noticed that Scene.Close() will only call Close on non-shared region
modules. i've now added code to SceneManager.Close() to collect all
shared region module from each scene before calling Scene.Close()
on it and then, once, all Scenes are closed, go through the list of
collected shared region modules and close them as well. SceneManager.Close()
is only called when we initiate a shutdown --- i've verified that a
Scene restart does not trigger the shutdown of shared modules :-)
also, this adds a couple of bug fixes to the IRCBridgeModule (which
after all didn't take kindly to being closed) as well as a check to
InterregionModule's Close() call.
finally, this fixes the RestPlugin's XmlWriter so that it no longer
includes the "xsd=..." and "xsi=..." junk.
* Added a Non-finite avatar position reset. This will either handle the <0,0,0> avatar gracefully, or send the avatar to 127,127,127 if that also doesn't work. ( I've only been able to reproduce this error once on my development workstation )
* Looks up UUIDNames for script time and colliders in a separate thread.
* Hopefully this'll allow you to look at top scripts on a region that has a lot of scripts without crashing your client thread.
* Some situations do not match the client's render of the tori, we know and are working on it. This is an initial support patch, so expect it to not be exact.
* Some tapers are acting slightly odd. Will fix.
There appears to be a problem with the mapping of scripts when an llHTTPRequest completes.
CheckHttpRequests() looks for a function that maps to the localID associated with the http
request. However, the only context in which it looks is that of the first region. That is,
m_CmdManager.m_ScriptEngine.m_ScriptManager is the same no matter where the script executed
that initiated the llHTTPRequest. Since scripts appear to be loaded into a region specific
scriptmanager on startup, the event handler is only found for requests coming from the first region.
* Linden uses a neutral height channel of 128.0 on their multiplier. OpenSim was using a neutral of 127.0 - this has been changed to 128.0, this may cause files exported to the .RAW format to look slightly different when loaded back in - it is highly recommended to use the R32 format instead which avoids these sorts of issues.
* Made a tweak to the Terrain Plugin loading process.
Attached is an initial implementation of llGetNotecardLine and
llGetNumberOfNotecardLines. I decided to go ahead an send these out for
comment while I continue to work on the second part of the proper
implementation. These functions work and return the values requested, as
initially defined in the code, but should be properly implemented to return
the requested information via a dataserver event. This
event will be added and these functions fixed and included in a second
patch shortly.
llLoopSound sends out one packet to clients in view, so it doesn't work anymore
when clients enter later on, or the prim is modified in any way.
Solution: Stored sound data on prim, send full update instead.
llStartSound and llLoopSound now accept both LLUUIDs to a sound as well as object
inventory sound names. llStopSound clears prim data and sends full update.
* It's verbose currently since this is new. You'll see: [CLIENT]: Warning Duplicate packet detected X Dropping. After this is sufficiently tested we'll remove that m_log.info line.
If a script updates an object to the same position or rotation offset,
the object triggers an update and storage of the object. This become
more prevalent in sensor and timer events which may be firing frequently.
This patch proposes a new function : osOpenRemoteDataChannel(key channeID)
that allow to open an XMLRPC channel for remote_data event. The difference
is that the channelID can be customized instead of being randomly generated.
Justin Clarke Casey