Open arcane86 opened 13 years ago
Not sure how we can protect against that. Any idea ?
We should look at the random ST we get some time during operation on bundle lisfe-cycle.
We should also review the code to find the spot where we do not test if a service is still there when we use it.
It seems also link to issue #71.
We need to refactor how ServiceImpl is getting its service instances from OSGi service registry. i'm working on a solution using dynamic proxies like for injected service.
CDI-OSGi do not test to much that services and bundles are still in the expected state during long processing (e.g when switching between a unregistered factory service and another one, entire CDI-OSGi framework restarts with the assumption that this second factory service is still there).