so i would like to know devs opinion if they are interested. Packaging current boinc is quite big pita so anything that improves it will make me happy.

[Eric Korpela]

"but definetly i wont test all features" is probably the reason I'll say no. You also probably aren't willing to test it on flavors of BSD, Solaris, and other assorted operating systems. If you were volunteering to take over maintenance of the build system, I'd say go for it. As proposed, it's just throwing an incomplete build system I don't have any experience with in my lap and calling it an improvement.