Chat log from the meeting on 2014-05-27

From OpenSimulator

(Difference between revisions)
Jump to: navigation, search
(Created page with "[11:00] Region found! [11:00] Region found! [11:00] Mata Hari: hi Dan [11:00] Teleport completed from https://osgrid/region/Sandbox%20Plaza/203/38/22 [11:00] Connecting t...")

Revision as of 14:53, 3 June 2014

[11:00] Region found! [11:00] Region found! [11:00] Mata Hari: hi Dan [11:00] Teleport completed from https://osgrid/region/Sandbox%20Plaza/203/38/22 [11:00] Connecting to in-world Voice Chat... [11:00] Connected [11:00] Richardus Raymaker: hi all [11:00] dan banner: hey rich [11:01] Connecting to in-world Voice Chat... [11:01] Connected [11:01] Dev Random: How may avs does it take to overheat my laptop? Fan quit.... 73 deg now.... [11:01] dan banner: better get some canned air [11:01] TBG Renfold: not many by the sounds of it :P [11:01] Mata Hari: I could pile on a ton of mesh if you'd like to really roast it [11:02] Dev Random: brb with Ragegast [11:03] Justin Clark-Casey: hi folks [11:03] Mata Hari: hi Justin [11:03] TBG Renfold: hiya [11:03] Justin Clark-Casey: this is one of those days where I'm being pulled in at least 2 different directions at once [11:03] Sarah Kline: hiyas [11:03] dan banner: heya justin [11:03] Mata Hari: competing teams of horses.....being drawn and quartered? [11:03] Richardus Raymaker: hi justin [11:03] Justin Clark-Casey: mentally, it sometimes feels like it :) [11:04] Nebadon Izumi is Online [11:04] TBG Renfold: had that last week on a project, everyone wanted a piece of me at once [11:05] Nebs Metal Bar Stool v1.5 (w/sit & launch): Hello Nebadon Izumi, enjoy your sit.. [11:05] Nebadon Izumi: hello everyone [11:05] Mata Hari waves [11:05] TBG Renfold: hi [11:05] dan banner: hey neb [11:05] Sarah Kline: hello [11:06] Richardus Raymaker: hi neb [11:06] dan banner: had an issue crop up last night, not sure how isolated it is but it probably needs to be fixed before the release [11:06] dan banner: robert is aware of it [11:06] dan banner: [11:06] Justin Clark-Casey: oh man, that's so not what I like to hear ;) [11:07] dan banner: ya me either [11:08] Mata Hari: what's the issue? [11:08] Justin Clark-Casey: yeah, I think it was always going to be the case that as more people tested 0.8, bulletsim issues were going to come up [11:08] Mata Hari: lag..... [11:09] dan banner: mata from what robert said it has to do with linked mesh/sculpted objects [11:09] Justin Clark-Casey: what I may well do is put RC2 out soon and then just add in necessary bulletsim fixes [11:09] Justin Clark-Casey: otherwise this just drags out indefinitely [11:09] dan banner: ya [11:09] dan banner: RC2 would be good [11:09] dan banner: and by a little time for robert [11:09] Mata Hari: yeah....I meant "lag" in that I asked what the issue was, then you posted the link, then my chat message got printe [11:09] dan banner: buy* [11:10] dan banner: oh your lag got lagged? [11:10] Mata Hari: yes "p [11:10] dan banner: ok [11:10] Justin Clark-Casey: other than buletsim issues, I think we're pretty much there but I might not get time to push it through today [11:11] Mata Hari: there was a patch that someone posted in my Mantis that has a fix for debug line numbers....might be worth including it if it's "good" [11:11] dan banner: the user that had that exception was last attempting to track down which object it was so perhaps we can use it to help reproduce the exception [11:12] Justin Clark-Casey: yes, I will look at that. Really want to avoid much more since every bug fix is unfortunately also a potential regression [11:12] Justin Clark-Casey: which might not get picked up until months down the line [11:12] Mata Hari nods [11:13] Mata Hari: and since Diva seems to be becoming somewhat active again, it would be really nice if she'd take a moment to fix all those HG friends issues [11:13] Richardus Raymaker: you better hang this somewhere around. 99-bugs-in-the-code.jpg [11:13] dan banner: for the most part i think bullet is doing pretty well [11:14] Dahlia Trimble is Online [11:14] Justin Clark-Casey: yes, it' also has vastly better performance from my experience [11:14] dan banner: yup [11:14] Justin Clark-Casey: with ODE, if you make a couple of moderately complex objects physical the physcis frame times leap up [11:15] Justin Clark-Casey: but of course, ODE will be present for a long, long time as a fallback, if not forever [11:15] dan banner: the only other major issue i have seen is the collision bug [11:15] Justin Clark-Casey: yeah, I've seen discussion of that on and off [11:16] Dev Random: actually, I have 100MB of physics logs for radams from a regions where the collisions have stopped working [11:16] Justin Clark-Casey: I'm still confused as to whether it was isolated to bullet only or is actually a general issue [11:16] dan banner: ya it kind of ruins things like teleporters, etc [11:16] Mata Hari: oh....that reminds me.....@Robert, if you want some fun tp to Sandbox Plaza II and *walk out....the orange bits of the landing platform are very slightly raised (maybe 0.1m) and unless you approach them at 90 degress you can get stuck [11:16] dan banner: i switched lbsa back to ode and the teleporters have not failed [11:17] dan banner: on bullet they would break after 12 to 24 hrs [11:17] Mata Hari belatedly realizes that Robert isn't here :p [11:17] Dahlia Trimble waves [11:17] Justin Clark-Casey: heh [11:17] Justin Clark-Casey: hi dahlia [11:17] TBG Renfold: lol [11:17] TBG Renfold: there was me looking around as well lol [11:18] Mata Hari: I was there yesterday and wondered if Dan had added a force field there or something [11:18] dan banner: there where? [11:18] Mata Hari: Sandbox II [11:18] dan banner: ahh [11:18] dan banner: the teleporter? [11:19] dan banner: oh [11:19] dan banner: nvm [11:19] Mata Hari: yes, I just tped in and tried to walk out onto the stuck on the orange thingies [11:19] dan banner: ya [11:19] Dev Random: flypaper [11:19] Mata Hari: and they're barely most 0.1m [11:19] dan banner: small edges are still a little problematic [11:20] Richardus Raymaker: hmm [11:20] dan banner: just need bigger steps lol [11:20] Sarah Kline: wear flat shoes instead of stillettos [11:20] Mata Hari: I was barefoot at the time :p [11:21] Mata Hari: @Justin: I've been watching the inventory fetch closely the last little while and it might be just a coincidence but if I expand all folders when the fetch gets stuck the ones that are invariably incomplete are the V3 ones.....the #Firestorm AO folder and the Outfit folder.....ones where the contents are entirely "links" to other inventory items.....maybe that will help to track it down when you tackle that [11:21] Sarah Kline: i cant say I've had a problem though [11:21] Justin Clark-Casey: mata: It would surprise me if there's a correlation but I'll bear it in mind [11:22] Justin Clark-Casey: it is true that such requests generate a few more inventory calls but that should be swamped by all the others going on [11:22] Mata Hari: maybe a race condition? [11:23] Mata Hari: and that's how it gets "stuck" [11:23] Justin Clark-Casey: My chief susppicion is that the number of inventory calls is overwhelming the inventory service, making it either reply very slowly or some requests timeout [11:23] Justin Clark-Casey: then the throttling signals the server would give to the viewer don't really work properly, so the re-requests are never made [11:24] Justin Clark-Casey: there's a complex mechanism behind this, which is why it's not an easy fix [11:24] Justin Clark-Casey: though I could be wrong, but won't know until I've inveistgafed this possibliity [11:24] Justin Clark-Casey: in other situations, the inventory service may not be so heavily loaded and so others don't see ths issue [11:24] Mata Hari: seems like a distinct possibility [11:24] Justin Clark-Casey: or I might be barking up the wrong tree entirely [11:25] Mata Hari: I find it curious that my AO folder needs to be fetched at all, though, since I haven't made any changes to its contents in months [11:25] Justin Clark-Casey: I believe some viewres are very aggressive with inventory fetch [11:26] Justin Clark-Casey: for AO viewer reasons, I've heard - though I don't know the fine detail [11:26] Justin Clark-Casey: something to do with them requiring a certain asset which they know is linked in the inventory somrewhere but they don't know where, so they have to fetch the whole thing.... [11:26] Dahlia Trimble: Im not all that familiar with inventory, but I thought part of http inventory was to do bulk transactions; that many folders and items could be sent to the viewer in a single request? [11:26] Justin Clark-Casey: dahlia: yeah - not sure how much that facility is used [11:27] Mata Hari: there is a timing component to it never happens in a Plaza (or in SL) [11:28] Justin Clark-Casey: could be because they have very good connections to inventory service [11:28] Justin Clark-Casey: but I'm theoriizing [11:28] Mata Hari: some in some way the region server is involved [11:29] Mata Hari: (which surprises me a little....I would have thought inventory fetch was a transaction between the viewer and the login or asset server rather than passing through the region server [11:29] Justin Clark-Casey: I think it's to do with the requets from the region simulator to the service [11:29] Justin Clark-Casey: which is why in stats you see the number of requests pile up [11:30] Justin Clark-Casey: mata: THey are proxied throug hthe simulator, llike many requests [11:30] Mata Hari: ah [11:31] Dahlia Trimble: I wonder if there is any real need to proxy them at all [11:31] Dahlia Trimble: why would the region need to be involved? [11:31] Dahlia Trimble: other than legacy code [11:32] Justin Clark-Casey: there are experimental ways to go direct to inventory service, I can't remember the security implications of such [11:32] Justin Clark-Casey: at least, I think there are, right Neb? [11:32] Richardus Raymaker: technical you would say its safer if the viewer conenct straight to ribust. but... [11:32] Justin Clark-Casey: it also puts more load on the inventory service directly since theat needs to handle more logic, CAPS, etc. [11:32] Billy.Bradshaw mata: you could try switching off HTTP inventory on the viewer, see if that makes a difference [11:32] Richardus Raymaker: thats the bad side [11:33] Justin Clark-Casey: mata: do you have http inventory disabled? [11:33] Mata Hari: or at least maybe add some way to re-start it via a console command if it gets stuck...anything to not get into perpetual mess since you can't change outfits until it finishes loading that folder [11:33] Mata Hari: I have tried both ways [11:33] Dahlia Trimble: the inventory server needs to be involved anyway unless the region is able to cache somehow [11:33] Mata Hari: currently I have it enabled [11:33] Billy.Bradshaw mata: noted [11:34] Dahlia Trimble: I believe the region needs to be involved for UDP inventory or for setting up CAPS [11:34] Dahlia Trimble: but once the CAPS are set... [11:35] Dahlia Trimble: anyway I'm just curious, not suggesing a rewrite [11:36] Justin Clark-Casey always treats all suggestions as demands for a rewrite ;) [11:36] Mata Hari: lol [11:36] Dahlia Trimble: ok well then I expected it committed by next tuesday :P [11:37] dan banner: with viewer source too ;p [11:37] Richardus Raymaker: 3.. 2.. 1.. [11:37] Mata Hari: and fully tested and debugged by Thurdays :p [11:38] Justin Clark-Casey: ha [11:38] Mata Hari: in general, though, once you manage to get logged in with inventory it's pretty darn stable and smooth [11:39] Dahlia Trimble: sometimes takes several minutes for my AO to start up [11:39] Annabelle Fanshaw is Online [11:40] Caro Fayray: i always log into the sandbox then it loads fast and once done is fine [11:40] Mata Hari: though I think FS has a new bug.....if you log in very high in the air your legs get stuck in "flying" animation [11:41] Mata Hari: I do too, when I didn't last log out stark naked :p [11:41] Justin Clark-Casey: I think these are also issues with a "p2p" approach. You have to somehow avoid one "bad" node from causing problems with the session [11:41] Caro Fayray: ah lol [11:41] Justin Clark-Casey: and the LL protocol wasn't built for that scenario, as they have all their services in the same high performance data centres. I should thinik [11:42] Andrew Hellershanks is Online [11:42] Andrew Hellershanks: Almost missed the meeting [11:43] Dahlia Trimble: Andrew, its ok, as long as you have it implemented, tested, and committed by next tuesday :) [11:44] Andrew Hellershanks: hehe [11:44] Justin Clark-Casey: yep [11:44] Andrew Hellershanks thinks he still has some fine tuning to do on ossearch for event listings. [11:46] Andrew Hellershanks: Should I leave so you can continue talking about me? It seems awfully quiet in here.  ;) [11:47] Caro Fayray: listening:) [11:47] TBG Renfold: < same [11:47] Justin Clark-Casey is debugging something else at the same time [11:47] Mata Hari: don't think there's much else....kind of in a holding pattern until 0.8 I assume [11:47] Dahlia Trimble: we were talking about invenory glitches [11:47] Andrew Hellershanks: I was distracted by the major downpour and minor thunderstorm or I would have been here earlier. [11:48] Kayaker Magic: I'd like to bring up a mantis #6743 [11:48] Caro Fayray: we had that too [11:48] Justin Clark-Casey: it's far easier not ot have big changes right now. Once 0.8 is out it will be more open season again, [11:48] Justin Clark-Casey: though not everybody pays attention to that ;) [11:48] Andrew Hellershanks: I need to go over my pending bug reports to see if there are ones I need to mantis. [11:48] Mata Hari: yeah, totally understand that [11:48] Kayaker Magic: There are complex rules about what llGetRot returns, and most of them work in OpenSim, [11:48] Kayaker Magic: until the avatar sits down, then it returns the wrong thing. [11:48] Andrew Hellershanks: How close are we to 0.8 now? [11:49] Justin Clark-Casey: I am very close to putting out 0.8 rc2 but probably will not have time today [11:49] Andrew Hellershanks: ok [11:49] Justin Clark-Casey: a lot of regression have now been fixed. [11:50] Simulator Version v0.5 shouts: OpenSim 0.8.0 Dev 9bae636: 2014-05-24 00:12:23 +0100 (Unix/Mono) [11:54] Justin Clark-Casey: any other topics right now? [11:54] Andrew Hellershanks: Is the bug re: line number reporting for LSL errors completely squashed? [11:55] Richardus Raymaker: good question to know [11:55] Justin Clark-Casey: Alerci put an additioanl patch in mantis 6764 [11:55] Justin Clark-Casey: would be great if I could have confirmation that it improves matters further [11:55] Andrew Hellershanks: Right. I haven't had a chance to run some of my own tests. Hope to do that soon. [11:57] Mata Hari doesn't know how to or I would [11:57] Andrew Hellershanks: I have large scripts I can use for testing. [11:57] Kayaker Magic: People have differing ideas about that "large" means in a script... [11:58] Andrew Hellershanks: Almost 1,900 lines [11:58] Richardus Raymaker: 10 goto 10 [11:58] Mata Hari: I can test it on one that is 5 scripts working together with a total of around 4000 lines [11:58] Andrew Hellershanks: I have two scripts working together that are almost 3,000 lines [11:59] Richardus Raymaker: thats big [11:59] Justin Clark-Casey: I am surprised anbody can write such a big script in LSL without their eyeballs popping out [11:59] Andrew Hellershanks: Richardus, yup. And it was a pain when the line numbers for errors were out by what seems like hundreds of lines. [11:59] Mata Hari: or more :p [12:00] Andrew Hellershanks: justin, not much different than writing a program in some other language. [12:00] Mata Hari: other than being really, really clumsy [12:00] Justin Clark-Casey: yeah, but you don't get nice stuff like splitting it up into logical files, object oreintation, etc [12:00] Justin Clark-Casey: instead you have to do insane things with strided lists, etc. [12:01] Kayaker Magic: Then there are the fireStorm pre-processor issues, doubles the size of your script and fails to display errors on lines above around 1900 [12:01] Mata Hari: and passing strings back and forth [12:01] Andrew Hellershanks: I did split it up. [12:01] Andrew Hellershanks: OO is for the weak. ;) [12:01] Richardus Raymaker: yes. lsls is not the mosty nice lanbgauge for complex things [12:01] Justin Clark-Casey: I guess you cna split it up, then you have to have some bespoke inter-script comms [12:01] Mata Hari: assembler! [12:01] Justin Clark-Casey: I do think it's a fine language for small things but I wouldn't want to write anything big in it... [12:02] Dahlia Trimble: OO? meh. Replace LSL with Forth :P [12:02] Andrew Hellershanks: yeah. My two scripts communicate with each other at times. [12:02] Andrew Hellershanks: Dahlia, that would be neat. [12:02] Richardus Raymaker: pain full andsrew, especiall if iths physic object and use sensors. that combination is so problematic. or where [12:02] Dahlia Trimble: or better yet, List :) [12:02] Dahlia Trimble: errr Lisp [12:03] Kayaker Magic: listhp [12:03] Justin Clark-Casey: ok, I need to go and attend to business [12:03] Andrew Hellershanks: hehe [12:03] Justin Clark-Casey: thanks for the meeting, folks [12:03] Mata Hari: just write it straight in binary..... [12:03] Sarah Kline: byes [12:03] Dahlia Trimble: cya jcc [12:03] Kayaker Magic: bye! [12:03] Andrew Hellershanks: ok, tc, Justin [12:03] Mata Hari: bye Justin [12:03] Caro Fayray: bye justin and ty [12:03] Annabelle Fanshaw: waves [12:03] dan banner: bye all [12:03] Justin Clark-Casey waves [12:04] dan banner is Offline

Personal tools
About This Wiki