SmartThreadPool code comes from http://www.codeproject.com/Articles/7933/Smart-Thread-Pool
This version implements thread abort (via WorkItem.Cancel(true)), threadpool naming, max thread stack, etc. so we no longer need to manually patch those.
However, two changes have been made to stock 2.2.3.
Major change: WorkItem.Cancel(bool abortExecution) in our version does not succeed if the work item was in progress and thread abort was not specified.
This is to match previous behaviour where we handle co-operative termination via another mechanism rather than checking WorkItem.IsCanceled.
Minor change: Did not add STP's StopWatch implementation as this is only used WinCE and Silverlight and causes a build clash with System.Diagnostics.StopWatch
The reason for updating is to see if this improves http://opensimulator.org/mantis/view.php?id=6557 and http://opensimulator.org/mantis/view.php?id=6586
TimeSpan.Milliseconds is an int64. However, STP casts this to an int (32-bit).
If TimeSpan.MaxValue is given then the casting results in an invalid value for the SDK WaitHandle.WaitAll() call.
This was causing the co-op script termination regression tests to fail on Windows but not Mono 2.10.8 (which is perhaps not strict in the negative values that it accepts).
Solution here is to use the int millisecondsTimeout STP call rather than the TimeSpan one.
This also allows us to more clearly specify Timeout.Infinite rather than TimeSpan.MaxValue
Thanks to Teravus for this spot.
Moves the noise co-op start/stop debug log messages to only display if xengine debug level >= 1
Logs which stop strategy is being used (abort or co-op)
Adjusts some other logging to remove not very useful stuff
This is to eliminate disruption until co-op termination has been well-tested.
In non co-op mode, XEngine will continue to load DLLs of the existing Script class and the new XEngineScript class.
Moving to co-op mode still requires existing script DLL deletion to force recompilation, either manually or by setting DeleteScriptsOnStartup = true for one run.
This change also means that scripts which fail to initialize do not still show up as running scripts.
This involves inserting opensim_reserved_CheckForCoopTermination() calls in lsl -> c# translation at any place where the script could be in a loop with no wait calls.
These places are for, while, do-while, label, user function call and manual event function call.
Call goes through to an XEngineScriptBase which extends ScriptBase.
IEngine is extended to supply necessary engine-specific parent class references and constructor parameters to Compiler.
Unfortunately, since XEngineScriptBase has to be passed WaitHandle in its constructor, older compiled scripts will fail to load with an error on the OpenSim console.
Such scripts will need to be recompiled, either by removing all *.dll files from the bin/ScriptEngines/<region-id> or by setting DeleteScriptsOnStartup = true in [XEngine] for one run.
Automatic recompilation may be implemented in a later commit.
This feature should not yet be used, default remains termination with Thread.Abort() which will work as normal once scripts are recompiled.
This makes use of EventWaitHandles since various web references indicate that Thread.Interrupt() can also cause runtime instability.
If co-op termination is enabled, then termination sets the wait handle instead of waiting for a timeout before possibly aborting the thread.
This allows the script to cleanly terminate if it's in a llSleep/LL function delay or the next time it enters such a wait without any timeout period.
Co-op termination is not yet testable since checking for termination request within loops that never trigger a wait is not yet implemented.
This commit, unlike 1b5c41c, passes the wait handle as an extra parameter through IScript.Initialize() instead of passing IScriptInstance itself.
This makes use of EventWaitHandles since various web references indicate that Thread.Interrupt() can also cause runtime instability.
If co-op termination is enabled, then termination sets the wait handle instead of waiting for a timeout before possibly aborting the thread.
This allows the script to cleanly terminate if it's in a llSleep/LL function delay or the next time it enters such a wait without any timeout period.
Co-op termination is not yet testable since checking for termination request within loops that never trigger a wait is not yet implemented.
This is to allow the future co-operative script thread terminate feature to detect and act upon termination requests.
This splits the assembly and state loading out from the ScriptInstance() constructor to a separate Load() method
in order to facilititate continued script logic regression testing.
As with script stop (via llDie()) aborting other scripts event threads, llResetOtherScript() can also abort any current event thread on another script.
On mono 2.6, 2.10 and possibly later this may cause locking problems in certain code areas.
This commit reuses the recently introduced [XEngine] WaitForEventCompletionOnScriptStop to make this a 1 sec timeout, rather than 0 secs.
This is for bug hunting purposes where thread aborts may be causing dangling lock issues and subsequent vm crashes on mono (with ReaderWriterLockSlim, etc.)
This commit also adds script name, part name, uuid, etc. for later identification.
This information has been sent to console since 2009 but may be turned down if it proves too noisy.
However, I still currently need it to investigate some region problems probably triggered by scripting.
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.
These behave identically to llAttachToAvatar() and llDetachFromAvatar() except that they do not enforce the PERMISSION_ATTACH check
Intended for use in completely controlled dedicated environments where these checks are more a UI hinderance than a help.
Threat level high.
The previous lines-per-second measurement used for top scripts report was inaccurate, since lines executed does not reflect time taken to execute.
Also, every fetch of the report would reset all the numbers limiting its usefulness and we weren't even guaranteed to see the top 100.
The actual measurement value should be script execution time per frame but XEngine does not work this way.
Therefore, we use actual script execution time scaled by the measurement period and an idealised frame time.
This is still not ideal but gives reasonable results and allows scripts to be compared.
This commit moves script execution time calculations from SceneGraph into IScriptModule implementations.
This seems to be a particular problem with ReaderWriterLockSlim, though other locks can be affected as well.
It has been seen to happen when llDie() is called in a linkset running more than one script.
Alleviation here means supplying a ScriptInstance.Stop() timeout of 1000ms rather than 0ms, to give events a chance to complete.
Also, we check the IsRunning status at the top of the ScriptInstance.EventProcessor() so that another event doesn't start in the mean time.
Ultimately, a better solution may have to be found since a long-running event would still exceed the timeout and be aborted.
On resume, we need to place requeue the script for event processing if there are any events on the queue.
Also need to do this under m_Script lock in order to avoid a race
These aim currently to suspend and resume all scripts.
However, resume isn't currently working due to what looks like a bug in resume functionality itself.
The only times when ParentGroup might be null is during regression tests (which might not be a valid thing) and when scene objects are being constructed from the database.
At all other times it's not possible for a SOP not to have a SOG parent.
is fully rezzed and all scripts in it are instantiated. This ensures that link
messages will not be lost on rez/region crossing and makes heavily scripted
objects reliable.
pass script state and assembly again properly. Reintroduce respecting tht
TrustBinaries flag. Changes the interregion protocol! No version bump
because it was broken anyway, so with a version mismatch it will simply
stay broken, but not crash. Region corssing still doesn't work because
there is still monkey business with both rezzed prims being pushed across
a border and attached prims when walking across a border. Teleport is
untested by may work.
- Faster prim inventory actions. Multiple threads can read at once.
- Fixes the known prim inventory thread locks
- In the event of a thread lock occurring, it will usually self heal after sixty seconds with an error message in the console
Melanie