Guillaume Nodet created FELIX-3806:
--------------------------------------
Summary: Some bundle events are sent to SynchronousBundleListeners
without holding the bundle lock which can cause the event to be delivered out
of order
Key: FELIX-3806
URL: https://issues.apache.org/jira/browse/FELIX-3806
Project: Felix
Issue Type: Bug
Components: Framework
Reporter: Guillaume Nodet
This can be very troublesome for extenders which react to the STARTED /
STOPPING events for example.
If two threads are starting / stopping the bundles, those two events can be
delivered out of order.
The spec mentions that pathological listeners can lead to such results, but
that's only because the listener does change the state by calling back the
framework.
The above out of order delivery is very difficult to overcome when writing
extenders because the bundle can be stopped concurrently while processing the
STARTED event.
I've seen the purpose of this behavior is to enable bundles to modify the
lifecycle of the bundles in reaction to the STARTED event.
However, I don't understand this objection. If the thread holding the lock
calls back into the framework to stop the bundle, the lock is reentrant and
that should work.
If the work is done in a separate thread, the thread will wait until the
STARTED event is processed and that should not be a problem.
So I suggest moving back the STARTED and STOPPED events inside the bundle lock.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
