Commit Graph

16931 Commits (8c8e6220fbfbcd4b1441f66ceaa119d74ecc2825)

Author SHA1 Message Date
Justin Clark-Casey (justincc) f2f8dcd65c Add regression test for client logout due to ack timeout. 2012-06-14 02:50:02 +01:00
Justin Clark-Casey (justincc) 332f8b6623 Remove null checks at top of LLUDPServer.ProcessInPacket(). Neither packet nor client are ever null. 2012-06-14 02:49:55 +01:00
Justin Clark-Casey (justincc) 4eda679e12 Store already retrieve IClientAPI in IncomingPacket structure for later use rather than doing another retrieve on dequeue.
Instead of checking whether the client still exists by trying to retrieve again from the client manager, this patch gets it back from IncomingPacket and checks the IClientAPI.IsActive state.
2012-06-14 02:49:39 +01:00
Justin Clark-Casey (justincc) 498154af80 Don't make duplicate call to ScenePresence.Close() separately in ETM.DoTeleport() if an agent needs closing.
This is always done as part of Scene.RemoveClient()
Also refactors try/catching in Scene.RemoveClient() to log NREs instead of silently discarding, since these are useful symptoms of problems.
2012-06-14 02:49:31 +01:00
Justin Clark-Casey (justincc) ba0ebe6d75 Go back to calling IncomingCloseAgent() in the "kick user" command for consistency instead of IClientAPI.Close() directly.
This no longer double counts child agent removals
2012-06-14 02:49:22 +01:00
Justin Clark-Casey (justincc) 28f93512bc Remove duplicate update of user count in Scene.IncomingCloseAgent()
This is already done in Scene.RemoveClient() which IncomingCloseAgent() always ends up calling.
2012-06-14 02:49:14 +01:00
Justin Clark-Casey (justincc) 78c2ef2346 Fix bug with "kick user" reducing agent counts by 2 instead of 1.
This is done by making the kick user command call IClientAPI.Close() rather than routing through Scene.IncomingCloseAgent(), which also called IClientAPI.Close()
DisableSimulator for child agents is moved from IncomingCloseAgent() to RemoveClient(), this is not a functional change since IncomingCloseAgent() always ends up calling RemoveClient()
2012-06-14 02:49:05 +01:00
Justin Clark-Casey (justincc) d30d68657e Record the fact that child agents can have asset transactions.
Also change code to grab the agent asset transaction module once.
2012-06-14 02:48:56 +01:00
Justin Clark-Casey (justincc) 3a27f656b3 Don't send kill object messages to clients when a child agent is closed. 2012-06-14 02:48:48 +01:00
Justin Clark-Casey (justincc) cc27a6cb84 Log warning if we try to remove a UDP client that has already been removed. 2012-06-14 02:48:36 +01:00
Justin Clark-Casey (justincc) ed21576ce0 Allow the thread watchdog to accept an alarm method that is invoked if the timeout is breached.
This alarm can then invoke this to log extra information.
This is used in LLUDPServer to show which client was being processed when incoming and outgoing udp watchdog alarms are triggered.
2012-06-14 02:48:29 +01:00
Talun 9d9e042b4c Mantis 4597 AgentPaused packet is ignored.
The packet was actually being handled but not acted on.
This change extends the default timeout for paused clients to 5 minutes
and makes both the paused and non-paused timeout periods configurable.
2012-06-14 02:47:21 +01:00
Justin Clark-Casey (justincc) 4fb8d5cfba Remove unused ScenePresence list structure in llGetAgentList() 2012-06-06 01:40:45 +01:00
Justin Clark-Casey (justincc) 896c4b6248 Fix build break whree accidentally did inv.Folders rather than inv.Folders.Count in a minor change. 2012-06-04 21:17:43 +01:00
Justin Clark-Casey (justincc) f6730da13a minor: tidy up some comments 2012-06-04 21:17:23 +01:00
Justin Clark-Casey (justincc) 41d98916df Fix various issues with http inventory
1) The return messages were being wrongly populated with the names of asset, inventory and sale types when their corresponding integers should have been used instead.
2) Folders with links were including the linked items in the descendents figure, when only the links should be included.
3) Links and linked items in link folders were not being included in the return data, and not in the correct order.

Now that these issues have been addressed, outfits and attachments appear to work consistently when HTTP inventory is enabled (as is now the default).
2012-06-04 21:17:15 +01:00
Justin Clark-Casey (justincc) 771539a4e0 Instead of updating sim stats root agent, child, objects and scripts accounts every single scene frame, update in the once every 3 seconds SimStatsReporter run 2012-06-04 21:16:46 +01:00
Justin Clark-Casey (justincc) 27c62bba99 Add optional total avatars, total prims and active prims stats to ODE plugin.
These will act as a sanity check with the main scene stats, to show that physics scene entities are being managed properly.
Total prims will not match scene total prims since physics total does not include phantom prims
2012-06-04 21:08:28 +01:00
Justin Clark-Casey (justincc) 93e053a122 If OdeScene.Near() returns no collision contacts, then exit as early as possible. All subsequent code is only relevant if there are contacts. 2012-06-04 21:08:21 +01:00
Justin Clark-Casey (justincc) eb022f4cc1 Add optional stat for the other collision time per frame not spent in ODE native spaces or geom collision code 2012-06-04 21:08:12 +01:00
Justin Clark-Casey (justincc) 83542034dd Add avatar forces calculation, prim force and raycasting per frame millisecond optional stats 2012-06-04 21:08:02 +01:00
Justin Clark-Casey (justincc) 627382f702 Collection optional avatar and prim taint frame millisecond times 2012-06-04 21:07:51 +01:00
Justin Clark-Casey (justincc) 2eb563b3bb Remove recent optional native collision frame milliseconds stat
Unnecessary since this has now been broken down into space collisions and geom collisions
2012-06-04 21:07:41 +01:00
Justin Clark-Casey (justincc) 54a23f14d5 Add optional stat that records milliseconds spent notifying collision listeners in physics frames 2012-06-04 21:07:32 +01:00
Justin Clark-Casey (justincc) e8059b74f8 Add avatar and prim update milliseconds per frame optional stats 2012-06-04 21:07:23 +01:00
Justin Clark-Casey (justincc) 075909520a Add option native step frame ms stat 2012-06-04 21:07:16 +01:00
Justin Clark-Casey (justincc) 808bf12cd5 Add total ODE frame time optional stat, as a sanity check on the main scene physics stat 2012-06-04 21:07:07 +01:00
Justin Clark-Casey (justincc) 0f39f41317 Break down native ODE collision frame time stat into native space collision and geom collision stats 2012-06-04 21:07:00 +01:00
Justin Clark-Casey (justincc) f23b7ae3e9 Rename new collision stats to 'contacts' - there are/can be multiple contacts per collision and this is what is actually being measured. 2012-06-04 21:06:52 +01:00
Justin Clark-Casey (justincc) 5e4b09fc22 Stop adding an unnecessary duplicate _perloopcontact if the avatar is standing on a prim.
This has already been added earlier on in the method.
2012-06-04 21:06:42 +01:00
Justin Clark-Casey (justincc) 25ab7841b7 minor: comment out currently unused OdeScene.sCollisionData 2012-06-04 21:06:32 +01:00
Justin Clark-Casey (justincc) 0c0e575379 Add ODE avatar and prim collision numbers if extra stats collection is enabled. 2012-06-04 21:06:23 +01:00
Justin Clark-Casey (justincc) 68946bffae Fix OdeScene.GetTopColliders() to return the top 25 colliders rather than the first 25 that had non-zero collision scores.
Also zeros collisions scores on all prims after report collection, not just the top 25.
As before, this collision scores are only reset after a report is requested, which may give unrealistic numbers on the first request.
So to see more realistic scores, ignore the first report and then refresh the request after a couple of seconds or so.
2012-06-04 21:06:13 +01:00
Justin Clark-Casey (justincc) 2fc461d9ab Add an optional mechanism for physics modules to collect and return arbitrary stats.
If active, the physics module can return arbitrary stat counters that can be seen via the MonitoringModule
(http://opensimulator.org/wiki/Monitoring_Module)
This is only active in OdeScene if collect_stats = true in [ODEPhysicsSettings].
This patch allows OdeScene to collect elapsed time information for calls to the ODE native collision methods to assess what proportion of time this takes compared to total physics processing.
This data is returned as ODENativeCollisionFrameMS in the monitoring module, updated every 3 seconds.
The performance effect of collecting stats is probably extremely minor, dwarfed by the rest of the physics code.
2012-06-04 21:05:57 +01:00
Justin Clark-Casey (justincc) 59a48d9ebb Add console command "teleport user" to allow teleport from the region console
See "help teleport user" on the console for more details
2012-05-31 02:28:30 +01:00
Justin Clark-Casey (justincc) 533d1ea20c refactor: factor out entity transfer state machine into a separate class to make code more analyzable 2012-05-31 02:28:20 +01:00
Justin Clark-Casey (justincc) dca1ca1d07 If handling the failure of teleport, move agent state to CleaningUp when we start the handling.
Also fixes the log warning from ResetInTransit() if the state is cleared direct from Transferring or ReceiveAtDestination, as pointed out in mantis 5426
2012-05-31 02:25:07 +01:00
Justin Clark-Casey (justincc) 20a3907e86 If restating a region, clean up the physics scene after the main scene has been closed not before.
If this is done before then on ODE agent update calls still incoming can fail as they try to use a raycastmanager that has been disposed.
Bullet plugin does nothing on Dispose()
However, I wouldn't be at all surprised if individual region restarting was buggy in lots of other areas.
2012-05-31 02:24:36 +01:00
Justin Clark-Casey (justincc) 7358e5748d Add Blake/Techplex to CONTRIBUTORS. Thanks! 2012-05-31 02:24:31 +01:00
Blake.Bourque a1b64db942 One can now get hyoergrid region co-ordinates with llRequestSimulatorData 2012-05-31 02:23:00 +01:00
Justin Clark-Casey (justincc) 4d44f2d248 Use GetInventoryItem() in LSL_Api.InventoryKey(string name, int type).
Also removes small bug where calling this method would add 1 to LPS, evne though all callers already did this.
2012-05-31 02:22:53 +01:00
Justin Clark-Casey (justincc) 67abbcf269 Use SceneObjectPartInventory.GetInventoryItem() in OSSL.AvatarStopAnimation instead of searching the task inventory manually. 2012-05-31 02:22:45 +01:00
Justin Clark-Casey (justincc) 6b819a9032 refactor: replace LSL_Api.InventoryKey(string) largely with SceneObjectPartInventory.GetInventoryItem(string)
Also gets llStopAnimation() to call KeyOrName rather than duplicating logic.
2012-05-31 02:22:27 +01:00
Talun 2021a8aedb Mantis 6028 osAvatarStopAnimation not stopping animations via UUID
Corrected to stop animations using the animation UUID similar to llStopAnimation.
See http://opensimulator.org/wiki/OsAvatarStopAnimation
2012-05-31 02:18:55 +01:00
Justin Clark-Casey (justincc) 58dc175ae3 on agent cross, remove from physics scene after its been placed in transit, not before. 2012-05-31 02:18:27 +01:00
Justin Clark-Casey (justincc) 4ad6763956 refactor: make ETM.CrossAgentToNewRegionAsync neighbourRegion == null check return earlier to simplify method 2012-05-31 02:18:20 +01:00
Justin Clark-Casey (justincc) 702826b850 Fix bug where a failed QueryAccess to a remove region would always have the reason "Communications failure" no matter what the destination region actually returned 2012-05-31 02:18:11 +01:00
Justin Clark-Casey (justincc) 8d30a1f74b Stop it being possible for an agent to teleport back to its source region before the source region has finished cleaning up old agent data and structures.
If this is allowed, then the client usually gets forcibly logged out and data structures might be put into bad states.
To prevent this, the binary state machine of EMT.m_agentsInTransit is replaced with a 4 state machine (Preparing, Transferring, ReceivedAtDestination, CleaningUp).
This is necessary because the source region needs to know when the destination region has received the user but a teleport back cannot happen until the source region has cleaned up.
Tested on standalone, grid and with v1 and v3 clients.
2012-05-31 02:18:03 +01:00
Justin Clark-Casey (justincc) 1a988ba835 In remote QueryAccess, also receive the actual status (true|false) instead of always true no matter what the callee actually returned.
This was due to two things
1) SimulationServiceConnector.QueryAccess was always looking to the outer result["success"].
But if a "_Result" map is returned (which is certainly the case right now), then the true success is _Result["success"], result["success"] is always true no matter what
2) If QueryAccess was false at the destination, then AgentHandlers.DoQueryAccess() was never putting this in the result.
The default action of SerializeJsonString() is not to put false booleans in the JSON!!!, so this has to be explicitly set.
2012-05-31 02:17:55 +01:00
Justin Clark-Casey (justincc) c422f852a6 Don't actually proceed on a within-region teleport if another is already taking place, rather than just (falsely) logging that we're not going to proceed.
An oversight from recent commit 9ab0c81
2012-05-31 02:17:49 +01:00