Chat log from the meeting on 2014-12-09

From OpenSimulator

Jump to: navigation, search
[11:06]  [2014/12/02 12:05]  Justin Clark-Casey waves
[11:06]  Nebadon Izumi: this is an old bug
[11:06]  Connecting to in-world Voice Chat...
[11:06]  Connected
[11:06]  Nebadon Izumi: it has to do with AppDomainLoading and attachments with scripts
[11:06]  Aine.Caoimhe kk...well once this number gets down to -1 the region seems to go haywire
[11:06]  Shez.Oyen Hidie-ho Justin
[11:06]  Aine.Caoimhe and all scripts in the region cease to work
[11:06]  Nebadon Izumi: that is likely coincidental
[11:06]  Richardus.Raymaker Oh hi justin. hard to see its justin in this suite
[11:06]  Aine.Caoimhe only a restart will fix it
[11:06]  Nebadon Izumi: what is your AppDomain setting set too?
[11:07]  Aine.Caoimhe let me double check that
[11:07]  Nebadon Izumi: this is more suggestive of a bad scrpt
[11:07]  Nebadon Izumi: in your attachment
[11:07]  Nebadon Izumi: if it starts going haywire
[11:07]  Nebadon Izumi: the OSgrid plazas and even these regions
[11:07]  Nebadon Izumi: pretty much always show a negative number
[11:07]  Aine.Caoimhe set at default value of true
[11:07]  Nebadon Izumi: because of the high volume of traffic
[11:07]  Seth.Nygard AppDomainLoading = false, the default setting
[11:07]  Nebadon Izumi: it never seems to make much of a difference
[11:07]  Nebadon Izumi: however
[11:08]  Nebadon Izumi: in Linux we are forced to use false for app domain
[11:08]  Aine.Caoimhe er...sorry, yes, false
[11:08]  Nebadon Izumi: meaning all scripts are a single app domain
[11:08]  Nebadon Izumi: which means scripts can never be unloaded
[11:08]  Nebadon Izumi: so it eventually leads to bloat
[11:08]  Nebadon Izumi: and a bad script can be problematic
[11:08]  Aine.Caoimhe it's an insanely simple script so I'm 99.99999% sure it's not a faulty sscript
[11:08]  Nebadon Izumi: as it will never be unloaded
[11:09]  Nebadon Izumi: what does the script do?
[11:09]  Aine.Caoimhe tint on touch
[11:09]  Nebadon Izumi: can you share it?
[11:09]  Aine.Caoimhe not in my suitcase
[11:09]  Nebadon Izumi: ok no worries
[11:09]  Aine.Caoimhe but it's touch....opens dialog...pick a tint uses llSetLinkPrimParamsFast to change it and closes the listener
[11:10]  Nebadon Izumi: my guess is its causing thread to stop or something
[11:10]  Aine.Caoimhe so it's super super basic
[11:10]  Nebadon Izumi: i wonder if you let it go for like an hour
[11:10]  Nebadon Izumi: if it would recover
[11:10]  Billy.Bradshaw Aine said AppDomainLoading = true,
[11:10]  Seth.Nygard As a side effect we have noticed that after enough times the Active Script count goes negative after which all scripts in the region start to encounter problems.
[11:10]  Aine.Caoimhe no, I was wrong's false (the default)
[11:10]  Billy.Bradshaw kk
[11:10]  Nebadon Izumi: the default is true
[11:10]  Nebadon Izumi: i beleive
[11:11]  Justin Clark-Casey: there is a bug with the script engine in current master if AppDomainLoading = false
[11:11]  Nebadon Izumi: oh really
[11:11]  Nebadon Izumi: when did that start?
[11:11]  Seth.Nygard ;; However, setting this to false will also prevent script DLLs from being unloaded from memory if the script is deleted.
 ;; This may cause an OutOfMemory problem over time when avatars with scripted attachments move in and out of the region.
 ;; Some Windows users have also reported script loading problems when AppDomainLoading = false
 ; AppDomainLoading = true
[11:11]  Nebadon Izumi: something new?
[11:11]  Justin Clark-Casey: one moment, I'm in two conversations here
[11:11]  Justin Clark-Casey: yes, relatively now
[11:11]  Nebadon Izumi: k
[11:11]  Justin Clark-Casey: new
[11:11]  Nebadon Izumi: i can tell you in Encitra grid
[11:12]  Nebadon Izumi: we have a way point generator that rezzes spawn points for our traffic simulator
[11:12]  Nebadon Izumi: i can only use it about 60-70 times before the script engine completely dies
[11:12]  Nebadon Izumi: so every 60-70 waypoints i have to restart the region
[11:12]  Nebadon Izumi: total pain in the ass
[11:12]  Richardus.Raymaker linux i guess ?
[11:12]  Nebadon Izumi: yea
[11:13]  Nebadon Izumi: somtimes our models have 100s of waypoints
[11:13]  Aine.Caoimhe AppDomainLoading = true
[11:13]  Nebadon Izumi: so you can imagine how painful that is
[11:13]  Aine.Caoimhe triple-checked it
[11:13]  Nebadon Izumi: running under windows Aine?
[11:13]  Nebadon Izumi: if your under windows leave it at True
[11:13]  Aine.Caoimhe that's the opensimdefaults.ini value
[11:13]  Nebadon Izumi: if its Linux it absolutely must be false
[11:13]  Nebadon Izumi: or it will crap out relateively quickly from my experience
[11:14]  Justin Clark-Casey: It is and probably has been around since Jul 2014
[11:14]  Aine.Caoimhe kk....that region is linux and I think it may be set to true unless Seth changed it
[11:14]  Justin Clark-Casey: whether it is cayusing other problems I can't say. I actually have a fix but I want to write a rgression test for it too
[11:14]  Seth.Nygard It is the default value, I just checked the ini file
[11:14]  Aine.Caoimhe kk...and default is true
[11:14]  Nebadon Izumi: ok thats bad
[11:14]  Nebadon Izumi: in Linux
[11:14]  Nebadon Izumi: expect major problems
[11:15]  Aine.Caoimhe so that needs to be changed for that region
[11:15]  Nebadon Izumi: yes
[11:15]  Aine.Caoimhe and probably explains a few other regions that are choking on scripts
[11:15]  Nebadon Izumi: that will probably resolve your issue
[11:15]  Seth.Nygard That needs to be changed then for all the regions
[11:15]  Nebadon Izumi: which i suspect is different than what Justin posted
[11:15]  Nebadon Izumi: Justins bug doesnt sound like a region killer
[11:15]  Nebadon Izumi: but more of just an attachment killer
[11:15]  Justin Clark-Casey: There shouldn't be a problem if AppDomainLoading = true
[11:15]  Justin Clark-Casey: it should just make initial script load very slow
[11:15]  Aine.Caoimhe on my self-hosted regions it's .NET.....should I leave it true for them?
[11:15]  Justin Clark-Casey: on linux
[11:15]  Nebadon Izumi: not from my experience justin
[11:15]  Nebadon Izumi: from My experience it kills the engine completely
[11:16]  Nebadon Izumi: after a few 100 scripts
[11:16]  Aine.Caoimhe that's what we've been seeing Neb
[11:16]  Nebadon Izumi: I am not surprised
[11:16]  Andrew.Hellershanks I was working on documentation and almost forgot about the meeting.
[11:16]  Aine.Caoimhe so is this "WAD" behaviour or a bug?
[11:16]  Nebadon Izumi: totally unusuable for me under linux if its set to true
[11:16]  Nebadon Izumi: infact i can make it fail with just 1 script
[11:16]  Nebadon Izumi: if its set to true
[11:16]  Nebadon Izumi: its a nasty script, the Klein Bottle rezzer
[11:17]  Nebadon Izumi: Melanie had the same reaction as you Justin
[11:17]  Nebadon Izumi: she seemed to think it would be fine for up to around 500 scripts
[11:17]  Nebadon Izumi: but my testing fails way short of that
[11:17]  Seth.Nygard So yes, I just confirmed the loaded configuration and AppDomainLoading = true, it is running under Linux
[11:18]  Nebadon Izumi: she of course uses a custom version of mono
[11:18]  Nebadon Izumi: which is probably why she is mistaken
[11:18]  Justin Clark-Casey: hum
[11:18]  Nebadon Izumi: the OSgrid release is false by default
[11:18]  Nebadon Izumi: because we got tired of helping people
[11:18]  Aine.Caoimhe in .NET should I leave it true? and (2) is that something that needs to be logged in Mantis of is it intended behaviour?
[11:18]  Justin Clark-Casey: yeah, really there should be an auto setting or similar - I actually set AppDomainLoading = false in the releases
[11:18]  Justin Clark-Casey: manually
[11:19]  Nebadon Izumi: ya that would be nice
[11:19]  Nebadon Izumi: should be doable
[11:19]  Nebadon Izumi: that would be a bandaid though
[11:19]  Nebadon Izumi: i can tell you
[11:19]  Nebadon Izumi: xengine blows under linux
[11:19]  Nebadon Izumi: even with False
[11:19]  Nebadon Izumi: .net destroys mono for xengine
[11:19]  Nebadon Izumi: xengine in linux is really quite bad
[11:19]  Richardus.Raymaker Blame mono
[11:19]  Nebadon Izumi: yes
[11:20]  Nebadon Izumi: its totally mono
[11:20]  Nebadon Izumi: threading
[11:20]  Aine.Caoimhe so if you're hosting regions with scripts, host them on .NET
[11:20]  Nebadon Izumi: its just not good at it
[11:20]  Nebadon Izumi: for intensive scripts
[11:20]  Nebadon Izumi: you would do better under windows yes
[11:20]  Aine.Caoimhe or that people might visit while wearing scripts
[11:20]  Nebadon Izumi: like lots of timers and sleeps
[11:20]  Nebadon Izumi: just clobbers mono
[11:20]  Nebadon Izumi: but .net its a breeze
[11:20]  Nebadon Izumi: you can probably run 20-30 times as many timers and sleeps in .net than mono before it fails
[11:21]  Nebadon Izumi: but even .net falls very short of what SL can do
[11:21]  Seth.Nygard That perhaps leads to my question. I am currently using Mono 3.2.8, would it be better to use Mono 3.10.0?
[11:21]  Aine.Caoimhe that's a flaw in mono itself?
[11:21]  Nebadon Izumi: the only flaw is Xamarins lack of funding compared to microsoft
[11:22]  Nebadon Izumi: .net has about 10 years on mono
[11:22]  Nebadon Izumi: in terms of development time
[11:22]  Aine.Caoimhe kk....let me rephrase....that's a mono issue not an opensim one?
[11:22]  Nebadon Izumi: well
[11:22]  Nebadon Izumi: perhaps we could accomidate mono better
[11:22]  Nebadon Izumi: but ultimately yes, i think the failure lies at monos feet
[11:22]  Nebadon Izumi: we know it sucks at heavy threading
[11:22]  Justin Clark-Casey: it might be eeven utlimately architectural os issues
[11:23]  Justin Clark-Casey: but tbh, the opensim approach to running concurrent work blows
[11:24]  Nebadon Izumi: id be surprised if the problem stems all the way back to the linux kernel
[11:24]  Nebadon Izumi: because i suspect running mono under windows
[11:24]  Nebadon Izumi: you would expereince the same failures
[11:24]  Nebadon Izumi: i recall trying that years ago
[11:24]  Nebadon Izumi: and it ran just as bad under windows
[11:25]  Nebadon Izumi: hopefully with .net being semi-open source now
[11:25]  Nebadon Izumi: and Xamarin scrambleing to grab that code
[11:25]  Nebadon Izumi: that maybe in the next year we will see massive threading improvement in linux
[11:25]  Nebadon Izumi: but im speculating
[11:25]  Jak Daniels: hi all
[11:25]  Nebadon Izumi: and being super hopeful
[11:25]  Nebadon Izumi: hello Jak
[11:25]  Justin Clark-Casey: I think I read stuff that impleied Microsoft were gong to release their stack on Linux
[11:25]  Justin Clark-Casey: which would make a very itneresting point of comaprison
[11:25]  Shez.Oyen Hi Jak
[11:25]  Nebadon Izumi: where did you read that
[11:25]  Dahlia.Trimble is Online
[11:25]  Nebadon Izumi: i recall reading exactly the opposite
[11:25]  Nebadon Izumi: that microsoft has no intention of doing the work
[11:25]  Justin Clark-Casey: I don't hav ethe link I'm afraid
[11:26]  Justin Clark-Casey: yeah, that's what I would have though, which is why what I read surprises me
[11:26]  Nebadon Izumi: but who knows, i read words from Miguels blog
[11:26]  Nebadon Izumi: that completely say the opposite of what the microsoft blog read
[11:26]  Nebadon Izumi: totally contradictory statements
[11:26]  Nebadon Izumi: which isnt good
[11:27]  Andrew.Hellershanks Perhaps you are talking about this ->
[11:27]  Dahlia.Trimble is Online
[11:28]  Nebadon Izumi: I think there has been a lot of misunderstanding and misinformation put out there by people assuming that microsoft open sourcing their stuff meants they intended on pushing it to other platforms
[11:28]  Nebadon Izumi: but my understanding was that Xamarin and Microsoft have been partnered for the last year
[11:28]  Dahlia.Trimble hi
[11:28]  Billy.Bradshaw Maybe it is perspective, form mono they will include the opensource .net and Microsoft will release .net on Linux
[11:28]  Shez.Oyen Hi Dahlia :)
[11:28]  Nebadon Izumi: and that Miguel was a big part of the negotiations to get microsoft to open the source in a manner that xamarin could share code
[11:29]  Andrew.Hellershanks The article says "...Microsoft is not planning to open source the client side .NET stack..."
[11:29]  Dahlia.Trimble I heard m$oft has an internal team working on a linux port
[11:29]  Nebadon Izumi: though i suspect most of the sharing will be unidirectional
[11:29]  Nebadon Izumi: interesting
[11:29]  Nebadon Izumi: do you have a link to blog or article about that Dahlia?
[11:30]  Dahlia.Trimble not any more
[11:30]  Nebadon Izumi: maybe that is why Miguel seems grumpy about this partially
[11:30]  Nebadon Izumi: his blog seemed a bit negative at times
[11:30]  Nebadon Izumi: which surprised me
[11:30]  Dahlia.Trimble it does make sense for them to do it considering azure is a big part of their future revenue and linux is a big part of azure
[11:30]  Richardus.Raymaker this one ?
[11:31]  Billy.Bradshaw Yes, Microsoft want people to use Azure for Linux
[11:32]  Dahlia.Trimble anyway I'm not holding my breath. Even if they are actively working on a port these things have a habit of not reaching completion
[11:32]  Nebadon Izumi: that seems to contradict their overall business model
[11:32]  Nebadon Izumi: sounds like microsoft is giving up on Windows Server
[11:32]  Nebadon Izumi: other than for hosting linux in the cloud
[11:32]  Nebadon Izumi: god that sounds so horrible to me
[11:32]  Dahlia.Trimble their new business model is exploiting cloud services Nebadon
[11:32]  Nebadon Izumi: why anyone would want to run linux virtually under windows un a production environment
[11:32]  Richardus.Raymaker typ. hi dahlia
[11:32]  Nebadon Izumi: you gotta be nuts
[11:33]  Richardus.Raymaker oh. enough people are nuts nebadon. or you would stop useing office
[11:33]  Nebadon Izumi: I seriously hate the cloud
[11:33]  Nebadon Izumi: i hope it dies
[11:33]  Nebadon Izumi: lol
[11:33]  Jak Daniels: perhaps its more to do with software written in .net running on linux and using microsoft's cloud services
[11:33]  Nebadon Izumi: worst thing ever in my book
[11:34]  Dahlia.Trimble lol I dont see it dieing any time soon
[11:34]  Andrew.Hellershanks I'm waiting to hear the first major breach of a "cloud" site. I think it is only a matter of time given how often people crack their way in to sites to steal passwords.
[11:34]  Nebadon Izumi: yea that seems likely but that also seems like it could make a big dent in their server software business
[11:34]  Richardus.Raymaker the cloud like we use now. i se MeH , bhee bah.
[11:34]  Jak Daniels: but maybe they're after the existing large linux base
[11:34]  Billy.Bradshaw Ballmer also hated cloud services
[11:34]  Nebadon Izumi: to me cloud = all your eggs in 1 basket approach
[11:35]  Jak Daniels: at the moment they can't sell linux users any services
[11:35]  Richardus.Raymaker But first thing thats can explode for me ifs facebook
[11:35]  Nebadon Izumi: from what i read
[11:35]  Nebadon Izumi: Azure is primarly Linux
[11:35]  Nebadon Izumi: its like 10:1
[11:35]  Dahlia.Trimble I hear linux is big on azure
[11:35]  Nebadon Izumi: over windows servers
[11:35]  Andrew.Hellershanks I recently discovered one site that provides free cloud storage only does so for the first year (or two?). After that it seem you have to pay for it.
[11:36]  Richardus.Raymaker hi andrew
[11:36]  Andrew.Hellershanks That limitation wasn't exactly all that noticeable
[11:36]  Andrew.Hellershanks Hey, Richardus
[11:36]  Jak Daniels: maybe they finally realised... if you can't beat 'em, join 'em
[11:36]  Nebadon Izumi: well it will certainly be interesting to see if microsoft does push their own .net to linux
[11:37]  Nebadon Izumi: that wreaks of desperation to me though
[11:37]  Jak Daniels: hehe yeah
[11:37]  Andrew.Hellershanks How many bugs and security holes will it have? ;)
[11:37]  Nebadon Izumi: not a good sign that Microsoft is a healthy company
[11:37]  Dahlia.Trimble the new ceo was the guy who was in charge or azure so he probably has a different perspective than the old guard
[11:37]  Jak Daniels: probably about the same as mono ;)
[11:37]  Nebadon Izumi: even though it could be good for us
[11:37]  Nebadon Izumi: yea well Ballmer was not a good leader
[11:38]  Nebadon Izumi: he was a club swinging cave man
[11:38]  Andrew.Hellershanks Perhaps they are just really hoping the OpenSource community can fix their software bugs for them. :)
[11:38]  Dahlia.Trimble I think Ballmer was a good leader, but not a good visionary
[11:38]  Richardus.Raymaker If the just make windows desktop betetr and dump the whole app disaster it would work perfect on tablet.. The spend to much money and energy in soemthiing not userfriendly
[11:38]  Justin Clark-Casey: ms is an extremely profitable company
[11:38]  Justin Clark-Casey: a new ceo brings in a new direction
[11:39]  Nebadon Izumi: yea still, they are not profitable as they were at the peak of Windows 7
[11:39]  Nebadon Izumi: far from it
[11:39]  Dahlia.Trimble could be worse, m$oft cound have brought in someone from outside to be ceo
[11:40]  Nebadon Izumi: ya its hard to say right now, time will tell on that
[11:40]  Nebadon Izumi: hehe
[11:40]  Nebadon Izumi: Microsoft looks ugly to me right now though
[11:40]  vegaslon.plutonian they have done alot of branching of their business
[11:41]  Dahlia.Trimble dunno. I tend to like m$oft more than others do ;)
[11:41]  Sarah Kline:
[11:41]  Richardus.Raymaker Well, i dont like apple. so only onme option left and microsoft works best :O
[11:41]  Nebadon Izumi: for a long time I was an ardent microsoft supporter, but lately i feel have a hard time justifying any kind of interest in their current offering of products
[11:41]  Jak Daniels: I've had a love/hate relationship with them since about 1992
[11:42]  Richardus.Raymaker More hate then love. but last time its betetr since i discover that linux is hell.
[11:42]  Dahlia.Trimble oh they do annoy me now and then but there are a lot of companies out there I have much less respect for
[11:42]  Nebadon Izumi: ya that is true
[11:42]  Richardus.Raymaker Oh Dahlia. that wil stay
[11:42]  Nebadon Izumi: *cough* apple *cough*
[11:42]  Dahlia.Trimble lol
[11:43]  Andrew.Hellershanks All depends on why you need to use MS stuff. I do most of my regular stuff without it.
[11:43]  Aine.Caoimhe PS 2014 is impossibly in Linux
[11:43]  Nebadon Izumi: I just wish Adobe would get on the Linux express
[11:43]  Aine.Caoimhe otherwise I would have switched years ago
[11:43]  Nebadon Izumi: that would definitley be bad news for microsoft i think
[11:43]  Jak Daniels: for a long time now, I've been windows desktop but all linux for servers
[11:43]  Nebadon Izumi: Adobe is going the other direction
[11:43]  Richardus.Raymaker Well, i used linux in the past. but problems styarted when desktops changed, after that hardware support did go fast down. and now its a big pain and nothing runs right.
[11:43]  Nebadon Izumi: I think adobe is even killing off Adobe reader for linux
[11:44]  Dahlia.Trimble I play a lot of games that just dont work well on linux. I also do a lot with graphics and linux is just way behind there
[11:44]  Andrew.Hellershanks I used to need MS for video editing but now, Lightworks has been release so I may not need MS for video editing any longer.
[11:44]  Nebadon Izumi: I tried Lightworks, didnt like it
[11:44]  Richardus.Raymaker i avoid adobe if possible , appliucations are anyway bad like adove with the customized bad UI
[11:44]  Nebadon Izumi: ya but Photoshop has no competitors
[11:44]  Richardus.Raymaker Andrew, never seen a good working editor on linux
[11:45]  Nebadon Izumi: gimp is like Windows Paint compared to Photoshop
[11:45]  Andrew.Hellershanks Nebadon, it is aimed at the professional viedo editing market.
[11:45]  Richardus.Raymaker hmm, paint shop pro works perfect for me. and its cheaper
[11:45]  Jak Daniels: adobe = bloatware... a 100+Mb application just to read PDF files ;)
[11:45]  Andrew.Hellershanks Richardus, I haven't either until I saw Lightworks.
[11:45]  Nebadon Izumi: dont get me wrong I like GImp too
[11:45]  Jak Daniels: I still uses Paint shop pro too lol
[11:45]  Aine.Caoimhe there are features in PS that just aren't anywhere else....and for people working heavily with textures they're kind of critical
[11:45]  Nebadon Izumi: but I absolutely cant live without photoshop myself
[11:45]  Aine.Caoimhe nor can I
[11:46]  Aine.Caoimhe I can do 70% of my stuff in Gimp if I have to, but the other 30% is a game-breaker
[11:46]  Andrew.Hellershanks Aine, maybe not critical but would take more work to accomplish in something else.
[11:46]  Richardus.Raymaker never used it, complex program cant read the menu to
[11:46]  Richardus.Raymaker gimp is good for some things yes
[11:46]  Dahlia.Trimble I think gimp is just as powerful but photoshop is a lot easier to use, probably just because I'm more familiar with it
[11:46]  vegaslon.plutonian adobe creative cloud does look promising, expecially since it works on chrome books
[11:46]  Andrew.Hellershanks GIMP isn't any more complex a menu system than PS. It is just different.
[11:47]  Aine.Caoimhe I need to be able to use layers to modify layers all the time = can't use gimp
[11:47]  Richardus.Raymaker chrome books = evil cloud
[11:47]  Aine.Caoimhe and GIMP forces a lot of destructive edits
[11:47]  Nebadon Izumi: yea perhaps dahlia, I have tried to switch over many times, but its very hard to switch over
[11:47]  Andrew.Hellershanks Aine, there is work under way that will change that.
[11:47]  Aine.Caoimhe sure
[11:48]  Aine.Caoimhe I figure anything in PS will be in Gimp about 4-5 years later
[11:48]  Dahlia.Trimble gimp has a lot of plugins for free that you have to pay extra for on PS
[11:48]  Aine.Caoimhe but but then PS will have something new and shiny I can't live without either :p
[11:48]  Richardus.Raymaker well, let PS first start with useing windows fonts and dpi more for the menu's etc.
[11:48]  Aine.Caoimhe if something saves me hundreds of hours of work, I'll happily pay for it
[11:49]  Andrew.Hellershanks There is always new and shiny stuff in programs that you wonder how you ever lived without it after you start using it.
[11:49]  Dahlia.Trimble shiny? *runs to PS camp*
[11:49]  Andrew.Hellershanks :)
[11:49]  Justin Clark-Casey: as there are 10 mins left, any other opensim issues?
[11:49]  Seth.Nygard just wants OpenSim to behave well under mono, can live with Gimp
[11:49]  Aine.Caoimhe they serve cookies and cake too, Dahlia
[11:49]  Dahlia.Trimble cookies? o.0
[11:49]  Seth.Nygard What version of Mono is best to use? I am currently using 3.2.8, but have seen some notes that perhaps 3.10.0 would be a better choice.
[11:49]  Richardus.Raymaker YEah, why do we still not ahve nice mesh trees and plants :O :)
[11:49]  Aine.Caoimhe 1: mesh isn't flexi
[11:50]  Nebadon Izumi: oh
[11:50]  Nebadon Izumi:
[11:50]  Nebadon Izumi: i didnt see this before
[11:50]  Dahlia.Trimble I prefer
[11:50]  Andrew.Hellershanks Does llSetTexture have the limitation it does in SL where it won't work with textures for which you don't have full perms? I think the main limitation is if you didn't have mod perms
[11:50]  Nebadon Izumi: Dated Nov 18th
[11:50]  Aine.Caoimhe 2: poly count would be insane
[11:50]  Justin Clark-Casey: Idon't know. 3.2.8 is quite a common mono version for dsitributions so tends to get used quite a lot
[11:50]  Justin Clark-Casey: andrew: I didn't know that was an sl limitation. From the code, it looks like there is no limitation in os
[11:50]  Andrew.Hellershanks Justin, ok.
[11:50]  Seth.Nygard I've seen some comments that 3.2.8 may still have some garbage collection issues that are better in 3.10
[11:51]  Dahlia.Trimble I have way more memory use when I use any mono beyond
[11:51]  Richardus.Raymaker Oh.. about scripting. why is opensim so easy with command errors what SL start to flip and throw error. like with KLeyFrameMotion you can use command where its not allowed. (but it still works in opensim)
[11:51]  Justin Clark-Casey: I have not had any experience with gc issues, seems fine in 3.2.8 to me
[11:51]  Dahlia.Trimble its like double or worse
[11:51]  Richardus.Raymaker Oh well, i know what justin is saying now. "scripting need much more love in opensim"
[11:51]  Seth.Nygard Also I have twice seen a console comment about nursery size, is there a recommended setting for the nursery size in GC_PARAMS for mono?
[11:52]  Justin Clark-Casey: richardus: script needs much more work in opensim :)
[11:52]  Richardus.Raymaker :)
[11:52]  Nebadon Izumi: i suspect its going to be a while before we see any .net stuff on linux from Xamarin or Microsoft
[11:52]  Nebadon Izumi: that road map sounds very long winded
[11:52]  Andrew.Hellershanks hopes he an get his main computer running again in the new yar
[11:52]  Justin Clark-Casey: xamarin are porting quite a lot of stuff but not all, esp where they arelady have better multi-platform support
[11:52]  Andrew.Hellershanks s/yar/year/
[11:52]  Justin Clark-Casey: but yeah, who knows when ms will publish anything
[11:53]  Justin Clark-Casey: on non-windows platforms
[11:53]  Nebadon Izumi: Seth > export MONO_GC_PARAMS=nursery-size=64m < is what i have been using
[11:53]  Nebadon Izumi: its what is in use here on this region
[11:53]  Seth.Nygard OK, thanks. I'll try than and see if it makes any difference
[11:53]  Nebadon Izumi: it can
[11:54]  Nebadon Izumi: i was having spurious crashes without that
[11:54]  Justin Clark-Casey: really? mono vm crashes?
[11:54]  Nebadon Izumi: yea
[11:54]  Justin Clark-Casey: urgh
[11:54]  Nebadon Izumi: the default is 4mb
[11:54]  Aine.Caoimhe would be nice if some of those tweaks were collected in a document somewhere
[11:54]  Nebadon Izumi: these were heavy regions
[11:54]  Nebadon Izumi: like Wright Plaza etc..
[11:54]  Dahlia.Trimble my little gridlet has been running very well lately, hardly ever a crash
[11:54]  Justin Clark-Casey: there is, the wiki performance document
[11:54]  Justin Clark-Casey: but I only document what I see or can test directly
[11:55]  Sarah Kline: Wrong Plaza is brill
[11:55]  Justin Clark-Casey: it's up to other people to document tweaks that they've tested
[11:55]  Aine.Caoimhe nods....was kind of hinting that that ;)
[11:55]  Nebadon Izumi: you can see my mono docs here
[11:55]  Nebadon Izumi: I tend to put tweaks and docs on my own archive forum
[11:55]  Justin Clark-Casey: hoarder! :)
[11:55]  Andrew.Hellershanks lol
[11:56]  Nebadon Izumi: some people may find my stuff a bit drastic / wreckless maybe
[11:56]  Nebadon Izumi: lol
[11:56]  Aine.Caoimhe how about a link to them from the main wiki
[11:56]  Nebadon Izumi:
[11:56]  Nebadon Izumi:
[11:56]  Nebadon Izumi: here is the whole archive
[11:56]  Justin Clark-Casey: yeah, can always put a reference to neb's stuff from the perf wiki page if anyone wants
[11:56]  Nebadon Izumi: ya please test my stuff
[11:56]  Nebadon Izumi: i tend to do very drastic / bleeding edge testing
[11:56]  Aine.Caoimhe I think the more information that's available and semi-easy to locate would be extremely helpful
[11:56]  Justin Clark-Casey: neb: eventually drastic/reckless stuff gets refined. HOwever, not everyone is comfortable with that kind of process
[11:56]  Nebadon Izumi: that may not be suitable for others
[11:57]  Nebadon Izumi: yes
[11:57]  Nebadon Izumi: thats only reason i avoid wikifying my knowledge most of the time
[11:57]  Nebadon Izumi: until others are testing and using it
[11:57]  Dahlia.Trimble I dont care for bleeding edge dependencies myself...
[11:57]  Seth.Nygard I am happy to help test things, especially if they serve the greater community
[11:57]  Aine.Caoimhe kk...but unless they know about it it's hard for them to test it :p
[11:57]  Nebadon Izumi: true
[11:57]  Nebadon Izumi: but i talk a lot about what i do in various places
[11:58]  Nebadon Izumi: mostly IRC
[11:58]  Justin Clark-Casey nods
[11:58]  Nebadon Izumi: not like i hide my stuff either
[11:58]  Nebadon Izumi: hehe
[11:58]  Justin Clark-Casey: more than most ppl
[11:58]  Andrew.Hellershanks Nebadon, topic 44 is about compiling mono
[11:58]  Nebadon Izumi: i tend to work on the extreme end of the spectrum
[11:58]  Nebadon Izumi: crazy big models
[11:58]  Nebadon Izumi: the current model im working on is 123 regions
[11:58]  Nebadon Izumi: in land size
[11:58]  Aine.Caoimhe I guess the issue is that when one is looking for information one might not have been logging all IRC conversations for the last 5 years (or be able to quick-search it for that tidbit of useful advise)
[11:59]  Nebadon Izumi: yea, im not sure having it on the wiki wold make finding it any easier to be honest
[11:59]  Nebadon Izumi: lol
[11:59]  Aine.Caoimhe much the same being true for Hiro's using IRC as a new dissemination method
[11:59]  Aine.Caoimhe *news
[11:59]  Nebadon Izumi: thats another reason i use my own forum
[12:00]  Nebadon Izumi: because i can never find anything on the wiki
[12:00]  Aine.Caoimhe hehe..well yeah, that can be a challenge too
[12:00]  Justin Clark-Casey: wiki does get crawled by search engines. I jnprinciple an external service could archive IRC but I haven't seen it done a lot
[12:00]  Justin Clark-Casey: possibly considered an invasion of privacy
[12:00]  Andrew.Hellershanks Forums are ok for archiving info but I don't like them for discussions
[12:00]  Dahlia.Trimble I use google to search the wiki
[12:00]  Andrew.Hellershanks Some IRC channels are archived
[12:01]  Justin Clark-Casey: wiki also tends to bre more organized info, whereas IRC is discussion
[12:01]  Aine.Caoimhe nods
[12:01]  Aine.Caoimhe and trying to locate stuff from old meeting logs is also very hit and miss
[12:02]  Aine.Caoimhe so "formalizing" some of the more important bits in one easy-to-find place would be handy
[12:02]  Justin Clark-Casey: doc writing takes a lot of work though. I think many ppl are uncomfortable doing it unless they are expert in the area.
[12:02]  Dahlia.Trimble I try to hang out on IRC and answer questions when I know the answer. Many of the channels have been quite dead for a while tho.
[12:02]  Nebadon Izumi: ya IRC is like a wave, lots of ups and downs
[12:02]  Andrew.Hellershanks doc writing takes a lot of time. I always remember that when i have to document the scripted gadgets I make.
[12:03]  Aine.Caoimhe I'm relatively expert in documentation but lack the expertise in the subject
[12:03]  Andrew.Hellershanks I wrote some documentation for a particle generator I made. Took 6 notecards.
[12:03]  Dahlia.Trimble my documentation expertise has waned quite a bit. "Use it or lose it" effect :/
[12:04]  Aine.Caoimhe so I can't document Opensim because I lack the opensim-specific knowledge to make it useful
[12:04]  Nebadon Izumi: yea from my experience anything i document i generally have to end up explaining to people again anyway
[12:04]  Richardus.Raymaker |Depends. if you have a good OSSL example you could add that.
[12:04]  Justin Clark-Casey: I used to think that people could make wiki changes anyway and that would get refined over time
[12:04]  Justin Clark-Casey: but I now think you need quite a high critical mass for that to happen, like wikipedia level
[12:05]  Nebadon Izumi: yea wikipedia has 1000s of editors
[12:05]  Andrew.Hellershanks Nebadon, No matter how good the documentation there are always people who won't read it.
[12:05]  Nebadon Izumi: and army of bots
[12:05]  Justin Clark-Casey: otherwise old info sits about for a long time, though that may still be better than nothing
[12:05]  Aine.Caoimhe it's also very hard to know what's "general" vs vase-specific
[12:05]  Aine.Caoimhe *case
[12:05]  Aine.Caoimhe or variance between OS
[12:06]  Andrew.Hellershanks I've occasionally updated small bits of the wiki but have too many other projects to devote a lot of time to reviewing and updating it.
[12:06]  Seth.Nygard I do have another question regarding .NET and linux then. has anyone tried to see how a .NET runtime works with OpenSim under wine? Although I am loathe to consider that a solution.
[12:06]  Richardus.Raymaker Yiks
[12:06]  Nebadon Izumi: not possible
[12:06]  Justin Clark-Casey: afaik, wine uses mono for its .net support
[12:06]  Nebadon Izumi: .net needs kernel support
[12:06]  Justin Clark-Casey: I don't know if you could get actual MS .net code running
[12:06]  Nebadon Izumi: i dont think you can
[12:07]  Nebadon Izumi: maybe like .net 1
[12:07]  Richardus.Raymaker sounds a bit bloated with overhead
[12:07]  Nebadon Izumi: but anything past that probably not
[12:07]  Nebadon Izumi: its very integrated into windows
[12:07]  Seth.Nygard Yea, I certainly would expect issues.
[12:07]  Justin Clark-Casey: but any more so than the rest of the windows code?>
[12:07]  Justin Clark-Casey: I'm sure wine has been dealing with those kinds of issues for a decade+
[12:08]  Nebadon Izumi: other than that Wine is actually quite good
[12:08]  Nebadon Izumi: i even got Grand Theft Auto 4 running pretty good in Wine
[12:08]  Dahlia.Trimble wine uses mono for windows to do .net compatibility
[12:08]  Justin Clark-Casey: I was playing diablo 3 on wine, no problems
[12:08]  Aine.Caoimhe so really the only choice is just restart regions a lot if they're running under mono and tell the region owners to avoid using scripts or having visitors who wear scripts (even when scripts are disabled in the region) or else run under .NET
[12:08]  Andrew.Hellershanks Still a apps that Wine won't run due to incomplete, or buggy, implementation.
[12:08]  Nebadon Izumi: i wouldnt say a lot
[12:08]  Dahlia.Trimble I have lots of problems with games on wine
[12:08]  Nebadon Izumi: ive had regions run for 100s of days
[12:09]  Nebadon Izumi: i just restarted avacon grid about a month ago it was running for 121 days
[12:09]  Justin Clark-Casey: really, there shoudl be an option so that attachments scripts are run under their own AppDomains and other scripts not
[12:09]  Nebadon Izumi: yea
[12:09]  Aine.Caoimhe a party with 30 HG visitors required 2 restarts in a 2 hour period
[12:09]  Nebadon Izumi: that would be nice Justin
[12:09]  Andrew.Hellershanks I have an App I'd like to run under Windows but it won't run because the licencing portion of the program fails.
[12:09]  Justin Clark-Casey: so that at least one could unload attachment scripts, but this is non-trivial work
[12:09]  Nebadon Izumi: Aine that would be unusual
[12:09]  Aine.Caoimhe but that's with apploading true
[12:09]  Nebadon Izumi: OSCC didnt ahve to restart hardly at all
[12:09]  Justin Clark-Casey: well, then you have a different problem I would say
[12:09]  Nebadon Izumi: we had 1 crash
[12:10]  Aine.Caoimhe so will be interstng to see how that changes with it false
[12:10]  Nebadon Izumi: yea it will be huge
[12:10]  Justin Clark-Casey: as I said , there is a bug currently and I will kbe fixing soon
[12:10]  Nebadon Izumi: you likely wont have to restart very often now
[12:10]  Nebadon Izumi: the biggest down side to false is memory bloat
[12:10]  Justin Clark-Casey: it may or may not have knock on effecst on other situations, though I would also only expect it to be an issue on simulators with more than one region
[12:10]  Nebadon Izumi: not being able to unload scripts leads to more memory that needed, but shouldnt lead to much instability
[12:11]  Nebadon Izumi: i never run more than 1 region Justin
[12:11]  Andrew.Hellershanks Has anyone tested Warp3D with large (or largish) var regions?
[12:11]  Nebadon Izumi: and its a problem almost 100% of the time
[12:11]  Nebadon Izumi: I have Andrew
[12:11]  Nebadon Izumi: works fine
[12:11]  Seth.Nygard At least the server has quite a bit of extra memory so a bit of bloat would be preferred over having the region hang
[12:11]  Justin Clark-Casey: warp3d is another bloody problem
[12:11]  Andrew.Hellershanks yeah.
[12:11]  Nebadon Izumi: works no different than a 256m region anyway
[12:11]  Justin Clark-Casey: stupid memory leaking
[12:11]  Nebadon Izumi: same limits and bugs
[12:11]  Aine.Caoimhe Freaky has fixed that in Ariba
[12:12]  Nebadon Izumi: i have gotten away from doing map tile generation though
[12:12]  Nebadon Izumi: i try to hard code tiles myself now
[12:12]  Aine.Caoimhe you might want to peek at his code since it's published, Justin
[12:12]  Nebadon Izumi: the region starts up so much faster
[12:12]  Richardus.Raymaker i use warp3d on 3x3 var. but i think its set to only generate at startup
[12:12]  Justin Clark-Casey: aine: the warp3d dll?
[12:12]  Jak Daniels: warp3d is ok if you only use it to generate a map at startup
[12:12]  Aine.Caoimhe yes
[12:12]  Andrew.Hellershanks I've had another thought how I might fix my MB (or what th eproblem might be). It is a pain wanting to work on some OS code issues and not being able to do so.
[12:12]  Aine.Caoimhe he was telling me that last Friday
[12:12]  Justin Clark-Casey: I mean, the source code for it
[12:12]  Nebadon Izumi: what the heck is Ariba?
[12:12]  Justin Clark-Casey: aine: ok, I will take a look somtime, thanks
[12:13]  Andrew.Hellershanks Jak, that is fine for regions that don't often change but not so good for those that do unless you do frequent restarts.
[12:13]  Richardus.Raymaker Never heared about aribva. some opensim clone. i would not use myself
[12:13]  Aine.Caoimhe Arriba is the fork of Opensim that Freaky is doing
[12:13]  Nebadon Izumi: oh
[12:13]  Aine.Caoimhe a lot of major optimizations in it
[12:13]  Nebadon Izumi: ya first I heard of that
[12:13]  Richardus.Raymaker more opensim fork sorry
[12:13]  Aine.Caoimhe and a lot of fixes
[12:13]  Jak Daniels: but the problems with warp3d was what got me to code up the static maptiles options
[12:13]  Andrew.Hellershanks Another fork of OS??
[12:13]  Aine.Caoimhe unfortunately he doesn't seem to think he is able to contribute code back into master because of the agreement or whatever
[12:14]  Nebadon Izumi: not sure why he would think that
[12:14]  Nebadon Izumi: heh
[12:14]  Aine.Caoimhe but there are a lot of people running it for higher traffic regions
[12:14]  Aine.Caoimhe and those ones tend to perform much, much better than ones that don't
[12:14]  Nebadon Izumi: i dont think i ever met or spoke with this person though, name doesnt seem familiar
[12:14]  Justin Clark-Casey: For performance claims I always want to see data
[12:14]  Aine.Caoimhe he used to come to dev meetings
[12:15]  Justin Clark-Casey: it's easy to make them on subjective feel but that's not scientific
[12:15]  Andrew.Hellershanks Freaky, as in Freaky Tech? That name seems familiar
[12:15]  Aine.Caoimhe ues
[12:15]  Nebadon Izumi: oh hrmm ok ya maybe
[12:15]  Nebadon Izumi: so many people
[12:15]  Nebadon Izumi: and i suck with names
[12:15]  Richardus.Raymaker I dont think ariba runs much betetr. sofar i have seen. maby on weak linux VPS systems with low memory ?
[12:15]  Aine.Caoimhe all of Akira's host regions
[12:15]  Andrew.Hellershanks I'm usually bad with names but that seems familiar for some reason.
[12:15]  Dahlia.Trimble ya without hard data performance claims are worthless
[12:15]  Aine.Caoimhe so that's 30+ in Metro alone
[12:15]  Nebadon Izumi: id be so much happier if everyone had borg designations
[12:15]  Nebadon Izumi: lol
[12:16]  Aine.Caoimhe and has one of the single busiest sims (Close Encounter on Fridays)
[12:16]  Nebadon Izumi: ya i mean, i can show you many people running opensim
[12:16]  Nebadon Izumi: that it runs terrible
[12:16]  Justin Clark-Casey: aine: so maybe you should run that for your hg stuff and tell us how it goes?
[12:16]  Aine.Caoimhe always 25+ people there and 99.9% of the time is smooth as silk
[12:16]  Nebadon Izumi: i take 5-10 minutes to review thier ini
[12:16]  Nebadon Izumi: make 2-3 tweaks
[12:16]  Nebadon Izumi: and their sim runs 100 times better
[12:16]  Nebadon Izumi: no code changes
[12:16]  Richardus.Raymaker But close encounter is running on linux sofar i know
[12:17]  Dahlia.Trimble sometimes a few config tweaks make a world of difference in performance
[12:17]  Nebadon Izumi: yes
[12:17]  Aine.Caoimhe I don't want to do that because if I run Arriba I can't provide any meaningful feedback on master
[12:17]  Nebadon Izumi: Moses grid was a good example
[12:17]  Nebadon Izumi: for years that grid ran quite bad
[12:17]  Dahlia.Trimble and the defaults aren't all that good
[12:17]  Nebadon Izumi: i spent 15-20 minutes tweaking their ini and its useable again now
[12:17]  Nebadon Izumi: they couldnt even make it through a 1 hour meeting with 15 avatars prior to my changes
[12:17]  Andrew.Hellershanks Nebadon, is it usually the same tweaks you make for people to make things run well? If so, did you document the common config issues?
[12:18]  Nebadon Izumi: OSgrid Release
[12:18]  Nebadon Izumi: is my ini
[12:18]  Nebadon Izumi: lol
[12:18]  Nebadon Izumi: there for all to see
[12:18]  Andrew.Hellershanks oh, rigith
[12:18]  Andrew.Hellershanks is a bit of a tpyo geneator today. ;)
[12:18]  Richardus.Raymaker Aine. yes thats one of my reason i would not use ariba to. its not default opensim
[12:18]  Aine.Caoimhe also things like suggested MySQL tweaks, mono tweaks, etc would be good
[12:18]  Nebadon Izumi: yea that is true
[12:19]  Aine.Caoimhe exactly, Rich...I'd rather try to do my little bit to improve the stock code
[12:19]  Nebadon Izumi: I thought that was on the wiki at one point though
[12:19]  Andrew.Hellershanks yeah. I've been running with some stats thing turned on for MySQL so I can run a report to indicate possible tunings for the DB.
[12:19]  Aine.Caoimhe but when I see the differences I thought I'd point them out in case something useful could be gleaned by looking at Freaky's adjustments
[12:19]  Andrew.Hellershanks I know there are settings I can change to let it use more RAM than it currently does.
[12:19]  Aine.Caoimhe and he does post the full source
[12:20]  Nebadon Izumi: yea I wasnt even aware of Ariba myself until a few minutes ago
[12:20]  Andrew.Hellershanks Aine, that works until they deviate too much from the base code of OpenSim (as happened with Aurora)
[12:20]  Dahlia.Trimble I had heard about it before but I havent looked at it
[12:20]  Aine.Caoimhe then you missed their booth at OSG7B
[12:20]  Seth.Nygard I can post the link to his GitHub repo if you want
[12:20]  Justin Clark-Casey: honestly, it's dificult to say without actual perf data. There are a huge number of fators that affect opensim performance
[12:20]  Nebadon Izumi: and when you look at code like Whitestar and Aurora, the changes are done in a way that makes using their code extreemely difficult
[12:20]  Justin Clark-Casey: and tbh, I have enough to deal with myself usually without going and inspecting someone else's code
[12:21]  Aine.Caoimhe nods
[12:21]  Nebadon Izumi: ya from my experience how good opensim runs
[12:21]  Nebadon Izumi: depends on the administrator of the grid / regions
[12:21]  Aine.Caoimhe I can sympathise with that
[12:21]  Nebadon Izumi: more than the code itself
[12:21]  Nebadon Izumi: even apache out of the box
[12:21]  Nebadon Izumi: isnt very robust
[12:21]  Justin Clark-Casey: but I'm sure if there are good changes hopefully they will surface over time
[12:21]  Nebadon Izumi: without a ton of undocumented tweaking
[12:21]  Nebadon Izumi: lol
[12:21]  Seth.Nygard So what can the Grid admin do to fix Mono other than move to Windows and run .NET?
[12:21]  Justin Clark-Casey: sometimes it's better to fork so you can do the changes you want without being bound by established opinion
[12:22]  Nebadon Izumi: the only thing you can do is 1, eat your own dog food
[12:22]  Justin Clark-Casey: seth: I don't see where Mono is an issue? Are you talking about script stuff?
[12:22]  Nebadon Izumi: a lot of admins dont actually use their grids or regions
[12:22]  Nebadon Izumi: 2 is lots of load testing
[12:22]  Nebadon Izumi: to verify what you are doing actually is not just in your head
[12:22]  Seth.Nygard Yes, regions that other people use will always have scritps
[12:22]  Andrew.Hellershanks Nebadon, I never had much trouble with Apache and I had it running a site with 100k+ views a week(?) or was a day. Can't remember now. It was a while ago.
[12:23]  Dahlia.Trimble I have code that blows mono up but runs flawlessly on .net. I had to port it to c++ to get around that problem
[12:23]  Justin Clark-Casey: seth: and you're worried about memory leakage?
[12:23]  Nebadon Izumi: oh ya Andrew APache is great
[12:23]  Dahlia.Trimble mono memory management is horrid
[12:23]  Nebadon Izumi: im just saying you cant expect apache to perform the same way out of the box that it does for major websites
[12:24]  Seth.Nygard I am worried about instances where we've had scripts stop running. We will try with the changes to AppDomain however and hopefully that addresses them.
[12:24]  Nebadon Izumi: alot of the time people change the source to get that extra bit of performance
[12:24]  Nebadon Izumi: or do things that are not documented
[12:24]  Justin Clark-Casey: seth: all scripts?
[12:24]  Nebadon Izumi: it really depends on what your doing with apache
[12:24]  Nebadon Izumi: the same goes for opensim
[12:24]  Nebadon Izumi: alot of the time i hear people bitching about performance
[12:24]  Nebadon Izumi: only to run out they are running 50 regions in a single simulator
[12:24]  Nebadon Izumi: with 10,000's of scripts
[12:24]  Nebadon Izumi: and wondering why it runs bad
[12:25]  Aine.Caoimhe all script slowly starve to death
[12:25]  Andrew.Hellershanks Nebadon it was a tourism website that was tied in to an Oracle database.
[12:25]  Nebadon Izumi: like seriously?
[12:25]  Seth.Nygard Yes, at times. We've seen the Active Scirpt count go to negative values, after which time other scripts that had been running OK either stop or have strange behaviour
[12:25]  Dahlia.Trimble many regions in SL had > 10k scripts
[12:25]  Nebadon Izumi: the reason it goes negative
[12:25]  Justin Clark-Casey: seth: there were some recent changes to master to actually show exceptions and stop them halting all futurenew script running
[12:25]  Nebadon Izumi: is because scripted attachments do not calculate corretly
[12:25]  Nebadon Izumi: when being detached
[12:25]  Justin Clark-Casey: but I wouldn't update to master until I've put in the fix for this particular script loading bug in mantis 7278
[12:25]  Nebadon Izumi: so you get wierd script counts eventually
[12:25]  Andrew.Hellershanks The use of var regions can make heavily scripted regions seem even worse.
[12:25]  Dahlia.Trimble almost every mainland region had at least 15k scripts back in the heyday
[12:26]  Nebadon Izumi: thats not actually why its a problem though
[12:26]  Justin Clark-Casey: which should only affect people running 1+ regions in a simulator with AppDomainLoading = false but might have other knock-on effects
[12:26]  Nebadon Izumi: the counter is just broken
[12:26]  Justin Clark-Casey: I mean >1
[12:26]  Aine.Caoimhe kk
[12:26]  Nebadon Izumi: doesnt work right for attachment scripts
[12:26]  Nebadon Izumi: throws the overall count off
[12:26]  Nebadon Izumi: even if its not negative
[12:26]  Nebadon Izumi: its likely wrong
[12:26]  Seth.Nygard In almost every case I have 1:1 ratio of regions to simulators
[12:26]  Justin Clark-Casey: I believe the code which stops scripts running in an HG region also puts that stat out of whack
[12:26]  Aine.Caoimhe well my observation is when this happens suddenly there's huge lag in scripts that otherwise were running flawlessly
[12:27]  Justin Clark-Casey: the script engine has deficiencies
[12:27]  Aine.Caoimhe like a dialog box that normally pops up in less than a second might take 10-15 seconds to pup up
[12:27]  Nebadon Izumi: ya
[12:27]  Nebadon Izumi: sounds like threads
[12:27]  Aine.Caoimhe yes
[12:27]  Nebadon Izumi: thread starvation
[12:27]  Aine.Caoimhe yet when you look at the report on threads it shows tons available
[12:27]  Nebadon Izumi: which is what happens under mono with AppDomain true
[12:27]  Aine.Caoimhe but that's exactly what's happening
[12:28]  Justin Clark-Casey: that doesn't apply to xengine
[12:28]  Andrew.Hellershanks Does the thread report include the ones used in the scripting engine?
[12:28]  Justin Clark-Casey: unfortunatley there are multiple threadpools in opensim, because of organic factors
[12:28]  Justin Clark-Casey: xengine uses its own threadpool separate from the other main threadpool
[12:28]  Nebadon Izumi: ya but xengine has its own thread settings
[12:28]  Nebadon Izumi: is suspect if you bump up the numbers in xengine
[12:28]  Nebadon Izumi: with it set to true
[12:28]  Andrew.Hellershanks Could be interesting to have some additional stats on xengine
[12:28]  Aine.Caoimhe there's no "debug show scripts threads" type of debug command
[12:28]  Nebadon Izumi: it wouldnt die so quickly
[12:29]  Nebadon Izumi: but it would eventually die
[12:29]  Justin Clark-Casey: and I believe xengine uses a couple of threads to run events - it doesn't launch new threads to do so
[12:29]  Aine.Caoimhe it just queues them all up
[12:29]  Justin Clark-Casey: I'm a little hazy on the details, but yes, I believe each script has its own queue of events
[12:30]  Nebadon Izumi: the Klein Bottle generator is a good example
[12:30]  Nebadon Izumi: i have yet to see it complete 100% under linux
[12:30]  Nebadon Izumi: but in windows its lightning fast
[12:30]  Nebadon Izumi: and never fails
[12:30]  Justin Clark-Casey: ah no, I'm probably talking rubbish
[12:30]  Nebadon Izumi: its very timery, sleepy though
[12:30]  Nebadon Izumi: bad combo for linux
[12:30]  Nebadon Izumi: sleep
[12:30]  Nebadon Izumi: just destroys xengine in linux
[12:31]  Nebadon Izumi: but without it the loop goes way to fast
[12:31]  Nebadon Izumi: and also clobbers it
[12:31]  Andrew.Hellershanks that sounds like a mono issue again
[12:31]  Nebadon Izumi: so there is no win
[12:31]  Aine.Caoimhe so changing app loading to false isn't going to fix that sort of issue
[12:31]  Nebadon Izumi: correct
[12:31]  Justin Clark-Casey: event work is being done by the treadpool I think. If you want to see xengine thread stats they are avialable via "xengine status"
[12:31]  Nebadon Izumi: generally with the Klein bottle
[12:31]  Nebadon Izumi: it doesnt tottally kill xengine
[12:31]  Nebadon Izumi: just that script
[12:31]  Nebadon Izumi: if i delete the script and run it again it works
[12:32]  Nebadon Izumi: so its not totally stopping xengine like App domain issue does
[12:32]  Andrew.Hellershanks nebadon, what if you just reset the script?
[12:32]  Justin Clark-Casey: sleep is implemented poorly and may even be worse with co-op
[12:32]  Nebadon Izumi: with AppDomain to true, xengine stops and never recovers
[12:32]  Justin Clark-Casey: which is another thing I have to do something about
[12:32]  Aine.Caoimhe yes under .NET there's no issue?
[12:33]  Aine.Caoimhe (regardless of false/true?)
[12:33]  Nebadon Izumi: well
[12:33]  Nebadon Izumi: with False windows can flounder
[12:33]  Nebadon Izumi: not as bad as Linux does with true though
[12:33]  Nebadon Izumi: that can depend on if your 64 or 32bit
[12:33]  Nebadon Izumi: in 32 with App DOmain false
[12:33]  Nebadon Izumi: you can quickly generate out of resources crash in .net
[12:33]  Nebadon Izumi: thats unrecoverable
[12:34]  Nebadon Izumi: because the app domain can swell to be quite large
[12:34]  Justin Clark-Casey: alright, I need to go and get sometihng done.
[12:34]  Aine.Caoimhe okies
[12:34]  Nebadon Izumi: ok Justin
[12:34]  Justin Clark-Casey: See you later, folks.
[12:34]  Nebadon Izumi: later man
[12:34]  Aine.Caoimhe well I guess we'll try that and see what happens
[12:34]  Andrew.Hellershanks bye, Justin
[12:34]  Jak Daniels: bye Justin :)
[12:34]  Seth.Nygard Have a great night Justin, take care
[12:34]  Aine.Caoimhe bye Justin
[12:34]  Richardus.Raymaker bye
Personal tools
About This Wiki