2008/4/27  <bryce at kampjes.demon.co.uk>:
>
>  Run the following in a 3.10 image:
>    100 timesRepeat:
>         [processes := (1 to: 10) collect: [:each| [[(MessageTally spyOn: [(Delay forMilliseconds: 500) wait]) ] repeat] fork ].
>         (Delay forSeconds: 1) wait.
>         processes do: [:each| each terminate].]
>
>  It locks up the image, debuggers will start popping up at some stage.
>  This works on 3.9. I've run the code above in a fresh 3.10 image and
>  reproduced the problem.
>
>  I'll try to find time to investigate further. It's probably the final
>  showstopper to me upgrading to 3.10. The code above is reworked from
>  on of Exupery's SUnits which locks up the image intermittently.
>

AFAIK, since 3.9 there was some fixes in Delay & Semaphores.
Its hard to say, if its a new bug in Delay or Process & scheduler or
this is remnants of code (in MessageTally) which was not fixed and
relies on old behavior.

As for me, its no wonder, that given code refuses to work correctly.
The implementation of Processes and all what is around them is very
brittle, and given stress test exposing this. As you know there are
different hacks & workarounds to keep things running without modifying
VM.
Thanks for giving another reason why squeak threading model requires
serious review. :)

>  Bryce
>
>



-- 
Best regards,
Igor Stasenko AKA sig.