Question on tree-walking and mutually-recursive types

[Paolo Bonzini]

> I don't see how this loop will work, except when heavily changed to 
> basically do the O(n*2) every-element to every-element test.

At a glance, Zack's code indeed looks like it's worst case O(n+m*2) 
where n is the length of the list and m is the length of the cycle.

Paolo