[ https://hibernate.onjira.com/browse/HHH-1570?page=com.atlassian.jira.plug... ] Yoann Antoviaque commented on HHH-1570: --------------------------------------- I looked at this issue, the experimental patch did not resolve the problem in our case, I had to modify it (will post it later today, once I have validated that it works completly for us). I still have a "regression" in FumTest#testKeyManyToOne() but I think it is not a regression but another bug that was hidden due to the absence of loading of linked tables: by adding those lines in FumTest#testKeyManyToOne(), i get a NonUniqueObjectException, failing exactly the same way, even with a non patched hibernate-core: // new test case showing non refreshed composite keys s = openSession(); InnerKey sid2 = new InnerKey(); sid2.setAkey("a"); sid2.setBkey("b"); Inner sup2 = (Inner) s.load(Inner.class, sid2); assertEquals("dudu", sup2.getDudu()); MiddleKey mid2 = new MiddleKey(); mid2.setOne("one"); mid2.setTwo("two"); mid2.setSup(sup); Middle m2 = (Middle) s.load(Middle.class, mid2); assertEquals("bla", m2.getBla()); d = (Outer) s.load(Outer.class, did); assertTrue( d.getId().getMaster().getId().getSup().getDudu().equals("dudu") ); s.delete(d); s.delete( d.getId().getMaster() ); //we do not get the master from session but the one from did, therefore fail with NonUniqueObjectException ! s.save( d.getId().getMaster() ); s.save(d); s.flush(); s.connection().commit(); s.close(); -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira