> I don't think this proof invalidates any of the excellent work that has been put into Meson. If anything, this shows that build configuration is a hard enough

> I don't think this proof invalidates any of the excellent work that has been put into Meson. If anything, this shows that build configuration is a hard enough problem that requires a turing-complete language. Most features have gone into Meson because somebody actually needed them. And for this reason, Meson serves as an estimate of the required complexity. I think it's true given the goal to support a specific set of requirements of modern build systems. The problem is that there is no separate dependency management protocol to target, and there is no separate projects description configuration format. That means build systems end up trying to interoperate with, or even implement large portions of, those kinds of subsystems.