This morning, i tracked strange bug on the Acceleo Hibernate module :

 __________________                 ______________________
| CLASS_MANY_01_BI |<-- * {-} 1 -->| CLASS_MANY_01_BI_END |
|__________________|               |______________________|

The generator created this bloc in Class_Many_01_BI_END.hbm.xml :

<set name="source" inverse="true">
        <key column="FK__TARGET_ID"/>            
        <one-to-many class="org.acceleo.hibernate.bug.entities.Class_Many_01_BI"/>

But the correct mapping seems to be :

<set name="source" inverse="true">       
        <key column="FK_CLASS_MANY_01_BI_END_TARGET_ID"/>            
        <one-to-many class="org.acceleo.hibernate.bug.entities.Class_Many_01_BI"/>

I wrote a basic unit test to detect the problem :

@Test public void testMany_01_BI() throws Exception{
        Class_Many_01_BI_END target = new Class_Many_01_BI_END();
        Class_Many_01_BI source = new Class_Many_01_BI();
        target = (Class_Many_01_BI_END)getSession().load(Class_Many_01_BI_END.class, target.getId());
        assertEquals("Bad mapping detected!", 1, target.getSource().size());

And after short debate on #acceleo IRC channel, thanks Gougou, i added my fix on trunk here :

svn diff -r1655:1656 svn://

This problem gives me to think how to improve this module. The "org.acceleo.module.pim.uml21.gen.jee.tests" non-regression plan is a good point to guarantee stability and the module consistency. Indeed, it's good tools but i think we could add a unit test suite to check all uses cases defines by the non-regression plan. And, we could detect others problems like my bug of the day :)