Simulator User Group/Transcripts/2011.09.20
| Prev 2011.09.16 | Next 2011.09.23 |
List of Speakers
| Andrew Linden | Fancy Detector | Flip Idlemind |
| Harlee Fallen | Imaze Rhiano | Joke Bubble |
| Kaluura Boa | Kelly Linden | Kennylex Luckless |
| Latif Khalifa | Leonel Iceghost | Liisa Runo |
| Moundsa Mayo | Moy Loon | MrTodd Barbasz |
| NeoBokrug Elytis | Pauline Darkfury | Rex Cronon |
| Sahkolihaa Contepomi | Simon Linden | Squirrel Wood |
| Stickman Ingmann | TankMaster Finesmith | Vincent Nacon |
Transcript
[12:00] Andrew Linden: That rock looks much cheaper than the other one.
[12:00] Sahkolihaa Contepomi: Haha, Simon said something somewhat on the same lines. xP
[12:00] Meeter: Welcome to the Server User Group
[12:00] Simon Linden: yeah, that's all we think about ... cost and throttles :)
[12:01] Andrew Linden: hrm... still not optimal
[12:01] Andrew Linden: unless you really want that particular shape
[12:01] Andrew Linden: at high rez it looks like a box with pieces cut out
[12:02] Andrew Linden: its lower LOD looks more like a "rock"
[12:02] Sahkolihaa Contepomi: I was just thinking that, heh.
[12:03] Stickman Ingmann: Is Squirel actually present?
[12:03] Andrew Linden: Alright, lets see if we have any news...
[12:03] Sahkolihaa Contepomi: Sitting to your left.
[12:03] Vincent Nacon: nice block
[12:03] Vincent Nacon: (not)
[12:03] Liisa Runo: hiyas ppls
[12:03] Latif Khalifa: heya
[12:03] Andrew Linden: I'm currently working on getting a few projects ready for testing...
[12:03] Andrew Linden: the "faster scripts" project had a new crash bug in it -- fixed that yesterday
[12:04] Fancy Detector: Simon Linden has arrived!
[12:04] Fancy Detector: Andrew Linden has arrived!
[12:04] Fancy Detector: Kelly Linden has arrived!
[12:04] Andrew Linden: LeTigre has a bug when rezzing objects in some regions... SVC-7303. I'm working on that right now.
[12:04] JIRA-helper: http://jira.secondlife.com/browse/SVC-7303
[#SVC-7303] borked LeTigre 11.09.09.240509 install, server placing anything rezzed with velocity at <0,0,0>
[12:04] Rex Cronon: hello everybody
[12:04] Flip Idlemind: ...My internet died just from hearing about that whisper thing
[12:04] Andrew Linden: And also trying to figure out why my maint-server project won't startup for Kelly
[12:04] Latif Khalifa: lol
[12:04] NeoBokrug Elytis: Faster scripts?
[12:04] Pauline Darkfury: lol
[12:04] Andrew Linden: (I'm working on that as I type -- I multitask)
[12:05] Sahkolihaa Contepomi: Yay multitasking.
[12:05] Latif Khalifa: that would explain at least some of the bugs ;)
[12:05] Andrew Linden: Yeah, the "faster scripts" project is just an optimization I made that should make a measurable difference, but isn't a Huge improvement.
[12:05] Andrew Linden: It will be up for test soon on aditi.
[12:06] NeoBokrug Elytis: What's it do? :o
[12:06] Andrew Linden: But at the moment it crashes... if you ever try to delete a script.
[12:06] Andrew Linden: (whoops)
[12:06] Sahkolihaa Contepomi: Nice. :p
[12:06] NeoBokrug Elytis: Awesome list processing time?
[12:06] Flip Idlemind: Who discovered that, and how much did they make
[12:07] Andrew Linden: Flip, that was a "new" bug when I cleaned up how some pointers are passed around.
[12:07] Andrew Linden: Maestro found it for his salary ;-)
[12:07] Pauline Darkfury: So, NULL doesn't make a good Mono binary? :p
[12:08] Andrew Linden: Actually this crash was for double delete on non-NULL pointer.
[12:08] Andrew Linden: My fault. I relearned a lesson about dtor's of base classes.
[12:08] Rex Cronon: no testing before del?
[12:08] Simon Linden: or setting to NULL after deleting
[12:08] Andrew Linden: Yeah, I didn't test it myself enough.
[12:09] Rex Cronon: if meant if(var!=null)...
[12:09] Rex Cronon: i meant*
[12:09] Kelly Linden: that isn't what a double delete is rex. :)
[12:09] Moundsa Mayo: "Pointers Considered Harmful" B^D
[12:10] Andrew Linden: lessee... any other news?
[12:10] Vincent Nacon: augh... stupid fashion expert blogs flooding my RSS feeds.
[12:10] Simon Linden: I don't have much going on
[12:10] Stickman Ingmann: I hear you, Vincent.
[12:10] Simon Linden: ... have been fixing a few crash bugs
[12:10] Andrew Linden: I can't think of any except the mesh-server project should have been promoted to the main channel today.
[12:11] Andrew Linden: Ok, table is open.
[12:11] Vincent Nacon: no it isn't
[12:11] Vincent Nacon: the table is blocked with something on it
[12:11] Vincent Nacon: :P
[12:11] Pauline Darkfury: PRIM_TYPE, PRIM_TEXTURE, and PRIM_COLOR are problematic with PRIM_LINK_TARGET
[12:11] Andrew Linden: hehe
[12:12] Rex Cronon: u guys find so many bugs, that u could become entomologists:)
[12:12] Pauline Darkfury: If you only want to set one of the multi-params set by those, you can't, without doing lots of work to read the stuff you don't want to change
[12:12] Flip Idlemind: Define "lots of work"
[12:12] Simon Linden: hmm, is that something that needs attention, Pauline, or is it just a "that doesn't make sense, don't do it" problem?
[12:12] Andrew Linden: oh yeah, Davido Chrome wanted me to talk about SVC-7250.
[12:12] JIRA-helper: http://jira.secondlife.com/browse/SVC-7250
[#SVC-7250] Collision detection has been seriously off since the Mesh servers were rolled out
[12:12] Pauline Darkfury: e.g. to change the color without the alpha, or the texture offset, without the complete set of texture params
[12:13] Vincent Nacon: how is it a problem?
[12:13] Moy Loon: Yeah, things that just effect the texture repeats, offsets, or just RGB, or A, has always been annoying
[12:13] Pauline Darkfury: It's a problem if you want your object to be as moddable as possible, so only want to change the things you actually want to change, and the the object owner change anything else as they see fit
[12:13] Liisa Runo: isnt that the same behavior we had even before PRIM_LINK_TARGET ?
[12:13] Flip Idlemind: I would certainly like to be able to change alpha without needing to change (or, specifically not change) color
[12:14] Pauline Darkfury: Yeah, it's an old issue, just prevents me converting e.g. 8x llSetLinkColor to a single llSetLPPF
[12:14] Pauline Darkfury: So, there's a new consequence from it in the world that has PRIM_LINK_TARGET
[12:14] Latif Khalifa: PRIM_TEXTURE is what is killing me, as llLinkTexture does 0.2 penalty
[12:14] Latif Khalifa: and I cannot use ALL_SIDES
[12:14] Simon Linden: Is there a jira for it?
[12:15] Pauline Darkfury: Yeah, I think there are old Jiras asking to split most of those out, don't have numbers to hand for them
[12:15] Latif Khalifa: there were several proposals how to fix it, like a constant NO_CHANGE or something
[12:15] Liisa Runo: sure would be usefull
[12:15] Kaluura Boa: SCR-4
[12:15] JIRA-helper: http://jira.secondlife.com/browse/SCR-4
[#SCR-4] Unbundle PRIM_* rules for llSetPrimitiveParams and similar functions to allow for individual parameter settings for all possible parameters.
[12:15] Flip Idlemind: This is probably obvious, but sometimes things are so obvious that you don't notice them: You can use PRIM_LINK_TARGET in llGETLinkPrimitiveParams as well, so you can get the params of multiple prims in a single call
[12:15] Pauline Darkfury: I'd prefer new PRIM_* flags to avoid list length bloat
[12:16] Latif Khalifa: so you could just specify texture UUID without having to set offsets for each side
[12:16] Squirrel Wood: llSetLinkPrimitiveParamsFast - no delay :)
[12:16] Pauline Darkfury: Yeah, I could do a multi-linknum get, but the saving on function calls is lost by all the list access
[12:16] Latif Khalifa: Squirrel, I want to change the texture, not mess with offsets and repeats, llSetLinkPrimitiveParamsFast() is useless in that case
[12:17] Leonel Iceghost: Latif, that would be very userfull for PRIM_TYPE, they are very long lists
[12:17] Latif Khalifa: yep
[12:17] Pauline Darkfury: PRIM_TEXTURE is particularly problematic, as you can't use it with a UUID and commercial textures
[12:18] Flip Idlemind: Idk...my giant cube here became much faster when I stopped using more than one llGetLinkPrimitiveParams call
[12:18] Pauline Darkfury: You also can't get the existing texture UUID
[12:18] Latif Khalifa: have you tried putting textures in a non-root prim?
[12:18] Rex Cronon: what u mean u can't get the uuid?
[12:19] Pauline Darkfury: If the object is not full perm, you get NULL_KEY back with llGet*
[12:19] Flip Idlemind: If its a non-full-perm texture, a script can't get the UUID
[12:19] Moy Loon: If it's not full perms, you can't get the UUID of a texture
[12:19] Flip Idlemind: Which is good, really, but then you can't change any of the texture params
[12:19] Pauline Darkfury: So, you can't use llSLPPF for texture offsets
[12:19] Pauline Darkfury: or repeats, or any other stuff that uses the existing texture
[12:19] Stickman Ingmann: I'm fully behind the points of separating these things out in llSetLinkPrimitiveParams. I've suffered a lot with these limitations. It's very frustrating.
[12:19] Rex Cronon: oh. ok. i always work onlywith my thins of things i have full perm
[12:20] Moy Loon: SRC-4 sounds great!
[12:20] NeoBokrug Elytis: Sorry about that. Left something silly on
[12:21] Object: Hello, Avatar!
[12:22] Pauline Darkfury: Yeah, SCR-4 would be my pref. If you're doing something complex, lots of NO_CHANGE would be undesirable, I think, both due to memory usage constructing the list, and extra processing to deal with it
[12:22] Latif Khalifa: Kelly commented that it would be a big project. I wonder if NO_CHANGE constant would be easier
[12:22] Latif Khalifa: I'd take NO_CHANGE over not having a solution any day :)
[12:22] Andrew Linden: It isn't clear to me how hard that project would be.
[12:23] Pauline Darkfury: Yeah, NO_CHANGE is better than NO_WAY ;)
[12:23] Latif Khalifa: hahaha
[12:23] Andrew Linden: Kelly, do you have any summary statement to make on SCR-4?
[12:23] Flip Idlemind: The "faster scripts" project specifically speeds up accessing prim parameters, if I understand it correctly?
[12:23] Latif Khalifa: How hard would it be to implement [PRIM_TEXTURE, some_id, NO_CHANGE, NO_CHANGE] where it would leave repeats and offsets alone
[12:24] Vincent Nacon: yeah I'd like NO_CHANGE better
[12:24] Liisa Runo: Kelly went poof
[12:24] Andrew Linden: oh
[12:24] Vincent Nacon: we don't need no silly Kelly
[12:24] Rex Cronon: u can have local vars that store the parameters u don't want to change...
[12:24] Stickman Ingmann: I vote for the feature. Don't care what it's called.
[12:24] Moundsa Mayo: Shorten to NO_OP?
[12:24] Flip Idlemind: Well looks like it's unanimous
[12:25] Vincent Nacon: or VOID ?
[12:25] Pauline Darkfury: I still prefer shorter lists as #1 preference. It's also shorter bytecode.
[12:25] Moy Loon: I'd prefer just the single setting things, that way there's not a big list of extra stuff
[12:25] Latif Khalifa: Rex, for textures you would have to read for each face. So cube will require to remember 6 faces repeats, uuids and offsets
[12:25] Moy Loon: I don't wanna have a list of NO_CHANGE, when all I want to do, is change the texture of a few dozen prims
[12:25] Vincent Nacon: yeah I cringed a bit when I saw that big list
[12:25] Rex Cronon: i didn't say would be nice. but u could work with it
[12:26] Moy Loon: It would be better than nothing for sure, by why go for the absoule minimum, why not go for something great!
[12:26] Latif Khalifa: well i can use llLinkTexture to do it, with need of 6 reads and one set 0.2 delay is what I get anyway :P
[12:27] Andrew Linden: I don't want to hazzard a guess on SRC-4. I can't tell how hard or easy it would be, so I'll just accept Kelly's statement of "large project".
[12:27] Andrew Linden: I don't see us getting to that before the end of this year. We've got so many other big projects we want to work on.
[12:27] Vincent Nacon: it is a large project because you'd have to label them with their own integers assignment
[12:27] Latif Khalifa: any you can talk about?
[12:27] Pauline Darkfury: It might be worth further analysis, to see if there's any "low hanging fruit" in a subset of the full list of desired params in there, perhaps?
[12:28] Pauline Darkfury: e.g. some of those would be a direct mapping to an existing llSetLink*
[12:28] Pauline Darkfury: Without seeing the backend, it seems like those cases should be "easy"
[12:28] Andrew Linden: I think I mentioned last week that I might start working on the "interest list" code, which is the part of the simulator that figures out what objects you can see
[12:28] Andrew Linden: and sends the geometry and transform info to the viewer
[12:29] Vincent Nacon: but I noticed that we don't have LSL function for Gravity, Friction, Density, and Rep. yet... unless I missed the memo?
[12:29] Andrew Linden: that is a big project, and we'll probably start it in a few weeks
[12:29] Latif Khalifa: Andrew. it would be a great thing to get a more oprimized intereset list
[12:29] Flip Idlemind: If Kelly were still here I would bring up SCR-91, but since he's not, I won't
[12:29] JIRA-helper: http://jira.secondlife.com/browse/SCR-91
[#SCR-91] llRequestAgentKey (llName2Key)
[12:29] Andrew Linden: I've also got all sorts of maintenance projects that people are asking me to do.
[12:29] Flip Idlemind: Oops, looks like I just did, guess we have to talk about it
[12:29] Latif Khalifa: It drives me mad that sim doesn't send small objects even close by
[12:30] Andrew Linden: yeah, the interest list (IL) really needs some attention
[12:30] Andrew Linden: and some debug hooks so we can measure how well it works as we change things
[12:30] Meeter: Timecheck : User Group is half over
[12:31] Andrew Linden: it is also one of those projects that would help everyone, not just scripters or builders
[12:31] Latif Khalifa: Andrew, will IL work also consider the order of object updates sent? Ie. to rez first stuff really close to you?
[12:31] Simon Linden: We made some improvements in it earlier this year, but it definitely can use some more work
[12:31] Andrew Linden: then, I wanted to take a swing at fixing region crossing bugs before the end of the year
[12:31] Rex Cronon: why can't the viewer say what it wants sent?
[12:31] Simon Linden: Latif - it already does that, but combines size and distance to figure out importance
[12:31] Andrew Linden: dunno if I'll be able to participate in that, but we'll probably have some dev's work on it
[12:31] Pauline Darkfury: Simon's change earlier in the year attempted to improve the sorting by distance, didn't it?
[12:32] Moundsa Mayo: Oooohhhh, that would be greeat for llTeleportVehicleWithPassengers B^D
[12:32] Andrew Linden: right Rex, a "viewer subscription" model has been talked about
[12:32] Simon Linden: I made a few changes ... the biggest was probably in teh way we packed "ObjectUpdateCached" messages
[12:32] Andrew Linden: we're considering that architecture
[12:32] Latif Khalifa: Simon, it seems to factor in other stuff, such as scripted and not scripted, which often results that big far objects get rezzed first
[12:32] Rex Cronon: it will free some time server side
[12:32] Vincent Nacon: I predict that Andrew claims to have made fixes for sim crossing on 12/21/2012
[12:33] Pauline Darkfury: llSetRegionCrossing(DONT_CRASH|NO_BUGS|NO_RUBBERBAND);
[12:33] Simon Linden: Yes - it calculates a score based on size, distance, being an AV ... and I think scripts or selection state
[12:33] Latif Khalifa: I think there might be some error in calculation how big an object would be on screen and that the absolute scale gets higher importance than it should
[12:33] Andrew Linden: Oh yeah, some news on prim encroachment... I spoke to Patch Linden and he's going to be working to get that out.
[12:33] Moundsa Mayo: Pauline, why not llSetSL(PERFECT)?
[12:33] Latif Khalifa: Andrew, for the mainland?
[12:33] Andrew Linden: The encroachment ball in in my court at the moment... I need to write up a summary email on the project to bring Patch up to speed.
[12:33] Pauline Darkfury: YAY! Thanks, Andrew. Any guesses at timescale?
[12:34] Andrew Linden: No guesses really. I'll try to have a guess later this week.
[12:34] Andrew Linden: Yes Latif, for the mainland.
[12:35] Pauline Darkfury: EMs can already enable it as desired
[12:35] Flip Idlemind: So yeah SCR-91 would be great for getting someone's key. Is there any reason it would be a "big project" as well?
[12:35] Andrew Linden: I turned on encroachment for Pauline in two regions last week
[12:35] Andrew Linden: so it is enabled in two mainland regions somewhere...
[12:35] Pauline Darkfury: :)
[12:35] Liisa Runo: is there something more to be done for mainland than just flipping the switch? or are you talking about the sim to another return?
[12:35] Vincent Nacon: Flip, don't you meant llUsername2Key ?
[12:36] Pauline Darkfury: It worked nicely in the cross-region case
[12:36] Andrew Linden: No Liisa, I just think the switch needs to be thrown.
[12:36] Andrew Linden: And the Lindens who work on land and content need to know about it.
[12:36] Flip Idlemind: Username or Legacy Name. I imagine it would be a nightmare to search by display name
[12:36] Vincent Nacon: send them a telegram.
[12:37] Leonel Iceghost: Andrew it would be nice to have a setting in the viewer to control how much object occlusion you use, not just an "on off".. I'm afraid everyone tests those IL things in fast computers, but many people have very slow ones
[12:38] Andrew Linden: Leonel... "object occlusion"? That sounds like a viewer feature. I agree, it should have a setting.
[12:38] Leonel Iceghost: object occlusion != IL?
[12:38] Andrew Linden: No, they are not equal... at least not how I read it.
[12:39] Pauline Darkfury: Yeah, I'd guess that's mostly VWR. I'd have thought it would be a fair amount of pain for the server to figure out what's obscured vs. not (it changes constantly based on camera position & angle)
[12:39] Andrew Linden: IL is sorta like object culling... but it doesn't handle "occlusion"... which to me means: big ObjectA occludes ObjectB, so don't send or draw ObjectB
[12:40] Andrew Linden: But yeah, a smarter IL could help lower-end viewers
[12:40] Simon Linden: yeah, if we did that on the simulator it would just be too slow
[12:40] Andrew Linden: by just not sending too much data, improving cache, and packing it better
[12:40] Flip Idlemind: Occlusion probably shouldn't be done server side, because someone might derender on object viewer-side, and then they don't see anything behind it because the server doesn't know you can't see it
[12:41] Simon Linden: Right, and alpha textures would be pretty tough to check as well
[12:41] Object: Hello, Avatar!
[12:41] Pauline Darkfury: Yup, and consider someone slowly panning the camera around, the server would be constantly re-calcing the occlusion
[12:42] Andrew Linden: but we do have some ideas on how to make the IL smarter which will help everyone, including low-end viewers
[12:42] Andrew Linden: just need to get to work on it
[12:42] Latif Khalifa: yeah optimizing IL would help nicely
[12:42] Latif Khalifa: having written one such "low end" viewer hehehe
[12:42] Andrew Linden: however... a major overhaul of the interestlist probably can't help old viewers...
[12:42] Latif Khalifa: why?
[12:42] Andrew Linden: it may require a protocol change, or a new protocol that old viewers don't know
[12:43] Andrew Linden: er... what I meant was..
[12:43] Simon Linden: I put in some configurable parameters earlier this year, we might start experimenting with those and see if they help at all
[12:43] Vincent Nacon: another way to kill 1.23 viewers
[12:43] Liisa Runo: (Simon, turn off the porn setting from this parcel, not being able to see people outside this parcel is annoying and creepy)
[12:43] Latif Khalifa: IL just determines the order of what object updates you get. What api would that require?
[12:43] Pauline Darkfury: hopefully you'll be able to at least send approx optimal within the limits of the old protocols, as well as any newer shinier ones
[12:43] Simon Linden: ok, I guess I don't need to demo that any more
[12:43] Andrew Linden: yeah, the pressure to just port the 1.23 UI to the new codebase is growing
[12:44] Stickman Ingmann: :D
[12:44] Stickman Ingmann: Yes please.
[12:44] Andrew Linden: actually, that is a good question for you all... why hasn't the open source community backported the 1.23 UI to viewer-2 or -3?
[12:44] Simon Linden: OK, the it's-not-really-called-the-porn-setting is off
[12:44] Latif Khalifa: heh, better late than never :)
[12:44] Pauline Darkfury: lol
[12:44] Leonel Iceghost: because the UI
[12:44] Latif Khalifa: Andrew, they did it the other way around. Backported v2 renderer to v1
[12:44] Leonel Iceghost: and fps
[12:44] Andrew Linden: I don't follow the opensource mailing list closely... too busy with other stuff
[12:45] Vincent Nacon: because viewer source code is messy?
[12:45] Pauline Darkfury: Firestorm has done their approximation of a 1.x-ish UI
[12:45] Latif Khalifa: I'm on a viewer currently with v1 based UI and v2 rendering engine, so I can see mesh and stuff
[12:45] Rex Cronon: i think it has to do with lincences
[12:45] Leonel Iceghost: fps in 2.0 is really bad in low end computers.. overall is the same.. but those few frames that are slow make your gaming impossible
[12:45] Andrew Linden: Latif, what is that renderer backport project called?
[12:45] Pauline Darkfury: It's not the same, but it's much closer
[12:45] Sahkolihaa Contepomi: Well...the OpenGL 3 switch is coming very soon.
[12:45] Sahkolihaa Contepomi: Runitai is finishing off the last bits of it.
[12:46] Latif Khalifa: Andrew, I don't think there is a project name. But 3-4 viewers have the rendering engine from v2 in v1 already
[12:46] Vincent Nacon: and of course.... LL didn't make a proper UI layout structure for custom skinning for viewer 2/3
[12:46] Sahkolihaa Contepomi: https://jira.secondlife.com/browse/SH-2240
[12:46] Pauline Darkfury: It was Henri that did a lot of the work?
[12:46] Sahkolihaa Contepomi: Even a test build of OpenGL 3 is there.
[12:46] Vincent Nacon: I want to stressed that part on "Proper"
[12:46] Andrew Linden: ok, well if those projects are alive then they would be able to port any new IL protocol changes
[12:46] Latif Khalifa: Pauline, Henri and the people of the Singularity project
[12:47] Latif Khalifa: Anrew, I'm curious about what changes would that be?
[12:47] Pauline Darkfury: right, so there's your primary patch / source repos for it, Andrew
[12:47] Kaluura Boa: * PUUURRRRZZZZ *
[12:47] TankMaster Finesmith: the only thing we (ant TPVD viewer) dont yet have working is the ability to upload mesh
[12:47] Andrew Linden: I'm still trying to look over the IL code to understand it better
[12:48] Simon Linden: One protocol change I'd love to make is getting rid of the position quantization we do for messages to the viewer
[12:48] TankMaster Finesmith: thats still being worked on by a few
[12:48] Leonel Iceghost: when you make a new protocol, please make it small or compressed.. it really helps to minimize packet loss outside USA if you have less to download
[12:48] Andrew Linden: but I've got a few ideas on how to tweak the current protocol:
[12:48] Simon Linden: We send the info at lower accuracy to use fewer bits
[12:49] Squirrel Wood: quantization.. I'm really here, but I want you to believe I am way over there where the coordinates are much nicer :)
[12:49] Andrew Linden: (1) the IL currently sends a ObjectKill message for things that have moved out of your field of view
[12:49] Andrew Linden: that message really is indistinguishable from what we send for an object that is really being deleted
[12:50] Andrew Linden: er... we send the same message for both cases
[12:50] Pauline Darkfury: Mmm, and that's the one that leaves the client-ghost-objects
[12:50] Squirrel Wood: which is why the client has to reload it all over again..
[12:50] Andrew Linden: what it should do is send a message that means "the IL is no longer updating this object, so you probably don't want to render it, but you should cache it in case it comes back into view"
[12:50] Pauline Darkfury: Still seem to get those client ghosts fairly often, particularly with vehicles crossing a sim border at close to draw distance
[12:50] Latif Khalifa: Simon, I don't quite understand, aren't updates containging 32 bit floats already?
[12:51] Simon Linden: I'd have to look up the details, Latif, but I remember hitting a few bugs that came down to those limits
[12:51] Harlee Fallen: hey everyone
[12:51] Rex Cronon: tc
[12:51] Andrew Linden: (2) the IL currently doesn't send cacheable data for scripted objects that aren't changing their shape/position -- it should
[12:51] Andrew Linden: Hello Harlee
[12:52] Flip Idlemind: Oh, that reminds me. I'm not sure if this is "interest list" related but it seems like it might me. If I alt-cam far away from myself, objects on my HUD update much slower. Any ideas why that is?
[12:52] Latif Khalifa: Simon, how does sim track positions internally, they are floats not doubles I assume?
[12:52] Rex Cronon: i meant hi:)
[12:52] Andrew Linden: Flip, that might be an IL bug. Sounds like it.
[12:52] Andrew Linden: I'll keep an eye out for it.
[12:52] Simon Linden: We mostly use 32 bit floats
[12:52] Squirrel Wood: (has anyone noticed that far away avatars still animate at hyper speed?)
[12:52] Flip Idlemind: Yes I have
[12:52] Andrew Linden: I believe there are some bugs related to using the avatar position vs the camera position for IL culling calculations. I need to audit that stuff.
[12:53] Flip Idlemind: For, like, a long time now
[12:53] Kennylex Luckless: Yupp
[12:53] Latif Khalifa: Squirel, that' an ancient viewer bug
[12:53] Sahkolihaa Contepomi: 'Probably a JIRA collecting dust somewhere.'
[12:53] Squirrel Wood: must be one of the won't fix ones then :p
[12:53] Flip Idlemind: My guess is that, because HUDs are technically "on my avatar" (even though they're on my screen), if I zoom far away from my avatar, the server things my camera is far away from my HUDs as well. Sound reasonable?
[12:54] Simon Linden: That seems like it's worth fixing ... I'd guess something far away moving a lot would just burn up rendering time
[12:54] Andrew Linden: Oh... back to SVC-7250... before I forget: the new collision physics have definitely affected some vehicles for the worse
[12:54] Sahkolihaa Contepomi: Eep.
[12:54] Squirrel Wood: good :)
[12:54] Andrew Linden: I'm not sure what can be done there. I have to talk to Falcon about it and then play with some vehicles to see if I can do anything that help.
[12:55] Andrew Linden: I really don't know what can be done about SVC-7250. I have to poke around before I can say much more.
[12:55] Squirrel Wood: I can say with confidence that my roller coaster still does not work like it did with havok 1 :)
[12:55] Joke Bubble: hi
[12:55] Andrew Linden: Yes Flip, that is very likely the problem with HUDS.
[12:55] Pauline Darkfury: Very little does work like it did on Havok 1!
[12:55] Vincent Nacon: welcome
[12:55] Meeter: Timecheck : User Group is almost over
[12:56] Rex Cronon: that gizmo that was giving the url to jiras isn't working anymore?
[12:56] Rex Cronon: hi
[12:56] Joke Bubble: tyy vincentt ^.^
[12:56] Joke Bubble: hi rex
[12:56] Sahkolihaa Contepomi: Yeah, sims don't crash from physics within 10 seconds now. :p
[12:56] Andrew Linden: HUD's were added later to the attachment feature set, and I could easily see how the IL aspect of updates was not handled right.
[12:56] Vincent Nacon: reading the JIRA
[12:56] Simon Linden: It was working earlier, Rex ... it won't put out duplicates, so maybe it was mentioned earlier?
[12:56] Andrew Linden: Rex, that gizmo is supposed to NOT re-post jira links to issues that were mentioned already.
[12:57] Andrew Linden: Dunno what its memory timeout is.
[12:57] Leonel Iceghost: I don't know if it is a phoenix feature, but the SCR-number is already a link to me.. so I get it twise
[12:57] Vincent Nacon: guess I can't say much about the collision bug because I haven't come across it yet
[12:58] Rex Cronon: last time it said anything was at 12:29
[12:58] Andrew Linden: SVC-123 test test
[12:58] JIRA-helper: http://jira.secondlife.com/browse/SVC-123
[#SVC-123] llSensor should return the nearest 16 objects
[12:58] Andrew Linden: testing... SVC-123
[12:59] Andrew Linden: I think it is working,.
[12:59] NeoBokrug Elytis: I think it's working too
[12:59] Rex Cronon: u have to wait 20 minutes to test it again:)
[12:59] Vincent Nacon: one JIRA at a time
[13:00] Liisa Runo: testing testing: SCR-90 SCR-184
[13:00] JIRA-helper: http://jira.secondlife.com/browse/SCR-90
[#SCR-90] llGetBoundingBox() returns wrong values on Magnum RC
[13:00] JIRA-helper: http://jira.secondlife.com/browse/SCR-184
[#SCR-184] llMessageLinkedTo( "script name"...
[13:00] Pauline Darkfury: Mmm, that llSensor thing reminds me of a question. Do the various cases that return data via llDetected* use script memory for the detected things? I.e. if we're doing llSetMemoryLimit, do we need to allow enough memory for the invisible storage associated with them?
[13:00] Latif Khalifa: SVC-22
[13:00] JIRA-helper: http://jira.secondlife.com/browse/SVC-22
[#SVC-22] Vehicles crossing region borders aren't always treated as vehicles and can get incorrectly returned if the destination parcel is no-entry or parcel-full
[13:00] Kaluura Boa: (Not-so-subliminal advertizement)
[13:00] Latif Khalifa: had to be done :P
[13:00] Meeter: Thank you for coming to the Server User Group
[13:01] Flip Idlemind: So do you have enough room in your pile to make a note about HUDs when you start working on the interest list
[13:01] TankMaster Finesmith: tc andrew, simon
[13:01] Simon Linden: Thanks everyone for coming
[13:01] Moundsa Mayo: Andrew, Simon, and Kelly in absentia - thanks for your time and all your hard work!
[13:01] Pauline Darkfury: Thanks, Lindens. Have a good afternoon :)
[13:01] Sahkolihaa Contepomi: See you Andrew, Simon.
[13:01] Liisa Runo: thanks everybody
[13:01] Squirrel Wood: Keep on rocking things :)
[13:01] Imaze Rhiano: thanks lindens
[13:01] Kennylex Luckless: I has a question foor good Linens.
[13:02] Andrew Linden: Yeah, I've got to go. So much work to do.
[13:02] MrTodd Barbasz: I have a question
[13:02] MrTodd Barbasz: or a point
[13:02] Vincent Nacon: too late, Todd
[13:02] MrTodd Barbasz: damn
[13:02] Vincent Nacon: but go aehad
[13:02] Kennylex Luckless: How do that work?
[13:02] Pauline Darkfury: Anyone know if the llDetected stuff uses script memory other than when you actually pull the data in?
[13:02] Vincent Nacon: ahead*
[13:02] MrTodd Barbasz: well, i dont mind if a linden listens or not
[13:02] MrTodd Barbasz: I'm not techno boffin
[13:02] Kennylex Luckless: Is that a bad thing to do for it load longer?
[13:02] MrTodd Barbasz: but i've had an account for over 3 years
[13:03] MrTodd Barbasz: i went away from sl for about a year and a bit
[13:03] MrTodd Barbasz: recently come back
[13:03] MrTodd Barbasz: so my question is
[13:03] Vincent Nacon: welcome back!
[13:03] MrTodd Barbasz: whats changed?
[13:03] Pauline Darkfury: If your old account wasn't naughty, you can get it back for US$10
[13:03] MrTodd Barbasz: except there are less people
[13:03] Vincent Nacon: too much list to go on, Todd
[13:03] Rex Cronon: tc all those leaving
[13:03] Leonel Iceghost: search for "release notes second life" "server 11"
[13:03] Pauline Darkfury: everything has changed in some way, but much of it is at least similar in concepts
[13:03] Vincent Nacon: aye
[13:04] MrTodd Barbasz: well, it just strikes me that nothing really goes on anymore
[13:04] MrTodd Barbasz: adn that technology in sl is well behind modern gaming etc
[13:04] Flip Idlemind: We went from Viewer 1 to Viewer 2, then we went from Viewer 2 to Viewer 3. We got Mesh, we got mono2, we got display names, we got prim media, and we got sim versions named after Zoolander poses
[13:04] Sahkolihaa Contepomi: Simon's using an AO! What's the world coming to?!
[13:05] Pauline Darkfury: You can't compare SL to games or MMORPGs. They don't have free-form content which constantly has to be downloaded dynamically
[13:05] Simon Linden: I just want to feel your pain, guys
[13:05] Sahkolihaa Contepomi: xP
[13:05] Pauline Darkfury: hehe
[13:05] Vincent Nacon: no you don't
[13:05] Vincent Nacon: :P
[13:05] Simon Linden: note my nice high script count :)
[13:05] Vincent Nacon: not that you would want to
[13:05] Liisa Runo: SL is pretty much the same as always, == awesome, we have new features, script stuff you can easily find from the LSL wiki, otehr stuff you will just bump in to sooner or later. Prolly noticed we have new viewer, most people dont like it and prefer to use old viewer or some third party viewer.
[13:05] Kennylex Luckless: Only Torley needs a AO.
[13:05] Leonel Iceghost: I think we should compare, before some other company figure out how to do it better Pauline :P
[13:05] MrTodd Barbasz: well, i dunno, its just that sl seems on a major decline
[13:05] Sahkolihaa Contepomi: I'm using a client-side AO so, no scripts involved. :D
[13:05] Pauline Darkfury: Comparing SL to other games is like comparing a Ferrari to an Amtrak
[13:05] MrTodd Barbasz: along with all these avatar things
[13:05] MrTodd Barbasz: but ho hum
[13:06] Rex Cronon: u could take quake engine and make it to run as a sl viewr:)
[13:06] Pauline Darkfury: they both use fuel, travel from A to B, etc, but that's about it for similarities ;)
| Prev 2011.09.16 | Next 2011.09.23 |