On Sunday 05 July 2015 21:24:07 Jean-Baptiste Onofré wrote: > It could be a bug indeed. Let me take a look on SLING-4412.
Any ideas what goes wrong? > In the mean time, can you try to change the req enforcement mode in > etc/org.apache.karaf.features.cfg just to be sure ? Tested with all possible values, SLING-4412 updated. Regards, O. > Regards > JB > > On 07/05/2015 09:19 PM, Oliver Lietz wrote: > > On Sunday 05 July 2015 20:47:56 Jean-Baptiste Onofré wrote: > >> Hi Oliver, > > > > hi Jean-Baptiste, > > > >> if you want to migrate smoothly, you can still use features xmlns < > >> 1.3.0: it should work as before. > >> > >> If you want to use new style features repository XML (xmlns >= 1.3.0), > >> in that case you use the new features resolver. > >> > >> The dependency attribute is no more required with xmlns 1.3.0 as the > >> resolver deals with that. > > > > the XML namespace is still 1.2.0 - so maybe a bug? > > > > What happens on 3.x with new style features in degraded mode without > > dependency attribute? > > > > Regards, > > O. > > > >> Regards > >> JB > >> > >> On 07/05/2015 08:35 PM, Oliver Lietz wrote: > >>> hi, > >>> > >>> I'm trying to bring Sling's Karaf features to Karaf 4.0 but encounter > >>> some > >>> problems. So far there is a problem with dependency=true for bundles and > >>> the wrap protocol (may be related to Pax Exam?). > >>> > >>> I've documented my findings in SLING-4412 and added a patch (against > >>> trunk) > >>> which updates Karaf from 3.0.3 to 4.0.1-SNAPSHOT and removes the > >>> dependency > >>> attribute (workaround). > >>> > >>> Can someone with understanding of the new resolver have a look what goes > >>> wrong, please? > >>> > >>> Thanks, > >>> O. > >>> > >>> https://issues.apache.org/jira/browse/SLING-4412