This is to resolve previous build break.
This unnecessarily but harmlessly reads and sets the parameter multiple times - scene was doing the same thing.
This should really be happening for all console commands (though many don't).
However, things might get difficult if both a console command and other code invoke the same paths.
This increments a SlowFrames counter if a frame takes over 120% of maximum time.
This commit also introduces a generic OpenSim.Framework.Monitoring.Stat which is available to any code that wants to register a statistic.
This is more granualar than asking objects to create their own reports.
At some point this will supersede earlier IMonitor and IAlert facilities in MonitoringModule which are only available to scene code.
This controls how many undo steps the simulator will store for each prim.
Default is now 20 rather than 5 as it briefly was.
The default number could be increased through this is a memory tradeoff which will scale with the number of prims in the sim and level of activity.
This was present in the code but not enforced, which led to a memory leak over time as part properties were changed, whether by viewer, script or another source.
This commit enforces that limit, which will soon become configurable.
Regression test for undo limit added
Should help with http://opensimulator.org/mantis/view.php?id=6279
This was accidentally introduced in 4fc0cfb
This commit also consistently removes the AssetXferUploader when the transaction completes, no matter if it completed on asset upload or item operation.
The amount of data being retained was small, since this was clothing/bodypart metadata in the asset rather than textures themselves.
This was only used if none of new item, update item or update task item had been set.
But since all transactions go through these paths this old code is redundant.
This was preventing the previous race condition fix in 4fc0cfb from actually working.
This commit also removes some of the pointless transaction id checks - these conditions are already being enforced in AgentAssetsTransactions.
This is done for consistency and to allow removal or some access methods that increase code complexity.
However, this path has not been used for a long time, not even by LL 1.23 - viewers use caps http upload for this instead
On creating these items, the viewer sends a UDP AssetUploadRequest followed by a CreateInventoryItem.
It was possible for the CreateInventoryItem/UpdateInventoryItem to occasionally outrace the AssetUploadRequest and fail to find an initialized Xfer object, at which point the item create would fail.
So instead we always set up a Xfer object on either the asset or inventory item update request.
This does not introduce a new race because code already exists to delay the item operation until the asset is uploaded if necessary (but this only worked if the xfer object already existed)
This resends appearance uuids to avatars in the scene once a minute.
I have seen this help in the past resolve grey appearance problems where viewers have for unknown reasons sometimes ignored the packet.
The overhead is very small since only the UUIDs are sent - the viewer then requests the texture only if it does not have it cached.
This setting will not help with cloudy avatars which are usually due to the viewer not uploading baked texture data or uploading something that isn't valid JPEG2000
Make AssetServiceConnector return more useful data on failure, such as what DLL it was trying to load
Allow LocalAssetServiceConnector.GetData() to work without a cache present, as works for the other lasc Get* methods.
This is to resolve a problem where an asset marked as local but not temporary but still used in the scene would be removed.
The timed expiry scan no longer tries to refetch assets from the scene that are not currently in the cache - this is not helpful since it just drags a lot of data into the cache that may never be referenced.
This removes the DeepScanBeforePurge option since setting this to false will introduce the above problem. This previously had a default of true.
As far as I know, this was only used by the IBM Rest modules, much of which has been commented out for a very long time now. Other similar code uses HTTP or stream handlers instead.
So commenting this out to reduce code complexity and the need to make this facility consistent with the others where it may not be used anyway.
If this facility is actually being used then please notify me or uncomment it if you are core.
This allows us to associate debug logging messages with the right request.
It also allows us to put a request number on 'long request' logging even if other debug logging is not enabled, which gives us some idea of whether every request is suffering this problem or only some.
This is a separate internal number not associated with any incoming number in the opensim-request-id header, this will be clarified when logging of this incoming request number is re-enabled.
This commit also adds port number to HTTP IN logging to allow us to distinguish between different request numbers on different ports.
This is only printed if debug http level >= 4 and the request didn't take more than the time considered 'long', in which case the existing log message is printed.
This displaces the previous log levels 4 and 5 which are now 5 and 6 respectively.
This is for debugging purposes.
This is controlled via the "debug http" command which can already log incoming requests.
This now gains a mandatory parameter of in, out or all to control what is logged.
Log messages are also shortened and labelled and HTTP IN or HTTP OUT to be consistent with existing UDP PACKET IN and PACKET OUT messages.
This resends appearance uuids to avatars in the scene once a minute.
I have seen this help in the past resolve grey appearance problems where viewers have for unknown reasons sometimes ignored the packet.
The overhead is very small since only the UUIDs are sent - the viewer then requests the texture only if it does not have it cached.
This setting will not help with cloudy avatars which are usually due to the viewer not uploading baked texture data or uploading something that isn't valid JPEG2000
Make AssetServiceConnector return more useful data on failure, such as what DLL it was trying to load
Allow LocalAssetServiceConnector.GetData() to work without a cache present, as works for the other lasc Get* methods.
This is to resolve a problem where an asset marked as local but not temporary but still used in the scene would be removed.
The timed expiry scan no longer tries to refetch assets from the scene that are not currently in the cache - this is not helpful since it just drags a lot of data into the cache that may never be referenced.
This removes the DeepScanBeforePurge option since setting this to false will introduce the above problem. This previously had a default of true.
Viewer LL 3.3.4 and before sometimes fail to properly redisplay dynamic textures that have a small data length compared to pixel size when pulled from cache.
This appears to happen when the data length is smaller than the estimate discard level 2 size the viewer uses when making this GetTexture request.
This commit works around this by always regenerating dynamic textures that fall below this threshold rather than reusing them if ReuseDynamicTextures = true
This can be controlled by the [Textures] ReuseDynamicLowDataTextures config setting which defaults to false.
This is because recent viewers (3.2.1, 3.3.4) and probably earlier ones using the http GetTexture capability will sometimes make such invalid range requests.
This appears to happen if the viewer's estimate of texture sizes at discard levels > 0 (chiefly 2) exceeds the total texture size.
I believe this does not normally happen but can occur for dynamic textures with are large but mainly blank.
If this happens, returning a RequestedRangeNotSatisfiable will cause the viewer to not render the texture at the final resolution.
However, returning a PartialContent (or OK) even with 0 data will allow the viewer to render the final texture.
If true, this setting reuses dynamically generated textures (i.e. created through osSetDynamicTextureData() and similar OSSL functions) where possible rather than always regenerating them.
This results in much quicker updates viewer-side but may bloat the asset cache (though this is fixable).
Also, sometimes issue have been seen where dynamic textures do not transfer to the viewer properly (permanently blurry).
If this happens and that flag is set then they are not regenerated, the viewer has to clear cache or wait for 24 hours before all cached uuids are invalidated.
CUrrently experimental. Default is false, as before.
Intended for use if there are future issues with mono crashes whilst generate dynamic textures.
This test is triggered via a new test-stress nant target.
Not run by default.
Disabled (status quo) by default.
This flag makes the dynamic texture module reuse cache previously dynamically generated textures given the same input commands and extra params for 24 hours.
This occurs as long as those commands would always generate the same texture (e.g. they do not contain commands to fetch data from the web).
This makes texture changing faster as a viewer-cached texture uuid is sent and may reduce simulator load in regions with generation of lots of dynamic textures.
A downside is that this stops expiry of old temporary dynamic textures from the cache,
Another downside is that a jpeg2000 generation that partially failed is currently not regenerated until restart or after 24 hours.
This prevented more than one additional ordinary folder from being created in the base "My Inventory" user folder.
Added regression test for this case.
Switched tests to use XInventoryService with mostly implemented TestXInventoryDataPlugin rather than InventoryService
Disabled TestLoadIarV0_1SameNameCreator() since this has not been working for a very long time (ever since XInventoryService) started being used
since it doesnt' preserve creator data in the same way as InventoryService did and so effectively lost the OSPAs.
However, nobody noticed/complained about this issue and OSPAs have been superseded by HG like creator information via the --home save oar/iar switch.
This is to accomodate situations where viewers will create more than one 'type' subfolder (e.g. calling cards)
But at the same time to prevent multiple such 'system' folders (those in the base "My Inventory" user folder).
This also makes GetFolderForType() only return a folder in the base "My Inventory" folder, if such a type folder exists
This doesn't affect any core OpenSimulator code since all callers were passing true anyway
But it allows region modules to create objects that are never persisted.
In testing, it appears that if multiple threads dispose of separate GDI+ objects simultaneously,
the native malloc heap can become corrupted, possibly due to a double free(). This may be due to
bugs in the underlying libcairo used by mono's libgdiplus.dll on Linux/OSX. These problems were
seen with both libcario 1.10.2-6.1ubuntu3 and 1.8.10-2ubuntu1. They go away if disposal is perfomed
under lock.
This is to resolve a reported issue in http://opensimulator.org/mantis/view.php?id=6232
Here, the land management module is using OpenSim.Framework.Cache (the only code to currently do so apart from the non-default CoreAssetCache).
This is to allow a second attempt to remove an avatar even if "show connections" shows them as already inactive (i.e. close has already been attempted once).
You should only attempt --force if a normal kick fails.
This is partly for diagnostics as we have seen some connections occasionally remain on lbsa plaza even if they are registered as inactive.
This is not a permanent solution and may not work anyway - the ultimate solution is to stop this problem from happening in the first place.
This was working in 0.7.2 but was accidentally removed from 0.7.3.
The --publish option for "save oar" will now save oars stripped of parcel owner information as well as scene object info.
Please use the --publish option if you want to publish oars that may be later loaded by others to the same grid from which they were saved.
Add description and default of false (as before) to OpenSimDefaults.ini
If set to true, this config switch will resend avatar appearance information (a small amount of UUID data, not the baked textures themselves) to other avatars in the sim every 60 seconds.
For me, this has helped with situations where avatars appear persistently grey - the LL viewer sometimes did not appear to request assets the first time the appearance data was sent.
However, this switch will not help with other appearance failure situations (e.g. failure to bake assets).
This setting is experimental but will not have any significant impact on the simulator if turned to true.
The bug manifested as follows: a large world was saved. All the assets were found. But for some unknown reason, the timeout timer was restarted. So after 1 minute it closed the Archive Writer, because it didn't receive any more assets during that minute. That caused the OAR to become corrupted because ArchiveWriteRequestExecution.Save() was still running.
The convention is that if an object implements IDiposable() the code must explicitly call Dispose() or call it via the using statement.
This may be particularly important for GDI+ objects since they encapsulate native code entities.
themselves. For that, the physics module simply calls RequestAssetMethod, which
in turn points to Scene.PhysicsRequestAsset. This gives physics access to
the asset system without introducing unwanted knowledge of the scene class.
A deadlock was observed today where NPC removal on a script thread would lock the NPC list and then try to lock the sensor list via scripted attachment removal.
Concurrently, the sensor sweep thread would lock the sensor list and then try to lock the NPC list to check NPC status.
This commit resolves the deadlock by replacing the sensor list on update rather than locking it for the duration of the sweep.
c150320 started explicitly disposing of the physics scene and nulling it out on region shutdown.
However, the update loop may not have yet checked Scene.ShuttingDown, particularly if avatars were not in the scene, causing failure when it tried to lookup time dilation.
This commit moves the setting of m_shuttingDown above the existing 500ms pause to notify avatars that they are being kicked.
This should not affect the few other places that use this flag.
Possibly this was for a feature that was never implemented or was otherwise removed.
Thanks to SignpostMarv for the spot of the warning that shows this parameter was never changed.
and into its own class. The BulletSim data structures track
individual prims as linksets of 1 so most of the prim code is not
different between a linked and unlinked object.
If we copy the asset description then we will only ever replicate the very first description, if there was one, not any subsequent changes.
Thanks to Oren Hurvitz of Kitely for this patch from http://opensimulator.org/mantis/view.php?id=6107
I have adapted it slightly to change the order of arguments (name before description rather than vice-versa) and slightly improve some method doc.
This aims to capture the amount of memory that OpenSim turns over whilst operating a region.
This memory is not lost - apart from leaks it is reclaimed by the garbage collector.
However, the more memory that gets turned over the more work the GC has to do to reclaim it.
This is fired when all regions are ready or when at least one region becomes not ready.
Recently added EventManager.OnRegionReady becomes OnRegionReadyStatusChange to match OnLoginsEnabledStatusChange
There is less justification for this word arrangement (verb after noun) now that command help is categorized.
Also removes "image queues show" in favour of existing alias "show image queues".
justincc's original work. Sample scripts before doing so. Also refactor some
crucial common code and eliminate parameters that were only ever used with
the same constant value.
When linking, detach the no longer used SOG's from backup so they can be
collected. Since their Children collection is never emptied, they prevent
their former SOPs from being collected as well.
Physical linksets are fully functional.
Tweeking of the vehicle code to make it semi-work.
Utilize the new API2 for some setting operations.
Add GetOrientation() API call for proper reporting of children of linksets.
Changes the interface between C# and C++ code so old DLLs won't work!
with an ini configuration parameter.
Correct computation of relative offsets of children in a linkset.
Remove a prim from any link relationship before deleting it.
Minor code flow cleanups.
This same constant will later be used with llGetDetectedType().
This constant has a different name from NPC to avoid possible conflict with future LSL changes.
This constant has a different value to try and avoid unnecessary conflict with future constants that may use the same value.
Using the 'NPC' constant with llSensor() will remain valid but is deprecated.
Giving a large folder from one avatar to another was causing a long delay when handled synchronously, since it took some time to retrieve the necessary data from the inventory service.
Handling this asynchronously instead stops this delay from disrupting all avatars in the scene. This has been shown in OSGrid.
I see no reason for not handling all IM messages asynchronously, just as incoming chat is handled asynchronously, so this has been switched for all instant messages.
Thanks to Nebadon for testing this change out.
Even where checks are being made they aren't enough since they all assume that the Animator they just checked is still there in the next line, which is not necessarily the case without locking.
The memory used is small and these should be GC'd anyway when the SP is released. If this is not happening then the wider problem of old SPs being retained needs to be resolved.
This replaces EventManager.OnLoginsEnabled which only fired when logins were first enabled
and was affected by a bug where it would never fire if the region started with logins disabled.
This saves listeners from having to re-retrieve the scene from their own lists, which won't work anyway if multiple regions with the same name have been allowed
This reverts commit 59a29f5f22.
The original revert was committed by mistake - it turns out this was not the cause of Mantis 6089
Conflicts:
OpenSim/Region/ScriptEngine/Shared/Api/Implementation/LSL_Api.cs
This appears to be a regression from back in commit db91044 (Mon Aug 22 2011) where we started to send TaskInventoryOffered msg dialog rather than InventoryOffered dialog.
This is probably correct, but failed because the bucket was too large and because we wouldn't have handled the TaskInventoryDeclined option anyway.
This patch handles both of these and make llGiveInventoryList() use TaskInventoryOffered as well
Fixes http://opensimulator.org/mantis/view.php?id=6089
Note, this behaviour appears to be at variance with the ll grid as of Tues 17 July 2012, testing with viewer 3.2.1.
The item name in inventory does not change either at the point of detach or after a relog.
This doesn't seem to make any sense and probably stems from a period when this code was directly involved in attaching objects directly from the scene.
This message is already being sent by InventoryAccessModule code instead.
This was because teleporting to the new region invoked the new session setup code before the agent was removed from the old region, which then invoked the session teardown code.
Now, we only invoke the teardown code if the region ID occupied by the agent being removed is the same as the one registered for the current session.
This may well come back in the future when this subinterface is actually used but it currently isn't and I feel the name was poor.
Everything uses IClientAPI.RemoveEndPoint which also returned the full endpoint rather than just the ip address.
Also changes separator from comma to bar to match existing [LoginService] config features.
Divergence of config names for identical facilities in different places makes for an unnecessarily confusing user experience.
Also change to use standard table formatting
"show circuits" and "show connections" console commands are very similar but access different data structures.
The encodings are thread-safe and already used in such a manner in other places.
This isn't done where Byte Order Mark output is suppressed, since Encoding.UTF8 is constructed to output the BOM.
This is only currently meant for use by regression tests that don't have any issues if XEngine is started up quickly, since no other operations will be occuring simultaneously.
Therefore, this is not yet documented externally.
This also makes the detection in SP.FindNextAvailableSitTarget() and SendSitResponse() identical.
Previously they varied slightly (SendSitResponse didn't check for an older type of invalid quaternion) but the practical effect is most probably zero.
This is to fix a regression starting from 5301648 where attachments had to start being deleted before persistence in order to avoid race conditions with hud update threads.
This is to avoid the worst of the problems in mono 2.6, 2.10 where an aborted thread does not always release all its locks.
This very short grace period is identical to the existing behaviour when a script is removed from the scene.
This prevents a stack overflow where a get position on the avatar will refer to the attachment which will in turn refer back to the avatar.
This required recording of all sitting avatars on a prim which is done separately from recording the sit target avatar.
Recording HashSet is null if there are no sitting avatars in order to save memory.
This was causing 2 copies of the module to be created for each scene.
Probably no bad consequences other than a small waste of memory (both for the module and for the warp3D renderer it loaded)
This is always done later on in SceneGraph.AddSceneObject() if the call hasn't failed due to sanity checks.
There's no other purpose for this method to exist and it's dangerous/pointless to call in other conditions.
Cleanup preceeding major work. No major functional changes.
Collisions reported to simulator more efficiently.
BulletSim binaries updated using a more recent version of Bullet (v2.80-r2527).
This is to help detect situations where thread timeout warnings are being generated because of general machine issues rather than deadlock, network or other problems.
This works like osForceAttachToAvatar() but allows an object to be directly specified from the script object's inventory rather than forcing it to be rezzed in the scene first.
Still only attaches objects to the owner of the script.
This allows one to bypass the complicated co-ordination of first rezzing objects in the scene before attaching them.
Threat level high.
This occurred in b18c8c8 (Thu May 17 2012).
This was a cause of very occasional race conditions and likely memory leakage as clients came and went from the region.
null would only ever be returned if the item couldn't be located within inventory and this would happen immediately.
In this case, derezzing wouldn't work anyway since there is no item to derez.
This is to take account of situations where the user was intending to specify a script engine using colon using its default language.
This probably generates few false positive as scripts are less likely to end a first line colon with a comment for other purposes.
This works by only posting the "Selected engine unavailable" message if we're falling back on XEngine and the language is one handled by XEngine.
In cases where the language is not handled or not allowed, the user will still be notified by the later compiler error.
This avoids the overwhelming majority of false positives where the first line contains a : for other reasons (e.g. source control systems, vim settings, etc.)
Ultimately, I think it would be better to detect script language/engine with a mechanism that didn't just rely on : detection (e.g like #! in unix scripts).
This matches behaviour in fatpack crossing, where attachments are cloned before their properties are changed.
This only applies to crossings to simulators running code released before April 2011.
If delete doesn't occur first then the update thread can outrace the IsAttachment = false necessary to save attachments and send hud artifacts to other viewers.
Serialization of attachments requires IsAttachment = false so that correct positions are serialized instead of avatar position.
However, doing this when a hud is still attached allows race conditions with update threads, resulting in hud artifacts on other viewers.
This change sets SOG.IsDeleted before serialization changes take place (IsDeleted itself is not a serialized property).
LLClientView then screens out any deleted SOGs before sending updates to viewers.
This involves three steps
1) Return gracefully in UrlModule.HttpRequestHandler() instead of throwing an exception when the url cannot be found in its index
2) Return true instead of false in HasEvents() if no matching request is found in the map. This call will only happen in the first place for raced requests.
3) Return a 404 in GetEvents() if the request is not in the index, rather than a blank 200 OK.
Many thanks to Tom Haines in http://opensimulator.org/mantis/view.php?id=6051 for doing some of the work on this.
This is to provide an indication of what's happening now that the default isn't to report every single script start.
Changes XEngine logging level in OpenSim.exe.config from WARN to INFO.
Previously, setting IsAttachment to false then true was necessary to serialize the updated attachment object information.
However, UpdateAttachmentPosition no longer does this update.
Whilst IsAttachment is set to false there is a race condition where the update thread can wrongly send hud object updates to clients that do not own the hud, resulting in screen artifacts.
It's a valid configuration not to have these modules, but I think it's still worth logging the fact that certain permissions won't work (always return true)
Also removes superflous Close() commands for statements taking place within using() constructs
Also adds some comment out debug log messages for future use.
Accidentally make responseString null by default instead of String.Empty.
It needs to be something in case the XmlRpcRequest deserialize throws an exception due to bad xml (a failure which we silently swallow!)
If this happens, then the non-owners would see unremovable huds that they did not own until relog, and sometimes even beyond that.
This was due to a race between the entity update and the attachment code when moving an object from within scene to a hud.
Change SOP.SendFullUpdateToClient() and SoundModule.PlayAttachedSound() to use this rather than different magic number formulations.
This also corrects a bug in PlayAttachedSound() where the code assumed that all attachment points over 30 were HUDs.
It appears this is no longer true with Neck and Root (Avatar Center)
This is to avoid logging a 'slow' request when the source of delay is the viewer in processing a response.
This is not something we can do much about on the server end - it's server-side delay that we're interested in.
To ensure consistency, this commit also had to refactor and simplify inbound non-poll network request handling, though there should be no functional change.
IOSHttpResponse no longer exposes the Send() method, only classes in OpenSim.Framework.Servers.HttpServer should be doing this.
Only the GetTextureHandler was sending its own response. Now it leaves this to BaseHttpServer, like all other core handlers.
Added "timer_Interval" to the OpenSimDefaults.ini, leaving the default value set to 1200, as the previous default setting. The value represents seconds. To change the default, copy the entry to OpenSim.ini and multiply the number of minutes for the interval by 60.
Stop checking IsLoggingOut on these listeners, if called with a root agent then we always want to perform these actions.
This covers cases where the client is closed due to manual kick, simulator shutdown, etc.
IsActive is more appropriate since unack timeout is not due to voluntary logout.
This is in line with operations such as manual kick that do not set the IsLoggingOut flag.
It's also slightly better race-wise since it reduces the chance of this operation clashing with another reason for client deactivation (e.g. manual kick).
llTeleportAgentGlobalCoords. These do NOT use PERMISSION_TELEPORT like
their SL counterparts because that permission is not yet understood by TPVs
based on v1.x.
This is the same async behaviour as normal logouts.
This is necessary because the event queue will sleep the thread for 5 seconds on an ack timeout logout as the client isn't around to pick up the final event queue messages.
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.
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.
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()
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.
Adapted from a patch by Garmin Kawaguichi in http://opensimulator.org/mantis/view.php?id=6027
Garmin says that fix-phantoms allows one to reset objects when going back from megaregion to normal regions as well as the othe rway around.
Thanks!
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.
This makes frame time stats properly tally with fps, which saves confusion and makes it easier to interpret numbers.
In some ways this is not so artifical - physics FPS runs at the higher rate.
This is to make these statistics actually match their names (and also be more accurate as number of frames can vary under heavy load)
Currently using scene frames (11.23 every second) instead of physics frames (56.18 per second)
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).
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
Justin Clark-Casey (justincc)