User:Andrew Linden/Office Hours/2008 01 17

From Second Life Wiki
< User:Andrew Linden‎ | Office Hours
Revision as of 14:06, 18 January 2008 by Andrew Linden (talk | contribs) (New page: Transcript of Andrew Linden's office hours: {| |- style="vertical-align:top;background-color:#FFFFFF;" | [17:02] | Kitto Flora: | style="white-space:normal;"...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Transcript of Andrew Linden's office hours:

[17:02] Kitto Flora: Do we expect Lindens?
[17:02] Gaius Goodliffe: yup :)
[17:03] Andrew Linden: Hello. I was just trying to write up a little list of things to do tomorrow before I logged in
[17:03] Kitto Flora: Its only a little list?
[17:03] Andrew Linden: 7 items so far
[17:04] Andrew Linden: However, most of those have to do with the fact that I'm about to move into the other building
[17:04] Kitto Flora: Battery?
[17:04] Andrew Linden: so I've got lots of little projects to get done before theh
[17:04] Andrew Linden: then
[17:04] Andrew Linden: I think the preview was just updated this afternoon, finally
[17:05] Kitto Flora: It all got rebooted about 2 hours ago?
[17:05] Andrew Linden: Yeah, not too long ago. Lemme check to see if it is open...
[17:05] Gaius Goodliffe: It is.
[17:05] Kitto Flora: It was open after the reboot
[17:05] Gaius Goodliffe: Logged out about five minutes ago.
[17:05] Kitto Flora: What got updated?
[17:05] Andrew Linden: The Havok4 preview
[17:06] Arawn Spitteler: Script run-time error
[17:06] Arawn Spitteler: Stack-Heap Collision
[17:06] Andrew Linden: and also any Havok4 regions on SL proper. Er... they are in progress of getting updated.
[17:06] Ryozu Yamamoto: Heya Sidewinder, Arawn
[17:06] Kitto Flora: Hllo hello
[17:06] Andrew Linden: Sidewinder knows the status of any Havok4 updates to SL proper.
[17:06] Gaius Goodliffe: Oo! Do we have H4 regions back on the main grid?
[17:07] Sidewinder Linden: hi guys sorry...
[17:07] Ryozu Yamamoto: We did until the restart today =D
[17:07] Sidewinder Linden: the existing regions that had havok4 are in the process of being restarted with the new code as we speak
[17:07] Sidewinder Linden: i think most are done
[17:07] Gaius Goodliffe: Cool.
[17:07] Sidewinder Linden: and i'm about to hit publish on the blog post (thus my tardiness - finsihing things related to that)
[17:07] Kitto Flora: We are in one!
[17:07] Andrew Linden: Oh, so that is why the AGNI grid is in "startup" mode
[17:07] Sidewinder Linden: yes
[17:07] Sidewinder Linden: to double check
[17:08] Sidewinder Linden: go to "Help > About Second Life..."
[17:08] Sidewinder Linden: If it says "Havok4 Beta" on the fourth line then it's a Havok4 region
[17:08] Sidewinder Linden: the release number is to the right of that
[17:08] Sidewinder Linden: i'll brb - blogging
[17:08] Ryozu Yamamoto: Ah, sweet, we're on Havok4 again
[17:08] Gaius Goodliffe: Nice.
[17:09] Gaius Goodliffe: Can we use this region as a sandbox?
[17:09] Andrew Linden: I guess I have a few announcements to make...
[17:10] Ryozu Yamamoto: Thanks for pointing that out Sidewinder: Havok4 is getting close enough to H1 that it's not immediatly obvious
[17:10] Ryozu Yamamoto: Gaius: Yes, please do!
[17:10] Sidewinder Linden: ok blog post is up: http://blog.secondlife.com/
[17:10] Sidewinder Linden: might have to refresh since it's just posted
[17:10] Kitto Flora: Not appearing yet
[17:11] Andrew Linden: (1) Simon Linden says there is a bug in this latest update where objects can get their collision Level of Detail (LOD) dropped, and they won't recover to higher LOD until the region gets almost all the way back up to around 45 FPS
[17:11] Ryozu Yamamoto: Hehe
[17:11] Sidewinder Linden: oops :)
[17:11] Andrew Linden: (2) I've published some simplified C++ code that shows how llPushObject() works in Havok4 to the wiki, finally
[17:12] Andrew Linden: http://wiki.secondlife.com/wiki/LlPushObject/Havok4Implementation
[17:13] Andrew Linden: There is a gappnig mystery in that code... where objects push avatars, it just passes on the impulse to another class that is very complicated and does some magic, and can totaly override the impulse provided
[17:13] Andrew Linden: perhaps I'll fill in some of those gaps later
[17:14] Ryozu Yamamoto: =D
[17:14] Andrew Linden: (3) vehicles using banking still don't work right. I'm currently working on that bug now.
[17:14] Andrew Linden: That's all.
[17:14] Sidewinder Linden: rats... found a typo... blog post in the body specified the wrong version number - now fixed
[17:14] Ryozu Yamamoto: Anyone feel like Jira-ing that?
[17:14] Sidewinder Linden: the version number is really v77587, and that is what's in the subject line
[17:14] Andrew Linden: Uh.. is that object a demonstration of a bug? It looks like it has per-prim phantom set.
[17:15] Ryozu Yamamoto: Yep
[17:15] Ryozu Yamamoto: It's not per prim phantom though
[17:15] Gaius Goodliffe: Per-prim phantom sounds like a *feature* to me. :p
[17:15] Ryozu Yamamoto: The bounding box of the larger child sphere believes that it is 95% hollow and .9 start dimple
[17:15] Andrew Linden: Gaius, yes, but per-prim phantom requires some UI support, as well as legacy object support first.
[17:15] Gaius Goodliffe nods.
[17:15] Ryozu Yamamoto: After linking and setting physical, the bounding box does not update when parameters (Besides size) change
[17:16] Ryozu Yamamoto: For child prims
[17:16] Gaius Goodliffe: Ah
[17:16] Kitto Flora: The h4 blog has now appeard
[17:17] Andrew Linden: Is that partial phantom bug in jira yet? And do we have a repro recipe?
[17:17] Ryozu Yamamoto: I think it got Jiraed, Warkirby maybe?
[17:18] Andrew Linden: well, I took a copy just in case
[17:18] Ryozu Yamamoto: If you re-rezz it, the bounding box shoud update =)
[17:18] Andrew Linden: Ok, that was all the announcements I had.
[17:19] Andrew Linden: Yeah, I suspected that would happen.
[17:19] Andrew Linden: I can examine the asset, if necessary
[17:20] Kitto Flora: I tested the GSLR trolley car in Periwinkle today.
[17:20] Andrew Linden wonders how it worked.
[17:20] Kitto Flora: Results was a bit spectacular
[17:20] Kitto Flora: It rezzed and started ok
[17:20] Kitto Flora: But it goes nuts on the track
[17:20] Kitto Flora: when moving
[17:20] Kitto Flora: writhing and stwiching, flying off the track
[17:21] Andrew Linden: Really? How odd. I guess I'll just have to get my hands on a copy of your train and pay it some special attention.
[17:21] Kitto Flora: Its not my train
[17:22] Kitto Flora: Its Athos Mucrphys. I think
[17:22] Kitto Flora: Murphy
[17:22] Andrew Linden: You rez it by pressing some button?
[17:22] Kitto Flora: Its a 'slot car' design
[17:22] Kitto Flora: Yes
[17:22] Andrew Linden: oh, great
[17:22] Kitto Flora: You have to be in GSLR group then click the rezzer object
[17:22] Kitto Flora: It steers mechanically
[17:23] Ryozu Yamamoto: Ah, there we go: SVC-1127
[17:23] Kitto Flora: has 4 spheres on underside that act as clamps around the alpha rail
[17:23] Kitto Flora: the script just shoves it along
[17:23] Kitto Flora: Looks like the collision charactersitcs it sees are very different to H1
[17:24] Andrew Linden: I now have 9 items on my list for tomorrw
[17:25] Kitto Flora: I also tried my LSL steered train on the GSLR track, with a carriage, and crossing the border
[17:25] Kitto Flora: It actually worked
[17:25] Andrew Linden: wasn't that the one that had some problems with banking?
[17:25] Kitto Flora: The banking problem is not with that model
[17:26] Kitto Flora: at last as far as I know
[17:26] Kitto Flora: The GSLR version is BIG and theres no helical tracks for it
[17:26] Kitto Flora: baking problem is on the small engines
[17:27] Kitto Flora: I cant demo it as the trains wont run in Wish.
[17:27] Ryozu Yamamoto: Kitto: What do you need to get a train running say, here?
[17:28] Andrew Linden: Ok, well looks like I'm full up for about half a week at least anyway, unless I suddenly fix the various vehicle bugs I already have on my plate
[17:28] Kitto Flora: Here in this sim? Impossible. This sim is severly lagged for LSL
[17:28] Andrew Linden: but eventually I'll need to examine the buggy objects in detail
[17:28] Kitto Flora: Oh
[17:28] Kitto Flora: Hold on
[17:28] Kitto Flora: its gotten better
[17:28] Ryozu Yamamoto: Turned auto-return back on
[17:28] Ryozu Yamamoto: jcool left her UFO out, heh
[17:29] Ryozu Yamamoto: Amongst other things
[17:29] Kitto Flora: now at about 20% of topw speed here
[17:29] Kitto Flora: for LSL
[17:29] Kitto Flora: It may run an engine
[17:29] Kitto Flora: Is this class 5 here?
[17:30] Ryozu Yamamoto: Void sim, dunno about class, should be 5 thougj
[17:30] Kitto Flora: 3452? I think that is
[17:30] Sidewinder Linden: sry been multitasking - do you need to know what class a sim is?
[17:30] Ryozu Yamamoto: A good way to stress test stuff without having to load up a full sim -)
[17:30] Kitto Flora: In a class 5 one engine eats 3.5 mS/frame in physics
[17:30] Kitto Flora: so it will probably run
[17:31] Kitto Flora: In class 4, one vehicle uses ~18mS/frame for physics.
[17:31] Sidewinder Linden: wow
[17:31] Kitto Flora: which means that 1 vehicle will run.. 2 probably not
[17:32] Kitto Flora: In Wish there are 1-4 'vehicles' I cant stop - they are surfing waves -
[17:32] Kitto Flora: and that eats all the SIM cpu. Nothing much left for LSL
[17:33] Kitto Flora: Reall - forget the lean, forget the collisions. Fix the run time of the physics, especially in class 4
[17:33] Andrew Linden: I was trying to use google perftools to profile the vehicle performance today, but failed
[17:34] Andrew Linden: the results I was getting were bogus and I couldn't figure out why
[17:34] Kitto Flora: Looking at 'Sim Time (Physics) ?
[17:34] Gaius Goodliffe: Do the physics aside from collisions make that big an impact? We've had over a dozen planes in the sky in Dogfight Atoll before, even when it was back on class 4, without problems...
[17:34] Kitto Flora: Running H4 ?
[17:35] Gaius Goodliffe: Oh no that was the H1. Is H4 that much more intensive?
[17:35] Kitto Flora: H1 is fine
[17:35] Kitto Flora: This problem is in H4
[17:35] Gaius Goodliffe: Hmmm.
[17:35] Kitto Flora: Well something that ues 0.3mS physiccs in H1/Class 4 uses 18mS in h4/class 4
[17:35] Sidewinder Linden: there is some work going on w/r/t performance... and that should keep going as we find issues
[17:36] Andrew Linden: I'm going to try again to get some profiling working tomorrow or next week
[17:36] Ryozu Yamamoto: Eww, just thought of something
[17:36] Kitto Flora: The problem is this performance is so bad, the impact on LSL wrecks the vehicle - and you cant tell what else might be a real problem.
[17:37] Kitto Flora: Its H4 copiled with debugging on, no optimization?
[17:37] Ryozu Yamamoto: Fixing the bit where bounding box isn't updated is going to add a lot of cpu time I bet,
[17:37] Gaius Goodliffe: Hmm. I've never bothered looking while testing vehicle handling and stuff, I didn't realize that was something I should be looking out for... I'll have to starting looking at that while testing.
[17:38] Kitto Flora: have the stats panel open, and expand the Time(ms) panel.
[17:38] Gaius Goodliffe nods.
[17:38] Andrew Linden: ? What bounding box are you talking about Ryozu? The partially phantom object over there to my left?
[17:38] Ryozu Yamamoto: Yeah
[17:38] Ryozu Yamamoto: Actually, I suppose if it's just child prims with the issue it won't add that mu ch
[17:38] Andrew Linden: No, that shouldn't be expensive.
[17:39] Andrew Linden: We just need to make sure it is right when we build the object's collision shape
[17:39] Andrew Linden: so it is a single shot
[17:39] Ryozu Yamamoto: I figured it just wasn't rebuilding when parameters changed
[17:39] Ryozu Yamamoto: There's a PN Weapon where they shoot a bunch of torii that change prim params as fast as possible
[17:39] Andrew Linden: No, any params that change that affect the objects collision properties should trigger an update
[17:40] Andrew Linden: Yeah, we've made a few changes to make such scrambled tori less of a problem
[17:40] Ryozu Yamamoto: Ah, that's good to hear
[17:41] Andrew Linden: but eventually all objects's collision shapes get updated
[17:41] Kitto Flora: Oh - on possible collisions causing more cpu use - If I rez and fly one plane in Ahern - no collisions - its showing 15mS /frame physics time.
[17:42] Kitto Flora: Similar in Wish
[17:42] Kitto Flora: In a class 5 is more like 3.5mS hit
[17:42] Kitto Flora: On MG H1 I think its more like 0.1mS
[17:43] Andrew Linden: ok kitto, I'll be looking into that
[17:44] Kitto Flora: The better news was on the sim crossing with the bug engine + carriage - there was no case of them sticking together on sim crossing
[17:44] Kitto Flora: Only one time that carriage got stuck
[17:44] Kitto Flora: But if they crossed intersecting, the separated out again fine. Thats much better than MG H1
[17:45] Sidewinder Linden: well that's cool
[17:46] Andrew Linden: sooo many bugs! Makes me anxious to end office hours ;-)
[17:46] Ryozu Yamamoto: Hehe
[17:47] Kitto Flora: Hey - you fixed alot of bugs so far!
[17:47] Andrew Linden: it's closer than it's ever been and now it's even closer
[17:47] Ryozu Yamamoto: Even with the bugs that are left, things are looking really good to me
[17:47] Kitto Flora: I wish the MG got this many bugs fixed
[17:48] Ryozu Yamamoto: Haha
[17:48] Ryozu Yamamoto: Havok4 could be considered one huge bug fix
[17:48] Gaius Goodliffe: :) Now we if can avoid introducing too many new ones in the process...
[17:48] Sidewinder Linden: we're trying to get this as clean as reasonably possible before folding it back into release... although as you'd expect there will be some left on the table ;)
[17:48] Ryozu Yamamoto: Aside from vehicles, I'd say Havok4 does have less issues than Havok1
[17:49] Sidewinder Linden: cool!
[17:49] Sidewinder Linden: the initial reaction has been pretty good
[17:49] Andrew Linden: If only we had converted over to Havok2 before we made vehicles possible
[17:49] Andrew Linden: the original Havok2 effort started before the VolumeDetect feature too
[17:49] Sidewinder Linden: hopefully we can figure out the performance oddities... that is a bit worrying but ought to be resolvable
[17:49] Andrew Linden: the more features, the harder the transition
[17:50] Gaius Goodliffe: Oddly, most of the things that H4 breaks for me aren't even physics related.
[17:50] Ryozu Yamamoto: Oh?
[17:50] Redux Decosta: oh i had one question..... when i fly over about 150 meters, whether or not i have flight mode enabled i float like a brick. h1 gives you a little bit of a buffer when flight is enabled at very high areas, h4 has a very distinct "atmosphere" a hundred & some meters up
[17:50] Ryozu Yamamoto: Ah, there is that, almost forgot about it
[17:51] Andrew Linden: Oh yeah... I disabled the "height above the terrain" variation in flight speed
[17:51] Redux Decosta: so flight mode on or off above 200 meters causes you to behave exactly the same; you just drop like mad
[17:51] Ryozu Yamamoto: Flying is very strange
[17:51] Ryozu Yamamoto: At very high altitudes
[17:51] Andrew Linden: I was wondering if anyone would notice that.
[17:51] Redux Decosta noticed
[17:51] Sidewinder Linden: what are the reasons to have "height above terrain" enabled or disabled?
[17:51] Redux Decosta: it's very difficult to work on skyboxes or up in buildings for those of us that are unusually clumsy, by the time we get flight mode on we're already bouncing off the terrain & flying off the map
[17:51] Ryozu Yamamoto: Inconsistency
[17:51] Andrew Linden: yeah, who wants what?
[17:52] Ryozu Yamamoto: I'd say leave terrain based height limits disabled
[17:52] Andrew Linden: What you really want is to be able to control your flight speed with more ease, right?
[17:52] Ryozu Yamamoto: I think what Redux is implying is that the behavior above the flight threshold is too harsh
[17:53] Andrew Linden: What if it were just easy to decide to fly faster or slower?
[17:53] Ryozu Yamamoto: Not sure how that'd tie into the interface
[17:53] Andrew Linden: (Not that I have many bright ideas on how that input would be collected at this stage)
[17:53] Ryozu Yamamoto: Ah, actually
[17:54] Ryozu Yamamoto: Always Run
[17:54] Ryozu Yamamoto: doesn't change anything while flying, right?
[17:54] Gaius Goodliffe: Always Supersonic? ;)
[17:54] Andrew Linden: Oh... you mean "always run" also applies to flight speed?
[17:54] Ryozu Yamamoto: Feature Request: If Always Run is checked, double flight speed
[17:54] Andrew Linden: Well, it certainly is easy.
[17:55] Andrew Linden: If I did it now Dan would file it as a bug (it's different).
[17:55] Kitto Flora: Is Av fly speed 40kM/h?
[17:55] Ryozu Yamamoto: Hehehe
[17:55] Ryozu Yamamoto: Yeah, that's true
[17:55] Ryozu Yamamoto: the bug thing
[17:55] Redux Decosta: what i was implying is that in havok1 you can turn flight mode on & have a small degree of control, despite descending steadily. in h4 you get above that threshold and, flight mode or not, you drop straight down at the same rate
[17:55] Andrew Linden: Av fly speed is 16 m/sec or thereabouts
[17:56] Redux Decosta: i broke 100 meters per second falling straight down WITH flight mode enabled from 600 to 200 meters; took about 4 seconds
[17:56] Andrew Linden: Oh ok, I think I know what you're talking about Redux
[17:56] Redux Decosta: i can move the table up there & return it & show everyone hehe
[17:56] Kitto Flora: 57.6kM/h
[17:56] Ryozu Yamamoto: lol, chairs won't follow
[17:56] Redux Decosta: lol
[17:56] Ryozu Yamamoto: I didn't use warpPos on them
[17:57] Kitto Flora: And it is!
[17:57] Ryozu Yamamoto: Speaking of, have we tested WarpPos or llSetLinkPrim Params on sitting avatars lately?
[17:58] Andrew Linden: Redux, you're saying that the horizontal fly speed is too attenuated (decreased) at high altitudes?
[17:58] Redux Decosta: i'm saying its 100% nonexistant
[17:58] Andrew Linden: Or are you talking more about the rate of vertical falling?
[17:58] Redux Decosta: vertical falling
[17:58] Redux Decosta: you have absolutely no control over 200 meters, whether or not flight mode is no
[17:58] Redux Decosta: 8on
[17:58] Ryozu Yamamoto: To put it another way, I think, non-boosted flight above the flight threshold drops you like you're wearing a parachute
[17:58] Andrew Linden: Ok, I'll revisit that code
[17:58] Andrew Linden makes a note
[17:59] Ryozu Yamamoto: doh
[17:59] Kitto Flora: Had leaks like a seive
[17:59] Kitto Flora: Head
[17:59] Redux Decosta: it holds a lot of water
[18:00] Andrew Linden: Oh... looks like the hour is up.
[18:00] Andrew Linden: I actually have to leave on time today.
[18:00] Redux Decosta: looks like ryo is under the throas of A.D.D. =)
[18:00] Redux Decosta: hehe thanks for your hard work everyone ^_^
[18:00] Kitto Flora: Byebye
[18:00] Redux Decosta: must... feeed...... O_O
[18:00] Gaius Goodliffe: What? Leave work on time? That's not allowed in IT. :)
[18:00] Kitto Flora: HAppy moving!
[18:00] Sidewinder Linden: hehe uh oh redux :)
[18:00] Ryozu Yamamoto: Oooh
[18:00] Sidewinder Linden: thanks all guys... look for some more interesting news soon ;)
[18:00] Ryozu Yamamoto: llSetLinkPrimParams on a sitting AV IS failing
[18:00] Arawn Spitteler: Under Havoc 1, starting my wings while sitting would've orbited me, up to the heights in question
[18:01] Ryozu Yamamoto: Or I might have the wrong prim
[18:01] Ryozu Yamamoto: I'll get back to you on all that
[18:01] Sidewinder Linden: yes they're held here