[dozzie]

> Doesn't Turing completeness require the ability to modify an arbitrary amount of memory?

Why would it? Lambda calculus is nothing about modifying anything.

And GNU make (though with GNU extensions) is certainly Turing-complete: https://github.com/shinh/makelisp

[Sharlin]

In case of lambda calculus the equivalent condition is being able to evaluate arbitrarily large expressions. This sort of a Makefile-based evaluator could only do strings up to some hardcoded finite length.