Simulator User Group/Transcripts/2012.10.05

From Second Life Wiki
Jump to: navigation, search

Simulator_User_Group

Prev 2012.10.02 Next 2012.10.09

List of Speakers

Andrew Linden Ardy Lay Barrington DeSantis
GEE McAuley Jane1 Bookmite Kennylex Luckless
MartinRJ Fayray RaithSphere Enyo Rex Cronon
Simon Linden TankMaster Finesmith Theresa Tennyson
Tiberious Neruda TonyaSouther Toysoldier Thor

Transcript

[16:02] Tiberious Neruda: what's the incoming character mixup gonna be?

[16:02] Rex Cronon: hello everybody

[16:02] Kennylex Luckless: Hia

[16:02] Simon Linden: Hello

[16:02] Andrew Linden: hello

[16:03] Rex Cronon: hi

[16:03] Rex Cronon: hi andrew

[16:03] Rex Cronon: hi simon

[16:03] Andrew Linden: I missed the last two user groups. One I was busy and the other I was so busy I forgot all about it.

[16:04] Rex Cronon: were u busy fixing things:)

[16:04] Theresa Tennyson: It's okay. I think you were busy fixing all the broken stuff I was going to yell at you about with the interest list sims, because they work fine now.

[16:04] Andrew Linden: Oz Linden contacted me to confirm that I was planning on removing the cloud layer data from ther server, and also the legacy map tile download system.

[16:04] Kennylex Luckless: A busy Linden is a good Linden.

[16:04] Andrew Linden: I confirmed that I am working on removing the cloud layer

[16:05] Andrew Linden: but am not doing anything with the legacy map tile download system

[16:05] MartinRJ Fayray: hello

[16:05] Rex Cronon: hi

[16:05] Andrew Linden: I just always mention that system as a good reason for us to no longer support viewer 1.23 protocols.

[16:06] Andrew Linden: The old map tile download system doesn't scale well.

[16:06] Andrew Linden: I was fixing bugs in my interestlist rewrite.

[16:06] Andrew Linden: Turns out my first version just wasn't any faster than the old system

[16:07] Andrew Linden: so I'm trying again.

[16:07] Andrew Linden: Simon, go ahead with any deploy news. I haven't been paying attention to the promotion discussions.

[16:07] Rex Cronon: u could put it on google map?

[16:07] Simon Linden: well, a few of you may have noticed a bit of excitement (in a bad way) on Wednesday

[16:08] Andrew Linden: er... not an interstlist *rewrite* at the moment... I'm working on an interestlist cleanup/fixup pass.

[16:08] Simon Linden: There was a problem with the prim accounting code that blew up on the RC rollout, and a bunch of regions had to get rolled back as we aborted

[16:09] Simon Linden: That once again stopped our maintenance release with Baker's large group work from getting out

[16:09] Simon Linden: I _doubt_ that's going onto the grid next week either ... I haven't fixed that bug yet, so we're running out of time

[16:10] Simon Linden: Next week there will be a new RC, however, that is mostly internal changes but does include a minor update for the physics engine library

[16:10] Tiberious Neruda: ...update?

[16:10] Tiberious Neruda: what's being changed?

[16:10] Simon Linden: The RCs on Wednesday

[16:10] Rex Cronon: it is so interesting when u receive messages like this: "Your ... objects have been returned to your inventory Lost and Found folder by (???) (???) near parcel '...' at ... because the parcel is full."

[16:11] Simon Linden: They will all be the same new version

[16:11] Theresa Tennyson: Will it cause Havok with sim crossings?

[16:11] Simon Linden: /rimshot

[16:11] Tiberious Neruda: see, speaking of physics...

[16:12] GEE McAuley: Linkability distance rules are broken

[16:12] Simon Linden: It's almost all updating libraries ... we've been using a fairly old set of compilers and such to make some of the development builds of the servers, and this brings us to more recent code

[16:12] Rex Cronon: could those messages be more informative. like who/what actually returned your objects and the names of those objects?

[16:12] Andrew Linden: Yeah, the linkability bug is on my plate. I need to look into that one.

[16:12] Tiberious Neruda: I'm running into a combination of low limits on llSetVelocity and inconsistent results upon application to the avatar

[16:12] Simon Linden: Don't build stuff that will rely on that in the future, btw

[16:13] Andrew Linden: Er actually, the bug about linking working for too large distances was fixed in some RC.

[16:13] TonyaSouther: Is there a reason for a 54m linkability limit with a max prim size of 64 meters?

[16:13] Andrew Linden: but there is another bug... another way to get too-large sets linked together

[16:13] Simon Linden: I think that will be fixed in the next one ... it should go out 1 1/2 weeks from now

[16:13] GEE McAuley: General consensus Keep it broke LOL, The region protection will kep things from over spil.

[16:13] Simon Linden: It affects the interest list, Tonya

[16:14] Simon Linden: We base the visibility on the root prim ... so if there are parts very far from that, it might be outside your draw distance but affecting where you are

[16:14] Andrew Linden: Right, there are a few reasons we prefer to keep the linkability distances limited, and the max prim sizes too...

[16:14] Tiberious Neruda: for example, this spring here causes my HUD to do a llSetVelocity on me with a force of <0,0,30>... but I have to add extra to amplify that (by a whopping -2.7-) if I'm on the ground

[16:14] Andrew Linden: There are a number of problems that are exacerbated by really large objects.

[16:15] Andrew Linden: (1) We don't collide objects across region boundaries -- they exist in two different physics engines if they are in two different regions

[16:15] Tiberious Neruda: unfortunately, I'm reaching some kind of limit there, and only if my llGetAgentInfo returns 6 and the force calculated as necessary is that <0,0,30>, do I get the kind oh height I'm intending

[16:15] Andrew Linden: so very large objects just makes the the problems that result for objects that extend into other regions worse

[16:16] GEE McAuley: Haven't we a region boundary setting that can keep the neighbors out?

[16:16] Andrew Linden: (2) The interestlist code (the server-side culling that figures out what objects need to be streamed to the viewer, or that need updates) does not handle very large objects well.

[16:16] Tiberious Neruda: why is it that a llSetVelocity() call on an avatar does not have a predictable outcome?

[16:16] TonyaSouther: The interest list code seems a bit fragile, too...

[16:17] Andrew Linden: Those are the two problems that remain. We used to have a problem with parcel encroachment, but that was mostly fixed.

[16:17] Andrew Linden: Definitely Tonya. I'm finding it hard to modify.

[16:17] GEE McAuley: Hmmmm A shame. Would be a nice feature for the builders and also less prim count on the larger builds.

[16:17] Toysoldier Thor sips his 100 proof whiskey and coke and watches the intellect text by his local chat screen......

[16:17] Rex Cronon: so. if u r in the next sim your prim can extend into the sim near u?

[16:18] TonyaSouther: yeah, if you look at the interest list code cross-eyed, it seems to break.

[16:18] Ardy Lay: Is the parcel object entry property checked on region crosing?

[16:18] Andrew Linden: Tiberious, the reason llSetVelocity() doesn't work well on avatars is because of the "avatar action"... called that because it is a C++ class that derives from the Havok::Action class

[16:18] Andrew Linden: it is basically a bit of logic that runs every physics engine step

[16:18] Andrew Linden: and computes forces and impulses on the avatar to keep it moving/hovering/stopped/etc accordingly.

[16:18] Simon Linden: The physics engine also gets worse with large objects .... it makes one object affect many others

[16:19] Andrew Linden: So the llSetVelocity() will slam the avatar's velocity at one moment, an then the avatar action will run the next physics timestep and compute new forces to override the velocity

[16:19] Tiberious Neruda: for some reason, I didn't get the huge bounce to ~75m I was expecting... LOL!

[16:19] Andrew Linden: to achieve the movement that it thinks is correct

[16:20] Ardy Lay: https://jira.secondlife.com/browse/SEC-720 -- In reference to my question

[16:20] Tiberious Neruda: (usually, after a couple smaller, up to only ~53m, I get one that goes much higher, and what I've intended to achieve)

[16:21] Andrew Linden: Ardy, apparently it is not being checked in that case.

[16:21] TonyaSouther: Holy cow,m Tiberius, are you testing flubber shoes?

[16:21] Tiberious Neruda: it's supposed to be a platformgame-type spring

[16:21] Andrew Linden: The region now is supposed to know some of the parcel settings, and shapes, of neighbors parcels.

[16:21] Ardy Lay: Andrew, I haven't tried lately. I guess I'll try it again.

[16:21] Rex Cronon: u could rez an object that pushes u?

[16:22] Tiberious Neruda: I've tried several different things, Rex

[16:22] Tiberious Neruda: a push, an applied impulse... anything I can think of to force the av to get the same kind of air every time

[16:23] Andrew Linden: However, there may be a code path that moves an object onto a neighbor parcel without first checking the parcel settings. I wouldn't be surprised.

[16:23] Ardy Lay: Let's play "Count the Code Paths"!

[16:23] Andrew Linden: I thought there was still a bug open for vehicles driven by avatars... we might not be properly checking the parcel settings in that case.

[16:23] Simon Linden: does llSetVelocity() work on avatars?

[16:23] Tiberious Neruda: yes

[16:23] Tiberious Neruda: but kinda unpredictably

[16:24] Simon Linden: that's almost always better than using an impulse

[16:24] Toysoldier Thor: Well I hate to be rude and butt in on a topic but I got inlaws showing up any time now... so... Simon... have you chaecked the logs on those Sim Disruption attacks or "Sim Stalings"

[16:24] Andrew Linden: the llApplyImpulse() is even more unpredictable

[16:24] Simon Linden: Toy - yes, I did, but didn't find any smoking gun in the logs about what was going on

[16:24] TonyaSouther: Is this where we say "ye cannae change the laws o'physics, Captain!" ?

[16:24] Andrew Linden: because it uses the legacy script "energy budget"

[16:25] Andrew Linden: which will attenuate the results if the scripted object doesn't have enough "energy" to execute the impulse that it wants

[16:25] Andrew Linden: also, that llApplyImpulse() has great griefing potential

[16:25] Ardy Lay: But.... can we change the gravitational constant of the Univ^H^H^H^H Simulator?

[16:25] Andrew Linden: so we hobbled it a long time ago with a very high energy consumption rate

[16:25] Tiberious Neruda: Simon, there seems to be some kind of disconnect between the magnitude of the velocity when the avatar is going from the ground, and when the avatar is in the air

[16:26] MartinRJ Fayray: By the way: We still can't turn avatars with a LSL script, without forcing the resident to sit on a poseball, right?

[16:26] Andrew Linden: That is right Tiberious, the "agent action" I was talking about has distinctly different code paths for on-ground and in-air.

[16:26] Toysoldier Thor: OK thanks Simon... Qiu gave me some extra things to look for if I witness it again.... but I havent seen this event since the horrid saturday.

[16:26] Simon Linden: hmm, well, they definitely are in a different state when on the ground vs. flying. The mass doesn't change, but maybe there's some friction on the ground

[16:26] Toysoldier Thor: Inspire Space port was down for over 24 hours

[16:27] Andrew Linden: That is right MartinRJ, as far as I know.

[16:27] Simon Linden: ugh, that's not good

[16:27] Tiberious Neruda: Andrew, -why- is that different?

[16:27] Tiberious Neruda: is there a reason it needs to be?

[16:28] Andrew Linden: Well... theoretically no.

[16:28] Toysoldier Thor: OK I must run back to RL.... but Andrew.... please send a message to the Commerce Team that seems to like hiding from their customers.... they better get their butts in gear and figure out how to stoip stealing Lindens from Merchants and start talking to us !! We are getting pretty ticked off with Brooke and company.

[16:28] Andrew Linden: But the agent action is trying hard to keep the avatar from sliding downhill when it is just standing there

[16:28] Tiberious Neruda: ah

[16:29] Andrew Linden: and also trying hard to make the avatar move forward when walking

[16:29] Tiberious Neruda: okay. That makes sense

[16:29] Andrew Linden: so the forces and frictions it applies to do so will tend to defeat any external push/bump/force

[16:29] Toysoldier Thor: good night all......

[16:29] Tiberious Neruda: I've noticed when you use llSetBuyoancy to make yourself 'lighter', you do slide along the ground

[16:29] Rex Cronon: tc

[16:30] Tiberious Neruda: err, well, you know what function I mean

[16:30] GEE McAuley: Night Toy

[16:30] Andrew Linden: If the agent action thinks the avatar is on a ballistic trajectory (not flying and not hovering) then it lets it go, and external things can push it around.

[16:30] Tiberious Neruda: ah

[16:30] Andrew Linden: ToySoldier, is Inspire Space port still down?

[16:30] MartinRJ Fayray: llSetBuoyancy

[16:30] Tiberious Neruda: yeah, I know

[16:30] MartinRJ Fayray: =)

[16:30] Andrew Linden checks...

[16:30] Meeter: Timecheck : User Group is half over

[16:30] Tiberious Neruda: I typoed it. I know I did

[16:31] Andrew Linden: Hrm... I don't see a region by that name, unless he was talking about a "Space Port" in the region "Inspire".

[16:32] Tiberious Neruda: ...that sounds like it makes more sense, Andrew

[16:32] Tiberious Neruda: I think i've been there before

[16:32] Theresa Tennyson: It's in Shinda.

[16:33] Barrington DeSantis: Yup, Shinda. Inspire space park

[16:33] Andrew Linden: Hrm... currently Inspire is empty, but it looks like people have been visiting it in the last 24 hours.

[16:33] Andrew Linden: Oh, in Shinda.

[16:33] Andrew Linden: 33 people in Shinda atm

[16:33] Barrington DeSantis: 33 ppl

[16:33] Tiberious Neruda: anyway... do you think there might be some way for the av to 'opt in' to the 'ballistic' state without giving up buoyancy?

[16:34] Andrew Linden: No, that would not be easy Tiberious.

[16:34] Andrew Linden: The avatar motion is pretty fragile (as some of the other systems I've mentioned)

[16:34] Andrew Linden: last time I messed with it various bits of content broke

[16:34] Theresa Tennyson: Could you create a sabot that the avatar could travel in, Tiberious?

[16:35] Andrew Linden: for example, someone somewhere has made some steps that just barely work with the current avatar motion system

[16:35] Theresa Tennyson knows an unusual amount about weaponry for a teenage girl.

[16:35] Andrew Linden: that is... the steps have placed accidentally (or tuned intentionally) to allow the avatar to climb them, but not if the steps were any higher

[16:36] Andrew Linden: very small changes in the avatar motion code would result in some steps that suddenly just don't work at all

[16:36] Andrew Linden: And will I hear about it?

[16:36] Andrew Linden: Oh yes.

[16:36] TonyaSouther: Yes, breaking content is usually a bad idea.

[16:37] Tiberious Neruda: I've noticed that when the avatar lands from a great distance, if there's any motion parallel to the horizon, they actually will slide if the prim's friction is low enough... but only until you get to the 'brush' animation

[16:37] Andrew Linden: Right. What we've talked about in the past, and what I'd like to do someday...

[16:37] Ardy Lay: I showed my ladder to Falcon. He told me not to do that.

[16:37] Tiberious Neruda: is this when the standing action takes over?

[16:37] Andrew Linden: is to make a completely different avatar motion pipeline, that takes different data packets from the viewer

[16:37] Andrew Linden: and then viewers (and individuals) could opt in or not

[16:38] TonyaSouther: Wonder if it'd be practical to add an LSL call, llTakeOff(), to become ballistic?

[16:38] Andrew Linden: This idea has come up when people ask for joystick controls for the avatar (instead of key presses)

[16:38] Andrew Linden: and for configurable walk/run speeds

[16:38] Tiberious Neruda: Andrew: if you did that, I'd sign up in an instant

[16:38] Tiberious Neruda: want, want WANT

[16:39] Andrew Linden: Yeah, its still a good idea.

[16:39] Theresa Tennyson: CONROLLABLE vehicles? SQUEEEE!!!

[16:39] TankMaster Finesmith: have I missed anything? :P

[16:40] MartinRJ Fayray: hi Tankmaster

[16:40] Andrew Linden: everything!

[16:40] TankMaster Finesmith hates work

[16:40] RaithSphere Enyo: you broke the rules Tank

[16:40] Tiberious Neruda: ....maybe -that-'s what I can do... I've tried a 1.0 buoyancy... maybe I need to actually make the av float -up- a tiny bit before applying that velocity

[16:40] TankMaster Finesmith: which one are you referring to?

[16:40] TankMaster Finesmith: theres so many

[16:40] RaithSphere Enyo: xD

[16:40] TonyaSouther: llTakeOff(integer you_hoser)...

[16:41] Andrew Linden: Tiberious, you're trying to make a platform that throws the avatar somewhere?

[16:42] Tiberious Neruda: yeah. Ever play a platform game with springs that cause a high bounce?

[16:42] Tiberious Neruda: it's like one of those

[16:42] Andrew Linden: hrm.... bouncing straight up? or bouncing laterally as well?

[16:43] Tiberious Neruda: rezzed object communicates to HUD with its rotation and force, HUD applies velocity using the rotation as its direction

[16:44] Andrew Linden: the avatar has a little state machine on the server

[16:44] Andrew Linden: so the server keeps track of whether it thinks the avatar is walking, jumping, falling, hovering, flying, etc

[16:44] Tiberious Neruda: right. And I'm guessing llGetAgentInfo returns that value from the machine?

[16:45] Andrew Linden: so basically you want to be in the "falling" state before you apply the force that will provide the ballistic path

[16:45] Tiberious Neruda: correct

[16:45] Andrew Linden: so maybe you want to first bounce the avatar straight up a little bit, to get it off the ground

[16:45] Tiberious Neruda: except I -have- tried that too

[16:45] Andrew Linden: and then apply the real force (with lateral components) to get it to where you want

[16:45] Andrew Linden: oh hrm...

[16:46] Tiberious Neruda: I've tried a llSetVelocity(<0,0,.1>,FALSE)

[16:46] Tiberious Neruda: and then my main SetVelocity

[16:46] Andrew Linden: what is the avatar standing on when the force happens... whatever it happens to be on? or specific platforms?

[16:46] Tiberious Neruda: the object itself is phantom via VolumeDetect

[16:46] Tiberious Neruda: so if there's ground or prim under it, the av might land on that

[16:47] Tiberious Neruda: and because of possible delay in communication, the av almost always hits

[16:47] Andrew Linden: you might be able to create a prim under the avatars feet that pushes the avatar up a bit

[16:47] Andrew Linden: then delete/remove that prim... the avatar will go into falling state

[16:47] Andrew Linden: then push them

[16:48] Tiberious Neruda: that's... too slow

[16:48] TonyaSouther: Hm. Try shrinking the prim via llSetLinkPrimitiveParamsFast() ?

[16:49] Tiberious Neruda: I already have objects rezzing from the HUD, and even that is slow

[16:49] Andrew Linden: well, you have to wait at least a frame to get the avatar into the falling state

[16:49] Tiberious Neruda: eep

[16:49] Andrew Linden: if you push while they're standing the avatar action may win the pushing contest

[16:49] Tiberious Neruda: .02 second...

[16:50] Tiberious Neruda: maybe that's why it's not going...

[16:50] Andrew Linden: Yes, perhaps you need to delay a little bit.

[16:51] Andrew Linden: Push the avatar up, wait a frame or two, then slam the avatar's velocity for the trajectory you want.

[16:51] Tiberious Neruda: dumb question...

[16:51] Andrew Linden: Might work.

[16:51] Tiberious Neruda: can internal animations trigger 'state' changes? Or are those animations triggered by the state change?

[16:52] Andrew Linden: The animations are triggered by the state changes.

[16:52] Andrew Linden: The state changes can be triggered by packets from the viewer

[16:52] Andrew Linden: but I don't know of any LSL calls that can trigger state changes, except through bumps and pushes.

[16:53] Tiberious Neruda: I know I can stop, for example, 'standup', and regin control of my avatar sooner

[16:54] Andrew Linden: You mean by pressing the spacebar?

[16:54] Tiberious Neruda: no

[16:54] Tiberious Neruda: the 'splat' animation

[16:54] Andrew Linden: Spacebar triggers a "stop". So if you're falling and hold it down you'll fall slower.

[16:54] Andrew Linden: Or if you're sliding down a very steep hill, spacebar may slow you down.

[16:55] Andrew Linden: I think... haven't tried it in a long time

[16:55] MartinRJ Fayray: yes that's true

[16:55] TankMaster Finesmith: ugg, hitting space in firestorm sets focus tot he chat bar at the bottom

[16:55] Tiberious Neruda: I'm using scripts to stop the animation 'standup', which is the one where you fall face-first

[16:55] Andrew Linden: ah

[16:55] Meeter: Timecheck : User Group is almost over

[16:56] Tiberious Neruda: and that lets me take control again instantly

[16:56] Kennylex Luckless: Why is it so hard to ALT-zoom on mesh avatars, I can not get hold on them; Is that server or content creation question?

[16:56] Tiberious Neruda: which is why I wonder

[16:56] Andrew Linden: Kennylex, that would be a viewer bug/behavior.

[16:56] Andrew Linden: The viewer is in complete control of the camera position.

[16:57] Theresa Tennyson zooms in on Widely.

[16:57] TankMaster Finesmith: the viewer isn't properly handeling mesh bounding box, and instead focuses through it

[16:57] TonyaSouther: The entire avatar is mesh, or just is wearing something rigged mesh clothing, like my dress?

[16:57] Tiberious Neruda: it only happens to rigged mesh, doesn't it?

[16:58] Andrew Linden: Ah, I think I witnessed that today when I tried to focus on Kennylex's deep-underwater-monster avatar.

[16:58] Tiberious Neruda: Tonya, yours doesn't use offsetting, does it?

[16:58] MartinRJ Fayray: I request a Ctrl+Shift+Y (Sun -> always Midday) option for reallife.

[16:58] TonyaSouther: Nope.

[16:58] Kennylex Luckless: Yes, hard to get good snapshots if you want to try to put focus on the eye.

[16:59] TankMaster Finesmith: id rather it be always midnight irl, but then again, plants would probbaly die, and that wouldnt be good

[16:59] Tiberious Neruda: ....if I had to guess, I'd say it has something to do with the avatar's collision skeleton (Develop > Avatar > Show Collision Skeleton) being what's actually being focused on

[16:59] Tiberious Neruda: (this's just based on my observations)

[16:59] MartinRJ Fayray: humm you can import them from my side of the planet. We will be rich!

[17:00] TankMaster Finesmith: just send it to my rl inventory

[17:00] MartinRJ Fayray: lol

[17:00] Meeter: Thank you for coming to the Server User Group

[17:00] TankMaster Finesmith: ill rezz it later

[17:01] Tiberious Neruda: it seems like the viewer is assuming things about the collision skeleton shape that aren't necessarily so with a mesh avatar with offsets

[17:01] Tiberious Neruda: like I say, though, I'm -just- guessing

[17:01] Theresa Tennyson: Andrew, is the interest list code on Balkanski on Aditi new or did you just back everything out?

[17:01] Rex Cronon: i g2g. tc everybody

[17:02] Andrew Linden: Yeah, I'm going to get back to my intrerestlist bugs...

[17:02] Simon Linden: I have to as well ... thanks everyone for the discussion today

[17:02] Andrew Linden: must fix!

[17:02] MartinRJ Fayray: Goodbye!

[17:02] TankMaster Finesmith: your undesired features :D

[17:02] Theresa Tennyson: Bye!

[17:02] Jane1 Bookmite: Thanks SImon and Andrew

[17:02] Simon Linden: See you next time

[17:02] TankMaster Finesmith: thanks for your time, simon and andrew

[17:02] TonyaSouther: Have fun< Andrew.

[17:02] Andrew Linden: the good thing about interestlist bugs is that they don't translate well to misfeatures

[17:03] TonyaSouther: heh!

[17:03] Andrew Linden: *everybody* will want them fixed

[17:03] TankMaster Finesmith: true

[17:03] Barrington DeSantis: Enjoy the weekend and thanks

[17:03] Andrew Linden: cheers

[17:03] TankMaster Finesmith: to bad baker's group list loading code was bundeled with it

[17:03] Kennylex Luckless: Random crash

[17:03] TankMaster Finesmith: that seemd to work fine



Simulator_User_Group

Prev 2012.10.02 Next 2012.10.09