geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Karl Pauls" <karlpa...@gmail.com>
Subject Re: OSGifying org.apache.geronimo.specs
Date Fri, 21 Sep 2007 12:59:18 GMT
> So I have something more consistent, but I can really put it as a
> patch as it involves copying from tags to trunk, so I guess I'll just
> commit it for review.
> The only problem I have comes from the commonj spec: the package is
> java.commonj and it seems the felix plugin does not really like java.*
> packages and consider them as reserved... So I doubt this one can be
> OSGified right now.

This is because bundles don't need to import java.* packages. They are
boot delegated automatically. So you don't need to add them as an
import.

In this case, this presents a problem because you are trying to
provide a bootclasspath extension. There is a way that this can be
done in OSGi by making the bundle an extension bundle (see 3.15 in the
spec). Basically, you would need to add the following header:

Fragment-Host: system.bundle; extension:=bootclasspath

for a bundle that exports java.* packages (extension bundles are not
allowed to have imports so you would maybe need to split the bundle in
two: one extension bundle with only java.* and one normal one with the
rest).

However, I don't think there are many OSGi frameworks that support
bootclasspath extension bundles atm (neither Felix nor Equinox does).

regards,

Karl

> On 9/21/07, Guillaume Nodet <gnodet@gmail.com> wrote:
> > For ServiceMix 4.0, which will be based on OSGi, I will need to have
> > OSGified versions of some of the spec jars that geronimo provides.
> > It's quite easy to do in ServiceMix (see
> > http://svn.apache.org/repos/asf/incubator/servicemix/branches/servicemix-4.0/bundles/
> > for servlet, j2ee-management, jms mainly), but I think it would be
> > more useful for other projects if the specs jars were bundles
> > themselves.
> >
> > This is quite a simple process that can be done incrementally without
> > any real side effect and low risk of regression.  So unless someone
> > objects, I'd like to start working on that.
> >
> > --
> > Cheers,
> > Guillaume Nodet
> > ------------------------
> > Blog: http://gnodet.blogspot.com/
> >
>
>
> --
> Cheers,
> Guillaume Nodet
> ------------------------
> Blog: http://gnodet.blogspot.com/
>


-- 
Karl Pauls
karlpauls@gmail.com

Mime
View raw message