Segfault when wrapping an anon obj in more than one layer

[lkuper]

The call to my_c.foo() in this program segfaults:

fn main() {

    obj a() {
        fn foo() -> int {
            ret 2;
        }
    }

    auto my_a = a();

    auto my_b = obj() {
        with my_a
    };

    assert (my_b.foo() == 2);

    auto my_c = obj() {
        with my_b
    };

    assert (my_c.foo() == 2);
}

[lkuper]

Poking around in gdb reveals that a forwarding fn is being called in an infinite loop. Two forwarding fns are being generated. In the executable I'm looking at now, the one for my_b.foo() is called forwarding_fn1 and the one for my_c.foo() is forwarding_fn13.

First, the call to my_b.foo() takes us to forwarding_fn1, which takes us to my_a.foo(). This works fine. Then, the call to my_c.foo() takes us to forwarding_fn13, which takes us to fowarding_fn1. But then we get stuck on forwarding_fn1.

So, we don't always get stuck on forwarding_fn1. We only seem to get stuck when we arrive at it via another forwarding fn. We call my_c.foo(); my_c protests and says "oh, no, I'm not self, my_b is!"; we head to my_b.foo(), but since my_c has already insisted that my_b is self, we don't hear my_b telling us that actually my_a is self. I'll look at this some more when I get a chance to talk about it with graydon.