Simulator User Group/Transcripts/2011.09.09
| Prev 2011.09.06 | Next 2011.MM.DD |
List of Speakers
| Al Supercharge | Andrew Linden | Arawn Spitteler |
| Ardy Lay | Ashiri Sands | Draconis Neurocam |
| Flip Idlemind | Leonel Iceghost | Liisa Runo |
| Methias Kira | Montecore Babcock | Neit Tackleberry |
| Opensource Obscure | Rex Cronon | Sahkolihaa Contepomi |
| Script Counter | Sniper Siemens | Squirrel Wood |
| Techwolf Lupindo | Jionk | PogoJump |
Transcript
[16:00] Sahkolihaa Contepomi: Hey Andrew.
[16:01] Andrew Linden: hello
[16:01] Methias Kira: Lindens falling from the sky! Quick, get your umbrella.
[16:01] Ardy Lay: God Mode 200 meter drop?
[16:01] Montecore Babcock: hello Andrew
[16:02] Andrew Linden: Simon is out today.
[16:02] Andrew Linden: Let's see... what I've been working on this week...
[16:03] Andrew Linden: I've been wrapping up my "faster scripts" project (DRTSIM-96 on aditi).
[16:03] Draconis Neurocam: what all does that involve out of curiousity?
[16:03] Andrew Linden: I actually reviewed it with some of the other devs and they made some suggestions on how to make it better.
[16:04] Andrew Linden: Mostly about syntax changes... use static_cast<derived_class*> instead of (void*) for some pointers.
[16:04] Draconis Neurocam: oooo, still nice
[16:04] Rex Cronon: hi everybody
[16:04] Flip Idlemind: Oooh what are we talking about
[16:04] Flip Idlemind: Slash me lost track of time
[16:05] Andrew Linden: Meanwhile, I've been fixing a few bugs that our QA team has found in the mesh maint-server project (DRTSIM-85 on aditi)
[16:05] Draconis Neurocam: andrew's faster scripts project flip
[16:05] Andrew Linden: I'm listing what I've been doing this week
[16:05] Andrew Linden: really it is just my preamble to justify my concluding statement ("nothing very interesting to report")
[16:06] Andrew Linden: Finally, we had a little automation/unit-test hackathon in the company this Thurs and Friday, and I've been participating in that.
[16:06] Andrew Linden: The effort is all over our codebase (viewer, server, web, tools, QA)
[16:07] Andrew Linden: but I've been working on unit tests in the server code
[16:07] Rex Cronon: if u want something interesting to look at, u could try the recent ways to crash others. i just experienced that at rausch:)
[16:07] Andrew Linden: not very interesting work, and it always takes longer than I expect to write unit tests
[16:08] Andrew Linden: Ok that is all. See? Nothing interesting to report.
[16:08] Andrew Linden: I did not get around to benchmarking the "faster scripts" project.
[16:08] Rex Cronon: isn't there software that generates unit tests automagicallly:)
[16:08] Squirrel Wood: faster terraforming? :p
[16:08] Andrew Linden: Nor did I get around to what I was talking about on Tues: ways to try to fix SVC-5927
[16:08] JIRA-helper: http://jira.secondlife.com/browse/SVC-5927
[#SVC-5927] Temp on Rezzed objects get queued
[16:09] Montecore Babcock: I have a question when you open up the Q and A section
[16:09] Andrew Linden: I haven't heard of that software Rex.
[16:10] Andrew Linden: I'm willing to believe someone has soemthing that works for code... as you write it.
[16:10] Rex Cronon: i think for java there i somethng that does that
[16:10] Rex Cronon: there is something*
[16:10] Andrew Linden: But software that cleans up un-tested spagetti code and writes unit tests for it... that is a ways off.
[16:10] Ardy Lay: Andrew, SCR-133 is getting kinda old. Some of us worry we will see limits before the metrics display is fixed.
[16:10] JIRA-helper: http://jira.secondlife.com/browse/SCR-133
[#SCR-133] Script Info reporting on discontiguous parcels is not working as expected.
[16:11] Draconis Neurocam: haha, andrew
[16:12] Ashiri Sands: oh, that's interesting
[16:12] Jionk: what?
[16:12] Andrew Linden: Ardy, is that a bug about something that worked once and is broken? or something that never worked right?
[16:13] Ashiri Sands: Sometimes i get odd results with trying to get prim counts on separated parcels
[16:13] Ardy Lay: I have never seen it work right on my land. I have seen it work right elsewhere at the same time.
[16:13] Liisa Runo: (looks like viewer issue to me)
[16:13] Andrew Linden: Ah ok, I think I get it now... I had to read it more carefully.
[16:14] Ardy Lay: VWR-23393
[16:14] JIRA-helper: http://jira.secondlife.com/browse/VWR-23393
[#VWR-23393] Viewer reports script info for the wrong parcel, in some cases
[16:14] Andrew Linden: I think I know what is going on there... and I would guess it is a server-side problem.
[16:14] Ardy Lay: It's been batted back and forth for a while now.
Total Avatar Scripts - 662
Highest - Psi Merlin: 115 (17%)
Lowest - Andrew Linden: 0 (0%)
Total Avatar Scripts - 662
Highest - Psi Merlin: 115 (17%)
Lowest - Andrew Linden: 0 (0%)
[16:15] Andrew Linden: The script resource pools should really group by parcel owner, and something is probably being grouped by parcel.
[16:15] Andrew Linden: er... something is being done per-parcel and should be done per-parcel-owner
[16:16] Andrew Linden: I've worked in the other code that does that kind of stuff... parcel accounting of prim resources
[16:16] Andrew Linden: so I think I could fix that one... if only I could find the time
[16:17] Andrew Linden: I've got all sorts of people asking for my time on projects, meanwhile I've got all the other projects of my own that I'd like to get done.
[16:17] Draconis Neurocam: i should think you would have people to delegate to you, as busy as you are
[16:18] Draconis Neurocam: err to delegate to*
[16:19] Andrew Linden: well, I'll write the jira number down... but don't know when I (or someone else) could get to it.
[16:20] Leonel Iceghost: what kind of projects would you really want to work on Andrew?
[16:20] Andrew Linden: Anyone know if it is still possible to create new accounts with the old "First Last" name system?
[16:20] Neit Tackleberry: I don't know how to.
[16:20] Leonel Iceghost: it is still possible yes
[16:20] Draconis Neurocam: i think all the old regapis have since died
[16:21] Opensource Obscure: that depends on 3rd party registration providers
[16:21] Draconis Neurocam: i could be wrong
[16:21] Andrew Linden: For a while we had that new account pipeline left open... dunno if it is still there.
[16:21] Leonel Iceghost: I saw a new last name four days ago
[16:21] Rex Cronon: now we have only resident 4 last name. right?
[16:22] Leonel Iceghost: Romeo Sport was the name
[16:22] Leonel Iceghost: sorry
[16:22] Andrew Linden: Ok, so rumor has it there may still be a way to create such accounts. Hrm...
[16:22] Leonel Iceghost: Romeo Spot**
[16:22] Sahkolihaa Contepomi: 'Resident' is only the backwards compatibility mode.
[16:22] Rex Cronon: u pay a linden...
[16:22] Rex Cronon: ;)
[16:22] Leonel Iceghost: why you want to block them?
[16:22] Opensource Obscure: yeah, by using a 3rd party provider using regapis
[16:23] Leonel Iceghost: last names are fun
[16:23] Andrew Linden: ok thanks Opensource. I don't know much about "3rd party providers". What are those?
[16:24] Liisa Runo: (if anyone know some regapi that still make intresting last names, tell me, all i know are dead or only giving resident)
[16:24] Andrew Linden: You mean like... very large estate owners who try to funnel new residents to arrive in their regions?
[16:24] Jionk: perch sono stati tolti i cognomi?
[16:24] Opensource Obscure: no i mean universities which provided a way to create SL accounts
[16:24] Opensource Obscure: let me find you a link to the SL wiki page.
[16:25] Opensource Obscure: http://wiki.secondlife.com/wiki/Linden_Lab_Official:Registration_API
[16:25] Arawn Spitteler: Large corporate accounts, like Linden or IBM?
[16:26] Jionk: whi the last names were removed?
[16:26] Opensource Obscure: foe example http://sl.nmc.org/join/
[16:26] Jionk: why*
[16:26] Draconis Neurocam: http://slregapi.com/mainframe/listsurnames.php apparently there arnt many options left
[16:27] Andrew Linden: yeah, thanks
[16:27] Andrew Linden: Ok thanks for that info Opensource
[16:27] Helen I - Blood: Hello, Avatar!
[16:28] Andrew Linden: I think Neit Tackeberry had a question. Go ahead Neit.
[16:28] Andrew Linden: jionk, last names were removed because SL was weird in that regard... no other online environments used it
[16:29] Neit Tackleberry: Thanks, Andrew.
[16:29] Andrew Linden: and the old blood of Linden Lab had been diluted enough that there werent' enough people passionate about it left.
[16:29] Neit Tackleberry: I really need to find the issue that is causing our hockey pucks to 1) go through the boards like they were cheese and 2) split when they come off the hockey stick. Neither the script nor the arena have changed since June, but the physics is totally hosed!
[16:29] Rex Cronon: some might say the right word is "unique" not weird:)
[16:29] PogoJump: so you removed something that made SL unique?
[16:29] Andrew Linden: the "First Last" name system was a social experiment/engineering effort.
[16:29] Leonel Iceghost: I would said SL was ahead on time.. now it went backwards
[16:29] Montecore Babcock: Just as a little back story on Neits question, I run a hockey league in SL for the past 3 years
[16:30] Leonel Iceghost: with less options
[16:30] PogoJump: Agreed
[16:30] Andrew Linden: Hrm... Neit. What kind of features do your hocky pucks use to move around?
[16:30] Leonel Iceghost: and crappy Name1234
[16:30] Neit Tackleberry: Monte is the scripter - and he's back with us now.
[16:30] Meeter: Timecheck : User Group is half over
[16:30] Andrew Linden: Surely they are scripted... but, do they use the llVehicle stuff?
[16:30] Montecore Babcock: It's a simple gun script
[16:30] Montecore Babcock: that shoots a prim puck
[16:31] Andrew Linden: puck is scripted? or no?
[16:31] Rex Cronon: hockey sticks shoot pucks:)
[16:31] Montecore Babcock: the puck has scripts yes to die when the puck reaches anotehr palyer and are "captured"
[16:31] Andrew Linden: what shape is used for the pucks? cylinder? or something else?
[16:31] Montecore Babcock: the pucks are physical
[16:31] Andrew Linden: (how big ar they?_)
[16:32] Montecore Babcock: a small cylinder
[16:32] Montecore Babcock: about .295 big
[16:32] Andrew Linden: oh really? hrm.
[16:33] Montecore Babcock: .295 by .295 by .095
[16:33] Lightsaber: releasing controls
[16:33] Kithylin's Jeweled Soul Staff: Touch staff for instructions
[16:33] SoulStaff: Free memory after load-state: 6562 kb, After typical data load: 5592 kb
[16:33] Andrew Linden: ok so something did change a couple months ago... when the mesh project went into Release Candidacy (RC)
[16:33] Montecore Babcock: the pucks normally slide around the ice and hit off the walls of the rink
[16:33] Leonel Iceghost: I saw 0.4 big things go through things too since mesh
[16:33] Montecore Babcock: but now they have been increasingly going through the prim boards
[16:33] Andrew Linden: What "channel" are your hockey regions on?
[16:34] Draconis Neurocam: i think it is also worth noting the shape of the boards and their dimensions
[16:34] Draconis Neurocam: ?
[16:34] Rex Cronon: maybe they have become lighter and now go faster?
[16:34] Montecore Babcock: I'm not sure what channel
[16:35] Andrew Linden: are the boards around the edge of the rink concave? or are they all convex?
[16:35] Montecore Babcock: the pucks seem to be the same speed
[16:35] Andrew Linden: Montecore, what is one of the region names?
[16:35] Montecore Babcock: the ones where they are going through are rectangular
[16:35] Montecore Babcock: Virtual Hockey League region
[16:36] Andrew Linden: ok, that region is on the main channel "Second Life Server"
[16:36] Montecore Babcock: one of the prmis is .240 by 7.931 by 1.51
[16:36] Andrew Linden: we have three other channels that we call LeTigre, BlueSteel, and Magnum
[16:36] Draconis Neurocam: ah thats very thin
[16:36] Andrew Linden: those are the RC channels where future updates are staged for a few weeks before they make it to the main channel
[16:36] Montecore Babcock: well we experimnetd with a very large prim
[16:36] Neit Tackleberry: It held 'em fine last season :)
[16:37] Montecore Babcock: one that was 6 meters thick
[16:37] Montecore Babcock: aand it still happened
[16:37] Andrew Linden: I was asking to try to get a better idea of when the mesh changes may have reaqched your region
[16:37] Andrew Linden: you said the problems started in June?
[16:37] Andrew Linden: lemme figure out when mesh went live everywhere...
[16:37] Montecore Babcock: it seems to have started in the past month
[16:37] Neit Tackleberry: Between last season and now, and the season ended in June...that's why I mentioned that date.
[16:37] Montecore Babcock: I cant say a certain day because we have bee taking a summer break after the season was over
[16:38] Andrew Linden: Mesh went live everywhere on Aug 24th, I think.
[16:38] Neit Tackleberry: Not very long ago.
[16:38] Montecore Babcock: and now we are playing games again and ite become a big hassle
[16:38] Andrew Linden: Anyway, this is what happened when mesh went out...
[16:38] Andrew Linden: the shapes of most objects subtley changed...
[16:38] Montecore Babcock: and I would be happy to show you the rink so youj could see what I mean
[16:38] PogoJump: has mesh been implimented into all viewers?
[16:39] Andrew Linden: These two boxes are "dynamic" (they fall under gravity)
[16:39] Andrew Linden: notice that there isn't much of a gap between the two boxes
[16:39] Sniper Siemens: only viewer 3 and kirsten i think...and Firestorm 3 Beta version
[16:40] Andrew Linden: it used to be the case that they would "hit" with a gap of 10cm
[16:40] PogoJump: ah, ok then.
[16:40] Draconis Neurocam: generally if i am making something like a bullet i want the length that it has to be a ratio of the speed it travels to the amount of server frames per second
[16:40] Andrew Linden: there also isn't much of a gap between the bottom box and the surface of the table
[16:41] Andrew Linden: that is what changed when mesh went live... we started shrinking the shapes of convex objects so that they would appear to stack correctly
[16:41] Andrew Linden: so... suppose you had a wall like this...
[16:42] Andrew Linden: if you butt the bottom of the wall to the table... there is actually an invisible "gap" between the box and the table as far as the physics collision shapes are concerned
[16:42] Andrew Linden: the details are a little complicated as to why this gap exists
[16:42] Montecore Babcock: ok, a lot of times thats where its happening now, but not all
[16:43] Andrew Linden: but it is a "quirk" of how the Havok physics engine tries to achieve speed and avoid object penetration
[16:43] Opensource Obscure: ahaha.
[16:43] Opensource Obscure: that's unlinked even
[16:43] Opensource Obscure: hooooo
[16:43] PogoJump: ....
[16:43] Opensource Obscure: i love this
[16:43] Flip Idlemind: Large house .()
[16:43] PogoJump: Agreed
[16:43] Techwolf Lupindo: Sorry, I though it was smaller.
[16:43] Opensource Obscure: you should come more often at LOL Techwolf ;)
[16:43] Ashiri Sands: hehe
[16:43] Andrew Linden: So Montecore, I haven't seen your rink
[16:44] Neit Tackleberry: We can fix that ;)
[16:44] Andrew Linden: however I'd suggest you try making your walls a bit longer on the bottom
[16:44] Montecore Babcock: I can ake you or give you a landmark
[16:44] Montecore Babcock: ok
[16:44] Andrew Linden: that is, try extending the walls into the ice at least 10cm
[16:44] Neit Tackleberry: Ah - make sure they go through the ice.
[16:44] Montecore Babcock: the basic point being that they overlap any gaps
[16:44] Andrew Linden: I can't look at it today, but if you give me a link I might be able to examine the content later
[16:44] Andrew Linden: I've got the region name
[16:44] Andrew Linden: so I'll make a note
[16:45] Montecore Babcock: thank you very much
[16:45] Neit Tackleberry: Oh excellent!
[16:45] Neit Tackleberry: Thank you.
[16:45] Neit Tackleberry: We do have a pre-season game tonight too ;)
[16:45] Leonel Iceghost: thanks for that info Andrew
[16:45] Montecore Babcock: Like I said, w've been doing this for three years and we have a good floowing of people who play
[16:45] Montecore Babcock: and I want to keep the experience a good one
[16:46] Andrew Linden: Now the shape shrinking starts to break down for small shapes
[16:46] Andrew Linden: objects that are smaller than about 0.2 meters on a side are adjusted differently
[16:46] Andrew Linden: because we can't shrink them much more before we go into negative numbers
[16:47] Andrew Linden: so small objects may still collide with a gap
[16:47] Rex Cronon: u get negative prims?
[16:47] Andrew Linden: and also the shrinking doesn't work for dynamic concave objects
[16:47] Neit Tackleberry: And antimatter :)
[16:47] Montecore Babcock: one moe thing....sometmes pucks can go through the middle of a prim, always been a problem, is there a fix for that or thats just a perfomance issue?
[16:48] Andrew Linden: when you say "through the middle of a prim"... how thick is the prim (a wall?) and how fast is the puck moving (about)?
[16:49] Montecore Babcock: maybe a speed of 15
[16:49] Montecore Babcock: .24 thickness
[16:49] Andrew Linden: the physics engine suffers from the "tunneling" problem
[16:49] Andrew Linden: the objects don't actually move smoothly, but basically teleport short distances every frame
[16:49] Rex Cronon: i don't think their pucks go fast enogh
[16:49] Rex Cronon: enough*
[16:49] Neit Tackleberry: You should come watch a game then :)
[16:50] Andrew Linden: which means fast moving objects can "arrive" into a state of penetration with another object, or even arrive on the other side if they are moving fast enough
[16:50] Andrew Linden: usually fast moving bullets suffer from this problem
[16:50] Montecore Babcock: but thats a sim preformance issue then right?
[16:50] Neit Tackleberry: That I have seen.
[16:50] Draconis Neurocam: 15 m/s at 44 fps average (lower probably) i would say the pucks should be about .34 meters in diameter, but thats jsut me
[16:50] Andrew Linden: and one solution is to make the bullets long in the direction which they travel
[16:50] Andrew Linden: right, the physics engine actually steps at 45fps
[16:51] Montecore Babcock: most of the time its not a problem but there's "weak" spots it seems in a prim
[16:51] Andrew Linden: er that is... it moves forward in time by 1/45th of a second... even if it only gets 10frames in a second
[16:51] Andrew Linden: so if the simulator is running slow (10fps) then as far as the phys engine is concerned it is only moving forward 10/45ths of a simulation second for every real second
[16:52] Neit Tackleberry: Good.
[16:52] Montecore Babcock: well i will try the overlap fix
[16:52] Andrew Linden: anyway that means that your puck is moving bout 0.33 meters per frame at 15m/sec
[16:52] Andrew Linden: so here is a way to try to reduce the problem...
[16:53] Leonel Iceghost: Andrew, that remembers me, the spead of an avatar being llImpulsed goes mad when the sim has less fps than normal.. something there is not taking fps based time
[16:53] Andrew Linden: you can stack the walls of your rink double or triple deep
[16:53] Andrew Linden: staggered every 0.1m
[16:53] Neit Tackleberry: Ah.
[16:53] Andrew Linden: since 0.1m is the "convex radius" of buffer zone that the physics engine uses
[16:54] Andrew Linden: so like this...
[16:54] Montecore Babcock: ould that be three walls of the same thickness or three walls that added tothe same distance?
[16:55] Draconis Neurocam: i would wonder if giving it an arbitrary convex surface to collide against would fix or just make more problems
[16:55] Draconis Neurocam: err concave
[16:55] Meeter: Timecheck : User Group is almost over
[16:55] Andrew Linden: ok so your puck would hit from this side
[16:56] Andrew Linden: and you stack your walls such that each inner face is 0.1 m behind the one in front
[16:56] Montecore Babcock: does it matter how thick each prim is?
[16:57] Andrew Linden: this will make the puck likely to hit the walls inside, even if it tunnels through the first or second layer
[16:57] Andrew Linden: you probably only need three layers max since your puck is moving about 0.33 m each step
[16:58] Andrew Linden: it is best to make your walls at least 0.1m thick however...
[16:58] Montecore Babcock: ok
[16:58] Montecore Babcock: I think that helps head me in the right direction
[16:58] Andrew Linden: on the far side the walls should mostly end around the same place
[16:58] Andrew Linden: if your first wall is too thin and its far side is before the near side of a later wall
[16:59] Andrew Linden: then it is possible for the puck to tunnel through the first, then get trapped for a while between the two inside faces
[16:59] Flip Idlemind: Any idea what's likely to get promoted / RC'd next week? Or too early to tell?
[16:59] Andrew Linden: this technique also works well for bullet shields, but you need more layers for the faster bullets
[17:00] Montecore Babcock: thank you Andrew
[17:00] Leonel Iceghost: thanks a lot for the explanation Andrew
[17:00] Meeter: Thank you for coming to the Server User Group
[17:00] PogoJump: i ask the same Flip does
[17:00] Andrew Linden: You're welcome.
[17:00] Montecore Babcock: we've been banging our heads about it
[17:00] Montecore Babcock: lol
[17:00] Flip Idlemind: Fingers crossed that it'll be what's on LeTigre
[17:00] Neit Tackleberry: Yes!
[17:00] Neit Tackleberry: And having mad hockey players is quite dangerous :)
[17:00] Andrew Linden: I *think* the most likely candidate for a promotion is a maint-server RC... not sure what channel it is on...
[17:00] Montecore Babcock: they have rabies
[17:01] Neit Tackleberry: And veins in their teeth :)
[17:01] Leonel Iceghost: and that remembers me, the spead of an avatar returned by llGetVel that is being llImpulsed while running goes mad when the sim has less fps than normal.. something there is not taking fps based time
[17:01] Andrew Linden: looks like LeTirgre would be my guess
[17:01] Flip Idlemind: Yayy
[17:01] Andrew Linden: it isn't guaranteed
[17:02] Rex Cronon: u become like a flash if u push yourself too hard:)
[17:02] Leonel Iceghost: the avatar almost doesn't move, but llGetVel gives 30m/s when it should give ~6 or 7m/sec
[17:03] Andrew Linden: oh really Leonel? Huh... that should be easy to fix.
[17:03] Andrew Linden: I wonder what content I'd break if I fixed it.
[17:03] Ashiri Sands: How is llGetVel calculated anyway?
[17:03] Leonel Iceghost: it only happens when the sim is blocked
[17:03] Draconis Neurocam: wouldnt that be related to velocity and or ping interpolation andrew?
[17:03] Leonel Iceghost: by physics, so not much content break..
[17:04] Andrew Linden: No, the code is probably using the wrong time (realtime instead of virtual, or virtual instead of real) for some velocity calculation
[17:04] Draconis Neurocam: interesting
[17:05] Neit Tackleberry: Developers would never do *that*.
[17:05] Andrew Linden: no no...
[17:05] Andrew Linden: (in fact, I'd guess I wrote that bug years ago)
[17:05] Methias Kira: Haha
[17:05] Neit Tackleberry: <--- developer by experience, too :)
[17:05] Draconis Neurocam: take care andrew, and thanks for everything
[17:05] Andrew Linden: I have to go. Thanks for coming everyone.
[17:05] Flip Idlemind: Baiii
[17:05] Opensource Obscure: thank you Andrew !
[17:06] Ashiri Sands: also when is the llGetVel called in relation to the impulse?
[17:06] Rex Cronon: tc andrew
[17:06] Methias Kira: Take care andrew
[17:06] Ashiri Sands: Thanks Andrew
[17:06] Rex Cronon: tc all those leaving
[17:06] Opensource Obscure: take care everybody
[17:06] Al Supercharge: is the BASIC SHADERS issues all Viewer?
[17:06] Sahkolihaa Contepomi: See you Andrew.
[17:06] Sahkolihaa Contepomi: No, it's more of a Windows problem.
[17:06] Sahkolihaa Contepomi: Because it works perfectly for me on Linux.
[17:06] Montecore Babcock: see you Andrew
[17:06] Neit Tackleberry: Thank you, Andrew.
[17:07] Al Supercharge: no its not on viewers < 2.7
[17:07] Neit Tackleberry: Neit Tackleberry
[17:07] Al Supercharge: but started w 2.7 and still on 3.0
| Prev 2011.09.06 | Next 2011.09.13 |