Re: Bad news on the scheduling front

[Doug Donalson]

I'm stumped as to where to go from here.  I used to have a procedure that
allowed me to print out the event schedule but that was a LONG time ago and
I can't find it.  I know that you don't like comparrisions but until a light
bulb goes on for me that's about all I have.

First a description of the system.

At time zero, I initialize the system with 40 agents.  Each agent has die
event scheduled in the future.  It also gets a "doNothing" event that is
given a longer time than the death event so that the agent, when it dies,
must remove the doNothing event from the event schedule.  Finally, a single
event is scheduled at a future time to the modelSwarm that adds a new agent
to the system.  When that addNew Agent procedure is serviced, a new agent is
created and a new addNewAgent porcedure is added to the schedule.

I compile the code in Swarm 2.1 with all patched dll's added but the new
swarm.dll.  I then run the code.  The code should kill off the 40 original
agents but at the same time add new agents through the addNewAgents events.
This is what I see, with of course, the memory leak.  I have a print
statement in the addNewAgent procedure that allows me to see when it is
called.  The system dynamics should and do increase from 40 to 100
individuals and then wander about 100 with some stochasticity.

Now, I add the new swarm.dll, no recompile (although I tried that as well.)
I see the first call to addNewMussel from the print statement.  I never see
another call to addNewAgent, the 41 individuals just die off exponentially.

The only difference is the new swarm.dll.  Even the random number sequence
should not change, so the initial ordering of the events should be the same.
Do you have an idea as to where I can track it from here?  I have no way to
see whether that event is on the event schedule but not executed, or never
gets inserted.  (Note that the code is identical and it is clear that the
new event does get inserted with the original swarm2.1 swarm.dll.  If it
didn't, future calls to addNewAgent would not occur.)

I will keep thinking on this.  Also, specifically, how can I make this
problem clearer for you?

Cheers?

   D3
----- Original Message -----
From: Marcus G. Daniels <address@hidden>
To: <address@hidden>
Sent: Saturday, March 25, 2000 3:48 PM
Subject: Re: Bad news on the scheduling front


> >>>>> "D3" == Doug Donalson <address@hidden> writes:
>
> D3> That program actually would serve quite nicely
> D3> as a starting point I think, It has all the components necessary
> D3> to test the asynchronous/concurrent stuff.  I'll work on it.  I'll
> D3> also look over your bug report suggestions and see what I can do
> D3> about writing one.
>
> Cool, the programs tests/activity may have some utility as examples.
> (Batch-mode simulations that return 0 for success or 1 for failure.)
>
> Btw, I can't think of a reason the last fix would cause events not to run.
>
>                   ==================================
>    Swarm-Support is for discussion of the technical details of the day
>    to day usage of Swarm.  For list administration needs (esp.
>    [un]subscribing), please send a message to <address@hidden>
>    with "help" in the body of the message.
>


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.