Re: [Swarm-Support] Lisp saving objects subclassed from Arguments_c

[Marcus G. Daniels]

Paul Johnson wrote:
> I have wondered about the following problem for some time.  When a 
> model is started with initSwarmArguments, and the user supplies the 
> argument/parameter class "Parameters", there is trouble when one tries 
> to save that argument object with lisp archiver.  
[..]
> That causes a segmentation fault and the backtrace points at the 
> lispOutVars method.
I suspect it is because there is a function variable in Arguments_c, 
that serialization code doesn't know how to handle.   I haven't checked 
that is indeed where the crash is, but it's the only difference I can 
think of offhand.  Swarm should really just skip over those. 

With Java, only public variables are serialized, which makes it easy to 
accommodate situations like this.   For Objective C, I think your 
approach of having a separate object to save is cleaner anyway. 

Incidentally, in true Lisp, it is possible to serialize functions and 
methods (and also to mutate them as data).  This makes it easy to move 
code around in a distributed simulation environment.