RE: [Tm4j-developers] TMAPI tm.getObjectById (id_association_role) -> StackOverflowError

[Kal A. <ka...@te...>]

Thanks for the patch Lars. I've now applied a modified version to CVS.
Handling 0 players is just a case of creating a wrapped AssociationRole
object with a null player. However it is not really possible for the Wrapper
to handle the case where a Member has >1 player, so the code now throws a
TMAPIRuntimeException with a suitable message - not a perfect solution, but
I think it's the best one given the constraints of TMAPI.

Cheers,

Kal

> -----Original Message-----
> From: tm4...@li... [mailto:tm4j-developers-
> ad...@li...] On Behalf Of Lars Heuer
> Sent: 21 December 2005 09:56
> To: George Tryfon
> Cc: tm4...@li...
> Subject: Re: [Tm4j-developers] TMAPI tm.getObjectById
> (id_association_role) -> StackOverflowError
> 
> Hi,
> 
> [...]
> >> org.tmapi.core.TopicMap tm;
> >> AssociationRole ar = (AssociationRole)tm.getObjectById(idAssocRole);
> 
> >> java.lang.StackOverflowError
> > [...]
> >> if the idAssocRole is not correct, the call gives null,
> >> but if the idAssocRole is correct, StackOverflowError
> 
> I think I've found the failure.
> The method
>     Object wrap(Object obj, TMAPITopicMapImpl tm)
> 
> forgets an argument (the player) if the "obj" is a Member instance and
> calls itself again and again. It should call
>     AssociationRole wrap(org.tm4j.topicmap.Member obj,
>                          org.tm4j.topicmap.Topic player,
>                          TMAPITopicMapImpl tm)
> 
> Maybe the attached patch helps.
> Note, that the patch is only a hack (see the TODO annotation) and is
> not tested.
> 
> Best regards,
> Lars
> --
> http://semagia.com