[kaffe] Bug Report

Timothy Stack stack@cs.utah.edu
Thu, 25 Apr 2002 12:46:45 -0600 (MDT)


> > will break because the top of the stack doesn't change between the first
> > and second lock.  Therefore, the first unlock will completely free the
> > lock since the vm can't tell the difference between the two.  So, other
> > than pushing something on the stack on every synchronized, i don't think
> > theres much you can do without rewriting the whole locking system.
> 
> Wow, that sounds too bad to be true. Can we solve this with
> a recursion counter or something?

Yeah, but that ends up being so much like the latte/ibm locking approach
it seems silly to not just go all the way.  Also, it wouldn't fix any
problems the stack approach has with optimizations (delayed pop and
so on).

> -Archie

tim stack