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...
This is so that other callers (such as SceneCommunicationService.SendCloseChildAgentConnections() can perform all closes asynchronously without pointlessly firing another thread for local closes).
No functional change apart from elimination of unnecessary chaining of new threads.
Unfortunately fails on Nebadon's system right now. Needs investigation. May put in a temproary option for experimentation soon.
This reverts commit d87ddf50fc.
On my own system, I can now eliminate the pause entirely and the reteleport happens whilst the teleport screen is still up.
Trying this change to see if this is true for other people.
This looks like an off-by-one bug since the view distance was already only 256 on the west and south sides.
This reduces the number of child agents being logged into regions neighbouring a megaregion.
This was because the calculation as to whether a new agent was needed in the receiving region did not take megaregions into account,
unlike the original calculation when the user first teleported into the region.
This meant that on teleport, entity transfer would create a new CAP but this would be ignored by the viewer and receiving region, meaning that the EQ could no longer be used.
This would prevent subsequent teleport, amongst other things.
Currently, regions up to 512m from a megaregion are considered neighbours.
This aims to reduce any side effects if the process tries to complete after the client has logged back in (e.g. it was delayed due to a slow destination region response).
This introduces a new Aborting entity transfer state which signals that the teleport should be stopped but no compensating actions performed.
If we do this after TeleportFinish, then it's possible for a neighbour destination to request the source to create a child agent whilst its still treated as root.
This closes the original presence which we don't really want to do.
This is probably okay (albeit with warnings on the console) but afaics there's no reason not to move the child agent signal.
Also adds regression test for the case where the viewer couldn't connect with the destination region.
Also refactoring of regression test support code associated with entity transfer in order to make this test possible and the code less obscure.
This option allows the simulator to specify that the cancel button on inter-region teleports should never appear.
This exists because sometimes cancellation will result in a stuck avatar requiring relog.
It may be hard to prevent this due to the protocol design (the LL grid has the same issue)
In small controlled grids where teleport failure is practically impossible it can be better to disable teleport cancellation entirely.
Previously, hitting the cancel button on a teleport would cancel on the client side but the request was ignored on the server side.
Cancel would still work if the teleport failed in the early stages (e.g. because the destination never replied to early CreateAgent and UpdateAgent messages).
But if the teleport still completed after a delay here or later on, the viewer would become confused (usual symptom appears to be avatar being unable to move/reteleport).
This commit makes OpenSimulator obey cancellations which are received before it sends the TeleportFinish event queue message and does proper cleanup.
But cancellations received after this (which can happen even though the cancel button is removed as this messages comes on a different thread) can still result in a frozen avatar.
This looks extremely difficult and impossible to fix.
I can replicate the same problem on the Linden Lab grid by hitting cancel immediately after a teleport starts (a teleport which would otherwise quickly succeed).
This means that if such an exception does occur, the region does not need to be reset before that user can teleport from it again.
This is all Oren's code from his patch in http://opensimulator.org/mantis/view.php?id=6374 but I've chosen to split it in two.
I believe UseCircuitCode is sent on EnableSimulator EQ message, rather than EstablishAgentCommunication
At least with LL 3.3.4, EstablishAgentCommunication appears unnecessary in the teleport context -
viewer still requests it though possibly only after TeleportFinish(). However, we will continue to send it.
The viewer would not see the folder move without this, either on accept or decline.
This commit also updates the TaskInventoryOffered message to better conform with the data LL uses
Changes are, agentID is prim owner rather than prim id, agent name is now simply object name rather than name with owner detail,
message is just folder name in single quotes, message is not timestamped.
However, folder is not renamed "still #RLV/~<name>". Long term solution is probably not to do these operations server-side.
Notes will be added to http://opensimulator.org/mantis/view.php?id=6311
