User:Andrew Linden/Office Hours/2008 04 24

Transcript of Andrew Linden's office hours:

[17:02]	Kitto Flora wonders where everyone else is
[17:02]	Arawn Spitteler wonders what happened to Everybody, now that we're here: Have you seen Bato's solution to Prim Bump?
[17:02]	Kitto Flora:	Well here's Andrew
[17:02]	Kitto Flora:	And Bato's solution, no
[17:02]	Kitto Flora:	What's he do?
[17:03]	Andrew Linden:	Nice Raven avatar Arawn.
[17:03]	Arawn Spitteler:	Appearently, Prim Seam Bump is cured, when the Scriupt sets a Zen Taper
[17:03]	Kitto Flora:	Thats a prop I never seen
[17:03]	Kitto Flora:	Ak
[17:03]	Kitto Flora:	We are outnumbered!
[17:04]	Arawn Spitteler:	It might have slipped the Havok Techies, because it makes no sense, or it mightbe a bad idea, somehow.
[17:04]	Kitto Flora:	Ah thats better
[17:04]	Arawn Spitteler:	Hi, Cyn
[17:04]	Kitto Flora:	Whats a Zen Taper?
[17:04]	Cinthya Loveless:	mpve your feet lose your seat
[17:04]	Cinthya Loveless:	omg
[17:04]	Cinthya Loveless:	that was totaly bad
[17:04]	Arawn Spitteler:	The Tapered Flag is set, by script, but the taper doesn't have to be other than zero
[17:04]	Sidewinder Linden:	ummm a taper that goes "ommmm"?
[17:05]	Kitto Flora:	Set tapered flag on what?
[17:05]	Andrew Linden:	huh?
[17:05]	Arawn Spitteler:	Just about. Set the Flag as Tapered
[17:05]	Sidewinder Linden:	(nvm bad bad joke about "zen taper"
[17:05]	Arawn Spitteler:	I haven't tried it myself.
[17:05]	Simon Linden:	Interesting - Andrew, that might make the difference between having a havok box vs. a mesh object
[17:05]	Andrew Linden:	That just sounds odd.
[17:06]	Andrew Linden:	Bato has been struggling with golf balls that hit prim seams.
[17:06]	Arawn Spitteler:	Zen Taper is already a joke, so a Zen Taper Joke is redundant, making it a joke which is no joke, and therefore the perfect humor
[17:06]	Kitto Flora:	Set tapered flag on what? Like.. golf ball prims? Or the 'ground' prims?
[17:06]	Arawn Spitteler:	The rail prims. Do your tracks have prim bump, Kitto?
[17:06]	Andrew Linden:	He came up with a workaround where a floor made of tapered boxes (with the pointy sides down) would eliminate the problem
[17:06]	Arawn Spitteler:	Then he found, you don't need the point
[17:07]	Arawn Spitteler:	The prim only thinks it's tapered
[17:07]	Sidewinder Linden:	that's a cute solution
[17:07]	Kitto Flora:	None of my vehicles show any appreciable bump motion. They do show excessive collisions.
[17:07]	Sidewinder Linden:	@arawn - that sounds like the beginning of a therapist joke :p
[17:08]	Kitto Flora:	So - does H4 do FFTs on prim shapes?
[17:08]	Rex Cronon:	hello everybody
[17:08]	Arawn Spitteler:	FFTs?
[17:08]	Andrew Linden:	No Kitto, not that I know of.
[17:08]	Andrew Linden:	FFT = fast fourier transform
[17:08]	Kitto Flora:	Or - digital filtering?
[17:09]	Creem Pye:	youmean to use that as a lowpass filter for making a simplified representation?
[17:09]	Andrew Linden:	No, not digital filtering either.
[17:09]	Andrew Linden:	They do use a simpler frequency filter. Don't know the details of it though.
[17:09]	Kitto Flora:	Well - its like a rectangular prim has spurious ridges around the 'corners' which is also a charateristic of some filtering algos.
[17:10]	Andrew Linden:	Sorry, the filter doesn't block real collisions, but you can filter the number of "collision event callbacks" for some callback hooks.
[17:10]	Cinthya Loveless:	my most recent complants are with rendering but I know thats not with this project
[17:10]	Andrew Linden:	Yeah, I don't have much to offer regarding the rendering.
[17:11]	Andrew Linden:	Well, shall we get started? Announcements first...
[17:11]	Cinthya Loveless sniffles
[17:11]	Andrew Linden:	Sidewinder, do you know the status of our hopeful server udpate?
[17:11]	Kitto Flora:	Does it make a rectangular prim look like that? For collisions?
[17:12]	Andrew Linden:	No Kitto, the rectangular prim actually has rounded corners... but within the collision tolerance there is some play between the two objects
[17:12]	Kitto Flora:	Well its *acts* like it looks like this
[17:13]	Cinthya Loveless:	I like what you did with the feet placment on the avatars in the Havok 4 Andrew its so much better then before
[17:13]	Andrew Linden:	also... contact points are "coherent" in Havok4... they are kept around even when the two objects separate by a few collision tolerances.
[17:13]	Andrew Linden:	But they don't necessarily act in that state.
[17:14]	Sidewinder Linden:	sry andrew... missed that question
[17:14]	Andrew Linden:	yeah Kitto the normals that are picked up on the sides fo the box appear to have really significant impulses... too much.
[17:14]	Sidewinder Linden:	afaik, one rack (~160 regions) were deployed earlier today
[17:14]	Sidewinder Linden:	with no apparent issues
[17:14]	Kitto Flora:	Mm
[17:14]	Kitto Flora:	Impuls
[17:14]	Kitto Flora:	e
[17:14]	Andrew Linden:	which is why I could see you visualizing that shape with burrs along the edges
[17:14]	Sidewinder Linden:	i believe that we'll be deploying another 10 racks toinght
[17:15]	Sidewinder Linden:	and then if that holds, the rest of the system tomorrow
[17:15]	Andrew Linden:	however you missed the other two sides
[17:15]	Kitto Flora:	Yeah
[17:15]	Kitto Flora:	I just modeled the ends here
[17:15]	Arawn Spitteler:	Are there some good rail sims, deployed?
[17:16]	Andrew Linden:	Ok cool. So this next update has a bunch of fixes. I suspect they were mentioned in a blog post yesterday... before we had to cancel the deploy for yet another day.
[17:16]	Andrew Linden:	I'm not going to list all of the bugs or fixes.
[17:16]	Sidewinder Linden:	btw... someone just may have found the "magic cause" of rotations stopping across a region
[17:17]	Sidewinder Linden:	one that coincides with the fix that we did, so iam pretty hopeful that the llTargetOmega fix wil be a good one
[17:17]	Kitto Flora:	What I see is a excessive collision where the red ball is
[17:17]	Kitto Flora:	And it seems others see a bump there
[17:17]	Andrew Linden:	I have one announcement. I've posted the llPushObjectLinear() proposal by Seifert Surface and Lex Neva on the wiki:
[17:17]	Andrew Linden:	You can find a link via my User:Andrew_Linden page
[17:17]	Gaius Goodliffe:	I did notice last time I fixed my rotating objects that each time I deleted a broken one, the "selected" count in the land dialog went down.
[17:18]	Sidewinder Linden:	andrew, would it make sense to post selkit's as well for comparison?
[17:18]	Cinthya Loveless:	thats a problem i've been having for a bit now every so often i have to restart my islands to get the stufff like dolphins back moving again
[17:18]	Kitto Flora:	I hope that T-Omega fix also fixed the dead vehicle issue
[17:18]	Sidewinder Linden:	cinthya, do you know if they use lltargetomgea?
[17:18]	Sidewinder Linden:	if so then it is likely that this update will cure that
[17:18]	Andrew Linden:	Here is the link... it is just a proposal at this stage, and is not in the schedule:
[17:18]	Andrew Linden:	http://wiki.secondlife.com/wiki/User:Andrew/ProposedLlPushObjectLinear
[17:18]	Sidewinder Linden:	are these the jumping dolphins?
[17:18]	Cinthya Loveless:	Im not 100% sure on that they may as they are not using physics to move
[17:19]	Sidewinder Linden:	oh btw - here is the fix list that is in the update being deployed today and tomorrow (same as this beta preview update) : http://blog.secondlife.com/2008/04/19/second-life-simulator-v121085362-update-is-on-the-beta-preview-with-12-physics-fixes-2008-04-19/
[17:19]	Sidewinder Linden:	ok
[17:19]	Rex Cronon:	is this how a cylinder looks like to h4 engine?
[17:19]	Andrew Linden:	No Sidewinder, this proposal is much more detailed than Selkit's linear dependency
[17:19]	Sidewinder Linden:	well i'd say - let's see if this update cures them, and if it doesn't then we can circle back and take a look, ok?
[17:19]	Sidewinder Linden:	ok
[17:19]	Cinthya Loveless:	okay
[17:20]	Andrew Linden:	the update will not fix prim seams for non-vehicles
[17:20]	Kitto Flora:	Does this llPushObjectLinear() proposal deal with the present permission problems?
[17:20]	Andrew Linden:	also, someone was testing their bugged cycles on the Beta Preview and said that there was little improvement for their prim seam problems.
[17:20]	Andrew Linden:	So, I know my fixes help some vehicles, but probably not all.
[17:21]	Andrew Linden:	I'll be needing more copies of bugged vehicles after this update.
[17:21]	Andrew Linden:	but have not yet returned to the vehicle code.
[17:21]	Andrew Linden:	I've been working on llPushObject() and (unsuccessfully) on some avatar motion problems.
[17:22]	Creem Pye:	so the impulse of llPushObjectLinear() would attenuate linearly with distance from the target object?
[17:22]	Andrew Linden:	BTW, there is a chance that I won't be able to attend office hours next Tuesday. I'm not sure what my schedule interference will be
[17:22]	Andrew Linden:	however, I think Sidewinder and/or Simon may be able to make it.
[17:22]	Simon Linden:	I should be around
[17:23]	Sidewinder Linden:	i will try...
[17:23]	Sidewinder Linden:	i am moving on may 1 as well, so "all may be up in the air" :)
[17:23]	Andrew Linden:	Yes, Creem that is the proposal... and I think it also only pushes linearly (llPushObject() takes a second angular vector)
[17:24]	Andrew Linden:	Ok then. The topics are open, we can discuss whatever.
[17:24]	Arawn Spitteler:	Have you found what makes a Choo-Choo fly?
[17:24]	Kitto Flora:	Does this llPushObjectLinear() proposal deal with the present permission problems?
[17:24]	Andrew Linden:	yes Arawn. I commented on that bug
[17:24]	Cinthya Loveless:	were the jumping dolphins target omega?
[17:25]	Andrew Linden:	I had to close it for "won't finish". The problem with the choo choo was a remnant llSetStatus(STATUS_ROTATE_X, FALSE) bit from a previous script
[17:25]	Andrew Linden:	I don't know about the jumping dolphins
[17:26]	Andrew Linden:	Kitto, no the llPushObjectLinear() does not have any permissions changes.
[17:26]	Creem Pye:	so I noticed that in H4, you can change the sculpt map on an object an avatar is sitting on, and the avatar sitting on it won't be unseated (even if the object is dynamic)
[17:27]	Andrew Linden:	Seifert and Lex just took the C++ outline of llPushObject() and tried to replace it with more predictable behavior that was still grief resistant as far as magnitudes and distances go.
[17:27]	Simon Linden:	Creem - that sounds like a good bug, can you file a jira?
[17:27]	Creem Pye:	it's pretty useful, but I'm wondering if other types of prim transformations are possible now that interrupted physics in the past
[17:27]	Simon Linden:	Is that different behavior than Havok1?
[17:27]	Andrew Linden:	Creem, I don't understand the "bug". Could you elaborate on the expected behavior?
[17:27]	Creem Pye:	yeah I think H1 would unseat the avatar
[17:27]	Kitto Flora:	OK. I dont know what problems S&L are trying to deal with. None that I have, I think.
[17:28]	Creem Pye:	it's positive to me... so I'm not enthusiastic about having it "fixed" ;)
[17:28]	Andrew Linden:	Oh right. Yeah, changing the shape of a seated object no longer boots the seated avatars.
[17:28]	Gaius Goodliffe:	Oh cool!
[17:28]	Creem Pye:	(because sculpt maps shouldn't affect physics anyway, right?)
[17:28]	Andrew Linden:	Also, it is possible to link objects together while the avatar is seated
[17:28]	Creem Pye:	ah cool
[17:28]	Andrew Linden:	I was wondering when someone would notice that.
[17:28]	Andrew Linden:	Dan Linden noticed it months ago.
[17:29]	Creem Pye:	but I see that llSetScale() doesn't execute in dynamic objects (I guess that's because it changes the mass)
[17:29]	Andrew Linden:	I'm not planning on "fixing" that.
[17:29]	Creem Pye:	great =)
[17:29]	Andrew Linden:	I considered it a "bug fix" in the first place.
[17:29]	Gaius Goodliffe:	I can pull a few prims out of some vehicles because of that. Before I'd hide or show different shapes, where now it sounds like I can just alter the cut
[17:29]	Andrew Linden:	It should work Gaius. If it doesn't let me know.
[17:29]	Creem Pye:	yeah exactly, gaius. just make sure your sculpts all use the same bounding box I guess
[17:30]	Andrew Linden:	I know you can link while an avatar is seated. Not so sure about unlinking.
[17:30]	Gaius Goodliffe:	I'm not sure you can actually change the shape of a sculpt, Creem. Just the illusion of shape.
[17:30]	Creem Pye:	right
[17:31]	Andrew Linden:	llSetScale() doesn't execute on dynamic objects?
[17:31]	Andrew Linden:	That kinda rings a bell....
[17:31]	Creem Pye:	I don't be lieve it did when I tried it the other day..
[17:31]	Andrew Linden:	but I don't think it is for mass changing
[17:32]	Andrew Linden:	If anything it would be to avoid penetrations... but you can do that anyway with child prim position changes.
[17:32]	Creem Pye:	nope, doesn't scale
[17:32]	Andrew Linden:	Hrm... I'll have to look at the code later.
[17:32]	Andrew Linden takes a note.
[17:32]	Creem Pye:	and I guess changing PRIM_TYPE still unseats avatars?
[17:33]	Creem Pye:	(if it's to a different shape than before, at least)
[17:33]	Andrew Linden:	No, I wouldn't expect it to Creem.
[17:33]	Creem Pye:	ah ok, I'll have to look into that
[17:33]	Andrew Linden:	The prim linkage code suffered a major overhaul and cleanup in Havok4
[17:33]	Andrew Linden:	and I added proper support for seated avatars there.
[17:34]	Andrew Linden:	How about performance problems?
[17:34]	Andrew Linden:	I've been hearing rumors of some places suffering from occasional performance problems, or even constant.
[17:35]	Creem Pye:	Les White would probably be the person to answer that question, with his motorcycle races..
[17:35]	Andrew Linden:	Other than the interpenetrating volume-detect piles
[17:35]	Sidewinder Linden:	in particular cyclic time dilation drops
[17:35]	Andrew Linden:	and "void region" periodic lag
[17:35]	Creem Pye:	ah
[17:35]	Yuu Nakamichi:	yes
[17:35]	Sidewinder Linden:	every few minutes from 0.8-0.9+ to 0.4-ish range
[17:35]	Kitto Flora:	Most noticeable performance problem in one of those sims that a void sim? Or reduced prim-count sim.
[17:35]	Yuu Nakamichi:	I haveseen that in voids
[17:35]	Sidewinder Linden:	then recovers on its own
[17:35]	Yuu Nakamichi:	even after the 1.20.1
[17:35]	Andrew Linden:	anybody notice other performance problems
[17:36]	Yuu Nakamichi:	I thought the RAM allocation was fixed
[17:36]	Yuu Nakamichi:	mostly TD
[17:36]	Sidewinder Linden:	it is possible to likely that there is more than one bug and more than one solution to this one
[17:36]	Yuu Nakamichi:	and performance problems that follow from that
[17:36]	Yuu Nakamichi nods
[17:36]	Cinthya Loveless:	I've noticed some sims like Sanctuary Rock have time diolation drop alot when 30 or so people are in it but that could be form the scripted stuff in it as well
[17:36]	Andrew Linden:	Simon, there are changes in 1.21 that may affect void regions, right?
[17:37]	Cinthya Loveless:	drop to where its like walking against gale force winds when you try to move
[17:37]	Kitto Flora:	I see OK TD, but very poor LSL speed - runing at 1/15th top speed
[17:37]	Simon Linden:	I don't think so Andrew, I believe they went out earlier
[17:37]	Sidewinder Linden:	they were i the build before that
[17:37]	Sidewinder Linden:	but there is a "top colliders" fix - so hopefully that list is now complete
[17:37]	Sidewinder Linden:	maybe that will give us better visibility
[17:37]	Andrew Linden:	Right Kitto. The script engine runtime is sacrificed the most when other prtions of the simulator take up too much time.
[17:38]	Kitto Flora:	Andrew: I understand that is the case. What I copuld no see was what it is thats eating up all the CPU time
[17:38]	Andrew Linden:	Understanding the time spent in scripts is an art.
[17:38]	Rex Cronon:	is it normal that when u r very close to a prim, and u push yourself very hard, and u can't tunnel through it?
[17:38]	Ellla McMahon:	I was just at Sine Wave island ... which is always 'difficult' but today even more so ... but then that place is ALL scripts !!! :)
[17:39]	Andrew Linden:	On a perfectly healthy region with very little physics or network costs... the script engine will show a lot of runtime...
[17:39]	Rex Cronon:	if u r 1 or 2 m awy it works
[17:39]	Andrew Linden:	because the script engine absorbs all the runtime it can get, so when there is a surplus it gets it all.
[17:39]	Rex Cronon:	away*
[17:39]	Yuu Nakamichi:	I am interested what other causes for TD in voids may have
[17:40]	Kitto Flora:	I measure LSL execution speed, not runtime.
[17:40]	Andrew Linden:	However, if the other parts of the simulator are using up a lot of time, then the script engine is preferentially starved.
[17:40]	Yuu Nakamichi:	given that there aren't many objects or scripts around by definition
[17:40]	Creem Pye:	so perhaps measuring the script time when a scripted object is alone in the sim isn't representative of its true "lag" factor?
[17:40]	Andrew Linden:	However, if it has too many scripts... it still gets a sizable chunk of time.
[17:40]	Kitto Flora:	Yes Andrew: We are painfully aware of that :)
[17:41]	Kitto Flora:	What I was seeing is not a lot of scripts, not a lot of pysics or images time. Yet LSL is executing very slowly.
[17:41]	Kitto Flora:	But TD is good
[17:41]	Cinthya Loveless:	I think Mono is something like that
[17:41]	Yuu Nakamichi:	do you have suspects for the TD other than what has been fixed in 1.20.1?
[17:42]	Andrew Linden:	Hrm... don't know what is up with that Kitto, unless another region on the same host were somehow slowing the machine down.
[17:42]	Andrew Linden:	Kitto, it wasn't a void region was it?
[17:42]	Sidewinder Linden:	i suspect from subjective results that it might have something to do with teleports, or rezzes, or just now repoted by another beta tester - maybe flight transitions with avatars
[17:42]	Sidewinder Linden:	but that's "completely theoretical" at this point - just folks fishing around trying to reproduce the behavior on demand
[17:43]	Kitto Flora:	Well its one of these 'low prin allocation' sims. Do they also get low total CPU time?
[17:43]	Yuu Nakamichi:	right
[17:43]	Andrew Linden:	We're interested in any clues you might have regarding sources of the performance problems.
[17:44]	Andrew Linden:	I think Simon picked up some performance analysis tools today, or they were ordered anyway.
[17:44]	Yuu Nakamichi:	TD per se is always with us of course - voids sims see sporadic TD all the time
[17:44]	Kitto Flora:	Reproduction is no problem in that sim - its running slow all the time.
[17:44]	Simon Linden:	Yeah, they were ordered but I don't think we got the license. I should be able to spend some time on it tomorrow or next week
[17:45]	Yuu Nakamichi:	in my experience, which is why it is hard to pin down why there's subjectively more of it now
[17:45]	Kitto Flora:	HimandIandUs region.
[17:45]	Andrew Linden:	Well, it sounds like TD problems are one of the "more annoying problems" right now.
[17:45]	Yuu Nakamichi:	we also don't have the tools to peek beeond theregion itself
[17:46]	Andrew Linden:	I'm not going to ask for everyone's most annoying bug today.
[17:46]	Andrew Linden:	I'm noping that the recent update will fix some of them.
[17:46]	Yuu Nakamichi:	we can't positively say whether there is another region that hogs the CPU
[17:46]	Andrew Linden:	That is right Yuu
[17:47]	Gaius Goodliffe:	Would be nice if every estate owner would put SimNeighbours beacons out. :)
[17:47]	Yuu Nakamichi:	or if it the cause it somwhere in the region itself
[17:47]	Andrew Linden:	Hrm... it might not be too hard to open up some "top" info someday.
[17:47]	Kitto Flora:	I recall past claims that 'one region cannot affect another region, on the same CPU' - is that no longer true?
[17:47]	Andrew Linden:	That is, host load info through the simulator.
[17:47]	Yuu Nakamichi:	that would be great
[17:47]	Simon Linden:	Kitto - I think that was refering to direct interaction.
[17:48]	Gaius Goodliffe:	Yeah, it's never been that simple.
[17:48]	Simon Linden:	If one region hogs the CPU, memory or other resources (disk access), it can slow down other regions
[17:48]	Cinthya Loveless:	would be nice if you could your simso n same servers so you don't end up shairing with a resorce hog
[17:48]	Arawn Spitteler:	Dual Core doesn't seperate the clocks?
[17:48]	Kitto Flora:	Slowing down one region, due to tings going on in a co-hosted region is is pretty direct interaction?
[17:49]	Sidewinder Linden:	umm one region on a host where the other regions are on other cpu's likely have little interaction, until the memory bus or something else gets saturated
[17:49]	Andrew Linden:	multi-core does isolate the processes, but they share a pile of RAM, HD, and network connection.
[17:49]	Kitto Flora:	Simon: Are you sure? There were strong denials of that some years(?) back when residents complained about it.
[17:49]	Creem Pye:	I believe openspaces sims would share a core with 3 others
[17:49]	Sidewinder Linden:	but i don't think it's reasonable to say that four regions running on one cpu will never interact
[17:49]	Gaius Goodliffe:	Multiple cores doesn't give you multiple network I/O, for one thing -- if one sim saturates the line, everyone waits.
[17:50]	Sidewinder Linden:	btw from what i understand... there is nohing that guarantees how many partnered regions there will be on an openspace cpu
[17:50]	Sidewinder Linden:	i think kelly and i talked about this yesterday, and his belief was that it's up to 16 regions per host, with the os determingin which cpu's they get affinity to
[17:50]	Sidewinder Linden:	so it might not be 4/cpu at the end of the day
[17:50]	Cinthya Loveless:	i thought it was 4 on an open space cpu
[17:50]	Andrew Linden:	max of 4 per CPU yes
[17:51]	Simon Linden:	I suspect they were talking about stuff like a region crossing or TP - going from one region to another on the same computer is basically the same as going to another computer (as far as I know)
[17:51]	Cinthya Loveless:	and full prim sims have thier own cup form my understanding correct?
[17:51]	Andrew Linden:	yes
[17:51]	Cinthya Loveless:	cpu*
[17:51]	Sidewinder Linden:	non-openspace regions are one per cpu
[17:51]	Sidewinder Linden:	each server has four cpu's
[17:52]	Kitto Flora:	The slowness we were seeing is constant, and theres no great TP activity or rezzing in the region. Just a few avs standing around.
[17:52]	Andrew Linden:	I guess I should start using "openspace" instead of "void"
[17:52]	Gaius Goodliffe nods.
[17:52]	Kitto Flora:	How many prims allowed in an 'open space' region? (roughly)?
[17:52]	Creem Pye:	people who have builds rezzed in a "void" sim might get offended ;)
[17:52]	Gaius Goodliffe:	3750
[17:53]	Kitto Flora:	I guess thats what that region is
[17:53]	Cinthya Loveless:	3750 correct
[17:53]	Andrew Linden:	3750 = 15000/4
[17:53]	Sidewinder Linden:	have you looked at top colliders and top scripts? anything of note there?
[17:53]	Kitto Flora:	I cant - not the manager
[17:53]	Sidewinder Linden:	ahh
[17:53]	Sidewinder Linden:	if i'm on at some point when you are, feel free to ping me kitto and i'll take a look
[17:54]	Kitto Flora:	OK, Sidewinder
[17:54]	Gaius Goodliffe:	Top Scripts had been badly broken for as long as I've had a sim.
[17:54]	Andrew Linden:	Given than the reports of the problem have been widespread for openapce regions I suspect it is content independent.
[17:54]	Yuu Nakamichi:	is there a pjira where top colliders are collected for this issue?
[17:54]	Gaius Goodliffe:	Less than half the running scripts show up in it.
[17:54]	Kitto Flora:	What theye is is a lot of 'dumb' residents and visitors, who cant understand why 'stuff dont work well' :)
[17:54]	Yuu Nakamichi:	I'ld love to check once I see it happen and drop it there
[17:54]	Sidewinder Linden:	andrew - any idea why many scripts would not show up in top scripts? (he asks dumbly - i didn't realize it was that broken)
[17:55]	Kitto Flora:	Top Scripts seems to work ok on Wish Sim
[17:55]	Kitto Flora:	It takes a while for the list to build
[17:55]	Andrew Linden:	We split the scripts up into two categories... those that can go "inactive" and those that must stay "active"
[17:55]	Creem Pye:	scripts in avatar attachments aren't counted unless the avatar is sitting
[17:55]	Andrew Linden:	My first guess is that the inactive ones are never on the top scripts list
[17:56]	Gaius Goodliffe:	The numbers in Ctrl-Shift-1 are always at least double what shows up in Top Scripts, often much more.
[17:56]	Andrew Linden:	An "inactive" script is one that do not have: timer, listener, repeated scanner, or other routine that requires occasional runtime...
[17:57]	Andrew Linden:	scripts that rely entirely on user interaction... touch() for example, or collision() can also go "inactive".
[17:57]	Gaius Goodliffe:	In one case, I've seen 1900 "Active Scripts" in Ctrl-Shift-1, and 200 in "Top Scripts"
[17:57]	Andrew Linden:	er... scripts that rely on "outside" interaction"
[17:57]	Sidewinder Linden:	if that's the case, why would they show up as script time in the stats bar, but not in top scripts?
[17:58]	Sidewinder Linden:	or maybe i'm not understanding what i see sometimes..
[17:58]	Sidewinder Linden:	perhaps it's not hitting refresh on top scripts at "the right time"?
[17:58]	Andrew Linden:	Right, and finally... I don't think the "top scripts" report has enough room for all of the scripts in the region
[17:58]	Andrew Linden:	so it probably truncates at some max count
[17:58]	Gaius Goodliffe:	It also apparently doesn't show you the top of max count, then. You get a report of a random sample of scripts.
[17:58]	Andrew Linden:	Historically we packed info into UDP packets that had limited size
[17:59]	Cinthya Loveless:	from what I've seen the top scripts doesn't show all scripts in the sim just the top highest MS using ones and as you remove them the list keeps moving down to the next top speed
[17:59]	Kitto Flora:	Is 'Active Scripts' in the Stats panel 'ALL" the scripts of only the script that are 'active' ?
[17:59]	Andrew Linden:	Dunno whether the "top scripts" report has been migrated to TCP packet streams or not
[18:00]	Kitto Flora:	s/of/or
[18:00]	Gaius Goodliffe:	You can also get scripts to show up or disappear from Top Scripts by sitting on unsitting objects (it won't count scripts on avs that aren't sitting), but even in sims without no other avs, I've seen wildly off counts.
[18:00]	Andrew Linden:	I think "active scripts" in the stats panel lists the number of scripts in the "active" list
[18:01]	Andrew Linden:	not all of the scripts in the region
[18:01]	Kitto Flora:	OK
[18:01]	Andrew Linden:	Consider the default start script, "Hello Avatar" -- I think that is an "inactive" script
[18:01]	Andrew Linden:	since it requires an outside interaction to get any runtime once the state_entry() fires
[18:02]	Kitto Flora:	Looks like it is counted
[18:02]	Andrew Linden:	counted in the active scripts?
[18:03]	Gaius Goodliffe:	It takes time, too. Even scripts with no event handlers take away time -- put a script with just state_entry in a bunch of prims, and you can see the impack on sscript time in the stats.
[18:03]	Kitto Flora:	Yes
[18:03]	Kitto Flora:	A script with *NOTHING* in it is being counted in the stats panel
[18:03]	Andrew Linden:	Alright, apparently I don't know what I'm talking about.
[18:03]	Kitto Flora:	I suspect that 'ALL' scrips, nut 'Active' ...
[18:03]	Gaius Goodliffe:	Unless you turn off running.
[18:03]	Kitto Flora:	Stats panel is an antique hack?
[18:04]	Andrew Linden:	Hrm... I wonder if "active" in this case means "enabled".
[18:04]	Gaius Goodliffe:	Could be.
[18:04]	Kitto Flora:	Looks like it
[18:04]	Andrew Linden:	We definitely have an "active list" and an "inactive list" for those scripts that must have repeated runtime.
[18:05]	Kitto Flora:	Which would expain the discrepancey between Stats panel count and Debug count
[18:05]	Gaius Goodliffe:	Actually, it doesn't.
[18:05]	Gaius Goodliffe:	Did the math on that one with Kyle Linden once.
[18:05]	Creem Pye:	by the way, it would be nice if LSL could enable/disable scripts in other links of the same object =)
[18:05]	Gaius Goodliffe:	Actally, he did the math.
[18:05]	Andrew Linden:	That is, internally we have two lists of scripts we maintain.
[18:05]	Gaius Goodliffe:	The sum total of the script time of all the scripts not showing up in Top Scripts failed to add up right.
[18:06]	Kitto Flora:	Creem: Add controller scripts...
[18:06]	Gaius Goodliffe:	Things were being excluded that took more time than items on the list.
[18:06]	Andrew Linden:	and there is a method of the form: LLScript::canGoInactive() or something like that.
[18:06]	Creem Pye:	yeah, but then you have running controller scripts =P
[18:06]	Kitto Flora:	Not runnng much. Listening for alink message. And will control in-prim scripts
[18:06]	Creem Pye:	I guess you could inject the scripts into the other prims, and set them to running
[18:07]	Creem Pye:	(and have them dsiabled in the original prim)
[18:08]	Andrew Linden:	Well, if I find the time I'll try to investigate the script system more and answer some of the questions here.
[18:08]	Andrew Linden:	I'm going to have to head out soon.
[18:08]	Gaius Goodliffe:	While you're at it, see if you can make scripts on avatar attachments show up in Top Scripts. That'd be a real nice plus.
[18:08]	Kitto Flora:	Dinner Time!
[18:08]	Ellla McMahon:	Thank you Andrew :)
[18:09]	Kitto Flora:	Byebye Andrew
[18:09]	Gaius Goodliffe:	take care
[18:09]	Simon Linden:	Thanks for all the feedback everyone
[18:09]	Rex Cronon:	bye andrew
[18:09]	Creem Pye:	and how's the new CEO?
[18:09]	Andrew Linden:	Gaius, I thought attachment scripts showed up in top scripts already.
[18:09]	Ellla McMahon:	and Side & simon :)
[18:09]	Sidewinder Linden:	heh creem - he starts may 15 ;)
[18:09]	Kitto Flora:	Byebye Sidewinder
[18:09]	Andrew Linden:	I haven't met the new CEO yet.
[18:09]	Sidewinder Linden:	so it's a bit early for that question
[18:09]	Creem Pye:	lol simon just got that dismounting bug
[18:09]	Arawn Spitteler:	Attachment Scripts are amongst the biggest users
[18:09]	Gaius Goodliffe:	Not last time I checked -- if they do, it's new.
[18:09]	Rex Cronon:	bye sidewinder
[18:09]	Sidewinder Linden:	oh actually ...
[18:09]	Kitto Flora:	Byebye Simon
[18:10]	Gaius Goodliffe:	If they SIT on something, they show in top scripts as what they're sitting on,
[18:10]	Rex Cronon:	bye simon, bye everybody
[18:10]	Sidewinder Linden:	andrew - is it a known thing to sometimes land in falling animation if you tp onto someone's head?
[18:10]	Gaius Goodliffe:	Otherwise they just don't show up at all.
[18:10]	Kitto Flora:	Oh yes - Avs with 2000 acctached scripts
[18:10]	Sidewinder Linden:	i've just realized i've seen that a few times - with only way out being to jump or fly off
[18:10]	Creem Pye:	is that without an AO?
[18:10]	Kitto Flora:	And those show up in 'Activ' in the Stats panel too