That looks a lot like what Zig does. One problem with the proposed defer there is that it has no way to interact with stack unwinding, which is part of the pla

That looks a lot like what Zig does. One problem with the proposed defer there is that it has no way to interact with stack unwinding, which is part of the platform ABIs for handling exceptions. Maybe some alternative syntax, such as "defer_finally" could make the defer block act as both a regular defer block, and also explicitly add that block to the stack unwinding chain.