Hi Andy, Are A, B, and C in the same class hierarchy? Did you re-enhance your entities with 2.0.0 (happens automatically with -javaagent)?
Seems like the field index is off somehow - leading to the wrong inverse being 'managed'.. -mike On Thu, Jun 3, 2010 at 3:49 PM, Andrew Thompson <[email protected]> wrote: > Hey Folks, > > I'm running into an issue upgrading from openjpa 1.2.2 to 2.0 when using > the inverseManager. > > The usecase is: > > object A has a lazy bidirectional relationship to object B and A also > has a lazy bidirectional relationship to object C. These are > bidirectional only because it was required to be that way by the jpa 1.0 > spec. we (unfortunately) depended on the inversemanager to deal > managing the inverse. > > what seems to be happening is: > an update to object B without properly managing the inverse object A > sets object C's foreign key relationship to to object A to null. which > orphans it. > > if possible, i'm looking for the minimal amount of code change. ie, > disabling the inversemanager and either updating the relationships to be > uni directional, or else properly managing inverses fixes the problem. > similarly having an actual foreign key where it should be means that at > least data isn't lost. it's just more labor then i was hoping to > allocate currently - especially when i haven't heard many other users > running into issues upgrading. > > thanks, > -Andy > > >
