Rob Davies escribió:
ActiveMQDestinations methods for equals() and hashCode() have always
been there - so yes it will always be removed from the concurrent hashmap
True. I'm afraid I shooted too early. I thought I'd found the problem.
Sorry for the false alarm.
Regards.
cheers,
Rob
On Jul 6, 2007, at 12:15 PM, Manuel Teira wrote:
I have observed that the Consumer advisory topics (at least for
temporary queues) are not disposed, at least in svn 4.1 branch. I
think that this is caused by the way they are tried to be deleted
from the AdvisoryBroker removeDestinationInfo method:
try {
next.removeDestination(context,
AdvisorySupport.getConsumerAdvisoryTopic(info.getDestination()), -1);
} catch (Exception expectedIfDestinationDidNotExistYet) {
}
try {
next.removeDestination(context,
AdvisorySupport.getProducerAdvisoryTopic(info.getDestination()), -1);
} catch (Exception expectedIfDestinationDidNotExistYet) {
}
But that helper method,
AdvisorySupport.get[Consumer|Producer]AdvisoyTopic is always creating
a new Topic:
public static ActiveMQTopic
getConsumerAdvisoryTopic(ActiveMQDestination destination) {
if( destination.isQueue() )
return new
ActiveMQTopic(QUEUE_CONSUMER_ADVISORY_TOPIC_PREFIX+destination.getPhysicalName());
else
return new
ActiveMQTopic(TOPIC_CONSUMER_ADVISORY_TOPIC_PREFIX+destination.getPhysicalName());
}
So, what we are feeding to next.removeDestination, where next is
eventually the RegionBroker chained instance, is a different
ActiveMQTopic than the original one that we were trying to delete. As
RegionBroker destinations is a ConcurrentHashMap, and the object we
are passing to be removed is different from the original one, is it
going to remove anything?
The procedure followed in trunk seems to be the same.
Regards.
