Hi @Grzegorz, Well, JDK dropped JAXB and endorsing so it must be a bundle now, putting it in the classpath is a workaround but not the other way around regarding JRE rules now. Now one of the liked features of OSGi is to be dynamic and updatable and using the JRE breaks that by design and you don't have the OSGi integration (bundle activator quite often) since it comes with the JRE so has the lifecycle of the JRE. This is why for me, if the boot classpath is more than OSGi container and a small config reader utility (caricaturally karaf main), it is a design pitfall.
I'm also thinking to vendors doing custom karaf distros and on this side they must be able to be secured and use the same distro on multiple JRE and this is one issue which shouldn't require properties customizations IMHO. Hope it makes sense. Romain Manni-Bucau @rmannibucau <https://twitter.com/rmannibucau> | Blog <https://rmannibucau.metawerx.net/> | Old Blog <http://rmannibucau.wordpress.com> | Github <https://github.com/rmannibucau> | LinkedIn <https://www.linkedin.com/in/rmannibucau> | Book <https://www.packtpub.com/application-development/java-ee-8-high-performance> Le lun. 27 janv. 2020 à 09:01, Grzegorz Grzybek <gr.grzy...@gmail.com> a écrit : > Hello > > I didn't work much with JDK9 (though JDK15 builds are already > available[1]...). But maybe (if it's the only problem) `osgi.contract` can > be added to system bundle via `jre.properties`? > > I mean - we're ~10 years after Xerces hell already and I hope JAXB and > other "endorsed standards" can be handled at lowest possible level... I > admire what spec bundles do, but it still (IMO) look like a workaround of > some fundamental problem related to adjusting JDK itself to OSGi... > > regards > Grzegorz Grzybek > === > [1]: https://jdk.java.net/15/ > > pon., 27 sty 2020 o 08:38 Jean-Baptiste Onofré <j...@nanthrax.net> > napisał(a): > > > It makes sense to me. > > > > Let me create Jira and work on an improvement about that. > > > > Thanks for the proposal ! > > > > Regards > > JB > > > > On 27/01/2020 08:17, Romain Manni-Bucau wrote: > > > Yep, also means karaf.main must not depend on these ones but > technically > > it > > > sounds very feasible and saner in terms of architecture (launcher > > > responsability vs container like one). > > > > > > Le lun. 27 janv. 2020 à 08:14, Jean-Baptiste Onofré <j...@nanthrax.net> > a > > > écrit : > > > > > >> Hi Romain, > > >> > > >> So, basically, your proposal is to remove jdk9plus and "force" use of > > >> spec bundles, right ? > > >> > > >> It makes sense to me, but it means that any spec has to be a bundle > and > > >> started in early stage of the boot process. > > >> If it's possible, it makes sense. > > >> > > >> Regards > > >> JB > > >> > > >> On 27/01/2020 08:11, Romain Manni-Bucau wrote: > > >>> Hi all, > > >>> > > >>> Playing with the r7 branch i tried to build an osgi-cdi distro but > > >> stumbled > > >>> upon the fact jdk9plus folder breaks resolution chain quite easily > when > > >>> switching of jdk. > > >>> > > >>> Long story short, having annotation, activation (and potentially jaxb > > >> but i > > >>> didnt need this one ;)) does not enable to have them as bundle in the > > >> same > > >>> version - so to do dynamic updates too ;) - and they miss > osgi.contract > > >>> entry config. > > >>> > > >>> I wonder if there is any rational to have them at all, sounds like > > karaf > > >>> can boot without them and just move to bundles all the logic > > potentially > > >>> needing them so no need to patch the classpath for java >= 9 IMHO. > > >>> > > >>> Did I miss anything? > > >>> Is it something to plan to clean up for karaf 4.3.0? > > >>> > > >> > > >> -- > > >> Jean-Baptiste Onofré > > >> jbono...@apache.org > > >> http://blog.nanthrax.net > > >> Talend - http://www.talend.com > > >> > > > > > > > -- > > Jean-Baptiste Onofré > > jbono...@apache.org > > http://blog.nanthrax.net > > Talend - http://www.talend.com > > >
