User:Andrew Linden/Office Hours/2008 08 21
Jump to navigation
Jump to search
Transcript of Andrew Linden's office hours:
[17:02] | Sindy Tsure: | heya andrew |
[17:03] | Sindy Tsure: | arawn was just here, too.. dunno where he ran off to |
[17:03] | Andrew Linden: | ok |
[17:04] | Sindy Tsure: | how are ya? |
[17:04] | Sindy Tsure: | involved in the mono deply pause? sounds like they're scurrying to figure out their next step... |
[17:04] | Sindy Tsure: | *deploy |
[17:05] | Andrew Linden: | I'm recovering... from heads-down work. Takes a moment to get into the "office hour mood" |
[17:05] | Tizzers BnT: | heya Andrew |
[17:05] | Sindy Tsure: | no hurry - i have nothing to bug you about this week! |
[17:05] | Sindy Tsure: | hiya tizzers |
[17:05] | Tizzers BnT: | Hi Sindy |
[17:05] | Andrew Linden: | Yes, I think there is some scurrying going on with MONO... bad performance in a few regions I think. |
[17:05] | Tizzers BnT: | ah I was just going to ask |
[17:06] | Sindy Tsure: | think prospero mentioned a sorta-rare crash loop, too :( |
[17:06] | Andrew Linden: | but then, I haven't really been paying attention today |
[17:06] | Sindy Tsure: | hi rez |
[17:06] | Sindy Tsure: | *x |
[17:06] | Andrew Linden: | ah, that is not too surprising |
[17:06] | Tizzers BnT: | Hi Rex |
[17:06] | Rex Cronon: | hi sindy |
[17:06] | Rex Cronon: | hi simon |
[17:06] | Rex Cronon: | hi tizzers |
[17:06] | Andrew Linden: | if they only introduce one crash mode then I'd be surprised |
[17:06] | Rex Cronon: | hi andrew |
[17:06] | Andrew Linden: | hello |
[17:06] | Rex Cronon: | i thought i saw simn here:) |
[17:06] | Rex Cronon: | simon* |
[17:07] | Andrew Linden: | I've been working most of the week trying to build our server code on gcc-4.1 (we're currently using gcc-3.3) |
[17:08] | Tizzers BnT: | Andrew, the military groups have been having major problems for months n ow with bullets going right through 10 meter prims |
[17:08] | Rex Cronon: | tizzers u r making more smoke than a coal factory:) |
[17:08] | Tizzers BnT: | lol sorry my particles are off |
[17:08] | Andrew Linden: | That doesn't really have any impact for the residents, however it is a prerequisite for moving to some modern server hardware we are currently trying to get working for us |
[17:08] | Sindy Tsure: | why the compiler update? just keeping up or is there a reason? |
[17:08] | Rex Cronon: | i was jocking:) |
[17:09] | Tizzers BnT: | what sort of schedule do you think we can expect on the bullet / prim issue? |
[17:09] | Andrew Linden: | Tizzers, those bullets may or may not be fixed by some code I'm still trying to get into the 1.25 codebase |
[17:10] | Andrew Linden: | depends... there were two bugs associated with some bullets that were passing through avatars |
[17:10] | Andrew Linden: | (1) some avatar-specific code would sometimes cause the bullets to pass through avatars |
[17:10] | Andrew Linden: | and (2) the collision_start() events were not being handled in a timely manner |
[17:11] | Tizzers BnT: | yes I've also noticed getting killed in damage land by bullets but no hits/bumps/pushes registering |
[17:11] | Andrew Linden: | it is still possible to tunnel bullets through objects in Havok4, if they are moving fast enough |
[17:11] | Andrew Linden: | so I'd have to examine the particular bullets you're talking about |
[17:11] | Tizzers BnT: | yes we've noticed anything like 100 m/s or faster goes through |
[17:11] | Andrew Linden: | to see if they are fixed or not |
[17:11] | Rex Cronon: | is still possible to walk throught bullets:) |
[17:11] | Tizzers BnT: | these are simple unscripted physical bullets |
[17:12] | Andrew Linden: | hang on a minute... the fixed code I'm talking about was deployed to a test grid |
[17:12] | Andrew Linden: | lemme see if it is still up and whether it is accessible |
[17:12] | Tizzers BnT: | ok |
[17:12] | Sindy Tsure: | wsa that what we were messing with in xyz a couple weeks ago? |
[17:12] | Andrew Linden: | retesting of it happened today... |
[17:13] | Tizzers BnT: | hi Simon |
[17:13] | Sindy Tsure: | hi hi, simon! |
[17:13] | Rex Cronon: | hi simon |
[17:13] | Simon Linden: | Hi ! |
[17:13] | Andrew Linden: | Okay, so the test grid is 'radha' |
[17:13] | Andrew Linden: | and it is up and open |
[17:14] | Rex Cronon: | i had a premonition simon was comming:) |
[17:14] | Rex Cronon: | i just was a few minutes earlier:) |
[17:14] | Sindy Tsure: | reverse lag, rex? |
[17:14] | Andrew Linden: | You should be able to log into radha and test your bullets. |
[17:14] | Andrew Linden: | I'd be curious to know if they are fixed or not |
[17:14] | Sindy Tsure: | it's a whole grid or a region on the beta grid? |
[17:15] | Andrew Linden: | dunno when radha's database was last updated |
[17:15] | Tizzers BnT: | ok how do I do that? is this the main beta grid or a separate target for the client? |
[17:15] | Rex Cronon: | i said hi to him when he wasn't here:) |
[17:15] | Andrew Linden: | so it may have your old password or something |
[17:15] | Tizzers BnT: | ok |
[17:15] | Andrew Linden: | The way to do it is as follows: |
[17:15] | Andrew Linden: | before you press the [Connect] button... |
[17:15] | Andrew Linden: | press ALT+CTRL+G (I think that is the combo) |
[17:16] | Andrew Linden: | a little combo-box UI element should pop up to the right of all the buttons |
[17:16] | Sindy Tsure puts her money on ctrl-shift-g | |
[17:16] | Rex Cronon: | for me ctrl+alt+g works |
[17:16] | Andrew Linden: | you can then select the test grid you want to log into |
[17:16] | Tizzers BnT: | ok |
[17:16] | Tizzers BnT: | cool |
[17:16] | Andrew Linden: | and then press [Connect] |
[17:16] | Tizzers BnT: | ok thanks we;ll check it out |
[17:16] | Andrew Linden: | So it may fail... if you don't have an account on that test grid's database |
[17:17] | Andrew Linden: | if it does fail, then send me a message and I might have time to look into how to fix it for certain accounts before the next office hour |
[17:18] | Andrew Linden: | however... since my changes all passed QA radha will probably be re-purposed soon for other tests |
[17:18] | Andrew Linden: | so if you do try it out, try today |
[17:18] | Tizzers BnT: | ok theres another issue to do with buoyancy with attachment vehicles. The merczateers use an attachment aircraft, but it exhibits different bouyancy characteristics depending on what altitude you are at and whether you attached it while over land or over prims. Nexii Maltus was telling me this was not fixed yet a few weeks ago. Any news on that? |
[17:18] | Sindy Tsure: | wb, arawn |
[17:19] | Andrew Linden: | Tizzers, that bug is definitely not fixed yet. |
[17:19] | Tizzers BnT: | when attached over land, it seems to work fine at normal altitudes, but when attached while over prims, it exhibits a sink rate of over 1 m/s |
[17:19] | Andrew Linden: | I hadn't heard about it yet, but I've got a good idea what is going wrong there. |
[17:19] | Tizzers BnT: | ok |
[17:19] | Andrew Linden: | (And I wonder how easy it will be to fix) |
[17:19] | Sindy Tsure has never heard of this.. do you have to be close to the prims when you attach? | |
[17:20] | Andrew Linden: | actually, I think i can fix it |
[17:20] | Tizzers BnT: | the flight feather freebie seems to have very poor z vector strength control at high altitudes |
[17:20] | Andrew Linden: | I'm just sad that I'll have to dig into that bit of code again |
[17:20] | Tizzers BnT: | lol |
[17:20] | Andrew Linden: | how high is high when it comes to broken flight-feather behavior? |
[17:20] | Andrew Linden: | is there a jira for it? |
[17:20] | Tizzers BnT: | 600-700 meters |
[17:20] | Tizzers BnT: | I'm not sure |
[17:21] | Andrew Linden: | Is there a jira for the buoyant vehicle attachment? |
[17:21] | Andrew Linden: | if so, please post the jira numbers here so I can revisit my office hour transcripts to find it |
[17:21] | Tizzers BnT: | ok I'll ask the mercz |
[17:22] | Andrew Linden: | I know filing jira items is a hassle for some people, and there are a lot of jira items that haven't had updates in a long time, nevertheless |
[17:22] | Andrew Linden: | it is convenient to have a number to reference |
[17:22] | Tizzers BnT: | I'll have Nexii contact you when he's awake |
[17:22] | Andrew Linden: | and where comments can be recorded as more info is gathered |
[17:23] | Sindy Tsure: | http://jira.secondlife.com/browse/SVC-22 !! |
[17:23] | Andrew Linden: | and seriously, I often go back to the transcripts of these hours to rediscover bugs that we've talked about |
[17:23] | Sindy Tsure: | just in case nobody else mentions a jira this week.. |
[17:24] | Andrew Linden: | I've got so much stuff going on, I have trouble remembering individual jira nujmbers anymore |
[17:24] | Tizzers BnT: | SVC-1769 Vehicle sinks when VEHICLE_BUOYANCY set to 1 (neutral buoyancy) |
[17:24] | Sindy Tsure: | (the jira above has nothing to do witht he conversation) |
[17:24] | Sindy Tsure: | (*my jira..) |
[17:24] | Andrew Linden: | that SVC-22 number is familiar, I should know exactly which one we're talking about |
[17:24] | Sindy Tsure tries to poke you at every office hour about it.. | |
[17:24] | Andrew Linden: | but my brain is full... I'll have to revisit the page... |
[17:24] | Simon Linden: | I think it's come up before ... |
[17:24] | Rex Cronon: | regarding jira, i wonder if we could talk a little about: http://jira.secondlife.com/browse/SVC-2608 |
[17:25] | Sindy Tsure: | vehicle region crossing into a full parcel is bad.. |
[17:25] | Tizzers BnT: | search jira for buoyancy, apparently theres a number of complaints |
[17:25] | Sindy Tsure also got a bunch of hits on 'feather' tizzers.. none seemed quite right, though | |
[17:26] | Andrew Linden: | as to SVC-2608, I did ask Don and Babbage Linden about it. |
[17:26] | Andrew Linden: | Don confirmed that there is no way to fix it. A new call or param must be created. |
[17:26] | Rex Cronon: | if u want to rotate relative to root right? |
[17:27] | Andrew Linden: | Actually, I think SVC-93 is the athoritative jira on that rotation bug |
[17:27] | Andrew Linden: | Lex Neva provides some workarounds in SVC-93 |
[17:27] | Andrew Linden: | but they aren't very pretty |
[17:28] | Rex Cronon: | actually SVC-2608 is not about rotations relative to the root |
[17:28] | Rex Cronon: | the script is not trying to rotate the child relative to the root |
[17:28] | Andrew Linden revisits SVC-2608 | |
[17:29] | Rex Cronon: | i rezzed above my head an sample |
[17:29] | Rex Cronon: | the root has nr 1, child nr 2 |
[17:30] | Rex Cronon: | when root is at zero child rotates 360 with inrements of 90 degrees without problem |
[17:30] | Andrew Linden: | You are indeed rotating the child prim. That call on child prims is just broken |
[17:30] | Andrew Linden: | yes, because the root's rotation is the identity rotation |
[17:30] | Andrew Linden: | the identity rotation commutes! |
[17:30] | Rex Cronon: | when root is rotated using that function the child no longer rotates right |
[17:31] | Rex Cronon: | if i rotate the root manually the child rotates without a problem |
[17:31] | Andrew Linden: | yes, the edit rotations go through a different path |
[17:31] | Sindy Tsure: | you're trying to rotate the child to an absolute rotation? |
[17:31] | Andrew Linden: | it is the code that handles the llSetPrimitiveParams that is broken |
[17:32] | Rex Cronon: | yes sindy |
[17:32] | Arawn Spitteler: | I had a bit of trouble with my twistor, having set TargetOmega, in hopes of turning a seat. It might be handy, if child prims were to use the static version, of TargetOmega |
[17:32] | Sindy Tsure: | so.. er.. it's the rotation you want divided by the root rotation, right? |
[17:32] | Andrew Linden: | child prims do use the "static" version of TargetOmega |
[17:32] | Sindy Tsure: | rotations give me headaches.. |
[17:33] | Andrew Linden: | that is to say, they don't actually rotate, but only appear to roate on the client |
[17:33] | Rex Cronon: | anyway it is your call if u want to set it resolved, but at least provide a link to it from svc-93 |
[17:33] | Andrew Linden: | The two are already linked |
[17:33] | Andrew Linden: | that is SVC-2608 and SVC-93 |
[17:34] | Arawn Spitteler: | I've found this oddness of sitting behaviour |
[17:34] | Rex Cronon: | i thought that if u mark it as resolved is going to be deleted |
[17:34] | Andrew Linden: | No the jira items never get deleted that I know of. |
[17:34] | Andrew Linden: | SVC-2608 would stay in the database, but as 'resolved --> duplicate' |
[17:35] | Rex Cronon: | ok, as i said i thought it was going to be deleted |
[17:36] | Andrew Linden: | There is pressure building to fix some parcel permissions problems, where objects can cross region boundaries and parcel boundaries at the same time, and violate entry perms |
[17:36] | Podium: <-0.50002, -0.50002, 0.49998, 0.49998> | |
[17:36] | Andrew Linden: | I'm going to have to fix that in the next big bug hunt, and i'm going to try to also see if I can fix SVC-22 at that time |
[17:36] | Sindy Tsure: | it'd be great to have that suggestion to allow scripts to see across the border fixed first.. |
[17:37] | Sindy Tsure: | so they can see if they're going to end up somewhere they shouldn't be |
[17:37] | Sindy Tsure: | think we talked about that last month.. i'll dig up the jira |
[17:37] | Rex Cronon: | u mean like a script can look in the next sim and see if it can run there and if is allowed to rez things? |
[17:37] | Sindy Tsure: | or even if it's allowed to cross into that parcel |
[17:38] | Sindy Tsure: | scripts today can't see if there are ban lines across a region boundary |
[17:38] | Arawn Spitteler: | That would be Get Parcel Flags? |
[17:38] | Sindy Tsure: | think so.. lemme find the jira |
[17:38] | Andrew Linden: | Hrm... yes it may be possible to fix those two bugs at once... make the script scan work, and also make the simulator automatically enforce the parcel perms across the border |
[17:39] | Andrew Linden: | if anyone happens to find those two bugs at once in jira, feel free to link them up for me |
[17:39] | Andrew Linden: | but not too important. I should be able to remember that one is associated with another |
[17:39] | Arawn Spitteler: | 2608 relates to 22? |
[17:40] | Andrew Linden: | No, not those two |
[17:41] | Andrew Linden: | dunno the actual numbers of the two bugs I was talking about |
[17:41] | Sindy Tsure: | http://jira.secondlife.com/browse/SVC-221 : "llGetParcelFlags(), llScriptDanger(), and llGetLandOwnerAt() should be able to peek accross sim boundaries" |
[17:41] | Andrew Linden: | but I seem to recall that some scripted parcel-info scans are broken across region boundaries |
[17:42] | Andrew Linden: | and there are some "security" bugs pointing out that parcel-perms are not enforced on region crossings |
[17:42] | Sindy Tsure: | like 'can enter'? |
[17:42] | Andrew Linden: | hrm... come to think of it, the security bugs are probably not visible... unless you created it |
[17:43] | Andrew Linden: | but there may already be dupes in the main jira db, dunno |
[17:43] | Andrew Linden: | Sindy, thanks for that jira number lookup |
[17:43] | Sindy Tsure: | :) |
[17:43] | Sindy Tsure: | you found a sit bug, arawn? |
[17:43] | Arawn Spitteler: | I'm sitting to the right of this podium |
[17:43] | Arawn Spitteler: | llSitTarget(<0.0, -.5, 0.0>,< -0.5,-0.5,0.5,0.5>); |
[17:44] | Andrew Linden: | Simon, Babbage Linden was asking about tools to help analyze bad simulator performance in the mono branch. Were you able to help him out today? |
[17:44] | Arawn Spitteler: | I should be behind |
[17:44] | Arawn Spitteler: | I might have to find what I haven't read, or file a Jira |
[17:44] | Sindy Tsure: | if you edit the podium and change Ruler mode to Local, which color arrow points forward? |
[17:45] | Arawn Spitteler: | Green is forward; I'mk using path cut for the angle |
[17:45] | Arawn Spitteler: | I'm displaced about .4 meters in the positive Z direction |
[17:45] | Sindy Tsure: | er.. |
[17:45] | Andrew Linden: | Red is in in the X-direction |
[17:45] | Sindy Tsure: | in the local z? |
[17:45] | Arawn Spitteler: | Red is the up arrow, in this case |
[17:46] | Object: Hello, Avatar! | |
[17:46] | Andrew Linden: | based on how the box is cut, looks like it is at a non-identity rotation -- local-Z is horizontal |
[17:46] | Arawn Spitteler: | Red is down, Z is left, Y is forward |
[17:46] | Andrew Linden: | Red arrow is the local-X |
[17:47] | Arawn Spitteler: | Zis right |
[17:47] | Andrew Linden: | that is Red arrow is the local -forward, but if you rotate the box then local-forward may point world-up |
[17:48] | Arawn Spitteler: | World Down, in this case. I'm using Pathcut for the sloped top, in this one |
[17:48] | Andrew Linden: | so your coodinates in the llSitTarget() are in the local frame... X component changes in the local-Red axis |
[17:48] | Arawn Spitteler: | Are you using the line I gave? |
[17:48] | Sindy Tsure: | hiya scott |
[17:48] | Sindy Tsure is | |
[17:48] | Sindy Tsure: | the prim is rotation y=90 |
[17:49] | Arawn Spitteler: | I don't know where the Z Displacement comes from |
[17:50] | Sindy Tsure: | it's like the sit target isn't an offset from the center |
[17:50] | Andrew Linden: | Anybody have other issues to talk about, besides SVC-2608, -22, -93, and -222? ;-) |
[17:50] | Object: Hello, Avatar! | |
[17:51] | Sindy Tsure does not! just waiting for mono! | |
[17:51] | Sindy Tsure: | :) |
[17:51] | Rex Cronon: | isn't mono here? |
[17:51] | Sindy Tsure: | sorta.. |
[17:51] | Podium: <-0.50013, -0.49987, 0.50013, 0.49987> | |
[17:51] | Andrew Linden: | The MONO deploy started yesterday |
[17:51] | Andrew Linden: | they were investigating some problems today |
[17:52] | Sindy Tsure: | if you just make a plain cube and rotation it y=90, your script works fine, arawn |
[17:52] | Andrew Linden: | dunno if they continued to push it out this morning |
[17:52] | Arawn Spitteler wonders if that's what crashed the database. | |
[17:52] | Andrew Linden: | I heard some chatter about that problem. It wasn't MONO that hurt the db |
[17:53] | Object: Hello, Avatar! | |
[17:53] | Sindy Tsure: | latest from prospero is "Re: the 1.21 roll, as announced in the main blog post, we're on hold as we wrestle with a couple of issues that we detected during the pilot roll. We believe that we've fixed one (a bug that caused a very few regions to go into a crash cycle), but because it's a rare enough case we haven't been able to reproduce the crash on a test grid yet, and thus aren't completely sure we've fixed it. The others we're working on are more subtle." |
[17:53] | Andrew Linden: | but there were some changes to our "central servers" that were happening around the same time |
[17:53] | Andrew Linden: | Ok yup, that blurb from Prospero pretty much sums up my knowledge of the situation |
[17:54] | Sindy Tsure: | oh.. i did have one other thing |
[17:54] | Sindy Tsure: | something that came up ages ago.. physical/phantom vehicle unsitting the person riding it |
[17:55] | Sindy Tsure: | sometimes i seem to bounce off the vehicle and sorta move at an angle, sometimes not |
[17:55] | Tizzers BnT: | if the script resets or does any changes in prim parameters that change the volume of a prim, then the mass has to be recalculated and you get unsat |
[17:55] | Sindy Tsure: | hm.. shoulda prepped for the meeting.. i'll have props to show this next week |
[17:55] | Andrew Linden: | hrm... |
[17:55] | Arawn Spitteler: | I get the same problem, with an ordinary cube |
[17:55] | Andrew Linden: | the phantom vehicle actually affects your avatar... with colliions? |
[17:56] | Andrew Linden: | yeah, that sounds like a bug |
[17:56] | Sindy Tsure: | this is for a waterslide, tizzers.. it's the dismount at the end that sorta throws you in another direction |
[17:56] | Sindy Tsure: | no change in anything, except that it unsits the avatar sitting on it |
[17:56] | Sindy Tsure thinks she's tried llUnsit and llDie.. both have the problem | |
[17:56] | Tizzers BnT: | ah, well when you sit on something you are a llink in the set, if you get unsat, its like two physical prims overlapping: you get a repulsion |
[17:57] | Andrew Linden: | yes Tizzers, but if the seat is phantom that push shouldn't happen |
[17:57] | Rex Cronon: | yes, but if u r in a prim, u don't get pushed out |
[17:57] | Andrew Linden: | what do you mean exactly by "in a prim" Rex? |
[17:57] | Rex Cronon: | i rez a 10^3 |
[17:58] | Rex Cronon: | and drag it around me |
[17:58] | Rex Cronon: | i can move inside it |
[17:58] | Tizzers BnT: | hmm the previous model of the Mercz Uriel was a physical/phantom vehicle, when you unsit from it, you get a significant toss in the air |
[17:58] | Andrew Linden: | oh right, yes there is special code that will disable collision with avatars if they are penetrating too deep in an object (this is the code that is misbehaving and breaking SVC-2511) |
[17:59] | Rex Cronon: | how is that going to be "fixed":) |
[17:59] | Sindy Tsure doesn't have a recent slide.. i'll bring one next week | |
[17:59] | Andrew Linden: | ok Sindy |
[17:59] | Andrew Linden: | Rex, how is what going to be fixed? SVC-2511? or something else? |
[18:00] | Rex Cronon: | yes, svc-2511 |
[18:00] | Arawn Spitteler showed a fun slide, for Lummerland, last sunday, but doesn't have the problem, in dismounting from Soap | |
[18:00] | Arawn Spitteler: | It might be the overlap, of the extra collision space. |
[18:00] | Sindy Tsure: | i think i know how to work around it.. it's not a huge deal, just a bit more work |
[18:00] | Andrew Linden: | I tweaked the code to make the 'avatar penetration' stuff recognize when it was getting lots of distinct penetrations as opposed to repeated penetrations form one object |
[18:01] | Arawn Spitteler: | Prim Wise or Object Wise? |
[18:01] | Andrew Linden: | the logic that detects if the avatar is trapped relies on 'penetration events' + history about previous events |
[18:01] | Sindy Tsure: | oh! that i have a demo for.. :) |
[18:01] | Andrew Linden: | the history resets itself when getting penetrations from new objects |
[18:02] | Andrew Linden: | currently the penetration history phases with the bullets and the logic thinks it is one object |
[18:02] | Andrew Linden: | Arwan... er object-wise. Not exactly sure what you meant by that |
[18:03] | Arawn Spitteler: | Bullets often have two prims. If you're colliding with both those prims, are we talking about a single object, or two prims |
[18:03] | Sindy Tsure bets object | |
[18:03] | Andrew Linden: | one object |
[18:04] | Andrew Linden: | there can be multiple "collision points" between two objects though |
[18:04] | Sindy Tsure: | is this because of the bounding box stuff? "one object, regardless of prim count, gets one box" sorta stuff? |
[18:04] | Arawn Spitteler: | So, a tripode on a platform would be three collision points |
[18:05] | Andrew Linden: | yes Arawn, or more... depending on how it is colliding with something |
[18:05] | Andrew Linden: | each contact point will trigger "events" |
[18:05] | Object: Hello, Avatar! | |
[18:05] | Andrew Linden: | these events are handled, and sometimes fed into some LSL collision_start() event handlers... if there is a script that is interested |
[18:06] | Andrew Linden: | otherwise, there are other stuff that handles the collision events and do various stuff |
[18:06] | Andrew Linden: | including checks to see if an avatar is involved and whether it is "trapped" |
[18:06] | Arawn Spitteler: | The other day, I was leaning against a wall, and went through, when a collision start event shouldhave closed it. |
[18:07] | Andrew Linden: | leaning how arawn? Just standing nearby? |
[18:07] | Andrew Linden: | Or pushing up against the side of the wall? |
[18:08] | Scott Tureaud: | so an object sitting like that would always have two contact points? or more? |
[18:08] | Andrew Linden: | Yes Scott, that object would have two or four contact points, I would guess |
[18:09] | Andrew Linden: | Havok does some logic where it tries to collapse nearby collision points into one. So if two points are very close they sometimes get coalesced |
[18:09] | Sindy Tsure: | hiya kitto |
[18:09] | Kitto Flora: | Hello |
[18:09] | Rex Cronon: | hi |
[18:10] | Andrew Linden: | Kitto, btw I looked into the Cemetery region and couldn't notice the performance problems you were talking about... a few weeks ago |
[18:10] | Andrew Linden: | not that they don't happen, but they didn't happen while I was there |
[18:10] | Kitto Flora: | Wayyyy too late |
[18:10] | Kitto Flora: | The problem was a 'Crow Murder' product |
[18:11] | Kitto Flora: | Its rezzing crows - temps I Think |
[18:11] | Kitto Flora: | They fly around fast |
[18:11] | Andrew Linden: | Ah ok, good to know that a culprit was found and mystery was solved |
[18:11] | Kitto Flora: | The owner tossed it out |
[18:11] | Kitto Flora: | Why it causes lag may be of interest to you |
[18:11] | Andrew Linden: | Simon is interested, I'm sure |
[18:12] | Arawn Spitteler: | http://jira.secondlife.com/browse/SVC-2901 For my mysterious sit displacement. I think I had trouble as a newbi, but now know how to edit local |
[18:12] | Sindy Tsure sometimes checks out laggy region reports from the forums.. 1/2 the time it's somebody rezzing 500 temp prims every minute.. :\ | |
[18:12] | Kitto Flora: | Simon, thats a notecard about the laggy product 'Murder' from 'Blood' |
[18:13] | Kitto Flora: | I get around one laggy sim to debug a week |
[18:13] | Simon Linden: | Sounds like a classic rezzer problem - something creating temp objects that just suck up too much CPU time |
[18:13] | Arawn Spitteler: | I thought it was Asset Server Time |
[18:14] | Kitto Flora: | May residents are having to pick and choose what to rez, for what sim performance they want. IE - do steam trains work |
[18:14] | Kitto Flora: | I just discovered that I can walk through a table thats non-phantom |
[18:14] | Simon Linden: | Well, if the asset server is slow, that just introduces a lag between the time you request an object and when it arrives |
[18:14] | Kitto Flora: | I have an object thats phantom for Avs, but not for physical objects |
[18:15] | Simon Linden: | There's also a big (too big) delay on the sim when rezzing the object - parsing all the info about the shape, texture, position etc, and sticking it into the world |
[18:15] | Andrew Linden: | show me kitto |
[18:15] | Sindy Tsure: | that's the slowdown-when-people-tp-in problem, simon? |
[18:15] | Kitto Flora: | I'd have to TP you there, want to see? |
[18:15] | Kitto Flora: | Its in Remedy II sim |
[18:15] | Andrew Linden: | sure, the office hour is pretty much over |
[18:15] | Kitto Flora: | OK |
[18:15] | Simon Linden: | That problem -- it affects TP, rezzing, sim crossings -- is on our radar and hopefully we can find ways to make it faster |
[18:15] | Kitto Flora: | I'll go back recheck it |
[18:15] | Andrew Linden: | Thanks for coming everyone. |
[18:15] | Simon Linden: | I also have to go .. thanks everyone for coming |
[18:15] | Tizzers BnT: | temponrez is causing performance problems in a number of sims when people are overusing the temponrez scripts |
[18:15] | Tizzers BnT: | thanks Andrew |
[18:15] | Sindy Tsure: | ty andrew and simon! |
[18:16] | Sindy Tsure: | cyas, all! |
[18:16] | Rex Cronon: | bye everybody |
[18:16] | Rex Cronon: | have fun:) |
[18:16] | Tizzers BnT: | friend me Rex? |
[18:16] | Rex Cronon: | u did it just when i tp:) |
[18:17] | Tizzers BnT: | lol sorry |
[18:17] | Arawn Spitteler: | I've got a class to organize. At least I got the picture; nice shot of Kitto and Andrew |
[18:17] | Rex Cronon: | ahh, swt is down agan |
[18:17] | Kitto Flora: | No can demo |
[18:17] | Rex Cronon: | again* |
[18:17] | Kitto Flora: | Noiw its act8ing normally |
[18:17] | Kitto Flora: | acting |
[18:17] | Tizzers BnT: | Andrew, I've got the LL multichair on my main about done, should have it for you next week |
[18:17] | Andrew Linden: | ok, another time then ;-) |
[18:17] | Kitto Flora: | I'll keep an eye open for the effect |
[18:17] | Rex Cronon: | what is going on |
[18:18] | Rex Cronon: | i walk through floor? |
[18:18] | Kitto Flora: | Any word on when other sims get MONO? |
[18:18] | Andrew Linden: | Tizzers, a replacement for this conference table? |
[18:18] | Kitto Flora: | Haha! |
[18:18] | Kitto Flora: | Ok |
[18:18] | Kitto Flora: | so theres the effect? |
[18:18] | Rex Cronon: | sorry |
[18:18] | Andrew Linden: | Kitto, no I'm not informed on MONO. The blog updates are your best source of info. |
[18:18] | Tizzers BnT: | yes one built to look like the linden lab logo |
[18:18] | Andrew Linden: | Oh that would be cool |
[18:18] | Kitto Flora: | Blog didnt even mention that MG has mono - last I looked |
[18:19] | Kitto Flora: | No mono here |
[18:19] | Rex Cronon: | btw andrew, seems like sandbox-weapons testing is down |
[18:19] | Rex Cronon: | trying to tp there and i can't |
[18:19] | Andrew Linden: | Somone had posted an "update" by Prospero talking about the Mono situation, earlier in the hour |
[18:19] | Rex Cronon: | . |
[18:20] | Andrew Linden: | I think the status us currently... MONO is deployed to approximately 500 regions (of of 20k+) |
[18:20] | Andrew Linden: | further deploy is probably on hold (but not totally sure about that) while they look into some issues |
[18:21] | Andrew Linden: | at least one new crash mode was introduced, but rumor is that it is rare |
[18:21] | Kitto Flora: | Ah I see Babbage posted a MONO note last nite |
[18:21] | Scott Tureaud: | atleast we're playing with mono instead of hearing about it :) |
[18:21] | Rex Cronon: | lets see if rausch is up |
[18:22] | Kitto Flora: | I did see severe TD depression in MONO sims |
[18:22] | Kitto Flora: | Ahern and Morris |
[18:22] | Kitto Flora: | Down to 0.15 |
[18:22] | Andrew Linden: | Yeah, rumor is that some regions suffer performance in MONO. Babbage Linden was trying to track that down today, I noticed in some IRC chatter, but don't know what progress was made there. |
[18:23] | Kitto Flora: | OK |
[18:23] | Kitto Flora: | I also went to the big sandboxes |
[18:23] | Kitto Flora: | They seemed fairly normal |
[18:23] | Kitto Flora: | I recompiled a couple planes to MONO and they worked ok |
[18:23] | Kitto Flora: | Thats was around midnight :) |
[18:24] | Andrew Linden: | Yeah, Babbage claimed that it had been a month since a major bug was found in MONO... before it went into the final QA |
[18:24] | Andrew Linden: | they found some in the QA, and now they are finding a few more in deploy |
[18:24] | Kitto Flora: | Yes, but is that a month with no significant new testing? |
[18:25] | Kitto Flora: | I've never found a MONO setup where I could do any serious stuff |
[18:25] | Andrew Linden: | not surprising. That is often how it is... the deploy is actually required in order to bring some bugs to the surface |
[18:25] | Andrew Linden: | really, what do you mean by "serious stuff"? |
[18:25] | Kitto Flora: | Forr sure |
[18:25] | Andrew Linden: | and how did it fail? |
[18:25] | Kitto Flora: | Only on MG can one get the loads |
[18:25] | Kitto Flora: | Serious is a bunch of automatic vehicles running non-stop for days |
[18:26] | Kitto Flora: | Which esposes all the random timing bugs |
[18:26] | Kitto Flora: | Well - a lot of them :) |
[18:26] | Andrew Linden: | ah yes, well vehicle performance problems are independent of MONO |
[18:26] | Andrew Linden: | the vehicle "action" is created by a script call, but then that "action" sits in the physics engine and gets processed |
[18:26] | Kitto Flora: | How you recon 'independent'? They live/die on scripts |
[18:27] | Andrew Linden: | The script events create and destroy yes |
[18:27] | Kitto Flora: | Ah ok - the 'vehicle code' is unchanged? |
[18:27] | Andrew Linden: | Yes |
[18:27] | Kitto Flora: | Yeah but I have complex event (interrupt) action |
[18:27] | Andrew Linden: | legacy LSL has its own binary byte code |
[18:27] | Kitto Flora: | Thats what gets messed up when timing changes |
[18:27] | Andrew Linden: | and mono runs on a new byte code and virtual machine |
[18:27] | Kitto Flora: | Yep |
[18:28] | Andrew Linden: | so the scripts that experience the most speedup are those that are staying in the byte code |
[18:28] | Kitto Flora: | Yes |
[18:28] | Kitto Flora: | I havea big mix |
[18:28] | Kitto Flora: | Theers a lot of math - expect that to get faster |
[18:28] | Kitto Flora: | And theres the events - stays same speed? |
[18:28] | Andrew Linden: | computing fibbonocci numbers (or however you spell it) |
[18:28] | Kitto Flora: | Who uses that? |
[18:29] | Andrew Linden: | but if the script calls a major event... such as llRezObject() well that triggers a big chunk of code that is "outside" of the script engine |
[18:29] | Kitto Flora: | Yep |
[18:29] | Kitto Flora: | I use |
[18:29] | Kitto Flora: | Target() |
[18:29] | Andrew Linden: | all that does is push an asset_id up on a list of things to be downloaded, and the script engine continues |
[18:29] | Kitto Flora: | and Sensor() |
[18:30] | Andrew Linden: | and the rez happens later, once the asset is downloaded, parsed, and so on |
[18:30] | Kitto Flora: | and timer() |
[18:30] | Kitto Flora: | timer timing dont care |
[18:30] | Andrew Linden: | yes sensor sets a flag that is checked later, and events are processed |
[18:30] | Andrew Linden: | however, the byte code logic that switches between events is faster |
[18:30] | Kitto Flora: | Right |
[18:31] | Andrew Linden: | so most scripts should notice a general speedup |
[18:31] | Kitto Flora: | So the relative speeds wil lchange |
[18:31] | Andrew Linden: | and also, there are memory footprint issues that have improved |
[18:31] | Kitto Flora: | and thats what concerns me |
[18:31] | Kitto Flora: | Will that expose an error |
[18:31] | Andrew Linden: | yes, that is possible |
[18:31] | Kitto Flora: | And it often takes hours or even day to hit an error like that |
[18:32] | Andrew Linden: | yes, race conditions can be intermittent |
[18:32] | Kitto Flora: | In Remedy II theres now tow vehicles running non-stop |
[18:32] | Kitto Flora: | on one track |
[18:32] | Kitto Flora: | in opposite directions |
[18:32] | Kitto Flora: | With block control and passing zones |
[18:32] | Kitto Flora: | and across 2 sims |
[18:32] | Kitto Flora: | They have managed to run maybe 48 hours with no error so far |
[18:33] | Andrew Linden: | cool |
[18:33] | Andrew Linden: | so MONO may change things when it gets there |
[18:33] | Kitto Flora: | Thats a huge ammount of logic that has to work perfectly |
[18:33] | Kitto Flora: | asll async timing |
[18:33] | Andrew Linden: | but probably not |
[18:33] | Scott Tureaud: | hmmm speaking of multible SIMS the crossings have gotten better. |
[18:33] | Kitto Flora: | Well .. thats the BIG question :) |
[18:33] | Andrew Linden: | really Scott? since when? |
[18:34] | Scott Tureaud: | although with MONO can you remove some limits like some of the required sleeps? |
[18:34] | Kitto Flora: | I not so sure |
[18:34] | Kitto Flora: | sim crossing between Ahern and Morris has been terrible. (both MONO sims) |
[18:34] | Andrew Linden: | not sure about that Scott. Babbage Linden knows more about whether we'll remove some of the intentional bottlenecks on various script calls |
[18:34] | Andrew Linden: | I hope we do |
[18:35] | Scott Tureaud: | oh yester day one of my vehicles when berserk on a boarder that is normally kinda bad and I didn't crash and stayed caught up. though obviously not mono. |
[18:35] | Kitto Flora: | First airplane cross out of Gougen was very smooth |
[18:35] | Kitto Flora: | but subsequent ones were so-so - no big change |
[18:35] | Scott Tureaud: | kept going in circles right at the edge |
[18:35] | Kitto Flora: | So I dont think anything changed much with MONO |
[18:35] | Andrew Linden: | No, there were no changes there that I know about. |
[18:36] | Scott Tureaud shrugs | |
[18:36] | Kitto Flora: | I expect no change, border wise :( |
[18:36] | Scott Tureaud: | though removing some of the bottle necks would be nice. |
[18:36] | Kitto Flora: | Main Q is if'n vehicles can go a bit faster (track-following ones) |
[18:36] | Andrew Linden: | alright, I've got to run now |
[18:36] | Andrew Linden: | see you around |
[18:36] | Kitto Flora: | cause the calculations happen faster |
[18:36] | Scott Tureaud: | bye then. |
[18:36] | Kitto Flora: | OK |
[18:36] | Kitto Flora: | Byebye |