(2009) I have grown fond of approaches where at each call site, the caller can explicitly declare whether or not a call is a tail call. The writer of the code

(2009) I have grown fond of approaches where at each call site, the caller can explicitly declare whether or not a call is a tail call. The writer of the code knows best: whether the call is recursive and can consume potentially unlimited stack. If it does, use a slightly different syntax* to communicate the fact that the code is relying on tail calls. *: such as an attribute. See https://reviews.llvm.org/D99517 https://reviews.llvm.org/D99517