Subj : Re: Deadlock theorem To : comp.programming.threads From : David Hopwood Date : Sun May 08 2005 09:30 pm David Schwartz wrote: > "David Hopwood" wrote: > >>For the latter, what you really want is a static analysis, which by >>definition has no runtime cost (although it could be quite complicated). > > I don't believe that. Dynamic behavior is unpredictable. What a silly reason to disbelieve that static analysis of deadlock is possible. All nontrivial programs have dynamic behaviour, and that doesn't make static analysis or type checking in general impossible. I tried to work out in detail how to extend the Cyclone type system to prevent deadlock using heirarchical locking, but ran into (solvable, I think) technical problems due to the possibility of different lock names referring to the same lock. I'm on holiday for the next week, but I'll come back to this afterward. -- David Hopwood .