This check is necessary to close a race condition where the CompleteAgentMovement processing could proceed when the UseCircuitCode thread had added the client to the client manager but before the ScenePresence had registered to process the CompleteAgentMovement message.
This is most probably why the message appeared to get lost on a proportion of entity transfers.
A better long term solution may be to set the IClientAPI.SceneAgent property before the client is added to the manager.
This reverts commit c7ded0618c.
This proves not to be necessary - the necessary checks are already being done via QueryAccess() before cross or teleport
This occurs on v2 teleport since the source region now waits 15 secs before closing the old child agent, which could still receive chat.
This commit introduces a ScenePresenceState.PreClose which is set before the wait, so that ChatModule can check for ScenePresenceState.Running.
This was theoretically also an issue on v1 teleport but since the pause before close was only 2 secs there, it was not noticed.
This is necessary because the hypergrid groups checks (as referenced by estates) require an agent circuit to be present to construct the hypergrid ID.
However, this is not around until Scene.NewUserConnection(), as called by CreateAgent() in EntityTransferModule.
Therefore, if we're dealing with a hypergrid user, delay the check until NewUserConnection()/CreateAgent()
The entity transfer impact should be minimal since CreateAgent() is the next significant call after NewUserConnection()
However, to preserve the accuracy of query access we will only relax the check for HG users.
Also separates out log messages to distinguish between close not finding an agent and wrong auth token, and downgrades former to debug and latter to warn
This approach has problems if a client quits without sending a proper logout but then reconnects before the connection is closed due to inactivity.
In this case, the DoNotCloseAfterTeleport was wrongly set.
The simplest approach is to close child agents on teleport as quickly as possible so that races are very unlikely to occur
Hence, this code now closes child agents as the first action after a sucessful teleport.
IncomingCloseAgent() now sets the scene presence state machine properly, which is necessary to avoid races between multiple sources of close.
Hence, it's also necessary for everyone to consistently call IncomingCloseAgent()
Calling RemoveClient() directly is currently generating an attention-grabbing exception though this right now this is harmless.
This was occuring because teleport to B did not set DoNotCloseAfterTeleport on A as it was a neighbour (where it isn't set to avoid the issue where the source region doesn't send Close() to regions that are still neighbours (hence not resetting DoNotCloseAfterTeleport).
Fix here is to still set DoNotCloseAfterTeleport if scene presence is still registered as in transit from A
The root cause was that v2 was only closing neighbour agents if the root connection also needed a close.
However, fixing this requires the neighbour regions also detect when they should not close due to re-teleports re-establishing the child connection.
This involves restructuring the code to introduce a scene presence state machine that can serialize the different add and remove client calls that are now possible with the late close of the
This commit appears to fix these issues and improve teleport, but still has holes on at least quick reteleporting (and possibly occasionally on ordinary teleports).
Also, has not been completely tested yet in scenarios where regions are running on different simulators
- Child and root agents are only closed after 15 sec, maybe
- If the user comes back, they aren't closed, and everything is reused
- On the receiving side, clients and scene presences are reused if they already exist
- Caps are always recreated (this is where I spent most of my time!). It turns out that, because the agents carry the seeds around, the seed gets the same URL, except for the root agent coming back to a far away region, which gets a new seed (because we don't know what was its seed in the departing region, and we can't send it back to the client when the agent returns there).
This prevents an issue if the user teleports back to the neighbour simulator of a source before 15 seconds have elapsed.
This more closely emulates observed linden behaviour, though the timeout there is 50 secs and applies to all the pre-teleport agents.
Currently sticks a DoNotClose flag on ScenePresence though this may be temporary as possibly it could be incorporated into the ETM state machine
In this new protocol, and as committed before, the viewer is not sent EnableSimulator/EstablishChildCommunication for the destination. Instead, it is sent TeleportFinish directly. TeleportFinish, in turn, makes the viewer send a UserCircuitCode packet followed by CompleteMovementIntoRegion packet. These 2 packets tend to occur one after the other almost immediately to the point that when CMIR arrives the client is not even connected yet and that packet is ignored (there might have been some race conditions here before); then the viewer sends CMIR again within 5-8 secs. But the delay between them may be higher in busier regions, which may lead to race conditions.
This commit improves the process so there are are no race conditions at the destination. CompleteMovement (triggered by the viewer) waits until Update has been sent from the origin. Update, in turn, waits until there is a *root* scene presence -- so making sure CompleteMovement has run MakeRoot. In other words, there are two threadlets at the destination, one from the viewer and one from the origin region, waiting for each other to do the right thing. That makes it safe to close the agent at the origin upon return of the Update call without having to wait for callback, because we are absolutely sure that the viewer knows it is in th new region.
Note also that in the V1 protocol, the destination was getting UseCircuitCode from the viewer twice -- once on EstablishAgentCommunication and then again on TeleportFinish. The second UCC was being ignored, but it shows how we were not following the expected steps...
The UMMTGUN form of Unknown User seems to appear because a viewer sometimes sends a UUIDNameRequest UDP request that fails to find a binding.
However, in theory the incoming agent should have made that binding before any such request is triggered.
So moving this binding to an earlier point in the process to see if this makes a difference.
Unknown user name is also updated to UserUMMTGUN2 - if you see the old name then you need to clear your viewer cache.
This relates to http://opensimulator.org/mantis/view.php?id=6625
This is to reduce race conditions where neighbours may be responding erratically, thus mixing up create and close agent requests in time.
This mirrors OpenSimulator behaviour on enabling child agents where each region is contacted separately.
Justin Clark-Casey (justincc)