Who's disappointed? - Petition - Sign Here!

Octree leaf errors...

Mad crashes.......

Bug is submitted... please sign here (Torley - help us or let us help ourselves!)

My words are few........because I have no words left.

Tale in the Desert ....at least in runs flawlessly. Even if the plot is shite.

Must be my programming nature .. I would be more inclined to sign an NDA and contract then I would a petition. I like the *doing* of things more then the talking about it.

Are you still here........ I thought I told you. Back off, I am working with Torley and Don here to try and sort it behind the scenes. There are debug files and log files flying around faster than you can type one word on your crappy windows keyboard with those fat fingers of yours.

Quit winding me up. If anyone here is not doing anything about it, its gotta be you. 'SENIOR' my ass.

NOTE TO LINDENS: Excuse my language - specifix - provocation. (again).

Heh, maybe, though I somehow expect that would be a hasty statement. But what do I know, I am apparently not upset enough to have anything useful to say.

Btw .. incase you didn't notice, you can reduce the problem by lowering your allowed network bandwidth, as I pointed out a few ago. Also .. technically an octal tree shouldn't be code implemented in any platform specific way. i.e. the same code should technically work anywhere .. as I also pointed out. So I dobt it is the occlusion code itself that is the problem. .. But I digress with useless theories that only seem to reduce the problem and outline a possible cause. And I have already done this in other threads.

/263/bc/110764/2.html#post1068035
/263/bc/110764/1.html#post1067563
/263/bc/110764/1.html#post1067726
/263/bc/110764/1.html#post1067764

I know .. it isn't the same sort of thing people have been posting the last week, the lack of tons of logs and such. Trying to understand the problem without the source IS painfully difficult, but there are a variety of possibilities to be had that have as good a chance as being accurate as any other guess in the dark at this point.

Anyway, I could be wrong, but I still suspect it is a race between the object propigation coming in from the network, and the world rendering using the object data in the octree. Something, somewhere, attempting to modify a bit of it that the other half plucked away. Nice and vague that huh? It at least fits the behavior. At the same time .. it could just be a bunk pointer issue that exists on all clients and for some reason the differences in the VM's only makes the problem obvious on the Linux client .. improbable, though not impossible.

Anyone try the Windows client under wine with memory debugging turned on? I would, but wine and cedega don't get along with SL at all on my systems.

And if you wanna come visit me, I will be hanging out in Elpenor, standing around with all the buildings and stuff using a Linux client.

So like: ./secondlife -url secondlife://elpenor

From what is caught with a debugger there is a segfault returning with LLSpatialGroup::updateInGroup()

However octree errors occur at the same time as moving from one sim to another, octree never really ends anything itself. After a while I get an OCTREE IS STILL HUGE error scrolling by. Then the client continues. Lowering bandwidth seems to sort some of it out, also not moving from sim to sim. Looking from one sim to another seems to cause a problem too. Perhaps this bug is network related to transfering users from one to another. The client is just about constantly updating the network throttle as well. Perhaps something is stressing it. Network health appears to be healthy between myself and LL.

I experience similar behavior. I stay near the middle of a sim and I have basically little to no issues. One of the reasons I am using the -url option to the client to login, it always drops me near the center. Getting near the edge of a sim deffinately causes me issues, and yah, dropping down the allowed network bandwidth deffinately reduces the problem quite a bit. I ended up needing to fly around the sim to crash the last time. The UseOcclusion FALSE doesn't seem to do anything really, DebugPermissions TRUE seems to do more, and lowering network bandwidth more yet.

As a minor correction, this is little to no issues and regardless of object count in the area. Just an FYI.

I'm not particularly disappointed, I understand what an Alpha is.

We (the SL LUG) were working our way through the issue earlier.


I'm certainly not disappointed enought to start trolling about it

Being able to say sorry doesn't make it ok to be nasty.