User:Andrew Linden/Office Hours/2010 05 25

From Second Life Wiki
Jump to navigation Jump to search

Transcript

[10:58] xstorm Radek: as i have been looking at the bug i see some Physics errors in how the physics works
[10:59] xstorm Radek: first off prims moving from one point to another with no scrips in the prims
[10:59] Simon Linden: What's the jira number?
[10:59] xstorm Radek: 2 the spher ball of doom
[10:59] xstorm Radek: ok but it was closed hold on
[11:00] Welcome to Linden office hours
[11:00] xstorm Radek: SEC - 680
[11:02] xstorm Radek: there are like 4 key factors that shows there is data not getting looked at not only in testing but the server physics
[11:03] Simon Linden: Looks like the fixes there will be in the next 1.40 version we roll out
[11:03] xstorm Radek: i just have to ask why is the server when the prims are set in physics moving the set number of prims to another rezzer
[11:03] Simon Linden: I think that's happening today, I was helping with the code merge this morning
[11:03] xstorm Radek: bet not roll out to day
[11:04] Andrew Linden: Hi Kitto.
[11:04] Rex Cronon: hello everybody
[11:04] Kitto Flora: Hello
[11:04] Ardy Lay: Hi Simon and Andres
[11:04] xstorm Radek: the error is bigger in my feeling after testing
[11:04] Bronson Blackadder: meow
[11:04] Kitto Flora: Are you guys involved in the version(s) running on Aditi?
[11:04] xstorm Radek: i feel there needs to be better testing tool for Aditi
[11:04] Rex Cronon: wassup people:)
[11:04] Simon Linden: Well, at this point the best thing to do is check out the new version when it's pushed to the beta grid
[11:05] xstorm Radek: the old Statistics is not that good for showing some of the errors
[11:05] Simon Linden: I'm actually going to be talking with Oskar today about Aditi and beta testing since I'm visiting his office ... we all want to figure out how to make testing there more effective
[11:06] Rex Cronon: btw what is with havok and center of mass?
[11:06] Rex Cronon: does havok have problems with falling objects?
[11:06] xstorm Radek: may i help point out some steps i have used in testing in the past for another company
[11:07] Kitto Flora: H7 has a problem with the 'edge fin'. It appears to be worse than in H4
[11:07] Rex Cronon: shouldn't a falling object have the center of mass go first down?
[11:07] Rex Cronon: and the rest of the object rotate?
[11:07] xstorm Radek: but the way i see it there is a nasty little problem under it all with server 1.40 and Havok 7
[11:08] Kitto Flora: Rex: If you want to imitate RL in detail you really have to model the aerodynamics of the shape
[11:08] Andrew Linden: What is this phenomenon about the center of mass being wrong?
[11:08] Rex Cronon: i am trying with no success to make an object fall right and seems it is impossible
[11:08] Liisa Runo: Rex, that would be expected if we would have aerodynamics, but we dont
[11:08] Simon Linden: Rex - yes, but what you see is going to really depend on the actual shape used in physics, which may not be exactly the same as what you see
[11:08] Simon Linden: Right, there is no aerodyanics done
[11:09] xstorm Radek: there is a number of problems i see ing with physics
[11:09] Andrew Linden: Kitto, what is the "edge fin"?
[11:09] Rex Cronon: i tried kitto. aerodinamics don't seem to play any role:(
[11:09] Kitto Flora: Earthside rotation of falling bodies is rotation due to center of mass -v- center of drag
[11:09] Simon Linden: But if you hit the edge of something in a collision, it should behave the way you expect
[11:09] xstorm Radek: i bet even Kitto has seen it
[11:10] Simon Linden: The simulator will also simplify the shape if performance drops a lot, so your complex shape may suddenly behave like a box
[11:10] Arawn Spitteler: AErodynamics sounds like a server side option for the future.
[11:10] Kitto Flora: Andrew - edge fin is the obstructing projection that sticks out from the edge of a rectangular object in Havok. A sliding object hits it.
[11:10] xstorm Radek: you do not even need complex shapes
[11:10] Rex Cronon: well the object i made always falls bottom first:(
[11:10] Andrew Linden accepted your inventory offer.
[11:10] Andrew Linden: Ah, "edge fin" is the collision with hidden faces.
[11:10] xstorm Radek: simple basic shapes even have problems
[11:11] xstorm Radek: path data is wrong
[11:11] xstorm Radek: the steps that say what happens next is wrong much of the time
[11:11] Techwolf Lupindo: One thing that can help with testing is to have some way of auto-updateing to user controllled updating of inventory. right now, I have some beta test products that I can't test on the beta grid now due to different inventory now.
[11:11] Rex Cronon: what use is havok if have to compute the path of a falling object and its rotation:(
[11:12] Rex Cronon: if i have*
[11:12] Andrew Linden: Rex, what is this "rotation during falling" you're talking about?
[11:12] Andrew Linden: For an object falling through the air?
[11:12] Kitto Flora: Andred - edge fins are the red things on that model
[11:12] Andrew Linden: Or for an object falling and colliding against another?
[11:13] Kitto Flora: Its how a box appears in collision in H$, and it seems worse in H7.
[11:13] Rex Cronon: well. do u see this mortar round?
[11:13] Simon Linden: Right, so syncing your inventory is an issue. I can imagine importing accounts is as well, but we probably don't want to make that totally open and automatic ... I'm not sure why I'm paranoid, but it sounds like a possible source of problems if it was
[11:13] xstorm Radek: plus one problem with Aditi is Agni and Aditi is moving data in and out to each other
[11:13] Andrew Linden: Yeah I get it Kitto.
[11:13] Rex Cronon: it always falls tails first
[11:13] Rex Cronon: doesn't mater if i move the root prim to fron or back:(
[11:13] Rex Cronon: tothe front*
[11:15] Andrew Linden: What causes the morter to fall nose first xstorm? Via script behavior?
[11:15] Arawn Spitteler: If a new server version is insecure for Scripts, then automatic inventory import can save a lot of time for the copiers.
[11:15] Simon Linden: Right now it's always a one-way street - agni to aditi. I don't think we want to go the otherway, but I can see managing that and possibly keeping new test stuff built on Aditi is important
[11:15] Arawn Spitteler: Old scripts would also be insecure, in that case.
[11:15] Kitto Flora: Rex: Fake it with VEHICLE_LINEAR_DEFLECTION_EFFICIENCY and _ANGULAR_
[11:16] Rex Cronon: i have to make it a vehicle? ahh:(
[11:16] Andrew Linden: Er... I mean Rex. What causes the mortar to fall nose first?
[11:16] Kitto Flora: It you want fancy action - use a VEHICLE
[11:16] Arawn Spitteler: It might be, that forces only effect the root prim
[11:16] Andrew Linden: Without the atmospheric drag effects an object should fall straight down -- no deflection.
[11:16] Rex Cronon: i can't answer that question. i don't see your server code for havok
[11:17] Andrew Linden: No matter where its center of mass is located.
[11:17] Rex Cronon: only thing i know is that it falls tail first, no matter where i put the root prim
[11:17] Andrew Linden: Rex, you're saying it falls nose first in H4 but not H7? or the other way around? or something else?
[11:17] Andrew Linden: Oh... tail first?
[11:17] Rex Cronon: haven't tried it in h7
[11:18] Rex Cronon: always tail first
[11:18] Kitto Flora: Rex: do you rez the bomb from Inv?
[11:18] xstorm Radek: H4 and H7 have problems that look the same but they are not the same
[11:18] Rex Cronon: i have a gun that fires it
[11:19] Cummere Mayo: sorry this new viewr rezzes very very slowly
[11:19] Kitto Flora: Rex: do you rez the bomb from Inv?
[11:19] Rex Cronon: i don't rezz it. my gun fires/rezes it
[11:19] Arawn Spitteler: It would rez from the gon's inventory, where it might be tail down
[11:19] xstorm Radek: and we are back to rezzers yet again
[11:20] Techwolf Lupindo: heh, the object has vector data when rezed. YOu need to chage the oject in that gun to one that has no motion to it.
[11:20] Rex Cronon: it needs to move
[11:20] Kitto Flora: OK - your gun rezzez it from Inv: One of the rez parameters is the rotation of the object you are firing. You must calculate that rotation so the bom appears at the correct angle. If you wish the bomb to ratate during its trajectory you are ggoing to have to script it. Look at Arrow scripts.
[11:21] Techwolf Lupindo: Rex, no motion when rezzed without a script adding motion to it.
[11:21] xstorm Radek: the prims are holding set location data when placed in to the rezzers
[11:21] Rex Cronon: it t is rezzed at correct angle
[11:21] Techwolf Lupindo: I have a pie in inventory that when I rez it, will bounce and fly off due to motion data is stored in the object data in the assest server.
[11:22] xstorm Radek: the prims even with no script in them has a small amount of set location data in the prims
[11:22] Arawn Spitteler: I seem to recall, for physical objects and for vehicles, there's ways of angling objects.
[11:22] Rex Cronon: andrew, simon . cand i fire one?
[11:22] Simon Linden: As long as you promise not to crash the sim :)
[11:22] Rex Cronon: is not scripted
[11:22] xstorm Radek: no Simon im the sim crasher
[11:23] Arawn Spitteler: If it crashes the sim, I want Lindens to see.
[11:23] Cummere Mayo: o.O
[11:23] Simon Linden: You can aim at my neighbors, I guess...
[11:23] xstorm Radek: shoot cogs
[11:23] Arawn Spitteler: Drop it on Bronson
[11:23] Rex Cronon: sorry. this one has a rotation script in it
[11:23] Andrew Linden: Rex, when I was designing the vehicle system I had in mind such mortars, and figured people would use the various deflection coefficients to make such bombs or darts orient to point in the direction they are moving
[11:24] Liisa Runo: people are using those
[11:24] xstorm Radek: you know simon this sim can crash and it will not be to hard to do
[11:24] Andrew Linden: Rex, the rotation script appears to give the mortar rotation about its axis of symmetry
[11:24] Simon Linden: I'm sure it can, xstorm
[11:24] Rex Cronon: ok. i guess i have to make it a vehicle
[11:24] xstorm Radek: i see 3 faults with it
[11:25] Rex Cronon: it was a early attempt at making it point to the ground:)
[11:25] Andrew Linden: oh nevermind... the angular impulse is in world-coordinates rather than local
[11:25] Rex Cronon: i thought there was something i was missing when it comes to havok
[11:25] Andrew Linden: who knows what that would do ... depends on the object's world oreientation
[11:26] Rex Cronon: i thought it wouldn't need to script it in oder to make it fall raight
[11:26] Arawn Spitteler: So, it would have to be multiplied by the object's orientation, or the gun's, but this isn't commutative.
[11:26] Andrew Linden: Ok, so Kitto. you've noticed worse collisions with perpendicular faces in H7?
[11:26] Rex Cronon: ok. extra work involved
[11:27] Kitto Flora: Andrew: Yes
[11:27] Andrew Linden: Those collision problems... are they associated with an object that has some Vehicle LSL code? or non-vehicle objects?
[11:27] Second Life Stats Cube: I display SL statistics on top & also output to TAG (L-Tek)
[11:27] Second Life Stats Cube: You can get a FREE WIDGET version of me for your web site. Goto:
 http://l-tek.vistalogic.co.uk/resources.htm
[11:27] Second Life Stats Cube: (25163a3f-e49a-21a4-ea8c-34acdcc75e49) I also output to TAG (L-Tek). Take a FREE copy of me if you wish.
[11:27] Rex Cronon: thanks. for making itclear to me what i have to do:)
[11:27] Kitto Flora: In H7 at least one of my newer train carriages is stopped dead by fins on the track. This does not happen in H4.
[11:28] Kitto Flora: Andrew: They have VEHICLE code
[11:28] Simon Linden: Hmm, I wonder if you're getting different shapes in the physics engine
[11:28] Arawn Spitteler: These are virtual fins, added to blocks, by H7?
[11:28] Andrew Linden: Ok, that makes sense -- I'd expect the vehicle code to break before the physics engine collisions in general.
[11:28] Kitto Flora: Arawan thats what it appears to be - see the model on the floor
[11:28] xstorm Radek: Kitto my old trains fly off the track and off sim do to that
[11:29] Andrew Linden: No, the "fins" are only appearant in the effect from sliding across seams of objects -- that isn't the real collision shape of the object.
[11:29] Kitto Flora: Other than this problem - and the touch-passing bug, my vehicles in H7 are working fine
[11:29] Andrew Linden: However, Havok has a problem with picking up perpendicular collision normals form edges.
[11:29] Sindra Aquila: has someone looked at the solution of antibounce found out by some weaponmakers for this maybe same problem?
[11:29] Arawn Spitteler: Might not be shape, but the need to compute the end of an object.
[11:30] office hours is half over
[11:30] Sindra Aquila: there is a secret soulution for anti bounce
[11:31] xstorm Radek: thats more like the bullet hit data is coming in to fast so the server will not look at some hits
[11:31] Arawn Spitteler: The secret solutions for SVC-93 left it vital to creative content.
[11:31] ROTATION and llSetRot incorrectly implemented for child prims
[11:31] xstorm Radek: dropping data
[11:31] Arawn Spitteler extends a hand to svc-22, just to keep his hand in
[11:31] Vehicles crossing region borders aren't always treated as vehicles and can get incorrectly returned if the destination parcel is no-entry or parcel-full
[11:32] Andrew Linden: Kitto, give me a copy of your newer train and a track for it.
[11:32] xstorm Radek: child prim is broken
[11:32] Andrew Linden: I'll have to try to reproduce the problem on H7.
[11:32] Kitto Flora: Andrew: Here or in Aditi?
[11:33] Andrew Linden: Kitto, give it to me here. I can always import it to aditi.
[11:33] xstorm Radek: Andrew use a sim boarder for your test
[11:33] xstorm Radek: and the sim needs trees and other forms of data in the sim
[11:34] Liisa Runo: trees! ❤
[11:34] Kitto Flora: OK Andrew, I'll make you a track that should show the problem and get an engie and carriage for you
[11:34] Kitto Flora: Will drop it later.
[11:34] Andrew Linden: Ok
[11:35] Bronson Blackadder: has anyone made a chain link yet and tested it on aditi?
[11:35] zeusdinne Baroque: hellooooo evryone :)))
[11:35] Rex Cronon: hi
[11:35] Kitto Flora: Someone did in my region - Mounds I think
[11:35] xstorm Radek: Hi Welcome to second life if you need help please let us know.
[11:36] Kitto Flora: 3 links were still hanging last I looked
[11:36] Bronson Blackadder: that would be one hell of a test
[11:36] Bronson Blackadder: its all convex surfaces
[11:36] Bronson Blackadder: concave
[11:36] Bronson Blackadder: excuse me
[11:36] Kitto Flora: He was using holloed boxes for the links
[11:37] xstorm Radek: holloed ? thats not good
[11:37] Kitto Flora: iTS A TEST
[11:37] Bronson Blackadder: lol
[11:37] xstorm Radek: oh ok
[11:37] Kitto Flora: Other option is Torus
[11:37] Bronson Blackadder: then shoot at them with triangle bullets
[11:38] Kitto Flora: Poor test is welded boxes
[11:38] xstorm Radek: yes
[11:38] Kitto Flora: I didnt hear any gunfire :)
[11:38] Andrew Linden: Why do you call it a "poor test" Kitto?
[11:38] xstorm Radek: i can fix that
[11:38] Andrew Linden: Welded boxes should perform better than concave prims.
[11:39] Kitto Flora: If anything will work for a collision ist a plain simple box
[11:39] Andrew Linden: But yes, if you want to stress the physics engine use the concave shapes.
[11:39] Kitto Flora: Hollow in a box is harder
[11:39] xstorm Radek: they do not Andrew
[11:39] Kitto Flora: Torus is hard to model too?
[11:39] Andrew Linden: Really?
[11:40] Bronson Blackadder: more poys in a torus I think
[11:40] Bronson Blackadder: polys
[11:40] xstorm Radek: thats some of them little basic physic problems i have been seeing
[11:40] Simon Linden: Plus collision detection gets more complicated when it is a concave shape : more math to figure out what is hitting and what's not
[11:41] xstorm Radek: i have even crash a sim with a basic physics problem when poor Kitto was in the sim
[11:42] Kitto Flora: Mostly I have been looking at if exising stuff that works ok in H4 still works ok in H7
[11:42] Andrew Linden: Well, maybe I need to do some side-by-side testing on H4 and H7.
[11:43] Kitto Flora: Right - is what I have - duplicate regions
[11:43] Andrew Linden: I've been busy doing other stuff and lately.
[11:43] Bronson Blackadder: oops
[11:43] xstorm Radek: there is that error with that prim
[11:43] Cummere Mayo: sorry guys i need to go
[11:43] xstorm Radek: that will not work right
[11:44] xstorm Radek: the server no longer sees the cut center of a prim right
[11:44] Andrew Linden: Right now I've heard that some vehicles are hitting hidden faces more. And maybe welded box performance has inverted with that of concave shapes.
[11:44] Andrew Linden: What is the "error with that prim" you're talking about xstorm?
[11:45] Kitto Flora: Wow, that was laggy
[11:45] zeusdinne Baroque giggles
[11:45] xstorm Radek: that prim and shaft and hole is seen wrong by the server
[11:46] zeusdinne Baroque takes a antilag spray
[11:46] Simon Linden: Have you tried this on a larger scale?
[11:46] Andrew Linden: Oh I see, this pole through hollow boxes?"
[11:46] Bronson Blackadder: yep
[11:46] xstorm Radek: yes
[11:46] Simon Linden: I wonder if you'd get more detail or different physics shapes if it was bigger
[11:46] xstorm Radek: many people have
[11:47] Bronson Blackadder: I just thought to try it now
[11:47] Kitto Flora: Whats wrong with these hollow boxes? That they dont slide?
[11:47] Bronson Blackadder: is oatmeal open to all on aditi right?
[11:47] xstorm Radek: the gap from one prim to the other is now greater
[11:48] Bronson Blackadder: plus these do slide
[11:48] Bronson Blackadder: dont
[11:48] xstorm Radek: and the way the havok 7 sees it and use the physics is not right
[11:49] Bronson Blackadder: if you whack it hard enough
[11:49] Bronson Blackadder: the prim will break through
[11:49] Arawn Spitteler: Who's to be e-mailed, if Aditi still thinks you're blocked for non-payment?
[11:49] xstorm Radek: you do not even have to hit it hard
[11:49] Kitto Flora: Woah
[11:49] Kitto Flora: evil!
[11:49] Andrew Linden: Arawn, talk to Oskar Linden.
[11:50] xstorm Radek: the point on the shaft is seen weeker the longe the shaft becomes
[11:51] xstorm Radek: more like a old bar magnet
[11:51] Bronson Blackadder: yeah tahts almost what it looks like
[11:52] xstorm Radek: the physics we are seeing is like physics from a magnetic field from a bar magnet
[11:53] xstorm Radek: and the prims are holding data
[11:53] Andrew Linden: Alas, that gap is a "feature" of the Havok physics engine.
[11:53] Bronson Blackadder: magnetic fields!!!!! taht explains it all
[11:53] Bronson Blackadder: lol
[11:53] Simon Linden: Is that the collision tolerance?
[11:53] xstorm Radek: but the server is starting to see it wrong
[11:54] Andrew Linden: If you were making a game you would tweak the collision geometry to be a little smaller than the visible shape.
[11:54] xstorm Radek: but the hole in the box is not seen right
[11:54] Andrew Linden: We are nearly to a point where we could tweak the collision geometry for you automatically, so that surfaces would match up
[11:54] Bronson Blackadder: that would look better
[11:55] Andrew Linden: however... due to legacy content issues, I wonder if we couold ever ship such a feature
[11:55] Simon Linden: I can imagine how many builds that would break :)
[11:55] Kitto Flora: Andrew: Is that gap adjustable in Havok?
[11:55] Bronson Blackadder: floating surfaces dont look right
[11:55] Ardy Lay: Can you make it boolean option on the object?
[11:55] Andrew Linden: also... there is the problem of small gaps in floor prims that are supposed to butt up next to each other.
[11:55] office hours is almost over
[11:55] xstorm Radek: in side the box the ends of the hole is bigger then its center
[11:55] xstorm Radek: we just can not see it
[11:56] Andrew Linden: Kitto, the gap is adjustable, but it is there for a reason and the reason relates to performance
[11:56] Andrew Linden: Havok performance degrades when objects actually interpenetrate
[11:56] Bronson Blackadder: I just increases the size of eth rod
[11:56] Simon Linden: The stack of boxes here shows it
[11:56] Andrew Linden: because of this special magical GJK algorithm they use.
[11:56] Kitto Flora: OK
[11:57] Andrew Linden: Perhaps a per-object setting for getting rid of the "collision tolerance" could be done.
[11:57] xstorm Radek: ok hold on let me build what the server is doing to that hold
[11:57] Bronson Blackadder: now the prims are moving on thier own
[11:58] Bronson Blackadder: because teh collision gap is penetrating ?
[11:58] Andrew Linden: Hrm... if you have a big enough system it might go unstable like that.
[11:59] Simon Linden: The little red box is 0.1 m, which is the gap we get between the stacked boxes
[11:59] Simon Linden: This gap is down into about the same scale, so I think we're just seeing the same effect
[11:59] xstorm Radek: ok that box i made is how the sever sees that hole
[12:00] Simon Linden: I'm sure the physics shape for the inner hole is not as smooth as it appears visually
[12:00] Thank you for coming to Linden office hours
[12:00] xstorm Radek: see the shape of the hole is bigger on its ends then the center
[12:01] Bronson Blackadder: but that looks badass
[12:01] xstorm Radek: thats what the server sees
[12:01] Bronson Blackadder: how did you get that inner shape?
[12:01] xstorm Radek: lol magic lol
[12:01] Andrew Linden: Actually the Havok terminology for the effect is "collision radius".
[12:02] Bronson Blackadder: so in effect its teh prims bounding box
[12:02] Andrew Linden: You can imagine little "contact point" data structures being created whenever two objects collide.
[12:02] Andrew Linden: The collision points act like little spheres... about 0.1 meter in diameter.
[12:02] Sindra Aquila: also on land?
[12:02] Andrew Linden: The physics engine tries to keep the objects appart by the diameter of the sphere.
[12:02] zeusdinne Baroque thinks is cool inner shape and wants to know the magic
[12:03] xstorm Radek: so in a way im right ?
[12:03] Rex Cronon: nice why for making the shape, xstorm:)
[12:03] Andrew Linden: Yes, sorta xstorm.
[12:03] Andrew Linden: Ok I've got to run. I've got way too much work to do today.
[12:03] Rex Cronon: tc andrew
[12:03] Andrew Linden: Thanks for coming everyone.
[12:03] Sindra Aquila: thanks andrew
[12:03] Bronson Blackadder: ok thanks Andrew :)
[12:03] Opensource Obscure: have a nice day andrew!
[12:03] Simon Linden: select it and see :)
[12:04] Simon Linden: Thanks everyone for coming today
[12:04] xstorm Radek: then its point are starting to make a error when you add in the memory of the prims and points of the prims
[12:04] Opensource Obscure: take care simon
[12:04] Opensource Obscure: e ciao a tutti
[12:04] Arawn Spitteler: We'd rather come to Andrew, than to someone who can't think of anything to do.
[12:04] Rex Cronon: tc simon
[12:04] zeusdinne Baroque waves her fingers farewell
[12:04] xstorm Radek: thank you every one :-)
[12:05] Ardy Lay: Buy Andrew
[12:05] xstorm Radek: i love this meeting
[12:05] Rex Cronon: tc all those that r leaving
[12:05] Bronson Blackadder: ok I picked up my trash
[12:05] Arawn Spitteler: Ardy, is he for Sale?
[12:05] Ardy Lay: Bye... oops
[12:05] Bronson Blackadder: dont want X Aring me for prim littering
[12:05] xstorm Radek: oh do i tell simon i made a black hole ?
[12:05] Bronson Blackadder: muah hahahahahahaha
[12:06] xstorm Radek: Rex why you looking at that ?
[12:06] Simon Linden: Any stuff you leave behind will get auto-returned
[12:06] xstorm Radek: ill take ;-)
[12:06] Rex Cronon: i wanted to see if i could get through it, and meet alice:)
[12:07] Rex Cronon: u know. down the rabbit hole...
[12:07] xstorm Radek: Bronson thats a funny joke you made
[12:07] Bronson Blackadder: nope
[12:07] Bronson Blackadder: you get stuck
[12:07] Simon Linden: ok, now I'm really going ... bye, and see you next time
[12:07] xstorm Radek: about the hole in that prim
[12:07] zeusdinne Baroque:
[12:07] Rex Cronon: have fun
[12:07] Sindra Aquila: bye simon
[12:07] Ardy Lay: Bye Simon

Generated with SLog Wikifier