On Fri, Dec 4, 2020 at 12:55 PM Miro Hrončok <mhroncok@redhat.com> wrote:
> For what's it worth I think that packages that only use make via cmake should
> not have an explcit dependency on make. Packages that use make directly should
> have an explicit dependency on make (even if they already BR cmake).
Does that mean that if the requires: make that is currently
in the cmake package that was added due to rhbz#1862014
is removed (as has been proposed since ninja is a valid
alternative) that you are fine with packagers having to go
fix their packages? Or would you expect another pass
across all packages to add a BR: make to be done? If
the later, it makes sense to me to do it once (when
someone is willing to do the work) to prepare for any
cmake cleanup(s).
I think the CMake package should always provide at least one build system as a dependency, otherwise someone could install cmake and not be able to fully use it. This is a case where we can be "opinionated" in the CMake package and set a default build system for cmake in the package since it is trivial for the user to change the generator at runtime with the -G option.
Then any consumers of the %cmake_* macros should rely on the choice of the CMake package for the system unless they want to override it themselves.
-Ian