| Also, yes this CCE itself was fixed already in 5.2. And yes I believe hibernate-infinispan still maps these to the same underlying cache. That in-and-of-itself is not a problem. However it was compounded by the fact that hibernate-infinispan also tried to use the "slimmed down" cache key which removes the entity-name amongst other information, which causes problems when different entities are then stored in the same underlying cache. That was fixed by forcing it to use the full cache key. Again, all of this was done in 5.2, so functionally this should all be fine in 5.2. The continuing problem was getting back to the more performant solutions in such cases, which needs the changes in
 HHH-11356 Open . And since you asked in email... imo it is not a good idea to backport this to 5.x. It is a pretty disruptive change for cache integrators. Its required changes to:
- how Regions are defined and built
- how access to these Regions are built and cached
- how Statistics and other SF-level use of Regions (evict, etc) are handled.
This last one also makes it a potentially disruptive change for users depending on if/how they interact with the L2 cache. |
