Sounds good. Karl and I have been working on a patch for this issue from
the framework side of things. The patch will solve the immediate issue
of the registry holding a lock while calling a service factory and it
will also try to improve the situation further in general by adding a
more fine-grained locking protocol inside the registry.
-> richard
Felix Meschberger wrote:
Hi all,
Incidentally Rajini and one of our Sling users identified a similar
deadlock issue in the interaction between the framework and the
Declarative Services implementation at the same time. I agree with Karl,
that FELIX-489 may certainly be viewed from the perspective of the
framework calling out of it and releasing any locks during that time.
Therefore I created FELIX-490 which looks at this issue from the
perspective of Declarative Services and requiring this to revisit
synchronization and try to not hold locks while calling out of itself
(mostly into the framework).
Regards
Felix
