Yes, I verified that the problem is that the TxBeanProcessor check the transaction type in the following code
Interceptor interceptor = (getTransactionType(supplierProxy) == PersistenceUnitTransactionType.JTA) ? new XaTxInterceptorImpl(tm, supplierProxy) : new ResourceLocalTransactionalInterceptor(supplierProxy); cdr.registerInterceptorWithComponent(beanData, interceptor); Did you remove this part? 2015-04-08 22:00 GMT+02:00 Christian Schneider <ch...@die-schneider.net>: > Thanks. Will also look into it. > > It could also be related to the check if the EnityManager is > RESOURCE_LOCAL or JTA. Maybe we have to defer this check. > In a new change I did not yet commit I removed most of the differentiation > beetween JTA and RESOURCE_LOCAL in the jpa-blueprint module. > After this change it might be easier to get it working. > > Christian > > > On 08.04.2015 18:44, Giuseppe Gerla wrote: > >> Hi Christian >> today I make some integration test in Karaf 2.4.1 using last version of >> jpa-experiments and my bundles. >> I found a problem about EmSupplierProxy. When the class TxBeanProcessor >> try >> to getTransactionType call preCall method of EmSupplierProxy that try to >> retrieve the osgi service instance of EmSupplier. The problem is that >> this >> service is registered after, so I have a NullPointerException (after wait >> 10 seconds). >> I try to increase the wait period to 100 seconds, but it seems that the >> registration of EmSupplier service is synchronous and postponed to the >> wait. >> I'm thinking that the problem is linked to my implementation of >> configuration management, because I move the registration of emf to the >> updated method. >> I will make a more detailed analysis tomorrow to find a solution. >> >> >> Regards >> Giuseppe >> >> >> 2015-04-07 22:04 GMT+02:00 Giuseppe Gerla <giuseppe.ge...@gmail.com>: >> >> Hi all >>> I try to do some integration test deploying the new implementation with >>> eclipselink and my test bundle. >>> I make a change on the logic of PersistenceProvider service filter. To be >>> abstract I don't want define the class of implementation in >>> persistence.xml, so in this case I cannot filter using the class name. >>> I found and fix also some minor issues. >>> I open another pull request. >>> >>> However, I don't make several test (like transaction and so on), but from >>> my point of the new implementation is more easy to manage of actual aries >>> jpa. >>> >>> >>> Regards >>> Giuseppe >>> >>> >>> 2015-04-07 11:54 GMT+02:00 Jean-Baptiste Onofré <j...@nanthrax.net>: >>> >>> Hi guys, >>>> >>>> Giuseppe, I like your first implementation. I gonna take a deeper look >>>> this afternoon. >>>> >>>> Regards >>>> JB >>>> >>>> >>>> On 04/07/2015 10:23 AM, Giuseppe Gerla wrote: >>>> >>>> Hi Christian >>>>> I open a pull request with the first implementation. I also add 2 unit >>>>> test. >>>>> It seems an easy implementation. >>>>> Wdyt? >>>>> >>>>> >>>>> > -- > Christian Schneider > http://www.liquid-reality.de > > Open Source Architect > http://www.talend.com > >