--- In [email protected], "entropyreduction" 
<alancampbelllists+ya...@...> wrote:
>
> Problem with setting a vector might be: when do you look at it?
> If all goes well, all goes well.  If all goes badly and powerpro locks up, 
> there's no way to look at the trace vector.  
> 
> However, can't hurt to try.  Will do in a day or so.


True.  But when I posted, most people were saying the version they had worked 
once, but could not be run again.  It would be interesting to know if there is 
something consistently different about the order of events in the cases this 
happens.  Sheri also reported getting inconsistent results by starting the 
program a different way.  Not sure if that was truly the cause, or if it was 
just random luck in some race condition.  But I personally have more luck with 
debugging traces guided by theorizing, then from just theorizing.

For the case when it locks up, you could have a setup call like 
com.dump(stop#)
which would stop and dump at step number stop# if stop#>0.  Start with 0 and if 
it blocks, tell people to keep working backwards from n until you get a dump 
(where n is highest step you have set up in your code.)  Of course do this 
several times to find a pattern. 


Reply via email to