This week following the Thanksgiving holiday we move back to the regular server updates. The Main Channel gets the maintenance package that was running on all three release candidate channels. This package has the fix for linksets and objects larger than 64 meters. (BUG-166 – Something invisible pushes avatars around.)
While some saw linksets larger than 64m as a good thing, it was being used to grief people and also accidentally causing problems. With this release the problem is ‘fixed’.
BUG-311 – Server: Disabled scripts still use same CPU time as if they were enabled. This problem was a reporting issue. It has been corrected so the correct time is reported. Region owners should see their region’s script time decrease.
PATHBUG-198– Object set to incompatible Navemesh type shows child prim name, not root prim name in error popup. This is a problem where a script error was returning a useless object name. If you set a house to be a static obstacle for Pathfinding, a door or other scripted movable item would through an error and return the name of the prim containing the script. That was often OBJECT. Now the error returns the name of the root-prim of the item containing the script, which is usually more helpful as it is something other than OBJECT.
As usual a couple of region crash causes have been fixed too. These unspecified fixes are often security/exploit fixes, so we won’t hear anything about what they were.
Blue Steel & Le Tigre
These two channels get the same server maintenance package. The visible change is in the scripting function: llGetObjectDetails(). The function gets three new options:
- OBJECT_PHYSICS – no wiki description yet.
- OBJECT_PHANTOM – no wiki description yet.
- OBJECT_TEMP_ON_REZ – no wiki description yet.
These options return a TRUE/FALSE answer reflecting the nature of the object’s details. However, attachments will always return 0/FALSE.
The package also includes ‘numerous’ fixes centered around mesh and physics issues.
This channel will get the same package items as are in Blue Steel with the additional crash fixes the Lindens hope will improve stability.
One way or another the fixes in Blue Steel will likely make it to the main channel next week. Of course if they fail, it is likely nothing will make it to the main channel. I would guess from the way Magnum is being handled the Lindens feel confident the Blue Steel changes are good. And I guess a little less so about the changes in Magnum. If all goes well, Magnum will roll forward to the main channel and next week all three release candidate channels will get new packages.
There is what some think is a server side memory leak. As the server runs memory use builds from 50mb up to 230mb in about an hour… or longer. Around 230mb the region starts to try and conserve memory by refusing to allow attachments to rez.
The region Rainbow Cove is presently suffering the problem. Other regions that had the problem last week were moved to the Magnum channel and that seemed to resolve the problem. The hope is the roll out from today resolved this problem. But, the problem is continuing in Rainbow. Having a region that consistently reproduces the problem gives the Lindens a big helping hand in finding the problem.
Lolita Park, Zerango, and Jinxies regions have also been reported as having high memory use and problems. The Lindens are looking at those trying to find the problem.
If you are seeing some odd problems you think may be related have the region owner check memory use and report the problem referring to these JIRA items, if it is similar to them. If you see a similar problem and you are a region owner filing trouble tickets mention Derrick and Vix Linden and Rainbow Cove stating you believe it is a similar case.
BUG-888 – Every 12 hrs sim has to be restarted. Error messages come from items all ready out, (different all the time) when trying to put out an item or clothes from inventory. This continues until sim is restarted and then ok again for 12 hrs.
or BUG-772 – Simulator refusing to rez objects after 10 hour timeframe.
World Map Problems
The world map uses images collected by a backend service that scans the grid. That service has been having problems for a couple of weeks. It seems the problem is continuing. It is better than it was, but it is still there. If you are having a problem, file a JIRA bug report.