MantisBT - Issues
Updated: 45 min 3 sec ago
llUnSit doesn't work on ODE, if there is no llSitTarget. In BulletSim it works and in SecondLife it works too.
When a Avatar sits on a childprim of an Linkset, then his Regionposition is wrong by get it via the Linksetscript, or by llSensor from an autonomous script outside as well.<br /> <br /> This has maybe to do with Mantis 0007043 (<a href="http://opensimulator.org/mantis/view.php?id=7043">http://opensimulator.org/mantis/view.php?id=7043</a> [<a href="http://opensimulator.org/mantis/view.php?id=7043" target="_blank">^</a>])<br /> <br /> This problem only appear on BulletSim. It works now for ODE since 7043 is fixed.
Hello.<br /> <br /> I want to add functionality of client authentication to HTTPserver of OpenSim.<br /> With this capability, server (Robust or Money etc) can authenticate the Region Server.<br /> Currently, I am using this code at our DTL/NSL Money.<br /> <a href="http://www.nsl.tuis.ac.jp/xoops/modules/xpwiki/?OpenSim%2FMoneyServer">http://www.nsl.tuis.ac.jp/xoops/modules/xpwiki/?OpenSim%2FMoneyServer</a> [<a href="http://www.nsl.tuis.ac.jp/xoops/modules/xpwiki/?OpenSim%2FMoneyServer" target="_blank">^</a>]<br /> <br /> Modification of opensim-libs is required for this.<br /> I attach the patch code.<br /> <br /> Thanks.
I tried using llGetObjectDetails worn as HUD and i found it will NOT work 100%.<br /> <br /> I try with:<br /> OBJECT_POS is ok <br /> OBJECT_ROT remains 0 <br /> OBJECT_VELOCITY remains 0<br /> <br /> I tested these 3 parameters in SL and all works fine.<br /> So there is also a bug with llGetObjectDetails in OpenSim.
Since commit 36d146b sit camera settings no longer set the camera correctly when the prim that the camera parameters are applied to is not root.
I created a small hud with llGetVel() in order to get the speed of movement of my avatar. It works very well.<br /> Then I installed my avatar on a single vehicle and there I found that my HUD was not working. llGetVel() simply returns 0 all the time. So it does not work when the avatar is on a vehicle.<br /> I have been on SL for my test and normal HUD with a vehicle and it works fine.<br /> I therefore concluded that there is a bug about it in OpenSim.
0007174: [HYPERGRID LINKER] hop:// handling could be improved - Cannot hyperlink to regions on the same grid
hop:// handling now works reasonably well in recent viewers such as Firestorm 4.6.5 but addresses that contain the SAME grid one is already on, including those placed in the nearby chat window when you teleport WITHIN a grid, cannot be used, possibly because "same grid" addresses are not allowed. If the Hypergrid Linker could allow same grid teleports or just ignore the grid element if it is the same as the grid the avatar is on, hop:// handling would be more useful.<br /> <br /> 11:45:14 - [HYPERGRID LINKER]: Cannot hyperlink to regions on the same grid
A script that detects notecards in the inventory then shows then on a dialog popup comes up twice on the list for each notecard on the list making the list double the names on the list
While writing a region module i was using 'ScenePresence.IsDeleted' to verify the avatar had not left. I found that it was always returning false, even after the ScenePresence had left the sim.<br /> <br /> On checking further i see that the property is inherited from EntityBase so i wondered if it was not intended to be used with ScenePresence. I checked this and while most of the calls to IsDeleted are to SceneObjectGroup which also inherits from EntityBase, there are however two calls in the script engine that try to verify a ScenePresence using this property. The script engine methods that try to use it are llGetAgentList and SensorRepeat.<br /> <br /> I would expect that if this did cause any problems it would be if those script functions are triggered at the same time as an avatar leaves.
I was performing region to region TPs, every 30 seconds, and BulletSim oopsed:<br /> <br /> <a href="http://pastebin.com/iFrgETHc">http://pastebin.com/iFrgETHc</a> [<a href="http://pastebin.com/iFrgETHc" target="_blank">^</a>]
0007170: Every so often you get teleported to the wrong coordinates when teleporting between regions.
While trying to track down something else, I started doing lots of teleports between neighboring regions. I was reminded of a very old, long-standing issue where you get thrown to the wrong coordinates in the destination region.<br /> <br /> Most of the time, you go exactly where you're supposed to. Just once in a while, you don't.<br /> <br /> Here are the log entries for on such occurrence:<br /> <br /> <a href="http://pastebin.com/a2hZxJC1">http://pastebin.com/a2hZxJC1</a> [<a href="http://pastebin.com/a2hZxJC1" target="_blank">^</a>]<br /> <br /> The choice of physics engine (BS or ODE, seems unrelated, as it happens under either.
For a while I see something wrong with BulletSim <br /> on the attached picture you can see that <br /> <br /> Image 1: vehicle position is +/- normal, just a little levitating above the ground, +/- 0.1, 0.2 <br /> <br /> Image 2: vehicle is driven into the ground, +/- 0.5. This is not normal.
If you teleport an agent out of the region, via an osTeleportAgent call, from within collision_start, collision_end does not trigger once the avatar is no longer colliding with the object. This is true of both ODE and BS.<br /> <br /> On a related note, if the teleport is to another location within the same region, then collision_end triggers as expected, but after a 5 second delay (presumably caused by osTeleportAgent). This is also true of both ODE and BS.
I contate that llSetKeyframedMotion not yet operating at 100%.<br /> Indeed, if I create a rotating platform of +/- 5.0x5.0x0.1 and i set my avatar standing above my avatar should follow the rotational movement as is the case in Second Life but in Opensim my avatar remains in place on the platform. He does not move. I guess this is something that concerns the physic engine. Currently I use BulletSim.
llSetMemoryLimit is ignored in Osgrid and in Metropolis as well. It always fails <br /> (MONO-Script) and the resulted amount of Memory is always 16 k.<br /> <br /> In Avination the Function fails too. But resulted Memory there is always: 196604<br /> <br /> This Script works in SecondLife as expected. It returns TRUE and the ammount of Memory<br /> is adjustable.
This morning I noticed a number of new git updates so I made a fresh clean build using r/24732 opensim-484aa72. On start-up a large number of maptile and script errors are thrown and on avi login a further large number of errors are also thrown and no textures appear on any of my worn mesh. Most messages seem to relate to the changes made to [SynchronousRestObjectRequester].<br /> <br /> I re-downloaded the git, rebuilt it and repeated the test in case something odd happened the first time but the results were identical.<br /> <br /> I've attached a log with shows the full startup sequence followed by me logging in and the out again.<br /> <br /> I'll revert back to r/24717 opensim-a845c1a which I was using prior to this morning and showed no errors on startup or login.
For a long time when using ODE I had a personal rule of thumb to never have my step-to-step height difference greater than 0.25m and usually kept it at 0.2m or even 0.15m.<br /> <br /> After switching to Bullet almost a year ago these simple prim stairs worked fine (though Bullet would occasionally struggle with 0.25m steps); however a month or two ago the mesh shape was altered in Bulletsim (to help with the ground issue iirc) and now Bullet is having a very, very hard time with shallow stairs but not with steeper ones. I get noticeable "stuttering" in avi motion when attempting to walk up a simple straight flight of 0.15m or 0.2m stairs, but completely smooth movement up slightly steeper stairs at 0.25 or even 0.3m. I suspect it has something to do with the shape of the new bounding box.<br /> <br /> It would be nice not to have to build invisible ramps over absolutely every staircase just in the off chance that Bullet has trouble with it so anything you can do to smooth this out would be super
1) Chat across region borders<br /> <br /> Possition two avatars next to a region border of two regions simulated by the same OpenSim process. These avatars need to be in chat range, while both avatars are on different regions. Then let one or both avatars send public chat messages to each other.<br /> <br /> That works, but when these avatars move away from each other, the chat messages of these avatars are still hearable anywhere (!) on the other region.<br /> <br /> It looks like that the way cross region public chat is not correct. When avatars move away from the region border messages should not cross the region border anymore.<br /> <br /> 2) Script chat always hearable<br /> <br /> There seems to be a similar bug related to objects sending chat messages.<br /> <br /> Additionally llOwnerSay, llRegionSay, etc. are hearable on the same region and on neighboring regions, even if the recipient is not the owner, or if the channel is not zero, or if the recipient is out of the defined hearing range.
I have been testing BulletSim physics on a separate computer connected to our grid. Finding it stable enough, at least on 0.7.6 Dev (fc4f86a), I've decided to run it on our "main" grid (running Ubuntu 10.04 and Mono 188.8.131.52).<br /> <br /> I was astonished at the amount of CPU that this is currently consuming. Usually, I can easily run 6 instances on this old server, each with about 100.000 prims on several regions, and, when idle, CPU consumption is around 2%, peaking to 10% occasionally (e.g. when saving snapshots or doing something), but, of course, when avatars are around, the load increases when the avatar just logs in. The server has 6 GBytes of memory and rarely ever consumes more than 3-4.<br /> <br /> When switching to BulletSim, memory consumption went to twice normal (see resolved issue <a href="http://opensimulator.org/mantis/view.php?id=6599">http://opensimulator.org/mantis/view.php?id=6599</a> [<a href="http://opensimulator.org/mantis/view.php?id=6599" target="_blank">^</a>]) but CPU skyrocketed. Idle instances are routinely consuming 40-60% of all CPU — each. Which, of course, brings the server to its knees, and logging in, while it allows everything to be loaded relatively quickly, avatar movement is next to impossible.<br /> <br /> Switching back to ODE restores the server to its usual low CPU consumption.<br /> <br /> Now I understand that this might be just related to BulletSim parameters, with which I'm not familiar (yet). Maybe there is a recommended set of parameters to keep CPU consumption as low as under ODE?
I been talking with one of the developers who contributed a lot of the hypergrid fixes to simmian. I found out that I needed to point UserAgentServerURI at the GridLogin folder instead of Grid but opensimulator during start no matter how I set it give me this message and resets it to ().<br /> <br /> 23:03:14 - [USER AGENT CONNECTOR]: Malformed Uri : Absolute URI is<br /> too short<br /> 23:03:14 - [USER AGENT CONNECTOR]: new connector to ()<br /> <br /> This makes opensim instance default to /Grid/ instead of looking in /GridLogin/ where the hypergrid.php file has been moved