| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
|
|
|
|
| |
EventWaitHandles, I'm a bit concerned about what effect this might have on performance in a large grid. But I've spent long enough on this problem for one day and as its valentines day, I will be killed and tortured (most likely not in that order) if I don't stop work for the day soon.
|
|
|
|
|
| |
* Minor work towards abstracting terrain.
|
| |
|
|
|
|
|
|
| |
* Converted all instances of "256" spotted to use RegionSize instead. Some approximations used for border crossings (ie 255.9f) are still using that value, but should be updated to use something based on RegionSize.
* Moving Terrain to a RegionModule, implemented ITerrainChannel and TerrainModule - nonfunctional, but will be soon.
|
| |
|
|
|
|
|
|
|
|
|
| |
appears that in conjunction with MW's locking changes, that two threads may race to create the avatar
appearance in the TryGetAvatarAppearance() method. Moving off Add will mean the second replaces the first instead of throwing the exception. This may still not be a good
permanent solution - more extensively locking or a code rearrangement could be better. I'm going to leave this in MW's much more familiar hands now.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
all TerrainEngine.LoadFromFileX(string filename) methods don't set the dirty bits in heightmap.diff[,] in contrast to the TerrainEngine.LoadFromFileX(string filename, int dimensionX, int dimensionY, int lowerboundX, int lowerboundY) cousins. this has the unpleasant drawback that the Scene.SendTerrainUpdate(bool) will not send layer data to the client.
the attached patch fixes all LoadFromFileX(.) methods.
Thanks!
* Also small build fix for AvatarFactoryModule
|
| |
|
|
|
|
| |
did much anyway, so perhaps it's not currently in use.
|
|
|
|
|
|
| |
* Many thanks to alex_carnell for the necessary information for this
|
| |
|
|
|
|
|
|
|
|
|
|
| |
(though I still don't recommend it for welcome regions unless object build is off.
* Updated the ode.dll for windows with a more reasonable stack space reserve. Linux users will need to type ulimit -s 262144 before starting up OpenSimulator if using Physical Prim to protect against stack collisions. or run the included ./bin/opensim-ode.sh to start up OpenSimulator in ODE mode.
* Added internal collision score and am keeping track of 'high usage' prim.
* Tweaked collisions some more
* Tested up to 460 physical prim in extremely close quarters (which was previously impossible in OpenSim). After 460 in tight quarters, physics slows down enough to make it hard to do any moving, however.. non physics things still work, such as logging on to the simulator, etc.
|
| |
|
|
|
|
|
|
| |
* This may or may not help with the memory leak, need to assess
|
|
|
|
|
|
|
|
| |
* Push asset update (invoked when clothing is altered) down into AgentAssetTransactions from Scene.Inventory.cs to join others
* I've tested that clothing creation and update still works, but please let me know if it suddently breaks for you.
* Add/correct comments
|
| |
|
|
|
|
|
|
|
|
|
| |
align classes with file names
* Small amount of ndoc
* This will probably require a prebuild and nant clean
|
|
|
|
| |
of the permission masks or Locked status using the available checkboxes.
|
| |
|
|
|
|
| |
other side now, assuming the region on the other side has physical prim enabled.
|
|
|
|
|
|
|
|
|
|
|
| |
** Sending the actual TimeDilation to the client now instead of the 62455 constant. The client is *supposed* to use that value to sync with the simulator. (actually sending ushort.maxvalue * TimeDilation)
** Disabling prim that inter-penetrate instead of just not attaching a joint
** Reduced prim spin a 'little' bit, but not *enough*
** Tweaked the TimeDilation algorithm to be closer to 1.0 by default and various changes to the sim stats reporter
** Created a .SetValues method to PhysicsVector so we can simply call the setvalues function instead of .x, .y, .z sets.
** Experimented with a .GetBytes Method on PhysicsActor to be able to use the LLVector3.FromBytes() method.
** Upped the Inter-penetration depth to 0.25 instead of .08.
|
|
|
|
| |
object in ODE and you delete the object. The script is still running and trying to add force, but the object reference is null.
|
|
|
|
|
| |
* In ODE, disposing of all of the ODE objects and the ODE World to reclaim memory when the simulator restarts.
|
|
|
|
| |
never occurs anyway
|
|
|
|
|
|
|
|
| |
appearance persistence needs to set new params for it (check OpenSim.ini.example)
New keys "appearance_persist" and "appearance_connection_string" are now under [Startup].
* Reorganized OpenSim.ini slightly
|
| |
|
|
|
|
|
| |
* temporary re-introduced the weird 'flip-back' behaviour, but debugging it; will remove it if I don't find anything.
|
|
|
|
|
|
| |
* Turned on see_into_this_sim_from_neighbor by default.
* Fix Race Condition with parts being added to a group while the simulator is starting up.
|
|
|
|
|
| |
* Backup your database just in case.
|
| |
|
|
|
|
| |
bitmasking error of the objectflags! It's still a little bit wonky when you check the checkbox, however it 'takes' and doesn't break anything.
|
|
|
|
|
|
|
|
| |
Added experimental console command to:
* unload module (note: module probably doesn't support it)
* load module
Not visible in help (needs testing first).
|
| |
|
|
|
|
|
|
|
|
| |
in ClientView
* Changed some uint constants to Enum values
|
|
|
|
| |
It might be smarter to just rebuild the whole objectflags before sending it to the client >.<
|
|
|
|
|
|
|
|
|
|
| |
** Created SendFullUpdateToAllClientsExcept(LLUUID) so that permission updates /appear/ to apply immediately
** Separated out the ObjectFlags and the Permission Flags. They're related but not the same
** Added a hack routine to add *back* the objectflags to the client flags because the client hates the way we're doing object permissions
** Updated the clientflags routine to properly tell the client when they can't edit admin objects (objects owned by the sim administrator) even when they're an estate manager(why? >.< argh!)
** Fixed a null sim administrator/estate manager/user from causing permissions to return false even when it should return true.
** Re-added ObjectModify hack to allow collaboration with the allow anyone to move checkbox until we get group permissions done.
|
| |
|
|
|
|
| |
be sent.
|
| |
|
|
|
|
|
|
| |
* Made Estate tools work for estate managers without needing to request admin status First
* Added code to make the Simulator version to be reported in the About box of the client
|
|
|
|
|
|
|
|
|
| |
memory
* However, I'm no longer sure they were even a big contributory factor (to this particular leak, there are other causes of other leaks). I need better measurement techniques
* Removed most of my debugging gawp
|
|
|
|
|
|
| |
Just want to try this out on windows quickly.
|
|
|
|
| |
light verbosity to this end
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Attached is a patch for Mantis 25 - this fixes the problem of not having a
touch_start happen for all prims in a linked group. So, with this, large
builds can now have a single script in the base prim and it will run when
any prim in the linked build is touched. The problem was that the
objectgrab event was not being propagated to all the prims in the group.
|