This bug was spotted by clang on FreeBSD.

clusterGetRandomName() generalized into getRandomHexChars() so that we can use it for the run_id field as well.

Fixed a non critical bug signaled by clang static analyzer thanks to Mukund Sivaraman for reporting it: there was a not initialized field populating the cluster message header, but it is always fixed at later time before sending the packet.

Redis Cluster: process node to node CLUSTERMSG_TYPE_PUBLISH messages and send it to the local clients.

propagate PUBLISH messages using the redis cluster nodes bus. Still need to process the incoming packets of that type. Work in progress.

A node can populate a slot if a message from a trusted slot is received claiming ownership of this slot, that is currently empty or served by a node in FAIL state. However this feature was broken since calling clusterAddSlot() was not enough as the slot bit is already set in the node that pong us. We need to directly alter the table. This commit fixes the issue.

Rewrite MIGRATE as DEL for AOF/replication. Also increment the dirty counter to both force replication and persistence.

increment server.dirty in restore command to make sure it is replicated and increments the dirty count for persistence concerns.

when processing gossip packets make sure to also update the node bitmap when associating slots to nodes. Fixed simply using the appropriate helper function to add a slot.

Clear the importing status from the slot if a CLUSTER SETSLOT command permanently assigns the slot to the importing node.

rioInitWithFile nad rioInitWithBuffer functions now take a rio structure pointer to avoid copying a structure to return value to the caller.