Often, by the time the UDPServer realizes that an entity update packet
has not been acknowledged, there is a newer update for the same entity
already queued up or there is a higher priority update that should be
sent first. This patch eliminates 1:1 packet resends for unacked entity
update packets. Insteawd, unacked update packets are decomposed into the
original entity updates and those updates are placed back into the
priority queues based on their new priority but the original update
timestamp. This will generally place them at the head of the line to be
put back on the wire as a new outgoing packet but prevents the resend
queue from filling up with multiple stale updates for the same entity.
This new approach takes advantage of the UDP nature of the Linden protocol
in that the intent of a reliable update packet is that if it goes
unacknowledge, SOMETHING has to happen to get the update to the client.
We are simply making sure that we are resending current object state
rather than stale object state.
Additionally, this patch includes a generalized callback mechanism so
that any caller can specify their own method to call when a packet
expires without being acknowledged. We use this mechanism to requeue
update packets and otherwise use the UDPServer default method of just
putting expired packets in the resend queue.
when client and simulator throttles are set. This algorithm also uses
pre-defined burst rate of 150% of the sustained rate for each of the
throttles.
Removed the "state" queue. The state queue is not a Linden queue and
appeared to be used just to get kill packets sent.
This is in a very crude state, currently.
The LindenUDPModule was renamed LindenUDPInfoModule and moved to OptionalModules
OptionalModules was given a direct reference to OpenSim.Region.ClientStack.LindenUDP so that it can inspect specific LindenUDP settings without having to generalize those to all client views (some of which may have no concept of the settings involved).
This might be ess messy if OpenSim.Region.ClientStack.LindenUDP were a region module instead, like MXP, IRC and NPC
If an LL 1.23.5 client (and possibly earlier and later) receives an object update after a kill object packet, it leaves the deleted prim in the scene until client relog
This is possible in LLUDPServer if an object update packet is queued but a kill packet sent immediately.
Beyond invasive tracking of kill sending, most expedient solution is to always queue kills, so that they always arrive after updates.
In tests, this doesn't appear to affect performance.
There is probably still an issue present where an update packet might not be acked and then resent after the kill packet.
Setting this to true avoids a 500ms or so client freeze when the LLUDP server thread is taken up with processing a UseCircuitCode packet synchronously.
Extensive testing on Wright Plaza appeared to show no bad effects and this seems to reduce login lag considerably.
Of course, a lot of login lag is still coming from other sources.
Object updates are sent on the task queue. It's possible for an object update to be placed on the client queue before a kill packet comes along.
The kill packet would then be placed on the state queue and possibly get sent before the update
If the update gets sent afterwards then client get undeletable no owner objects until relog
Placing the kills in the task queue should mean that they are received after updates. The kill record prevents subsequent updates getting on the queue
Comments state that updates are sent via the state queue but this isn't true. If this was the case this problem might not exist.
DisableFacelights option to OpenSim.ini to finally kill those immersion-
breaking, silly vanity lights that destroy nighttime RP. Girls, you look
just fine without them. Guys, you too. Thank you. Melanie has left the building.
* Handle logout properly. This needed an addition to IClientAPI, because of how the logout packet is currently being handled -- the agent is being removed from the scene before the different event handlers are executed, which is broken.
* Make several packets not asynchronous (such as AgentUpdate). In theory, all fast returning packet handling methods should not be asynchronous. Ones that wait on an external resource or a long held lock, should be asynchronous.
* Fixing a bug where the max burst rate for the state category was being set as unlimited, causing connections to child agents to saturate bandwidth
* Upped the example default drip rates to 1000 bytes/sec, the minimum granularity for the token buckets
