RE: Query on stacktrace management logic

[David Holmes]

A while back I wrote:
> In particular now there is the
> Class/VMClass split there's an implication that VM's shouldn't
> need to customise Class - but the static initialization issue
> would thwart that in many cases.

I take this back. We had a cyclic dependency that forced us to treat Class
as special during the initialization sequence. That dependency was
unecessary and was actually broken a while back - but I hadn't realized it.
All we need is to disable class initialization early in the boostrap
process, but still construct Class objects as needed. Once class
initialization is enabled then Class can be initialized just like any other
class.

Cheers,
David Holmes