Re: RMI Interoperability Patches (Part2)

[Norbert Frese <postfach at nfrese dot net>]

Hi!

Thanks for applying my patches to Kaffe. I'm very busy with my job this
month, but i will give the patched kaffe a try as soon a can. I think i
won't need CVS-write access right now (to be honest: i'm new to CVS and
might mess things up in your repository ;-)).

Maybe we should ask the JBoss people to test the new RMI things in
kaffe/classpath/libgcj. What do You think?
Plus: Are there any other RMI experts out there - who want to have a
look at the current state of the RMI-libraries.

Yust to mention it: There seems to be another little garbage-collection
issue in the current RMI implementation: A RMI server object, which is
not referenced otherwise in the current runtime gets cleaned away by the
GC - although another RMI "partner" holds a reference to the object.
This is a rare situation, because RMI objects usually are referenced
somehow in their local environment... I noticed the problem in a
callback setting. Seems to be an issue with the DGC.

regards,

norbert




On Wed, 2004-03-10 at 18:41, Dalibor Topic wrote:
> Norbert Frese wrote:
> > Source of the new RMIIncomingThread class, which has to be 
> > put into gnu/java/rmi/server. 
> > 
> 
> Checked in into kaffe's CVS, sorry for the long delay. It should go 
> straight into GNU Classpath from there, now that your paperwork is cleared.
> 
> We could give you CVS access for Kaffe if you want to hack more on it. 
> RMI seems to be quite hot for OfBiz and JBoss type of applications, 
> which seems to be what everyone wants to run on free rutimes these days [1].
> 
> cheers,
> dalibor topic
> 
> [1] Beside eclipse. ;)
>