#fork and deterministic resumption of the resulting process

Andreas Raab andreas.raab at gmx.de
Tue Feb 5 22:52:30 UTC 2008


Louis LaBrunda wrote:
> I haven't shown any code because VAST and Squeak are different enough in this
> area that my #forkReady code would not be helpful but if there is still a race
> condition is should be greatly reduced because the #forkReady code puts the
> newly created process in the queue of ready processes and returns.  The setting
> of the instance variable would have to be interrupted immediately by a task
> switch, caused by a timer or some other interrupt.

Which is *precisely* the problem we're talking about. Can you imagine 
how annoying it is that the code works in all your tests and when you 
ship it to a couple hundred customers it blows up just for statistical 
reasons?

Cheers,
   - Andreas




More information about the Squeak-dev mailing list