Re: [Help-smalltalk] [PATCH] kernel: Check the Semaphore before queuing the interrupt

[Paolo Bonzini]

Il 20/01/2014 20:30, Holger Hans Peter Freyther ha scritto:
> On Mon, Jan 20, 2014 at 05:30:52PM +0100, Paolo Bonzini wrote:
> > The patch looks good.
> >
> > Alternatively, what about this?  It is a bit clearer to me, but the
> > logic is pretty much the same.
>
> Thanks!
>
> >                  (self timedWaitOn: sem) ifTrue: [
> > -                   proc signalInterrupt: (Kernel.TimeoutNotification on: self).
> > +                   timeout = 0 ifTrue: [
> > +                       proc signalInterrupt: (Kernel.TimeoutNotification on: 
> > self)].
> >                  ].
>
> If we ever get the execution of multiple native processes the timeout
> code and the aBlock ensure:[] could run at the same time. How would we
> make that mutual exclusive? add a lock/token then?

We probably would need syntax for atomic compare and exchange of OOPs.

Paolo