[aidenn0]

I've been following along the gc of oil.

It succeeds because it only solves the GC problems that Oil has (and I mean this as a high complement).

Oil is single threaded, code runs in loops that are well understood, and the code is generated from a strongly typed subset of Python.

The GC then is run only between loop iterations, so there is no need for stack scanning. You never have to worry about a root being in a temporary (from the point of view of the C++ compiler) since there are just a few locals in the function running the loop, and any variables local to the loop are logically dead between iterations.

Since a goal of Oil is portability, not scanning registers and stacks is very important. Getting this right when the GC could be invoked at any allocation is potentially intractable with mypy semantics at least.

[drmeister]

Ah - thank you. We are looking for a GC that supports (in no particular order): (1) conservative stack scanning, object pinning. (2) optionally conservative and precise on the heap. (3) compacting when precise. (3) weak pointers. (4) good multithreaded performance. (5) finalizers. (6) ability to create pools of different kinds of objects - especially cons cells (pairs).

[aidenn0]

Probably not coincidentally, that sounds like SBCL's cgc...