User:Andrew Linden/Office Hours/2008 03 06

From Second Life Wiki
Jump to navigation Jump to search

Transcript of Andrew Linden's office hours:

[17:03] Vincent Nacon: wb
[17:03] Yuu Nakamichi: yup hope soft's fix will make it thru today
[17:03] Enrico Genosse: wb, sidewinder!
[17:03] Les White: hiya Andrew
[17:03] Vincent Nacon: ahh another linden, heya Simon
[17:03] Sidewinder Linden: hi everyone
[17:03] Andrew Linden: Hi, I made it
[17:03] Vincent Nacon: and Andrew
[17:03] Sidewinder Linden: ha! andrew made it
[17:03] Sidewinder Linden: hiya
[17:03] Enrico Genosse: good!
[17:03] Yuu Nakamichi: hello :)
[17:04] Les White: hiya SImon
[17:04] Simon Linden: Hello everyone
[17:04] Andrew Linden: I've got to start covering Gridmonkey for Callum around 17:45
[17:04] Andrew Linden: Shouldn't have to do anything, but just in case I go unresponsive
[17:05] Andrew Linden: that's where I'll be
[17:05] Yuu Nakamichi wonders what happened to --> http://blog.secondlife.com/2008/02/28/beta-preview-database-refresh-scheduled-for-monday-march-3-2008/
[17:06] Gaius Goodliffe: Best laid plans of mice and men...
[17:06] Andrew Linden: Sidewinder and Simon, I'll let you start with the anouncements. I'm nost sure what the current state of the QA pass on the latest update is.
[17:06] Simon Linden: I'll duck and pass to Sidewinder...
[17:06] Sidewinder Linden: ahh oops
[17:06] Enrico Genosse: hehehe
[17:06] Sidewinder Linden: yes that has been postponed - my apologies for not posting an update about that
[17:07] Sidewinder Linden: the update had to be delayed to avoid impacting qa of the current build, which was held off by the 1.19.1 release server issues....
[17:07] Sidewinder Linden: i'll update that entry
[17:07] Sidewinder Linden: andrew - is it workable to copy individual new user accounts to the beta preview?
[17:08] Sidewinder Linden: or should i just say - this will have to wait until after our next deploy early next week?
[17:08] Sidewinder Linden: if the steps aren't too involved i can take care of those on a case by case basis
[17:08] Andrew Linden: I don't know how to do it Sidewinder. There may be a tool, or someone could write one.
[17:08] Sidewinder Linden: ok
[17:08] Andrew Linden: We'll have to ask around the dev pool
[17:08] Sidewinder Linden: ok i'll check around
[17:09] Sidewinder Linden: is anyone here held off from getting into the beta preview because their account is too new?
[17:09] Vincent Nacon: not that I know of
[17:09] Yuu Nakamichi: No but I would verymuch like to have my inv refreshed if atall possible
[17:09] Sidewinder Linden: ok i'll set that up during this meeting yuu
[17:09] Vincent Nacon: shouldn't it be easier to make him older?
[17:09] Yuu Nakamichi: great thanks
[17:09] Vincent Nacon: muhahahaha!
[17:10] Andrew Linden: there is a tool for updating inventory of an existing account, already
[17:10] Sidewinder Linden: right - that is what i'm doing for yuu
[17:10] Yuu Nakamichi: awesome, thanks :)
[17:10] Sidewinder Linden: and can do for anyone else that needs beta preview inventory to match their current scond life inventory
[17:10] Arawn Spitteler wonders if there's a work-about of creating an account in Beta, but wouldn't recommend it.
[17:11] Sidewinder Linden: right... unlesss someone has a really critical need... rather avoid that
[17:11] Shirley Marquez: wasn't the beta grid re-snapshotted on Feb 28 anyway?
[17:11] Shirley Marquez: so you'd have to be REALLY young not to get in
[17:11] Sidewinder Linden: right < 2 months
[17:11] Sidewinder Linden: about
[17:11] Vincent Nacon: yeah
[17:11] Sidewinder Linden: yuu you should get an inventory refresh on next login
[17:11] Yuu Nakamichi: k
[17:12] Sidewinder Linden: well... we have a build that is in qa with many fixes... i think about 16...
[17:12] Sidewinder Linden: it is likely to deploy early to mid next week
[17:12] Vincent Nacon: so about the fixes, have you checked with prim's mass issue?
[17:12] Sidewinder Linden: andrew - did you want to talk about recent work and upcoming fixes?
[17:13] Andrew Linden: No Vincent, I haven't been able to work on that yet.
[17:13] Andrew Linden: That was the "extremely too big mass for sculpties" right?
[17:13] Les White: no, it was related to cut and prim type
[17:13] Vincent Nacon: turn out we were wrong
[17:13] Les White: a prim would snap from 1.0 mass to 1500 mass as certain cuts
[17:14] Les White: i tested a bunch of sculpts and found the mass to be pretty good on every test
[17:14] Wolfhaven Teleportation HUD v2.0 Ready
[17:14] Vincent Nacon: it seem like the glitches would happen when you have a lot of dimple, cut and hollow in use
[17:14] Andrew Linden: ah ok, but you've figured out the repro for those cuts?
[17:14] Vincent Nacon: well Les believe it's the cuts input
[17:14] Les White: yes. i passed a folder with examples. here it comes again :)
[17:15] Vincent Nacon: but I'm kinda betting on ratio meshes volume or whatever
[17:15] Andrew Linden: ok well if you've got a reliable way to reproduce it then great. Just make sure it is in in the jira
[17:15] Les White gave you Crazy MASS bug example.
[17:15] Andrew Linden: ok a folder with the objects should also work
[17:15] Vincent Nacon: cause cut doesn't seem like the only thing that can trigger it, via for hollow and dimple
[17:16] Vincent Nacon: so it's making me think it's the ratio from a full prim to a cut prim in volume or whatever
[17:16] Vincent Nacon: and sizes doesn't matter of course
[17:16] Andrew Linden: the new "mass properties" code that we're using relies on the faces of the triangles to NOT invert such that the inside surface is exposed on the outside
[17:17] Andrew Linden: it seems to me that if that were to happen we might have some mass computation problems
[17:17] Vincent Nacon: yeah
[17:17] Andrew Linden: that's my first theory, anyway
[17:17] Andrew Linden: but we'll check it out
[17:17] Rex Cronon: hello everybody
[17:17] Les White: that sounds possible with these cut tube shapes
[17:17] Andrew Linden: cut tubes... *cry*
[17:17] Vincent Nacon: it affects on sphere, torus, tube, and ring
[17:17] Andrew Linden: ok
[17:18] Vincent Nacon: maybe cube.... I think
[17:18] Vincent Nacon: well... that is if you added dimple to sphere and switch it to cube, maybe
[17:18] Enrico Genosse: I'd like to clarify, there was talk Tuesday of a mass problem with multiple linked prims, that it was fixed and awaiting QA. Is that correct?
[17:19] Andrew Linden: more context please Enrico?
[17:19] Sidewinder Linden: https://jira.secondlife.com/browse/SVC-1680 yes enrico - this one is in qa
[17:19] Sidewinder Linden: and passed i believe
[17:19] Simon Linden: Yes, that fix is in the next version
[17:19] Andrew Linden: ah thanks Simon
[17:19] Sidewinder Linden: mass not working until compile right?
[17:19] Enrico Genosse: right, thanks. (whew!)
[17:19] Sidewinder Linden: i had hoped to have this all ready to go by this meeting, but got sidetracked... if you'd like i can summarize the fixes that are already set.
[17:20] Kitto Flora: Pse do
[17:20] Aybabtu Aabye: yes please
[17:20] Vincent Nacon: sidetracked? you do live up to your name afterall
[17:20] Sidewinder Linden: actually
[17:20] Enrico Genosse: ha!
[17:20] Sidewinder Linden: hehe vincent
[17:20] Sidewinder Linden: *rolls eyes*
[17:20] Sidewinder Linden: ok there's this one
[17:20] Sidewinder Linden: SVC-1700: Havok 4: llTargetOmega() in child prims of physical link sets behaves differently
[17:21] Vincent Nacon: locally or global?
[17:21] Vincent Nacon: or just different issue?
[17:21] Sidewinder Linden: now you get to see me wait for jira ;)
[17:22] Vincent Nacon: yeah... was wondering why Wiki an Jira gotta be slow...
[17:22] Sidewinder Linden: oh i think that was where a child prim would stop rotating when the link set goes physical...
[17:22] Andrew Linden offers $1k bounty for someone who can break SVC-1700 or some other llTargetOmega() update problem... so we can threaten Kelly Linden with it.
[17:22] Vincent Nacon: oh... hmm sounds like it's a normal thing in Havok 1
[17:22] Sidewinder Linden: one to fix a problem with objects above 768 going off world
[17:22] Les White: hehe
[17:22] Sidewinder Linden: wow that's evil andrew :)
[17:22] Gaius Goodliffe's Pandora II w/TCS: Welcome aboard, Gaius Goodliffe. You are now the pilot.
[17:23] Andrew Linden: Kelly has unfortunately been assigned fixing llTargetOmega() problems recently.
[17:23] Sidewinder Linden: i should offer a bounty for 10 more vehicle bugs ;)
[17:23] Rex Cronon: USD?
[17:23] Les White: yeah. seems to be a problem item
[17:23] Andrew Linden: oh... L$
[17:23] Les White: any other juicy ones?
[17:23] Sidewinder Linden: fixed the sit targeting differences - or at least got them much closer
[17:23] Gaius Goodliffe: okay that's weird
[17:23] Sidewinder Linden: there was a bug where prim.docker would not auto-align prim edges corretly - another "small gap" problem
[17:24] Vincent Nacon: well about the target omega.... can't we just change it now? let it branch out little more with physic states?
[17:24] Pandora II w/TCS: Have a nice day, Gaius Goodliffe!
[17:24] Sidewinder Linden: a further adjustment for huggers
[17:24] Vincent Nacon: so people like me can use it for cars
[17:24] Andrew Linden: actually, I'd probably try to fix TargetOmega myself if I could fit it in, but I'd be sure to dangle the jira item at kelly for fun
[17:24] Sidewinder Linden: vincent... i think if we ask kelly to add things to targetomega *he* will go off-world :)
[17:24] Vincent Nacon: oh pfft, I'd say fire him
[17:24] Vincent Nacon: let me take over
[17:25] Sidewinder Linden: <- sets vincent for a rolling peronal restart ;)
[17:25] Les White: nice
[17:25] Sidewinder Linden: push weapons had been nerfed.. and now work
[17:25] Les White: clear his cache while you are at it
[17:25] Andrew Linden: work better
[17:25] Sidewinder Linden: although they are now going to respect the push limits
[17:25] Sidewinder Linden: right - work better... don't want to claim 100% on that
[17:25] Andrew Linden: blitz modes should still be supressed
[17:25] Vincent Nacon: yeah and it's about time too :)
[17:26] Arawn Spitteler wonders if Vence is careful what he wishes for.
[17:26] Sidewinder Linden: ok this next one...
[17:26] Sidewinder Linden: either is known as "the bunny problem"
[17:26] Sidewinder Linden: or
[17:26] Sidewinder Linden: SVC-1542: Objects with lots of pending shape changes get physically stuck
[17:26] Yuu Nakamichi: ;p
[17:26] Shirley Marquez: pending shape changes? How does that happen?
[17:26] Rex Cronon: what will happen if somebody finds a way to bring blitz to life, after h4 is realeased on the entire grid?
[17:26] Vincent Nacon: such as interpoling objects?
[17:27] Sidewinder Linden: llMessageLinked(LINK_SET, 0, "pose : eat1", NULL_KEY); - you can see the bug report for the script...
[17:27] Aybabtu Aabye: does shape changes include modifying a linked prims position relative to the root prim?
[17:27] Andrew Linden: when an object changes shape (child prim movement for example) we now push it onto a queue to have its shape changed... should complete faster than you can blink, but in certain cases it may be descheduled for a little while
[17:27] Sidewinder Linden: if you queue up enough things to change fast enoguh
[17:28] Sidewinder Linden: enogh
[17:28] Sidewinder Linden: enough sheesh
[17:28] Andrew Linden: yes Aybabtu, also scale changes and PrimParams changes
[17:28] Sidewinder Linden: and this one.. which caused some interesting problems...
[17:28] Sidewinder Linden: SVC-1688: Simulator not recognizing various shape changes in child prims until change made to root object
[17:28] Shirley Marquez: llSetPrimParams() has a delay, which you would think would keep that from being a problem, but I guess if you have multiple scripts in the link set all doing it...
[17:28] Vincent Nacon: hmm other thing to keep in mind that using setlinkprim param on agent
[17:29] Sidewinder Linden: iirc it was a tight loop with only shape cahnges in it... but i might not be remembering right
[17:29] Vincent Nacon: or is that not limited to pos/rot setting?
[17:29] Aybabtu Aabye: not sure if this is related, sometimes in h1 when a ground vehicle is moving, and a child prim that touches the ground moves downwards, sometimes it will get stuck sometimes just slow oddly, sometimes nothing, is this what youre talking about?
[17:30] Sidewinder Linden: next up is one that got umm "a lot of viewer comments"
[17:30] Sidewinder Linden: oops aybabtu?
[17:30] Andrew Linden: yes Shirley. that shape-change queue was in direct response to some piles of grief objects
[17:31] Andrew Linden: Aybabtu, that Havok1 slowdown sounds like a mini "deep think" -- bascially heavey lag but not so heavy as to kill the simulator
[17:31] Sidewinder Linden: a *lot* of folks were talking about avatars slipping down slopes... that resolution is also in the new build...thanks to andrew for a creative solution :)
[17:31] Vincent Nacon: maybe assign bounding box via LSL?
[17:31] Les White: magnetic shoes
[17:31] Sidewinder Linden: basically hehe
[17:31] Sidewinder Linden: temp prims were counting against parcel limits, and that counting is now done correctly
[17:32] Rex Cronon: where can one buy mangnets?
[17:32] Sidewinder Linden: we're almost through the list...
[17:32] Vincent Nacon: naa no magnet shoes... just need a "falling down the hill" animation
[17:32] Les White: push on!
[17:32] Sidewinder Linden: lots of fixes this time
[17:32] Sidewinder Linden: llSetStatus(STATUS_PHYSICS, FALSE) on fast moving object causes discrepancy between the visible and collision locations
[17:32] Andrew Linden: I don't think they knew about that one. I discovered that and fixed it.
[17:33] Sidewinder Linden: heh oops
[17:33] Sidewinder Linden: right internal
[17:33] Les White: i didnt :)
[17:33] Les White: are you considering this build as a grid wide RC?
[17:33] Gaius Goodliffe eeps.
[17:33] Sidewinder Linden: this particular build will not be deployed to all of second life
[17:34] Sidewinder Linden: but we are working towards that goal... and believe that we are getting reasonably close... so if you know of someone or something
[17:34] Sidewinder Linden: that has not been tested
[17:34] Sidewinder Linden: i would recommend that umm
[17:34] Sidewinder Linden: TODAY would not be too early ;)
[17:34] Vincent Nacon: patient is all needed
[17:34] Les White: yeah. better get on the bus soon :)
[17:34] Sidewinder Linden: SVC-1307: PRIM_CAST_SHAOWS no longer a recognized paramater in Havok4
[17:34] Andrew Linden: uh... we think we're almost good enough to ship
[17:34] Sidewinder Linden: this one is basically a fix for a removal of a parameter... it was breaking some scripts
[17:34] Vincent Nacon: what about Havok + Mono?
[17:35] Sidewinder Linden: so tit'sthere with a zero return iirc
[17:35] Vincent Nacon: shouldn't it be a good idea to ship both of them at the same time?
[17:35] Aybabtu Aabye: are we thinking a week or so?
[17:35] Kitto Flora: What about the carriage tripping on the end/joint of prim track? Any fix for that to be deployed?
[17:35] Les White: no way vince
[17:35] Sidewinder Linden: aybabtu - i'm not talking about dates yet... it will not be going grid wide in a week
[17:35] Vincent Nacon: I mean for SL wide grid
[17:35] Andrew Linden: kitto, I've improved that in the next update
[17:35] Kitto Flora: ty Andrew
[17:35] Sidewinder Linden: we are talking about not that many more early adopter releases though... likely less than a handful
[17:35] Andrew Linden: however, I've discovered a mode where that can break for vehicles that change the rotation of their wheels constantly
[17:35] Sidewinder Linden: i will publish more information soon on the schedule once we clear up a few things
[17:36] Andrew Linden: and I know how to fix it, but haven't yet done it (just discovered that one)
[17:36] Aybabtu Aabye: is it still possible to convert a sim to early adopter?
[17:36] Les White: i've been waiting on a request to be added at mooz for a while now o-o
[17:36] Sidewinder Linden: yes... concierge is doing those conversions
[17:36] Sidewinder Linden: hm i think they've been pretty busy of late...
[17:36] Vincent Nacon: told ya, les
[17:36] Les White: yeah. they dont want to stop playing with 1.19 :)
[17:36] Vincent Nacon: wait in the line
[17:36] Sidewinder Linden: and i know today we ran out of havok4 spares, so the have to set up some more racks
[17:37] Sidewinder Linden: so a couple of other things is that simon has been working on a few crash modes that need to go away, and other internals
[17:37] Sidewinder Linden: and andrew has done work on vehicles (again)
[17:37] Kitto Flora's PC reset button is wearing out.
[17:37] Sidewinder Linden: so this release has a *lot* of cleanup in it
[17:37] Les White: it will be interesting to see the difference in crash rates from the havok4 sims
[17:37] Aybabtu Aabye: much appreciated
[17:38] Sidewinder Linden: i have some numbers from the very first early adopter deploy
[17:38] Les White: it must be lower by a lot
[17:38] Sidewinder Linden: for the week before and after that, it looks like havok4 simulators reduced the crash rate by well over 60% - depending on the particulars up to 90% or more
[17:38] Shirley Marquez: wow -- that's huge
[17:39] Sidewinder Linden: that stats run did not filter out some regions taht were being used for "target practice" so i'd expect the real numbers to be better
[17:39] Les White: yeah. this is with people trying to kill them :)
[17:39] Sidewinder Linden: i'm working on getting some more recent stats for the early adopters - because that was only the first set of about 175
[17:39] Sidewinder Linden: in a recent office hours here i asked "how often do you see h4 crashing" and no onein office hours had seen one crash in over a week
[17:39] Sidewinder Linden: which makes me suspect that the numbers are going to be very good
[17:39] Enrico Genosse: hear hear!!!
[17:39] Sidewinder Linden: BUT of course there is a caveat
[17:39] Sidewinder Linden: ;)
[17:40] Les White: to be honest i have a sim running it for a month or more and i haven't crashed it a single time
[17:40] Vincent Nacon: yeah I'm sure it's very stable
[17:40] Sidewinder Linden: there are two kinds of crashes
[17:40] Vincent Nacon: even on a Class 4 sim
[17:40] Jcool410 Wildcat: o.o
[17:40] Sidewinder Linden: there are what we're calling "incidental crashes" - the ones that happen when an animal hits a wall, or jumping dolphins do all the things they aren't supposed to do, but do :p
[17:40] Sidewinder Linden: or when a teleport hits a weird spot
[17:40] Vincent Nacon: animal?
[17:40] Enrico Genosse: ahem.
[17:41] Sidewinder Linden: or two things are interpenetrated accidentally the wrong way
[17:41] Sidewinder Linden: yours don't do that much enrico
[17:41] Gaius Goodliffe: Bloody dolphins. Never trust an animal that grins all the time. They're hiding something...
[17:41] Sidewinder Linden: i think andrew said that in one grid monkey tour he had to return waht was it 25 sets of dolphins?
[17:41] WarKirby Magojiro: w/me waves
[17:41] Sidewinder Linden: in any case - from what we can tell those types of crashes, that are accidental, should be almost entirely gone with havok4
[17:42] Andrew Linden: Yeah "Jumping Dolphins". I forget who the creator was, but I told him how to make a less laggy model.
[17:42] Sidewinder Linden: here's the other side...
[17:42] Sidewinder Linden: as some of you may know
[17:42] Sidewinder Linden: it is also quite possible to crash the current havok1-based simulator on purpose
[17:42] Les White: very easy
[17:42] Sidewinder Linden: and it doesn't take much work if you konw how
[17:42] Vincent Nacon: must be target omega dolphins
[17:42] Sidewinder Linden: we have removed *many* crash modes fromt eh havok4 simulator, but
[17:42] Sidewinder Linden: there are still ways to do it
[17:42] Shirley Marquez plugs her ears
[17:43] Sidewinder Linden: we have a few
[17:43] Sidewinder Linden: small number but they're there
[17:43] Sidewinder Linden: known ways to kill a havok4 simulator
[17:43] Aybabtu Aabye: should we be trying the ways we know work on h1, i feel icky intentionally crashing a sim
[17:43] Sidewinder Linden: we are working on plugging those holes and think that we will have some effective strategies
[17:43] Andrew Linden: there are at least two that are actually crashes in Havok4 (as opposed to a crash in the script engine or something)
[17:43] Rex Cronon: using physics only?
[17:43] Sidewinder Linden: all known h1-style crashes no longer work on h4
[17:44] Sidewinder Linden: so the good news is two -fold
[17:44] Sidewinder Linden: 1) these crashes tend to leave VERY big fingerprints
[17:44] Andrew Linden: er... H1 physics crashes have been fixed
[17:44] WarKirby Watanabe: I'm stilll there...
[17:44] Sidewinder Linden: ppl doing them will be identified in almost all cases pretty easily
[17:44] Sidewinder Linden: 2) we think we have some strategies for making them go away
[17:44] Sidewinder Linden: however
[17:45] Andrew Linden: Yeah, it takes a little bit more work, and significantly more objects, to crash Havok4 in the physics engine.
[17:45] Les White: people will always find away to grief if that's their please, as long as all the random crashing from impacts and interpolation are gone then it's much improved
[17:45] Sidewinder Linden: we've decided that we will release the new simulator with a couple of these present, to give everyone relief from the huge number of incidental crashes, and will continue to clean up these other modes shortly after release
[17:45] Enrico Genosse: I concur with that plan.
[17:45] Sidewinder Linden: right the "random interpenetration crash" is gone
[17:45] Sidewinder Linden: and frankly
[17:45] Vincent Nacon: to make them go away, is to become one of them, hate linden with passion and work from there
[17:45] Gaius Goodliffe: I'm less worried about intentional crashers than the sim coming down every time there's a mid-air collision or a six-car pileup or the like...
[17:46] Les White: right
[17:46] Sidewinder Linden: many folks have handed me "sim crashers" in the last couple of weeks that are really more like lag bombs than crashers... none of them has actaully killed a simulator recently
[17:46] Sidewinder Linden: right gaius
[17:46] Sidewinder Linden: so that's where we are with crash rates
[17:46] Vincent Nacon: just need more physic state changes via LSL to ensure
[17:46] Sidewinder Linden: i would like to find a way to measure incidental crashes alone for crash rates, but am pretty sure that won't be very easy
[17:47] Sidewinder Linden: oh andrew and simon - do you remember where we are with those elevator problems?
[17:47] Sidewinder Linden: i got asked again today and couldn't remember where the are in the queue
[17:47] Sidewinder Linden: no pun intended
[17:47] Andrew Linden: elevator problems? Avatar falling inside an elevator?
[17:47] Vincent Nacon: but that's one thing about Havok with Mono... gotta make sure physic can with stands an object influced by mono's speed
[17:47] Simon Linden: I think they dropped in priority ;)
[17:47] Les White: that projects in the basement
[17:48] Sidewinder Linden: yes vincent - but that's after we get havok4 released.. then the mono integration issue will be cleaned up
[17:48] Andrew Linden: I had "fixed" some elevator problems, but Dan Linden re-opened the bug. I haven't been able to get back to it yet.
[17:48] Sidewinder Linden: hmm ok simon
[17:48] Sidewinder Linden: ran into a few regions taht use them today - i guess we'll see after this next deploy what ppl think
[17:48] Yuu Nakamichi: are there any more known problems with flight assists at this time?
[17:48] Vincent Nacon: yeah, but then you may ended up having to redo havok side than mono side
[17:48] Vincent Nacon: I dunno what in a example but just thinking out lou
[17:48] WarKirby Watanabe: is the flight limit going to be staying?
[17:48] Vincent Nacon: loud*
[17:49] Sidewinder Linden: /vincent - both teams konw that we'll be passing things back and forth when we get there
[17:49] Sidewinder Linden: warkirby?
[17:49] Sidewinder Linden: speed or altitude?
[17:49] WarKirby Watanabe: altitude
[17:49] Andrew Linden: I'm not too worried about the elevator problems. I think I know how to make them better, or as good as Havok1. Now that vehicles and the sliding avatar problems are fixed there aren't very many bugs that I phear.
[17:49] WarKirby Watanabe: unassisted flight fails above a variable, inconsistent distance from terrain
[17:49] Kitto Flora: Any chance of getting SVC-141 fixed under the H4 umbrella, even though is a long-time bug in H1?
[17:49] Sidewinder Linden: what is that one?
[17:50] Vincent Nacon: oh about the elevator thing.... can you make avatar not have a walking/running/flying state animation when they are only standing on a physical object?
[17:50] Kitto Flora: The Restrict Pushing option was, if I recall correctly, originally designed to allow pushing in certain circumstances even when pushing is restricted. For individual-owned land, objects owned by the land-owner were allowed to use llPushObject() on people who are on the parcel. For group-owned land, objects SET to the group and owned by an officer were allowed to use llPushObject().
[17:50] Andrew Linden: WarKirby, i thought that was fixed. I'd like a copy of that failure attachment if it still fails in the next update.
[17:50] Kitto Flora: Basically group land / group builds are being discrimitnated against
[17:50] WarKirby Watanabe: no, not talking about a bug
[17:50] WarKirby Watanabe: I mean the concept of the altitude limit in general
[17:51] WarKirby Watanabe: it will not be removed, I take it
[17:51] Andrew Linden: oh... No we don't have plans to remove it in Havok4. Perhaps after.
[17:51] Sidewinder Linden: oh warkirby are you talking about 768 vs 4096 etc?
[17:51] WarKirby Watanabe: no..., one moemnt
[17:51] WarKirby Watanabe: https://jira.secondlife.com/browse/SVC-742
[17:51] WarKirby Watanabe: I made a jira about it
[17:51] Andrew Linden: ah yes... push in group owned land. No Kitto, we can't let that delay Havok4 release.
[17:52] WarKirby Watanabe: it seems like something that would have been neat to remove for h4
[17:52] Shirley Marquez: unassisted flight fails somewhere between 200 and 250m, doesn't it?
[17:52] Andrew Linden: If we happen to find time to fix it then gravy, but we're trying to be very focused on the blockers right now.
[17:52] WarKirby Watanabe: yes, shirley
[17:52] Aybabtu Aabye: if 2 void sims were running on the same physical hardware, does anyone know if theres a noticeable lag crossing the sim border between the two
[17:53] Andrew Linden: unassisted flight starts at around 75 meters above the terrain, I think, and is almost totally gone at about 125 or 150 meters above the terrain
[17:53] Arawn Spitteler wears a Flight Feather, which probably defeats the purpose of that limit
[17:53] Aybabtu Aabye: im not sure how to find a way to test that one
[17:53] Sidewinder Linden: it depends strongly on how many scripted attachments the avatars have
[17:53] Sidewinder Linden: on
[17:53] Sidewinder Linden: and load of the simulators at the time the crossing happens
[17:53] Aybabtu Aabye: so even if they are on the same physical machine, crossing still is lagged?
[17:53] WarKirby Watanabe: I think what he's asking, is will there be a noticeable performance increase with two sims on the same hardware ?
[17:54] Sidewinder Linden: it's still a transfer from one process to the other
[17:54] Shirley Marquez: I suspect that most long-time residents are wearing some sort of flight assist, so the limit only affects the newcomers
[17:54] Sidewinder Linden: but you don't know deterministically which regions will be on which machines
[17:54] Sidewinder Linden: that mapping changes aftera restart
[17:54] Kitto Flora suspects border crossing lag is unaffected by if'n the 2 sims running on same hardware.
[17:54] Vincent Nacon: about avatar standing on a moving object, is it possible to get animation only check on local position by collision mesh instead of global with the sim?
[17:54] Aybabtu Aabye: yes could the lag be reduced for crossing sim borders, if all 4 neighboring void sims were on the same machine
[17:54] WarKirby Watanabe: maybe that's an optimisaiton that could be made
[17:54] Sidewinder Linden: if the two machines are in different data centers the lag will be longer
[17:54] WarKirby Watanabe: having a sim aware of what other sims are on it's machine at any particular time?
[17:54] Andrew Linden: in theory that is possible, however the simulators do not currently optimize for that situation
[17:54] Sidewinder Linden: aybabtu - the grid purposefully does not "latch" regions to specific hosts
[17:55] Sidewinder Linden: at each restart they hunt for an available match of class and version
[17:55] Sidewinder Linden: so you should not assume anything about host-attachment of regions in your work
[17:55] Aybabtu Aabye: could a block of 4 be tied together in such a way that when one goes down all 4 go down, and they all come up on the same machine, whatever machine that might be?
[17:55] Sidewinder Linden: and actually the other reason to not make this assumption is that there are blocks of larger than four regions and those will by necessity be on more than oe host
[17:55] Aybabtu Aabye: so that sim crossing would be less unpleasant
[17:56] Sidewinder Linden: aybabtu we do not do that for resident regions
[17:56] Andrew Linden: Aybabtu, I don't see much reason to tie regions together like that
[17:56] Sidewinder Linden: you are just transferring the issue to different region boundaries
[17:56] Aybabtu Aabye: yes
[17:56] Shirley Marquez: Aybabtu's idea might make sense for places that are meant to be closely tied, like CSI or the Life 2.0 regions
[17:56] Aybabtu Aabye: that way you could plan a bigger region
[17:56] Sidewinder Linden: but again it would ONLY work for a quad of regions
[17:56] Aybabtu Aabye: yes
[17:56] Les White: a better solution would be larger sims...what is the reason to limit a sim to 256? (besides the whole silly land market stuff)
[17:56] Sidewinder Linden: any crossing outside that quad will be normal
[17:56] Yuu Nakamichi: how big an impact would this have ?
[17:57] Les White: couldnt you dial in a sim to 512 on the same amount of bandwidth and CPU alocation?
[17:57] Sidewinder Linden: lots of internal issues and tools and various other architecture questions...
[17:57] Les White: just more land to rendor
[17:57] Aybabtu Aabye: that would open up some opportunities
[17:57] Sidewinder Linden: ok guys... wait... please...
[17:57] Andrew Linden: we're planning a project for overhauling how regions communicate to each other, and to the clients (Object Transport)
[17:57] Sidewinder Linden: this is all speculation and not the topic of this meeting ;)
[17:57] Gaius Goodliffe nods.
[17:57] Les White: let's talk about target omega
[17:57] Les White: kids
[17:57] Aybabtu Aabye: sorry, wasn't sure, thought it might count as SL architecture, my bad
[17:58] Les White: we want bigger sims :)
[17:58] Vincent Nacon: about avatar standing on a moving object, is it possible to get animation only check on local position by collision mesh instead of global with the sim?
[17:58] Sidewinder Linden: ok les...but that's not part of the havok4 project... so it's another discussion :)
[17:58] Aybabtu Aabye: yes to what les said
[17:58] Sidewinder Linden: we WILL ship, and that means no new features :)
[17:58] Les White: indeed
[17:59] Enrico Genosse smiles
[17:59] WarKirby Watanabe: Is there a definite plan in the works for adding proper collisoin data to sculpted prims?
[17:59] Vincent Nacon: heheh
[17:59] Aybabtu Aabye: acceptable :)
[17:59] Andrew Linden: not definite plans for sculptie collisions Warkirby, but Qarl and I want to work on it
[17:59] Aybabtu Aabye: it doesnt affect me that much, but has there been any progress on the various vehicle motors strengths?
[17:59] Andrew Linden: that problem sorta blocks true "megaprim liberation" so hopefully we'll start on it this year
[18:00] Sidewinder Linden: oh - quick question
[18:00] Vincent Nacon: was gonna suggest that to Qarl to based on 16x16 instead of the whole 32
[18:00] Sidewinder Linden: has anyone seen problems with "fall through floor" on megaprims?
[18:00] Sidewinder Linden: at altitude?
[18:00] Vincent Nacon: rarely
[18:00] Kitto Flora: Linear motor strength appears same in H4 as in H1.
[18:00] Sidewinder Linden: have you seen it at all in the last four weeks on havok4 vincent?
[18:00] Aybabtu Aabye: not as of tuesday, kitto
[18:00] Andrew Linden: Aybabtu, the vehile strengths you saw on Tuesday will be the vehicle strengths. They should be close to Havok1
[18:00] Yuu Nakamichi: do you have a jira for this sw?
[18:00] Andrew Linden: except for the vehicle motor... it has a higher limit
[18:01] Sidewinder Linden: no... i'm asking because i saw it once, and could not reproduce it
[18:01] Sidewinder Linden: @yuu
[18:01] Andrew Linden: er... I mean the linear motor
[18:01] Yuu Nakamichi: k
[18:01] Vincent Nacon: well if youre not counting on this grid
[18:01] Yuu Nakamichi: I've heard it mentioned by others though
[18:01] Aybabtu Aabye: ok, that works, scripters will adjust to the new strengths
[18:01] Vincent Nacon: but.... not sure if I have fell through recently on this grid updates
[18:01] Vincent Nacon: so maybe not
[18:01] Gaius Goodliffe: I know you sink a bit on curved ones, but not fall through entirely.
[18:02] Sidewinder Linden: ok thanks
[18:02] Kitto Flora: Anyone here capable of making custom megaprims? I need one such.
[18:02] Vincent Nacon: don't look at me
[18:02] Sidewinder Linden: i suspect there's still an issue there... but again have only seen one case once
[18:02] Gaius Goodliffe: i believed they fixed the ability to make them.
[18:03] Enrico Genosse: thanks, sidewinder, good meeting!
[18:03] Sidewinder Linden: yup - the tools were "fixed"
[18:03] Sidewinder Linden: thanks enrico....
[18:03] Shirley Marquez: thank you, Sidewinder
[18:03] Sidewinder Linden: so i have one question though before we break
[18:03] Sidewinder Linden: which are the "key items that we haven't fixed yet" that if we ship without them fixed y'all gonna really yell about
[18:04] WarKirby Watanabe: You can shape them to some degree with prim torture
[18:04] Vincent Nacon: hmm...
[18:04] Sidewinder Linden: understanding that we will continue to do cleanup after the grid-wide release for some number of weeks and a few release cycles
[18:04] Gaius Goodliffe: Buoyancy
[18:04] Enrico Genosse: I'd say the performance problem I see when multiple dogs move.
[18:04] Vincent Nacon: for not having force/motor param yet
[18:04] Gaius Goodliffe: And flight assists maintaining altitute. But mostly buoyancy.
[18:04] Les White: hard to believe that i personaly dont have a showstopper
[18:04] Kitto Flora: CPU load of physics engine. Class 4 sim residents with multiple vehicles are gonna bitch.
[18:04] Sidewinder Linden: gaius - isn't that fixed now?
[18:04] Gaius Goodliffe: Not on the last version I've had access to. Maybe what's in QA?
[18:05] Sidewinder Linden: hmm
[18:05] Enrico Genosse: Kitto, I wonder we we have the same problem.
[18:05] Les White: class 4 with multi vehicles doesnt have a picnic with h1 sims now...
[18:05] Vincent Nacon: but yeah, nothing yet, maybe the avatar's animation on a moving object
[18:05] Kitto Flora: Enrico - are the dogs vehicles?
[18:05] Sidewinder Linden: ok so for those concerned about script performance i have a request...
[18:05] Enrico Genosse: nope.
[18:05] Andrew Linden: I've fixed buoyancy "script energy energy consumption rates", however it didn't quite make it for this next update
[18:05] Yuu Nakamichi: that's why i was asking earlier abotu it too, not sure if it had been fixed completely
[18:05] Sidewinder Linden: it would be VERY helpful if you could frame up a simple demonstrator for the calls that seem to be overly expensive with h4
[18:06] Sidewinder Linden: the team is highly loaded right now plowing through the last open issues, and if you can triangulate on the areaas that need work it will help us to get it done quickly
[18:06] Enrico Genosse: I'm working on a demo for that in Turing Isle. A couple of objects at a distance that will call the dogs to them.
[18:06] Sidewinder Linden: ummm sidewinder's version of loaded = working hard... nothing to do with pints at a bar!
[18:07] Andrew Linden: clear simple "repro recipe's" go a long way toward making bugs easy to fix
[18:07] Vincent Nacon: only think I got is avatar's animation on a moving object as you just stand on it
[18:07] Kitto Flora: Sidewinder - are you asking for a 'simple demonstrator' for the multiple vehicle loading problem?
[18:07] Vincent Nacon: thing*
[18:07] Les White: no idea what you are talking about vince
[18:07] Vincent Nacon: if you stand on a boat... not actually siting with the object
[18:07] Vincent Nacon: it triggers your animation to walk or run
[18:08] Kitto Flora: Sure.. any time
[18:08] Vincent Nacon: because the animation is based on global sim
[18:08] Andrew Linden: no kitto, if the recipe is really simple you can describe it. Some vehicles cause more load than others though... so copies of the particular affected objects is always helpful
[18:08] Kitto Flora: I also have an open SVC- in jira about the CPU load problem
[18:08] Les White: you want the SL animation state machine to detect when you are moving with something. ah.
[18:08] Vincent Nacon: yeah
[18:08] Sidewinder Linden: kitto - does that jira narrow down to which lsl functions are running slowly?
[18:08] Sidewinder Linden: i remember one about trains in general causing high load
[18:08] Andrew Linden: kitto, does that CPU load problem block Havok4 release, in your opinion?
[18:09] Vincent Nacon: so you don't have to do the "falling" animation on a elevator
[18:09] Kitto Flora: Sidwinder - No.. there is no way for us to narrow it down
[18:09] WarKirby Watanabe: it's not related to H4, is it?
[18:09] Kitto Flora: AND you need multiple vehicles for it to show up in the Stats Panel - unless you look closely.
[18:09] Sidewinder Linden: hmmm
[18:09] Andrew Linden: I think Kitto's load is showing up in physics rather than the script engine, isn't that correct Kitto?
[18:09] Kitto Flora: Yes
[18:09] WarKirby Watanabe: oh, then it is...
[18:09] Aybabtu Aabye: does it depend on the vehicles involved?
[18:10] Kitto Flora: The Physics time figure is about double now on H4 than it is in H1
[18:10] Enrico Genosse: same with me.
[18:10] Andrew Linden: Kitto, at one time you were claiming a 4X load increase. We did a few optimizations since then. Do you know what the X increase is now?
[18:10] Andrew Linden: ok 2X
[18:10] Kitto Flora: There was evidence that a few- prim vehicle has less load than a 31-prim + 25 script vehicle.
[18:11] Andrew Linden: Yes, it could very well depend on the prims used.
[18:11] Kitto Flora: Adrew - yes. It went from X4 to X2 when yoy did optimizations
[18:11] Enrico Genosse: The dogs are all physical, 31 prims and use llApplyImpulse() to move. Several of them moving create noticeable time dilation.
[18:12] Gaius Goodliffe: Are the Havok4 beta servers being compiled with all the usual optimizations, or does it have lower optimization and/or debug code in it?
[18:12] Andrew Linden: Enrico, how many concave parts on the dogs?
[18:12] Kitto Flora: Erico, Stats PAnel - expand Time (mw) - look at Sim Time (Physics) compare effect in H1 and H4 sims
[18:12] Enrico Genosse: very few, mostly sculpties or torus.
[18:12] Enrico Genosse: will do, Kitto.
[18:13] Andrew Linden: The release simulators are compiled with mild compile optimizations. We were talkinga about algorithm optimizations.
[18:13] Sidewinder Linden: enrico - could you check in with me afterwards so that we can set up a time to take a look at this?
[18:13] Andrew Linden: er... "release simulator" means compiled for deployment -- no debug libs
[18:13] Gaius Goodliffe nods.
[18:13] Kitto Flora: Nuther thought on that - is H4 sim code being compile optimized for the Class4 Hardware?
[18:13] Vincent Nacon: shouldn't it be easier to set your own bounding box via lsl?
[18:13] Sidewinder Linden: the dogs have somewhat of a "learning curve" and it would be nice to have a demo..
[18:14] Enrico Genosse: yes, I plan to have an easily repeatable demo set up. I'll let you know, Sidewinder, probably late tomorrow.
[18:14] Andrew Linden: what I would like to do is to provide a hint where the content creator could specify that their multi-prim object be replaced by a single "convex hull" collision shape
[18:14] Sidewinder Linden: oh ok great - thanks
[18:14] Andrew Linden: but we won't be working on that pre-Havok4 release
[18:14] Les White: yeah andrew. that would solve a lot of issues
[18:14] Andrew Linden: Enrico, does the dog performance problem block Havok4 release, in your opinion?
[18:15] Yuu Nakamichi: yes :)
[18:15] Enrico Genosse: Tough call. Most people only have one or two dogs. Many have more than ten dogs....
[18:15] Aybabtu Aabye: single prim shell for collisions, i like that
[18:15] Enrico Genosse: they would notice a lot of lag when they go for a walk for example.
[18:16] Kitto Flora: Everyone in the sim will get the lag
[18:16] Enrico Genosse: I know if my partner Blu were here, she'd say a definite 'yes!'
[18:16] Gaius Goodliffe: Would they? Does it ever cause human noticable slowdowns, or do you need the debug panel open to notice?
[18:16] Sidewinder Linden: enrico - does this provideenough load to increase time dilation noticeably, or is it a percentage physics time increase that you're woried could crate the problem?
[18:16] Arawn Spitteler imagines a rezzer, where the griefer gets to set his dogs on you.
[18:16] Enrico Genosse: oh int's noticeable, you get that walk through mud feeling.
[18:16] Andrew Linden: btw, Havok4 release would probably be Server-1-19-2, and we'd expect to be releasing a 1-19-3 in relatively short order as problems show up
[18:16] Enrico Genosse: *it's
[18:16] Sidewinder Linden: hm ok
[18:17] Les White: and it runs clean on h1?
[18:17] Andrew Linden: on the order or 1 week, maybe 2
[18:17] Kitto Flora: In a class 4 thats was near the limit in H1 (and there LOTS of those) then switching it to H4 and running several vehicles - the increas in lag is very noticable
[18:17] Enrico Genosse: yes. but more likely to crash sim.
[18:17] Enrico Genosse: I'd say the increased lag is worth the trade for decreased crashes.
[18:17] Enrico Genosse: those crashes are a big problem with dog owners.
[18:17] Kitto Flora: Esentially - 4 'trains' wont run on a class 4 H4. Whereas they run ok on H1.
[18:17] Sidewinder Linden: this brings up a really good point
[18:18] Sidewinder Linden: that is really what many people have said - yes we want it faster
[18:18] Sidewinder Linden: but the reduced crashes are critical
[18:18] Andrew Linden: Perhaps dog kennels could be convinced to run fewer dogs for a while until the performance problems could be solved
[18:18] Enrico Genosse smiles
[18:18] Sidewinder Linden: so unless this is really really a blocker, i think the bulk of folks would prefer to have the update out, and we will continue to work with you on performance...
[18:18] Enrico Genosse: could be. :)
[18:18] WarKirby Watanabe agrees
[18:19] Andrew Linden: unfortunately, the performance problems are not the sort that I know right away how to solve. They will require research and I'm not totally confident we'll discover an easy solution
[18:19] Gaius Goodliffe: I'm sure we can live with it not being faster -- it does become a problem though if it's significantly *slower*, enough to cause problems that weren't there before.
[18:19] Enrico Genosse: I would only say, let me demonstrate and let you make the call after you see for yourselves.
[18:19] Sidewinder Linden: as i said - the team will stay on the project after we integrate with release
[18:19] WarKirby Watanabe: I had some problems a few weeks ago with a rock rezzer, that accidentally became a sim crasher thnks to the general suckiness of H!
[18:19] Sidewinder Linden: ok sounds good
[18:19] Kitto Flora: Andrew, do you think that the Havok4 CPU load *can* be decreased, eventually?
[18:19] Gaius Goodliffe: Class 6? ;)
[18:19] WarKirby Watanabe: <.<
[18:19] Sidewinder Linden: oh noooooooooo
[18:19] Vincent Nacon: there is no 6
[18:19] Enrico Genosse: 8?
[18:19] Sidewinder Linden: noooooo gaius :)
[18:20] Arawn Spitteler: Charge extra for Physics
[18:20] Gaius Goodliffe: hehee
[18:20] Andrew Linden: Yes I think so Kitto. The "simplified collision hints" feature would probably do it for most cases.
[18:20] WarKirby Watanabe: bring on the class 10, I say
[18:20] Les White: i'd pay extra!
[18:20] Les White: (maybe)
[18:20] Vincent Nacon: only Philip Linden has his own Class 6 for himself
[18:20] Andrew Linden: Also, Havok5 is out and there is some evidence that it will help for these multi-prim objects
[18:20] Kitto Flora: A Class 5 sim handle h4 and multiple trains ok.
[18:20] Vincent Nacon: with... sim crossing?
[18:20] Yuu Nakamichi: creators would be ableto set a flag right?
[18:20] Vincent Nacon: or limit set?
[18:20] Kitto Flora: Its the Class 4's loaded up thats the problem.
[18:20] Andrew Linden: finally, I've got an idea on how to do it right... but I need significant time to do it... two weeks at least
[18:20] Gaius Goodliffe: Actually, if he's like some CEO's I've met, he's got a class 1 somewhere :p
[18:21] WarKirby Watanabe: how easy will it be to upgrade to H5 after this?
[18:21] Aybabtu Aabye: is there a way to limit script load on the av level? allow no one av more than X% of the script interpreter load?
[18:21] Sidewinder Linden: hmm kitto... given the progression to class 5's i wonder how much "heavy optimization" this is worth then...?
[18:21] Aybabtu Aabye: err sim level, but limit avs
[18:21] Les White: that would make me dance in the streets ayb
[18:21] Andrew Linden: The Havok5 upgrade will be much easier... two months max, maybe 1
[18:21] WarKirby Watanabe: oh, neat
[18:21] Sidewinder Linden: *covers his eyes*
[18:21] Kitto Flora: If LL gave everyone a free Class 5 upgrade it would get rid of the problem for a long time.
[18:21] Enrico Genosse: I love hearing this.
[18:21] Sidewinder Linden: *did andrew really just say that?* :)
[18:22] WarKirby Watanabe: heh
[18:22] Yuu Nakamichi: ;p
[18:22] Andrew Linden: btw, I'm terrible at making estimates
[18:22] Simon Linden: Aybabtu - no, not now, but we've been thinking (longer term) about resource caps like that
[18:22] Gaius Goodliffe: Save the transcript!
[18:22] WarKirby Watanabe: it does sound suspiciously short
[18:22] Vincent Nacon: better kill him, Side
[18:22] Enrico Genosse: that's more like it. :)
[18:22] Sidewinder Linden: noooooo oh nooooooo i thought we got through this :)
[18:22] WarKirby Watanabe: me too
[18:22] Les White: no kidding hehe
[18:22] Aybabtu Aabye: good, thank you for considering it
[18:22] Yuu Nakamichi giggles
[18:22] Sidewinder Linden: so .. i understand that andrew's holding off a family commitment to be here... are we done so he can go out?
[18:22] Enrico Genosse: sure!
[18:22] Andrew Linden: yeah, time to go
[18:22] Vincent Nacon: let him out!
[18:22] WarKirby Watanabe nods
[18:23] Enrico Genosse: great meeting, thanks!
[18:23] Les White: ok. thanks for the time guys :)
[18:23] Kitto Flora: Byebye Andrew
[18:23] WarKirby Watanabe: #
[18:23] Shirley Marquez: bye, Andrew!
[18:23] Vincent Nacon: go go physic guru!
[18:23] WarKirby Watanabe: bye for now everyone
[18:23] Sidewinder Linden: thanks all - keep those cards and letters coming
[18:23] Yuu Nakamichi: thanks very much :)
[18:23] Aybabtu Aabye: ill save it for next time :)
[18:23] Sidewinder Linden: yw
[18:23] Aybabtu Aabye: thanks guys
[18:23] Gaius Goodliffe: Wait a minute, an IT guy with a life? :o
[18:23] Vincent Nacon: oh sure
[18:23] Vincent Nacon: see his avatar?
[18:23] Simon Linden: I've got to run as well - thanks everyone for the help testing and the feedback
[18:23] Arawn Spitteler ha a class to teach: A two month upgrade, for H5, sounds like a really good idea. It would be like the scheduled upgrade to H2
[18:23] Vincent Nacon: he look like a newbie
[18:23] Sidewinder Linden: no no no no one said that! :)
[18:23] Vincent Nacon: so yeah... he got life
[18:23] Yuu Nakamichi: thanks simon :)
[18:23] Sidewinder Linden: oh boy
[18:24] Sidewinder Linden: thanks guys.. see you around
[18:24] Vincent Nacon: later all
[18:24] Les White: see you! :)
[18:24] Shirley Marquez: the chairy-go-round is making me dizzy!
[18:24] Gaius Goodliffe: Good evening, folks