Chat log from the meeting on 2015-03-17

From OpenSimulator

Jump to: navigation, search
[10:58]  Shez Oyen: Hey Neb
[10:58]  Freaky.Tech is Online
[10:58]  Nebadon Izumi is Online
[10:58]  Nebadon Izumi: hello
[10:58]  Connecting to in-world Voice Chat...
[10:58]  Connected
[10:58]  Marcus Llewellyn: Hey Neb.
[10:58]  BlueWall.Slade Hi
[10:58]  Connecting to in-world Voice Chat...
[10:58]  Connected
[10:58]  Alicia.Raven hi neb
[10:58]  Marcus Llewellyn: Hiua Hustin.
[10:59]  Marcus Llewellyn: Er, Hiya Justin
[10:59]  Nebadon Izumi: hello
[10:59]  Justin Clark-Casey: hello
[10:59]  Alicia.Raven hi justin
[10:59]  Marcus Llewellyn: Chiclet Keyboards. *sigh*
[11:00]  Richardus.Raymaker hi all
[11:00]  BlueWall.Slade Hello RiRa
[11:00]  Shez Oyen: Hey Rich :)
[11:02]  Nebadon Izumi: Looks like Moses guys are going to attempt to pull off integrating PhysX
[11:02]  Aine.Caoimhe wonders if I'll top last DFriday's "unable to fetch profile data at this time" error message count today
[11:02]  Alicia.Raven as its st patricks day i changed your flags to ireland neb :)
[11:02]  Nebadon Izumi: nice :)
[11:03]  Marcus Llewellyn: I'm mostly wondering if PhysX will work on Linux.
[11:03]  Nebadon Izumi: should
[11:03]  Nebadon Izumi: its going to require a wrapper no matter what
[11:03]  BlueWall.Slade is that some distributed thing, or does the server need a GPU?
[11:03]  Nebadon Izumi: nah it has a software version
[11:03]  Alicia.Raven i havnt had any profile popups today aine
[11:03]  Kayaker Magic: I think InWorldz is using PhysX, but I don't know what OS their servers is running.
[11:03]  Marcus Llewellyn: It can do CPU, GPU, or both. At least on Windows, where there's driver support for that. Linux, I don't know for sure if it can use the GPU.
[11:03]  Aine.Caoimhe I've had 19 so far
[11:04]  Aine.Caoimhe on Friday I got up to 130 before I crashed
[11:04]  Richardus.Raymaker i hope it's using hardware too. what what does that mean for the energy bi,ll :O
[11:04]  Aine.Caoimhe at Doro's party
[11:04]  Nebadon Izumi: Profile popups?
[11:04]  BlueWall.Slade I see them, dunno what that's about.
[11:04]  Aine.Caoimhe I'd hoped that clearing cache would solve it but apparently not
[11:04]  Richardus.Raymaker besdies. the server only have simple graphics cards onboard stuff
[11:04]  Nebadon Izumi: not expereinced that one
[11:04]  Nebadon Izumi: ya you would not need a GPu for PhysX
[11:04]  Aine.Caoimhe I get one for every agent that enters the region
[11:04]  Aine.Caoimhe and only in OSG
[11:05]  Nebadon Izumi: it would work the same as ODE or Bullet
[11:05]  Richardus.Raymaker But GPU would make it betetr :)
[11:05]  Alicia.Raven i was getting them for every person that entered lastweek, not a single one sofar today
[11:05]  Nebadon Izumi: if they can even make that work with opensim
[11:05]  Aine.Caoimhe you're lucky, Alicia
[11:05]  Billy.Bradshaw If you are from a grid using core profiles and visit a grid using the old one, then you get the profile warning
[11:05]  Aine.Caoimhe ah, so that's what it is
[11:05]  Nebadon Izumi: I think how they are going about it may not make it suitable for inclusion into opensim, we will have to see
[11:05]  Nebadon Izumi: i give it a 50/50 chance right now
[11:06]  Nebadon Izumi: I just wish they didnt want to really try to pull it of themselves
[11:06]  Richardus.Raymaker i think make it as module
[11:06]  Marcus Llewellyn: If they take Robert's route, and don't touch core, it shouldn't be too difficult to include.
[11:06]  Nebadon Izumi: yes lets hope
[11:06]  Richardus.Raymaker It's the only right one maarcus. best to maintain too
[11:07]  Marcus Llewellyn: The other thing, though, is PhysX's licensing. It may be free as in beer, but it's not libre free.
[11:07]  Nebadon Izumi: they also said they have no intention of keeping compatibility with SL stuff
[11:07]  Nebadon Izumi: so even if its a module
[11:07]  Nebadon Izumi: it might not be very useful for core
[11:07]  Nebadon Izumi: if nothing works like the other physics modules
[11:07]  BlueWall.Slade It's doubtful that they would maintain it too.
[11:07]  Richardus.Raymaker MeH, whats the use then for physy ?
[11:07]  Aine.Caoimhe it's also card-specific isn't it? it's anVidea-onbly thing?
[11:07]  Nebadon Izumi: no its not
[11:08]  Aine.Caoimhe oh...okies
[11:08]  Billy.Bradshaw I saw an update that addressses the script status when HG TP, for me it still fails
[11:08]  Nebadon Izumi: physx will run at the cpu level
[11:08]  Marcus Llewellyn: No. You can only use Nvidi a cards for GPU calculations, though, I think. But that isn't required.
[11:08]  Nebadon Izumi: not requiring any GPu at all
[11:08]  OtakuMegane Desu: Kinda that. I'm sure there are some who would have used for it but the vast majority I don't really see what use it'd be.
[11:08]  Richardus.Raymaker i think fixing HG friends etc. things like that is more important
[11:08]  BlueWall.Slade ok thatis good to know, I thought it needed a GPU too
[11:08]  Kayaker Magic: PhysX uses a GPU on the server side, if available?
[11:09]  OtakuMegane Desu: BulletSim is coming along nicely and certainly beats ODE, so why add another one?
[11:09]  Nebadon Izumi: only if its programmed to do so
[11:09]  BlueWall.Slade I guess they are trying to get the physics performance up so they can populate sims with active users
[11:09]  Nebadon Izumi: Bullet can use GPu too
[11:09]  OtakuMegane Desu: ^
[11:09]  Nebadon Izumi: but we have no support for it yet
[11:09]  Jak Daniels: I think it would only use the GPU in the client where physics is client based
[11:09]  Dahlia.Trimble hi
[11:09]  Nebadon Izumi: infact as of right now bulletsim cant even use more than a single cpu core
[11:09]  Nebadon Izumi: its single thread app
[11:09]  BlueWall.Slade Hi Dahlia
[11:09]  OtakuMegane Desu: Linux support for OpenCL is still kinda iffy, but improving.
[11:10]  Richardus.Raymaker Still wish opensim get its own viewer. there's now still enough time.
[11:10]  Nebadon Izumi: it may prove difficult to make the simulator physics use a gpu
[11:10]  Alicia.Raven hi dahlia
[11:10]  OtakuMegane Desu: It shouldn't be too hard.
[11:11]  BlueWall.Slade I doubt many datacenters have gpu in their boxes?
[11:11]  Nebadon Izumi: ya remember though everything passes through the simulator still
[11:11]  OtakuMegane Desu: At least I wouldn't think so. I'm not the expert on that though XD
[11:11]  Nebadon Izumi: which is C# and mono
[11:11]  Marcus Llewellyn: It's PhysX's license that worries me more than anything, really.
[11:11]  Richardus.Raymaker besides many datacenters are not have the power for the gpu cards.
[11:11]  Justin Clark-Casey: they're talking using the gpu server-side I expect
[11:11]  Nebadon Izumi: so the returns would be very limited i suspect
[11:11]  Nebadon Izumi: yes
[11:11]  OtakuMegane Desu: Bluewall: No, GPUs in servers are uncommon unless it's specialized stuff like bitcoin or scientific simulators
[11:11]  Dahlia.Trimble I thought Physx used Cuda, not OpenCL
[11:11]  Dahlia.Trimble could be wrong tho
[11:11]  Richardus.Raymaker im happy with bullet, and that can only get better
[11:12]  Nebadon Izumi: iI think PhysX is PhysX
[11:12]  Nebadon Izumi: its not cuda either
[11:12]  Dahlia.Trimble you sure?
[11:12]  Marcus Llewellyn: It would be interesting to hear why they feel it's better to implement PhysX rather than improve Bullet.
[11:12]  Nebadon Izumi: well if its using GPU it might be cuda
[11:12]  Nebadon Izumi: but if its only using the Cpu no
[11:12]  Nebadon Izumi: there would be no cuda
[11:12]  BlueWall.Slade who knows?
[11:12]  Kayaker Magic: Right now I don't use BulletSim because it doesn't work with bullets! (fails to fire collision events!)
[11:12]  OtakuMegane Desu: BulletSim 3 will be OpenCL so you can use anything that had a driver for it pretty much.
[11:12]  Dahlia.Trimble Physx is Nvidia and so is Cuda
[11:13]  OtakuMegane Desu: Bullet*
[11:13]  Nebadon Izumi: yes
[11:13]  Nebadon Izumi: but they are seperate things
[11:13]  Nebadon Izumi: PhysX may be able to take advantage of cuda
[11:13]  Nebadon Izumi: but its not cuda per say
[11:14]  Richardus.Raymaker your sure thats just not slow lsl and a busy simulator ?
[11:14]  Alicia.Raven it would be nice if bullets worked, i have a mantis open for it and i think there is another
[11:14]  Richardus.Raymaker
[11:14]  Marcus Llewellyn: Their statement that information would be shared at meetings on the MOSES grid is also kinda meh. If they seriously want core devs included in things, they won't make the mountain come to Muhammad.
[11:14]  BlueWall.Slade ++
[11:14]  Nebadon Izumi: ya the problem with moses grid
[11:14]  Dahlia.Trimble well I meant the GPU acceleration in Physx is implemented in Cuda
[11:14]  Nebadon Izumi: is its closed
[11:14]  Nebadon Izumi: its not open to anyone
[11:15]  Nebadon Izumi: which is kind of a big brick wall
[11:15]  Justin Clark-Casey: they produce a 'moses in a box' which I believe is the system running in a vm
[11:15]  Nebadon Izumi: im sure most core devs could get in
[11:15]  Richardus.Raymaker the SL viewer is already a brick thats big enough
[11:15]  Nebadon Izumi: but they arent going to open it to everyone
[11:15]  OtakuMegane Desu: Stack of bricks lol
[11:15]  Richardus.Raymaker lol
[11:15]  Dahlia.Trimble I used to have an account but they changed some policy and deleted it
[11:16]  Nebadon Izumi: yea it sounds like they want people to test it using MIAB and that if it ever gets into core
[11:16]  Nebadon Izumi: it will be totally on us to make their code work
[11:16]  OtakuMegane Desu: -_-
[11:16]  BlueWall.Slade It would be cool if they focused on making general improvements to core OpenSim
[11:16]  Nebadon Izumi: so I am not saying its not going to happen
[11:16]  Nebadon Izumi: but its not sounding ideal
[11:16]  Richardus.Raymaker MeH, dves have enough todo already
[11:16]  AlexRime SiLiSiLi is Offline
[11:16]  BlueWall.Slade Else, they are just diverging to a point where it doesn't resemble core anymore.
[11:16]  Nebadon Izumi: I think what happened to get this whole ball rolling
[11:16]  Dahlia.Trimble well getting into core usually implies some form of support
[11:16]  OtakuMegane Desu: Yeah. BulletSim is doing fine for the physics side. Good point of focus is the rest of core.
[11:16]  Marcus Llewellyn: ++ BlueWal
[11:17]  Dahlia.Trimble and meeting user expectations
[11:17]  Nebadon Izumi: was at FCVW InWorldz did a presentation that talked about their physics engine
[11:17]  Nebadon Izumi: and what i saw didnt seem much better than what ODE or Bulletsim offer
[11:17]  Nebadon Izumi: but i guess Doug got excited about it
[11:17]  BlueWall.Slade hha
[11:17]  Justin Clark-Casey: I thoguht that was havok though
[11:17]  OtakuMegane Desu: Functionally I dunno how much difference there really is among the major engines these days.
[11:17]  Nebadon Izumi: no
[11:17]  Marcus Llewellyn: That's what I wonder. Why not just improve Bullet? It has many of the same features. It's multi-threaded. It can do utilize a GPU.
[11:17]  BlueWall.Slade if people don't understand something - they jump into the most obscure corners to fix it.
[11:17]  Nebadon Izumi: InWorldz uses Nvidia PhysX
[11:18]  Nebadon Izumi: SL uses Havok
[11:18]  Dahlia.Trimble Ive used Physx with OpenSim content before, I wasn't all that impressed. It was also pretty easy to get it to crash
[11:18]  BlueWall.Slade Hi Robert Adams
[11:18]  Dahlia.Trimble but that was a few years ago
[11:18]  Richardus.Raymaker hi robert
[11:18]  Shez Oyen: Hey Robert :)
[11:18]  Marcus Llewellyn: Hi Robert :)
[11:18]  Dahlia.Trimble hi Robert
[11:18]  OtakuMegane Desu: Perfect timing :)
[11:18]  Nebadon Izumi: yea I unforutnately think its kind of a waste to start yet another physics engine
[11:18]  Robert Adams: hello all.
[11:18]  Justin Clark-Casey: hi robert
[11:18]  Nebadon Izumi: but I cant say with an certainty bullet is actually better than PhysX or Vice versa
[11:18]  Richardus.Raymaker Nebadon, its more something to think about after opensim 1.0
[11:18]  Dahlia.Trimble I dont necessarily think it's a waste
[11:18]  BlueWall.Slade Taxpayers foot the bill, so they have dep pockets :)
[11:19]  Nebadon Izumi: but what I have seen from InWorldz, and Havok in SL
[11:19]  Nebadon Izumi: its not a great deal better than what we can do with ODE and Bullet
[11:19]  Nebadon Izumi: sure its better
[11:19]  Nebadon Izumi: but it nots jaw droppingly better
[11:19]  Robert Adams: I haven't looked into phyx and such. I kept finding that the library binding were changing too quickly
[11:19]  Dahlia.Trimble better how?
[11:19]  AlexRime SiLiSiLi is Online
[11:19]  Nebadon Izumi: well I think SL is just maybe a bit more stable
[11:19]  Richardus.Raymaker Translated, its as good as the (script) programmer can use it
[11:19]  Nebadon Izumi: more complete
[11:19]  Dahlia.Trimble I see things we can do that fail in SL
[11:19]  Nebadon Izumi: there is some stuff in BulletSim that isnt complete
[11:19]  Robert Adams: But another physics engine is always a nice addition
[11:19]  Nebadon Izumi: same with ODE
[11:19]  OtakuMegane Desu: Opensim isn't going to use all the potential function any physics engine has.
[11:20]  Nebadon Izumi: ya that is also true however I think in SL case
[11:20]  Nebadon Izumi: they limit it greatly
[11:20]  Nebadon Izumi: for non technical reasons
[11:20]  BlueWall.Slade FWIW, I think Bulletsim is looking really good and that is where the effort should go for improving performance.
[11:20]  Marcus Llewellyn: ++
[11:20]  Dahlia.Trimble try making a physical torus chain in SL, it fails miserably
[11:20]  Nebadon Izumi: yea
[11:20]  OtakuMegane Desu: SL physics are eithe ra mess or a joke. Sometimes both depending how they "fix" it.
[11:20]  Nebadon Izumi: i would rather see MOSES grid push on Bulletsim
[11:20]  BlueWall.Slade thanks Robert Adams fro bring that to core :)
[11:20]  Nebadon Izumi: I don't know why they insist on going a different route always
[11:20]  Marcus Llewellyn: (borken record) They should just help improve Bullet.
[11:20]  Billy.Bradshaw ++
[11:21]  Nebadon Izumi: they have very different goals than core opensim
[11:21]  Justin Clark-Casey: moses will do what they want to do, there's no changing anybody;'s minds
[11:21]  Nebadon Izumi: not much we can do about that though
[11:21]  Aine.Caoimhe (broken record) what are the chances of the infinite inventory fetch bug getting tackled in 0.8.2?
[11:21]  Nebadon Izumi: indeed Justin
[11:21]  Richardus.Raymaker Bluewall, agree bullet is good. offcorse people still seems to have troubles with verhicles.
[11:21]  Nebadon Izumi: I think alot of the problems we see with BulletSim are not because Bullet isnt capable
[11:21]  Nebadon Izumi: its just because our implementation is not complete
[11:21]  BlueWall.Slade I would say if they want to push us a new physics engine w/o putting any effort into maintaining it or helping integrate it, then we should probably just pass it on.
[11:21]  Nebadon Izumi: and fully optimized
[11:22]  Robert Patric: hello everyone
[11:22]  Nebadon Izumi: has nothing to do with Bullet as an engine
[11:22]  Richardus.Raymaker Bluewall +1
[11:22]  Marcus Llewellyn: I suspect a main concern of theirs is performance with a large number of agents in a simulator. They've been chasing that for a very long timw now.
[11:22]  BlueWall.Slade if they publish patches, gthen users can do their own integration if they want.
[11:22]  OtakuMegane Desu: Definitely. BulletSim is still fairly young, really. Lots of potential and tweaking to be done I'm sure.
[11:22]  Justin Clark-Casey: hello robert
[11:22]  Nebadon Izumi: yea I think Moses may be interested in actually using GPu to accelerate physics
[11:22]  Robert Patric: any osgrid owner here? I need help
[11:22]  Justin Clark-Casey: performance is a tough issue and physics is not usually the bottleneck
[11:22]  Nebadon Izumi: but I think that while yes that might be possible
[11:22]  Robert Adams: And there's bullets 3 coming along that will have OpenCL acceleration
[11:22]  Nebadon Izumi: the C# simulator can only handle so much data
[11:23]  Jak Daniels: Justin, you fixed in git 30b786351ea46cf6640dc45fa84bb4a4da1ab6d8, but I think it broke the ability to fly up and down in third person mode :(
[11:23]  OtakuMegane Desu: Yeah. BUllet will support GPU anyway. And OpenCL, so no being locked to one brand hardware or something
[11:23]  Nebadon Izumi: your never going to be able to push a GPu to its full potential
[11:23]  Nebadon Izumi: and expect the simulator to stay alive
[11:23]  Dahlia.Trimble I am a bit nervous tha they will donate something that only implements what they want and then mantis fills up with a lot of bug reports for missing features
[11:23]  Nebadon Izumi: I dont think they intend on donating anything
[11:23]  Justin Clark-Casey: if spomething isn't supported by any core dev then it shouldn't beaccepted
[11:23]  Nebadon Izumi: it seems like the code will be available
[11:23]  Robert Adams: They can always use one of divas loadable modules to add the physics engine
[11:23]  Nebadon Izumi: and if we want it we can grab it
[11:23]  BlueWall.Slade Dahlia, that is why we should think carefully about whether we pull it in.
[11:23]  Marcus Llewellyn: ++jcc
[11:24]  Nebadon Izumi: i dont think they intend on providing patches to opensim directly
[11:24]  Nebadon Izumi: atleast that kind of what it sounds like
[11:24]  Dahlia.Trimble Bluewall, of course
[11:24]  Robert Adams: We can cross that bridge when we get to it
[11:24]  BlueWall.Slade yeah, they said it might take some smoothing to get it in.
[11:24]  Nebadon Izumi: yea
[11:24]  BlueWall.Slade we have a lot to smooth already, lol
[11:25]  Nebadon Izumi: unless what they do is phenominal id rather see us focus on improving bulletsim
[11:25]  Justin Clark-Casey: jak: ok, I reopened 6835, pls could yopui putthe details on there
[11:25]  Nebadon Izumi: from what i saw at InWorldz it doesnt seem a whole lot better
[11:25]  Jak Daniels: thanks jcc that what I was going to ask someone to do for me :)
[11:25]  Nebadon Izumi: I think the biggest advantage InWorldz has over us
[11:25]  Justin Clark-Casey: aine: I expect it will be fixed for 0.8.2
[11:25]  Nebadon Izumi: is the ability to drive a car over a border
[11:25]  Nebadon Izumi: other than that, it seems about the same
[11:26]  Nebadon Izumi: and technically that has nothing to do with physics
[11:26]  Aine.Caoimhe @JCC do you need more data or are the dumps you already have enough for now?
[11:26]  Richardus.Raymaker i think scripting speed is a bigger problem then physics.
[11:26]  Justin Clark-Casey: aine: honestly, I had to stop reading the mantis - there was far too much going on and I had to concentrate on manyh other things. I will need to catch up
[11:26]  Marcus Llewellyn: I"ve sort of had the impression for a long time that the main problem with our physics isn't the engine, per se, but manpower. When you have, maybe two devs who understand that part of the code well, you can't expect them to implement everything, or certainly not quickly.
[11:26]  Neovo Geesink: Does Mono not keep up with faster scripting?
[11:27]  Aine.Caoimhe okies...well if you need more let me know...I can reporoduce it at well
[11:27]  Dahlia.Trimble I was surprised that my test follow gizmo for llLookAt() could follow me across borders, even if the sims ran different physics engines
[11:27]  OtakuMegane Desu: Physics is not an easy thing to work with. :)
[11:27]  Nebadon Izumi: yea, I think moses will run into similar problems
[11:27]  Kayaker Magic: Yeah, I wish I could script faster! :)
[11:27]  Nebadon Izumi: I think they only have 1 or 2 people working on it
[11:27]  Nebadon Izumi: and a lot less testers than opensim does
[11:27]  Justin Clark-Casey: marcus; integrating a physics engine is a very complex task
[11:27]  Marcus Llewellyn nods.
[11:28]  BlueWall.Slade no, lol
[11:28]  Richardus.Raymaker Noo. dahlia. then opensim get less time
[11:28]  BlueWall.Slade then they would screw us
[11:28]  Nebadon Izumi: hehe
[11:28]  Dahlia.Trimble lol
[11:29]  BlueWall.Slade they hired someone to fit their requisition, I doubt OpenSim experience had vey much to do with it.
[11:29]  Kayaker Magic: Justin, you were looking at co-routines for the llSleep problem. Any progress on that?
[11:30]  Nebadon Izumi: it will be interesting to see what happens, I think we are going to have to sit back and wait a while before we can see anything
[11:30]  Dahlia.Trimble <3 coroutines
[11:30]  BlueWall.Slade ++
[11:30]  Justin Clark-Casey: kayaker: I have not had an opportunity, though I think it's possible to use them to capture script state and wtich between sleeping scipts
[11:30]  BlueWall.Slade would we need to bump the tools version to 4.5 to do that?
[11:30]  Nebadon Izumi: I am not interested in setting up Moses in a box to test it though
[11:30]  Justin Clark-Casey: not trivial though
[11:30]  Aine.Caoimhe also on my wish list: when declining a friend offer in a doreign grid is there some way it could pretty please stop spamming with a freind request for every single visit I ever make again to that grid
[11:31]  Dahlia.Trimble I thought the MRM stuff had some coroutine stuff in it? or did at one time
[11:31]  OtakuMegane Desu: But...friendship! D:
[11:31]  Marcus Llewellyn: ++Aine. That's kinda annoying. Heh
[11:31]  Justin Clark-Casey: co-routines only because possible in c# 4 I think
[11:31]  Justin Clark-Casey: and that's using a hack of sorts
[11:31]  Aine.Caoimhe it also happens if you crash before having time to accept it
[11:31]  Dahlia.Trimble no there is a trick to do coroutines in earlier c# with iterators
[11:32]  Aine.Caoimhe (or the region you're in crashes)
[11:32]  BlueWall.Slade 4.5 has the async functions
[11:32]  Aine.Caoimhe and can we pretty please join more than 5 groups
[11:32]  Justin Clark-Casey: jak: so this is changing fly look in third-person? I tested mouselook at that appeared to ascend and descned
[11:32]  BlueWall.Slade but, the old mono keeps us off that.
[11:32]  Justin Clark-Casey: async is just syntatic sugar
[11:32]  Robert Adams: If one was going to add a new scripting language to open simulator, what would be the right language that would sand box nicely
[11:32]  Marcus Llewellyn: I don't think our group implementation limits the number of groups you can join?
[11:33]  Justin Clark-Casey: robert: perhaps lua?
[11:33]  Jak Daniels: yes my height remains constant now for fly look
[11:33]  Dahlia.Trimble unity has had coroutines for a long time, even when it used mono 2
[11:33]  Aine.Caoimhe it does when you're HGed and using core groups
[11:33]  Justin Clark-Casey: jak: in first person?
[11:33]  Aine.Caoimhe except I think the group gets added but you get a message saying you can't
[11:33]  Marcus Llewellyn: How odd. I didn't know that. I seem to remeber diva's distro has or had that sort of limit, but not the core implementation.
[11:33]  Aine.Caoimhe it's there....try joining 6+ groups in grids other than your own
[11:34]  BlueWall.Slade that limit is probably pretty easy to up - but may be in there for a purpose
[11:34]  Marcus Llewellyn: Good to know, Aine.
[11:34]  Aine.Caoimhe I reported it a year or so ago
[11:34]  Jak Daniels: sorry not sure if first person or third person lol. When you grab your av using left mouse button and mouse pointer
[11:34]  BlueWall.Slade things might get hairy with more that that.
[11:34]  Justin Clark-Casey: jak: so not in mouselook?
[11:34]  Jak Daniels: no
[11:35]  Justin Clark-Casey: ok, mouselook wiould be 1st person but av grab 3 rd person
[11:35]  BlueWall.Slade I noticed that too Jak
[11:35]  Dahlia.Trimble I bet you could kill a sim pretty fast with too many HG friends and groups, what wil how HTTP request seem to shut everything else down
[11:35]  Justin Clark-Casey: I love that all these regression surface immediately after a release
[11:35]  Jak Daniels: ok 3rd person is broken in fly mode
[11:35]  Nebadon Izumi: heh
[11:35]  BlueWall.Slade flying with the mouse and using the was, etc. keys
[11:35]  Aine.Caoimhe hunh?
[11:36]  Richardus.Raymaker uhh ?
[11:36]  Nebadon Izumi: well testing will always increase immediately after a release
[11:36]  Richardus.Raymaker that would be bad.
[11:36]  Justin Clark-Casey: then people will starting waiting for the next point release, etc.
[11:36]  Aine.Caoimhe well, how often would you like me to bump old reports?
[11:36]  Jak Daniels: i only updated to git master recently so only noticed it the other day
[11:36]  Marcus Llewellyn: Erk. I hadn't noticed that. But yeah... when flying and steering with the held mouse pointer, I can't go up or down.
[11:36]  Jak Daniels: yes
[11:37]  BlueWall.Slade that is the way SL behaves
[11:37]  Jak Daniels: really?
[11:37]  BlueWall.Slade yes
[11:37]  Richardus.Raymaker i always use page up/down thats still working ?
[11:37]  BlueWall.Slade but I like ours better
[11:37]  Dahlia.Trimble mouselook steer while flying?
[11:37]  BlueWall.Slade yes Dahlia
[11:37]  Jak Daniels: hmmm, well i'd kinda got used to being able to steer around in fly mode
[11:37]  Jak Daniels: :)
[11:37]  BlueWall.Slade control altitude with the mouse
[11:37]  Marcus Llewellyn: I seem to remember it working. But I haven't been to SL in ages.
[11:37]  Jak Daniels: not mouselook
[11:38]  Richardus.Raymaker hmm, steering with the mouse in SL in 3rd person ?
[11:38]  Shez Oyen: I thought SL fixed that
[11:38]  OtakuMegane Desu: Page up/down seems fine. Never gotten switched away from that.
[11:38]  Dahlia.Trimble oh I only thought it worked in mouselook
[11:38]  BlueWall.Slade Shez, maybe so - I only get there about twice a year.
[11:38]  Jak Daniels: grab your av with the mouse pointer (left button hold)
[11:38]  Richardus.Raymaker need toi test the mouse fly/steering in sl
[11:38]  Jak Daniels: then fly....
[11:38]  Jak Daniels: left right works with mouse
[11:38]  Jak Daniels: up down used to but now doesn't
[11:38]  Justin Clark-Casey: it may be that it'simpossible to fix both
[11:39]  OtakuMegane Desu: Hmm
[11:39]  Marcus Llewellyn: If SL doesn't let you go up or down steering with the mouse, I kinda prefer the way we've always had it. :)
[11:39]  Richardus.Raymaker sprry but i cannot steer in SL at all with the mouse in 3rd person
[11:39]  Justin Clark-Casey: both correcting av walk and fly due to the fact that the viewer gives no indication of av steering with lmb held down
[11:39]  Jak Daniels: it used to be possible to dive into the sea :)
[11:39]  Richardus.Raymaker you use always up/down
[11:39]  Justin Clark-Casey: which is why it might work that was on ll grid if it does
[11:39]  Shez Oyen: yoes you can fly guided by mouse (up and down) on SL
[11:39]  Justin Clark-Casey: was/say
[11:40]  Justin Clark-Casey: way even
[11:40]  Shez Oyen: Just checked it
[11:40]  Dahlia.Trimble seems to work in mouselook
[11:40]  Richardus.Raymaker i use much mouselook that works fine
[11:40]  Richardus.Raymaker But its normal that it dont work in 3rd person
[11:40]  BlueWall.Slade yeah, this is 3rd person view.
[11:40]  Jak Daniels: yes mouselook works, 3rd person used to too
[11:40]  Richardus.Raymaker No Jak. its not working in SL, just tried it
[11:40]  Jak Daniels: but I can see how that is related to the walk look up/down bug
[11:42]  Jak Daniels: ok, well, would be nice if it was still possible to do that, we've had that ability for such a long time. But no worries if it's incompatible with the other walking bug
[11:42]  Marcus Llewellyn: It wouldn't suprise me if SL used to let you, and they broke it. They've broken mouse steering with help LMB before.
[11:42]  Marcus Llewellyn: held*
[11:43]  Justin Clark-Casey: it might be quite a simply change, I would have to try
[11:43]  Justin Clark-Casey: looking at 30b786 I could preserve the X and Y paramters if the av is flying, not just if it's in flying and in mouselook
[11:43]  Jak Daniels: ++
[11:44]  Jak Daniels: yes it was only walking that was the problem
[11:46]  Justin Clark-Casey: I didn't pickupon this in the release, but I find the [Const] parameters at the top of OpenSim.ini.example to be confusing
[11:46]  Justin Clark-Casey: particularyl if you're running a standalone where they dont' apply expect in unusual cases
[11:46]  Justin Clark-Casey: except
[11:46]  Nebadon Izumi: yea that was the only time i got confused
[11:47]  Nebadon Izumi: was setting up a standalone
[11:47]  Marcus Llewellyn: I choose not to use it myself. Environment variable support I loooove. But the Const option isn't as useful to me.
[11:47]  Nebadon Izumi: so basically i just set both variables the same
[11:47]  Richardus.Raymaker Is there a way to keep the running scripts outside of the bin folder ?
[11:47]  BlueWall.Slade I will try to get some time to go thorugh those and see if I can simplify them
[11:47]  Nebadon Izumi: yes Richardus
[11:47]  Nebadon Izumi: you mean for Scriptengine?
[11:47]  BlueWall.Slade RiRa, you can keep everything outside
[11:47]  Richardus.Raymaker script states if am right
[11:47]  Nebadon Izumi: its at the very bottom of the [Xengine] section
[11:48]  BlueWall.Slade That is one thing I tought about adding to the [Const] section is a BasePath
[11:48]  Simulator Version v0.5 shouts: OpenSim Dev        5b31bb9: 2015-03-16 23:48:16 +0000 (Unix/Mono)
[11:48]  BlueWall.Slade set it to "." as the default.
[11:48]  Richardus.Raymaker Ok. ill check that when i need to configure it. i copied my rc2 configs in for now
[11:48]  Nebadon Izumi: ScriptEnginesPath = "ScriptEngines"
[11:48]  Richardus.Raymaker cool.
[11:48]  Justin Clark-Casey: putting everything in bin is insane
[11:48]  Marcus Llewellyn: YEah. Thats in the [XEngine] section.
[11:49]  BlueWall.Slade I wish we had our libraries outside
[11:49]  BlueWall.Slade just create the bin when we build.
[11:49]  Richardus.Raymaker You mean config outside of the bin bluweewall ?
[11:49]  Justin Clark-Casey: yeah, libraries and caches, other data, etc.
[11:49]  Richardus.Raymaker cache you can set outside the bin
[11:50]  BlueWall.Slade it bites us if we don't clear it manyally at times.
[11:50]  Marcus Llewellyn: I think pretty much all the subdirectory paths can be changed.
[11:50]  BlueWall.Slade they can
[11:50]  Nebadon Izumi: yea that is one reason I wrote the update script for osgrid to pull fresh everytime
[11:50]  BlueWall.Slade even the inventory and assets
[11:50]  Marcus Llewellyn: log file too, although that's trickier.
[11:50]  BlueWall.Slade hmm, you can "archive" it
[11:50]  Dahlia.Trimble not sure if j2kdecodecache can move
[11:51]  Richardus.Raymaker thirst thing on my list is own grid with FAssets
[11:51]  BlueWall.Slade it can Dahlia
[11:51]  Marcus Llewellyn: There' an option for j2kdecodecache somethwre in config-include, I think
[11:51]  Dahlia.Trimble have you tested it?
[11:51]  BlueWall.Slade you can run 100% in R/O directory
[11:51]  BlueWall.Slade yesmam
[11:51]  Jak Daniels: any news on when FSassets might come into core?
[11:52]  Richardus.Raymaker sofar i know it is in core jak
[11:52]  BlueWall.Slade I thinkI added options for most of those.
[11:52]  Jak Daniels: Mel's new asset service?
[11:52]  Marcus Llewellyn: I haven't bothered trying to move it myself, Dahlia.
[11:52]  BlueWall.Slade nope, that is XAssets
[11:52]  BlueWall.Slade FAssets is not there yet
[11:52]  Richardus.Raymaker oh.. now im lost.. :O
[11:52]  Nebadon Izumi: ya I need to follow up with Melanie about fsassets
[11:52]  Nebadon Izumi: she was still cleaning it up
[11:52]  Richardus.Raymaker aww.
[11:52]  Nebadon Izumi: and we have been tweaking things as they came up on osgrid
[11:52]  Marcus Llewellyn: Hehe
[11:52]  Nebadon Izumi: its been a while since we had an issue
[11:53]  Nebadon Izumi: so that is good
[11:53]  Richardus.Raymaker so back to mysql then ?
[11:53]  Jak Daniels: how has it been holding up?
[11:53]  BlueWall.Slade it's good to let it cook a while.
[11:53]  Dahlia.Trimble j2kdecodecache is specified in several places in the code and it may just fail to find anything cached rather than show some feedback on the log
[11:53]  Nebadon Izumi: the last week has been pretty good
[11:53]  Marcus Llewellyn: OSgrid's asset dump is prolly one heck of a test of that code. lol
[11:53]  Nebadon Izumi: hehe ya
[11:54]  Nebadon Izumi: i'll follow up with her soon and try to urge her to get it into core asap
[11:54]  BlueWall.Slade Dahlia, I should search for it and make sure it uses the assigned path?
[11:54]  Jak Daniels: cool, thanks neb
[11:54]  Dahlia.Trimble Bluewall, perhaps
[11:54]  BlueWall.Slade What would happen if it cannot wright to it's cache?
[11:54]  Dahlia.Trimble not sure
[11:54]  BlueWall.Slade because I run as user opensim in a dire owned by root:root
[11:55]  BlueWall.Slade have not seen issues with it.
[11:55]  Dahlia.Trimble look and see if it contains anything
[11:55]  BlueWall.Slade all my run-time data lives in the /var
[11:55]  BlueWall.Slade it woldn't have perms
[11:55]  Dahlia.Trimble I just have a separate bin for each instance :)
[11:55]  BlueWall.Slade opensim couldn't not write to it
[11:55]  Dahlia.Trimble and I like it that way
[11:56]  BlueWall.Slade that works too.
[11:56]  Dahlia.Trimble its so flexible that way
[11:56]  BlueWall.Slade but, it is possible to control all the things associated withit to make it in a R/O place.
[11:56]  BlueWall.Slade then all the associated data can live across upgrades.
[11:56]  Dahlia.Trimble I dont need it to be RO
[11:57]  BlueWall.Slade ++
[11:57]  BlueWall.Slade I have seen issues with artifacts in the directory when you run multiple instances out of it thoug.
[11:57]  Justin Clark-Casey: jak: ok, I just pushed 310e44f. That restores ascend/descend with lmb held on avatar for me
[11:57]  BlueWall.Slade thanks jcc
[11:57]  Marcus Llewellyn: My sim's OpenSim.ini, Regions.ini, and log files are all outside bin. Makes for smoother upgrades.
[11:58]  Jak Daniels: thanks Justin. ++ I'll try it after the meeting
[11:58]  Justin Clark-Casey: any sane sceheme would have all that stuff outside bin
[11:58]  Justin Clark-Casey: but unfortunately it didn't start out that way
[11:58]  BlueWall.Slade maybe I'll try to do that one day.
[11:58]  Richardus.Raymaker opensim.ini seems the most hard one. unless you passed parameter with starting opensim.ini ?
[11:58]  BlueWall.Slade (I won't push it though :)
[11:59]  Marcus Llewellyn: Sharing the asset cache among instances makes sense in my case. And ScriptEngines segregates by region UUID anyway.
[11:59]  Dahlia.Trimble I dont have a problem with it all in bin. It's not something that is traditional but I dont see any benefit to tradition in that matter
[11:59]  Dahlia.Trimble Im not a firm believer of tradition for tradition's sake
[11:59]  Justin Clark-Casey: sharing the asset cache is problematic when two processes try to manipulate it at once
[11:59]  BlueWall.Slade I think the biggest advantage would be that every build would be a clean directory.
[11:59]  Dahlia.Trimble on windows, yes
[11:59]  Justin Clark-Casey: then they suffer sharing violations, though I believe even current code suffers this
[12:00]  Marcus Llewellyn: Yeah, I worried about that at first, Justin. But so far, I've gotten away with it without a hiccup. :)
[12:00]  Nebadon Izumi: those sharing violations happen on linux too
[12:00]  Nebadon Izumi: a lot infact
[12:00]  BlueWall.Slade I keep all of them separate
[12:00]  Dahlia.Trimble Ive never seen it on linux and I have many instances sharing a single cache
[12:00]  Justin Clark-Casey: it may be safe to supprress all such problems, but then you may end up hiding a genuine issue
[12:00]  Nebadon Izumi: at OSCC i had all the Keynotes share same cache
[12:00]  BlueWall.Slade Justin - a while back we were talking about the HG avatar stuff going to the asset server ...
[12:01]  Marcus Llewellyn: I wouldn't want them supressed. All I can really say is that OpenSim has never once thrown an exception over it.
[12:01]  BlueWall.Slade Do you remember that?
[12:01]  Justin Clark-Casey: bluewall: I don't
[12:01]  BlueWall.Slade Well, you know what I'm talking about?
[12:01]  Marcus Llewellyn: The only time I can remember seeing exceptions related to the asset cache cam from people running into an issue with their antivirus on Windows.
[12:01]  Justin Clark-Casey: bluewall: not exactly
[12:02]  Dahlia.Trimble Ive seen a single instance complain about sharing violations on windows
[12:02]  BlueWall.Slade I think we were discussing HG avatar issues and we said that the assets end up on the asset server - for the HG avatar/
[12:02]  Justin Clark-Casey: bluewall: ah right, yes
[12:02]  Dahlia.Trimble perhaps flotsam could be modified to not attempt exclusive access
[12:02]  BlueWall.Slade in cases like that, would it make sense to mark the assets as temporary?
[12:03]  Richardus.Raymaker That sounds a good idea bluewall. so it get deleted when avatar leaves grid ?
[12:03]  BlueWall.Slade after a sset time, yes.
[12:03]  BlueWall.Slade If it makes sense.
[12:03]  Justin Clark-Casey: bluewall: I don't think you can
[12:03]  Richardus.Raymaker i would say it make sense. it not b loat the grid assets
[12:04]  BlueWall.Slade ok
[12:04]  Justin Clark-Casey: bluewall: well, I guess if someone copied an attachment you could then mark tmpeorary as permanent instead
[12:04]  Richardus.Raymaker if im wronmg, you can now yell
[12:04]  BlueWall.Slade would be great to be able to mark anything that could be removed later.
[12:05]  Justin Clark-Casey: bluewall: it might be possible but you would have to deal with al the cases where a temporary asset would become permanent
[12:05]  BlueWall.Slade I see.
[12:05]  Marcus Llewellyn: The idea of ancient unused notecards in assets gives me the willies sometimes.
[12:05]  Dahlia.Trimble lol
[12:05]  Alicia.Raven me too, and scripts saved for ever change
[12:05]  Justin Clark-Casey: such as if something that was an attachment is rezzed in a region, etc.
[12:06]  Dahlia.Trimble old love leters? ;)
[12:06]  vegaslon plutonian: little libraries of congress
[12:06]  Richardus.Raymaker scripts is something that bugs me. save, urrgh another assets . correct , save. yuck
[12:06]  BlueWall.Slade If the user adjusts the parts by rewearing, it updates them, I guess?
[12:06]  Marcus Llewellyn: I know they're nothing individually, but in aggregate... I can easily create a lot of assets that never see the light of day when scripting something.
[12:06]  Dahlia.Trimble maybe for scripts it could be leveraged into some kind of version control
[12:07]  Justin Clark-Casey: marcus; it is aproblem. But unfortunately the whole system is designed for immutable assets
[12:07]  Justin Clark-Casey: it's hard to change that
[12:07]  Marcus Llewellyn nods.
[12:07]  Marcus Llewellyn: I understand. If wishes were fishes. ;)
[12:07]  BlueWall.Slade As it stands now, do we have a real issue with the 64 byte -vs- 128 byte assets?
[12:07]  Justin Clark-Casey: and it was done that way for good reasons - immutable stuff can easily be cached, etc.
[12:07]  Justin Clark-Casey: uni
[12:07]  Justin Clark-Casey: imo
[12:07]  Dahlia.Trimble SL has (had?) an asset GC system
[12:08]  Marcus Llewellyn: They have a cold storage system, I think, at least.
[12:08]  Justin Clark-Casey: sl can in theory work out if an asset is referecned at all
[12:08]  BlueWall.Slade Mic Bowman said he has one too
[12:08]  Nebadon Izumi: yea it sort of requires a grid where the grid operators run all the regions to do that
[12:08]  Nebadon Izumi: its still pretty intensive
[12:08]  Dahlia.Trimble ya they do a deep search for anything that resembles a UUID
[12:08]  Justin Clark-Casey: if you ignore the fact that someone could one day reference a long dead asset uuid in a new script, which I should think probably never happens
[12:09]  BlueWall.Slade texture uuid?
[12:09]  Justin Clark-Casey: for instance
[12:09]  Justin Clark-Casey: in a new script - you could always read through every old script checking for uuids
[12:09]  Justin Clark-Casey: which is what the uuid gatherer does
[12:09]  Dahlia.Trimble ya but theirs is grid wide
[12:10]  Dahlia.Trimble anyway anything that deletes assets makes me nervous
[12:10]  Justin Clark-Casey: the worst issue isprobably serialized objects
[12:10]  Richardus.Raymaker So, devs make you nervous dahlia :))
[12:10]  Jim Jackson: Hi Posh
[12:10]  Justin Clark-Casey: scripts and notecards are tiny in comparison to an object with many prims
[12:10]  Alicia.Raven instead of delete, it could be moved to another table, then cleaned out after a period of time
[12:10]  posh prim: hello
[12:11]  Dahlia.Trimble Some do, Richardus :P
[12:11]  Marcus Llewellyn: It would be nice if there was some sort of separate class of mutable asset, I guess.
[12:11]  Nebadon Izumi: the biggest asset I ever saw was around 45mb
[12:11]  Justin Clark-Casey: there was a proposal to datestamp assets every n accesses, which is what xassets does
[12:11]  posh prim: can you see me jim
[12:11]  Jim Jackson: no, you are a cloud
[12:11]  posh prim: ok
[12:11]  Justin Clark-Casey: then an asset that hasn't been accessed for, say 3 years youmight be able to safely delete
[12:11]  posh prim: thyy
[12:11]  Justin Clark-Casey: but even that seems risky
[12:12]  Richardus.Raymaker whats different on xassets compared to mysql database ?
[12:12]  Justin Clark-Casey: xassets uses mysql but performance deduplication
[12:12]  Justin Clark-Casey: so identical assets are only stored once
[12:12]  BlueWall.Slade in all the 64 byte -vs- 128 byte description discussing, the biggest thing was HG landmarks.
[12:12]  Alicia.Raven even after 3 years it could be in some inventory thats not been used, a periodic uuid gather would be better i think
[12:12]  Dahlia.Trimble every n accesses? where is it counted?
[12:12]  Justin Clark-Casey: performs
[12:12]  Richardus.Raymaker justin, but only assets that are not stored in uer inventory's
[12:13]  Dahlia.Trimble and if it's counted, why not just stamp every access instead?
[12:13]  Justin Clark-Casey: dahlia: ah good point, it's it the asset is accessed after n days
[12:13]  Justin Clark-Casey: dahlia: my mistake
[12:13]  BlueWall.Slade I think 128 bytes is a small space for that application.
[12:13]  Marcus Llewellyn: A last access timestamp makes sense for cold storage, but not for removal.
[12:13]  BlueWall.Slade But, I guess anything larger would require viewer mods.
[12:13]  Justin Clark-Casey: alica: right, you can gather if tyou control all simulators
[12:13]  Dahlia.Trimble some assets are quite tiny, like very small sculpt maps or some materials
[12:14]  BlueWall.Slade ++
[12:14]  Justin Clark-Casey: marcus: if you timestamp if an asset has been accessed but last access time was > 90 days
[12:14]  Justin Clark-Casey: then you know an asset with a last access time of 3 years hasn't been accessed for 3 years
[12:14]  Richardus.Raymaker if you coimpile opensim 0.8.1 the search module etc. are corretc too right ?
[12:14]  Alicia.Raven yes, i think most people do control all sims, open grids are the exception and there are less of them
[12:14]  Justin Clark-Casey: but it's still possible it's in some long=lived cache on the only simulator where it is used
[12:15]  Marcus Llewellyn: That's sort of what I meant by my probably imprecise use of the term cold storage.
[12:15]  Dahlia.Trimble I have persistent shared cache so it's quite unlikely my asset server gets multiple accesses for any asset
[12:15]  Jak Daniels: rira you need the modded ossearch/profile for osgrid with 0.8.1 and above
[12:15]  Alicia.Raven bluewall, the truncation warning is only for an asset description for landmarks, that data isnt even used so the warning doesnt matter, just makes people worry they lost information
[12:15]  BlueWall.Slade yes
[12:15]  Richardus.Raymaker i dont use it on osgrid. it would be for own grid. i just saw message on mailing list that makes the question. uhmm but is there not a V2 search buildin ?
[12:16]  Dahlia.Trimble in fact I doubt my asset server gets *any* read accesses
[12:16]  Dahlia.Trimble its more just a backup
[12:16]  Richardus.Raymaker anyway, i doint think i need search :O\
[12:17]  Justin Clark-Casey: ok, I need to get going. Goodbye folks
[12:17]  Alicia.Raven does the asset service even need to warn about a truncation? i understand big grids dont want to alter the assets table to remove that dead data because of downtime, why not just stop printing the warning as the fields arnt used?
[12:17]  Marcus Llewellyn: B'bye, Justin. :)
[12:17]  Richardus.Raymaker bye justin
[12:17]  BlueWall.Slade Well, good meeting today - thanks. I need to take off. Will see you all next week.,
[12:17]  Alicia.Raven bye justin
[12:17]  Nebadon Izumi: see you Justin
[12:17]  Jak Daniels: bye justin... Thanks
[12:17]  Dahlia.Trimble cya jcc
[12:17]  Marcus Llewellyn: Cya BlueWall.
Personal tools
About This Wiki