…canning and reflection calls * Allow fallback to JDK reflection if annotated method not found on a given component.
Fixed JGroups XML files as well.
…h joining even though its peer caches were not started yet. This broke some tests that were not starting all their caches at once.
… in case we got an APPY_STATE command for an obsolete view.
…ommands in the JMX statistics
… mode, the new state transfer doesn't support restarts. We'd need to solve the whole asymmetric caches issue first.
Moved the rehashing logic out of the DistributionManager to a new class DistributedStateTransferManager. Extracted the transaction lock out of the TransactionLogger into a new class TransactionLock. Removed the dependency on a coordinator, using instead a broadcast to signal to all nodes that push is complete. Added a new REQUEST_JOIN DistributedStateTransferControlCommand to make sure joiners are properly started before the old members start sending any commands. The user will still need to start all the caches upfront, as we don't support asymmetric caches.
…regardless of whether we were an owner with the old CH. This way if we got a key for an intermediate rehash that doesn't belong to us in the final rehash we will invalidate it.
…led tests (bin/failedTestLogs.sh)
…canning and reflection calls * Remove ReflectionCache and dependencies on it * Use Jandex to index annotations at build-time, and add indexes to distribution jar file * Use a new annotated method cache to read the index and build a map at boot time * Refer to this cache when scanning for annotated @Inject, @Start and @Stop methods * Use annotated method cache to scan for @DefaultFactoryFor annotated classes * Remove hard-coded factory list in AbstractComponentRegistry
…canning and reflection calls * Remove log in abstract class; move it to each subclass that needs it