• Bug
  • Status: Closed
  • 1 Critical
  • Resolution: Won't Fix
  • teck
  • Reporter: teck
  • August 10, 2009
  • 0
  • Watchers: 0
  • September 14, 2009
  • August 14, 2009

Description

I need to do some research, but I think we might have a subtle problem with how we load and start TIMs. When we search TIMs for required bundles (ie. dependencies) we end up just building an ordered list. That list is taken as is and defines the order in which we call start() on the bundles.

I *think* this is why I couldn’t use a class from tim-hibernate-concurrency in tim-hibernate-cache even though the correct Required-Bundle and import/export package attributes were in place. As a result of this issue we have some duplicated code (HibernateVersion class)

Comments

Tim Eck 2009-08-14

There isn’t a bug here really. The start order of TIM activators should be handled by OSGi based on the presence of proper import/export attributes