You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geronimo.apache.org by gianny DAMOUR <gi...@hotmail.com> on 2003/10/02 17:19:35 UTC
ClassSpace - parent CL
Hello,
I was just trying to solve the problem of Matt Kurjanowicz, which seems to
be a weird CL problem and during my investigations I have discovered three
points:
- the Eclipse set-up I use to simulate a running server is simply flawed
(grrr...);
- the Resource Adapter Deployer I have submitted does not work, when
deployed within a running instance launched via maven run:main (grrrrr...);
and
- the parent CL of a ClassSpace is the TCCL if defined or the CL of the
ClassSpace class.
Regarding this last point, I really find it weird for the following reason:
If one creates a ClassSpace for a deployment unit, say a resource adapter,
and if I want to define a specific parent for such a ClassSpace, I will need
to set the TCCL. However, because this TCCL is outside our control between
the moment this TCCL is set and the actual creation of my ClassSpace (other
deployment planners could wish to set their own TCCL), it is impossible to
know for sure which CL will be the parent of the deployment unit ClassSpace.
So, I would like to know if one can add to ClassSpaceMetadata an attribute
referencing the name of the parent ClassSpace of the ClassSpace to be
created via this ClassSpaceMetadata. This way, a deployment unit could
explicitely defines its direct CL and the parent of this CL.
I have applied this fix and it solved my resource adapter deployer problem.
More accurately, the parent CL of the ClassSpace containing the resource
adapter archives is the geronimo.system:role=ClassSpace,name=System
ClassSpace (defined during the boot).
Gianny
_________________________________________________________________
MSN Search, le moteur de recherche qui pense comme vous !
http://search.msn.fr/
Re: ClassSpace - parent CL
Posted by Dain Sundstrom <da...@coredevelopers.net>.
On Thursday, October 2, 2003, at 12:45 PM, Aaron Mulder wrote:
> Dain,
> Can you say how the revised code fixes the issue? Does it do
> something similar to what Gianny described, or just default to a
> different
> parent?
I don't remember. I wrote it last week and have been sucked into
writing slides. The idea I did have wast to make this work more like
JMX class loading instead of the traditional purely hierarchal
structure in the JDK. They are both good approaches for different
types of deployments, so I was working on a hybrid.
> Would it make sense to apply Gianny's change for now and then you
> can apply your update when you have time?
Yes. We can always change things later (the project is young).
-dain
Re: ClassSpace - parent CL
Posted by Aaron Mulder <am...@alumni.princeton.edu>.
Dain,
Can you say how the revised code fixes the issue? Does it do
something similar to what Gianny described, or just default to a different
parent?
Would it make sense to apply Gianny's change for now and then you
can apply your update when you have time?
Thanks,
Aaron
On Thu, 2 Oct 2003, Dain Sundstrom wrote:
> I have a new version of the Class Space code on my laptop that fixes
> this issue, but I'm preparing for our upcoming training course. I
> should have it checked in late next week or the week after.
>
> -dain
>
> On Thursday, October 2, 2003, at 10:19 AM, gianny DAMOUR wrote:
>
> > Hello,
> >
> >
> > I was just trying to solve the problem of Matt Kurjanowicz, which
> > seems to be a weird CL problem and during my investigations I have
> > discovered three points:
> >
> > - the Eclipse set-up I use to simulate a running server is simply
> > flawed (grrr...);
> > - the Resource Adapter Deployer I have submitted does not work, when
> > deployed within a running instance launched via maven run:main
> > (grrrrr...); and
> > - the parent CL of a ClassSpace is the TCCL if defined or the CL of
> > the ClassSpace class.
> >
> > Regarding this last point, I really find it weird for the following
> > reason:
> >
> > If one creates a ClassSpace for a deployment unit, say a resource
> > adapter, and if I want to define a specific parent for such a
> > ClassSpace, I will need to set the TCCL. However, because this TCCL is
> > outside our control between the moment this TCCL is set and the actual
> > creation of my ClassSpace (other deployment planners could wish to set
> > their own TCCL), it is impossible to know for sure which CL will be
> > the parent of the deployment unit ClassSpace.
> >
> > So, I would like to know if one can add to ClassSpaceMetadata an
> > attribute referencing the name of the parent ClassSpace of the
> > ClassSpace to be created via this ClassSpaceMetadata. This way, a
> > deployment unit could explicitely defines its direct CL and the parent
> > of this CL.
> >
> > I have applied this fix and it solved my resource adapter deployer
> > problem. More accurately, the parent CL of the ClassSpace containing
> > the resource adapter archives is the
> > geronimo.system:role=ClassSpace,name=System ClassSpace (defined during
> > the boot).
> >
> > Gianny
>
Re: ClassSpace - parent CL
Posted by Dain Sundstrom <da...@coredevelopers.net>.
I have a new version of the Class Space code on my laptop that fixes
this issue, but I'm preparing for our upcoming training course. I
should have it checked in late next week or the week after.
-dain
On Thursday, October 2, 2003, at 10:19 AM, gianny DAMOUR wrote:
> Hello,
>
>
> I was just trying to solve the problem of Matt Kurjanowicz, which
> seems to be a weird CL problem and during my investigations I have
> discovered three points:
>
> - the Eclipse set-up I use to simulate a running server is simply
> flawed (grrr...);
> - the Resource Adapter Deployer I have submitted does not work, when
> deployed within a running instance launched via maven run:main
> (grrrrr...); and
> - the parent CL of a ClassSpace is the TCCL if defined or the CL of
> the ClassSpace class.
>
> Regarding this last point, I really find it weird for the following
> reason:
>
> If one creates a ClassSpace for a deployment unit, say a resource
> adapter, and if I want to define a specific parent for such a
> ClassSpace, I will need to set the TCCL. However, because this TCCL is
> outside our control between the moment this TCCL is set and the actual
> creation of my ClassSpace (other deployment planners could wish to set
> their own TCCL), it is impossible to know for sure which CL will be
> the parent of the deployment unit ClassSpace.
>
> So, I would like to know if one can add to ClassSpaceMetadata an
> attribute referencing the name of the parent ClassSpace of the
> ClassSpace to be created via this ClassSpaceMetadata. This way, a
> deployment unit could explicitely defines its direct CL and the parent
> of this CL.
>
> I have applied this fix and it solved my resource adapter deployer
> problem. More accurately, the parent CL of the ClassSpace containing
> the resource adapter archives is the
> geronimo.system:role=ClassSpace,name=System ClassSpace (defined during
> the boot).
>
> Gianny