Chat log from the meeting on 2017-11-07

From OpenSimulator

Jump to: navigation, search

[11:07] Arielle.Popstar hey JR i was reading that page you have on unity/opensim and hypergrid......did you ever get that server running?
[11:08] JayR.Cela @Arielle / was your question directed at me ?
[11:08] Arielle.Popstar yes :)
[11:08] Arielle.Popstar sorry should have said JayR
[11:09] Ubit Umarov: ohh i was looking for JR ewing
[11:09] JayR.Cela answer / no I have not / :_(
[11:09] Ubit Umarov: hmm if that was the name..
[11:09] Arielle.Popstar aww ok
[11:09] JayR.Cela :_)
[11:09] Arielle.Popstar came across it this morning and wondered
[11:09] Leighton.Marjoram nods @ Ubit
[11:10] Sheera Khan: Ubit - do you happen to know if the setting for maximum bandwidth in the viewer accounts for UDP only, TCP only or both?
[11:10] Ubit Umarov: udp. singu did add a http part
[11:10] Sheera Khan: ok, ty :-)
[11:10] Ubit Umarov: but server side does look to udp bandwith to limit http
[11:11] Kayaker Magic: Wow! Everyone is early today! Daylight Wastings Time confusion is finally over?
[11:11] JayR.Cela :_)
[11:11] Sheera Khan: UDP bandwith controlls HTTP too in the server?
[11:11] Ubit Umarov: yeap in 0.9
[11:11] Andrew.Hellershanks Kayaker, I wouldn't count on it but it is a good turn out today.
[11:12] Sheera Khan: oh, that's a bit confusing then...
[11:12] Ubit Umarov: a reason why you do notice texts and meshs a bit slower version 0.8
[11:12] George Equus: No, confusion remain... but will be stable for around 6 months
[11:12] Ubit Umarov: ... than...
[11:13] Arielle.Popstar everyone will be eaRLY AN HOUR 6 MONTHS FROM NOW
[11:13] Ubit Umarov: well a area that can have a bit improvement
[11:13] George Equus: guess that is better than autumn when no one is on time lol
[11:14] Ubit Umarov: well what i said is valid for all opensim. details :)
[11:14] Andrew.Hellershanks Arielle, was it you that had previously mentioned having to restart Robust on a regular basis to keep from it eating up too much memory?
[11:14] Arielle.Popstar no wasnt me Andrew
[11:15] Andrew.Hellershanks Arielle, ok. I recall it was someone at this meeting that mentioned that.
[11:15] Sheera Khan: I'd like to ask again as I did quite some time ago... Would it be possible to add a permanent mute for other avatars to opensim for the next release?
[11:15] Sheera Khan: Or is that a commercial addon?
[11:16] Ubit Umarov: you mean suport viewers mutes ?
[11:16] Arielle.Popstar did test the opensim32.exe thing though and found a significant difference in usage
[11:16] Andrew.Hellershanks I'd say it is possible. It needs someone to work on it.
[11:16] George Equus: Is that that not only for duration of one session?
[11:16] Ubit Umarov: err in usage ??
[11:16] Arielle.Popstar 30-45% better memory utilization
[11:16] Sheera Khan: the viewer supports the muting.. it just doesn't get a mute list from the server so the muting isn't permanent
[11:16] Ubit Umarov: ahh that use  :)
[11:17] Sheera Khan: so after each relog the mute list is empty
[11:17] Ubit Umarov: guess there is no grid side support
[11:17] George Equus: OK, thanks Sheerah
[11:17] Sheera Khan: leading to a little bit of drama in my grid again ^^
[11:17] Ubit Umarov: hmm and/or standaloned.. guess dual code on that
[11:18] Arielle.Popstar doesnt derendering and blacklisting also mute?
[11:18] Andrew.Hellershanks I don't think it would be that hard to support. The first step is working out the messages between the viewer and grid. WinGridProxy can help with that.
[11:18] Ubit Umarov: tell yr ppl that mental mute is the best option really :)
[11:18] Andrew.Hellershanks :)
[11:18] Ubit Umarov: ahh good andrews did volunter to do it
[11:18] Sheera Khan: yes Andrew, you asked Melanie about it and you got the answer: "I've got to look, what I can share..."
[11:18] Arielle.Popstar :)
[11:18] Andrew.Hellershanks Ubit, I did start looking in to it some time ago. Haven't done anything further on it recently.
[11:18] Ubit Umarov: actually think she did it once
[11:19] Ubit Umarov: hmm don't remember avn had it actually
[11:19] Ubit Umarov: well protocol to viewers is the easy part
[11:20] Ubit Umarov: possible part done '??
[11:20] Andrew.Hellershanks The response I sort of remember getting from Melanie made me thing that she had implemented it for Avination.
[11:20] Ubit Umarov: but perm storage is the thing
[11:20] Sheera Khan: In principle Andrew already told me how it is supposed to work... I'm just no C# programmer :-(
[11:20] Andrew.Hellershanks There is a mutelist module. It might just need the right hooks to receive the messages.
[11:20] Ubit Umarov: and guess standalones need local thing...
[11:21] Ubit Umarov: or not having it ? :)
[11:21] Andrew.Hellershanks It seemed to me that mute information was being stored in a file. Perhaps that was only meant to be done on the viewer side.
[11:21] Sheera Khan: well, my guess would be a standalone wouldn't need that feature as dearly as a bigger grid ...
[11:21] Andrew.Hellershanks Right
[11:22] Ubit Umarov: yeap think viewer do that also yes.. well never looked
[11:22] Sheera Khan: If I remember right the viewer gets the mute list during login ...
[11:22] Ubit Umarov: i always use mental mute.. ( or eject/ban :p )
[11:23] Andrew.Hellershanks The mutelist code is in OpenSim/Region/CoreModules/Avatar/InstantMessage/MuteListModule.cs
[11:23] Andrew.Hellershanks It has the hook to handle the request for the mute list. Just no hooks for adding/removing entries.
[11:24] Ubit Umarov: m_RestURL = cnf.GetString("MuteListURL", "");
[11:25] Andrew.Hellershanks One important question is going to be whether the mute list info should be stored in a file or database. If in a database it will be easier to keep updated but would require a DB migration step due to adding a new(?) table.
[11:25] Sheera Khan: back :-)
[11:25] Andrew.Hellershanks If the module is asking for a URL it would be getting handled via an external PHP file.
[11:26] Andrew.Hellershanks I could use a version of the PHP code I used for maintaining requests to be notified of upcoming events.
[11:27] Ubit Umarov: hmm Sheera.Khan file mantis with a feature request on this
[11:29] Andrew.Hellershanks The mute list module needs to add hooks to handle UpdateMuteListEntry and RemoveMuteListEntry.
[11:30] Arielle.Popstar no other issues?
[11:31] Andrew.Hellershanks I just found the events that need to be hooked.
[11:31] Sheera Khan: It is the most wanted feature in Metropolis ;-)
[11:31] Kayaker Magic: Last week I reported a speedup in rezzing complex objects and border crossings by tuning the MySQL configuration, I have an update on that:
[11:31] Arielle.Popstar is metro on .9 now?
[11:31] Ubit Umarov: lludp is there
[11:32] Kayaker Magic: I had heard that setting autocommit=0 by default in MySQL would speed things up, but I'm told that is a BAD IDEA. instead,
[11:32] Ubit Umarov: region code seems there
[11:32] Ubit Umarov: ( unless ll did move it to http :p )
[11:32] Sheera Khan: @Arielle: not up to now :-(
[11:33] Arielle.Popstar kk Sheera thx
[11:33] Kayaker Magic: I am told innodb_flush_log_at_trx_commit=0 in MySQL is the thing to do.
[11:33] Bill.Blight It is a bad idea for anybody not running on enterprise class hardware for sure, as shutdown can cause issues
[11:35] Ubit Umarov: well mutes region code seems there and assumes a external php of opensim.defaults.ini
[11:35] Arielle.Popstar did you try it Kayaker?
[11:36] Kayaker Magic: Yes, but I am having trouble finding a good way to measure the improvement.
[11:36] Arielle.Popstar a party
[11:36] Bill.Blight I use jopensim, and the mute list works fine .. as it has always to my knowledge use a external php
[11:37] Kayaker Magic: Rezzing a complex build has two delays: getting it out of the asset server and getting it into the viewer cache.
[11:37] Andrew.Hellershanks Ubit, Right. Add two more event hooks and the code to pass data to the external PHP file.
[11:37] Kayaker Magic: I'm not sure how to separate the two.
[11:42] Arielle.Popstar any answers for Kayaker?
[11:42] Ubit Umarov: MySQL question? no clue.. bill said something
[11:43] Jak Daniels: I guess the only way to test it really is on a standalone where everything is self contained... no possible network issues etc
[11:43] Kayaker Magic: Is there a cache in the region memory? If I rez something, wait for viewer cache, delete object, then rez it a second time. Does the second rez have only the fetch-from-asset_server time in it? Or does that fetch run faster the second time because of a region cache?
[11:44] Andrew.Hellershanks Bill, what is jopensim?
[11:44] Ubit Umarov: assets are cached on disk
[11:44] Ubit Umarov: hmm assets on region..
[11:44] Kayaker Magic: but do items in that cache get removed when they are deleted?
[11:44] Bill.Blight It is a plugin for Joomla, that allows user creation, groups, offline IM, events, classifies, the full suite of features that most grids want
[11:45] Ubit Umarov: there are some config options to auto clean
[11:45] Andrew.Hellershanks I've just put in the initial framework for full support of mute list requests. I just have to add some code to make the call to the PHP file and update the PHP file I have to handle the requests.
[11:45] Leighton.Marjoram it is incredible I use it for my grid website
[11:45] Ubit Umarov: but best is you do it by hand possibel
[11:45] Leighton.Marjoram Jopensim*
[11:45] Ubit Umarov: hmm i did change that code on floatsam but forget details oops :p
[11:46] Ubit Umarov: btw floatsam is the only cache currently usefull
[11:46] Ubit Umarov: i mean for assets
[11:46] Andrew.Hellershanks Bill, if it is only a Joomla plug-in I don't see how it helps with mutes unless it is being used to maintain a file that is sent by the mutelist module.
[11:47] Bill.Blight yes
[11:47] Bill.Blight it is
[11:47] Andrew.Hellershanks Still not sure how it would help with marking objects/avatars to be muted via the viewer.
[11:47] Bill.Blight and does
[11:47] Bill.Blight <shrug>\
[11:48] Andrew.Hellershanks If the source is available for it we can see what it is doing.
[11:51] Ubit Umarov: well about dev state.. still like last week.. waiting for bug reports, to fix as possible in order to have a post fixes
[11:52] Ubit Umarov: i do have some code on a branch waiting to merge on master.. but that will make it again full dev ( ie less reliable )
[11:54] Ubit Umarov: ( silence... moments :p )
[11:54] Arielle Popstar: well if nothing coming in then maybe its time for post fixes?
[11:54] Ubit Umarov: seem anymore things on that mono 5.0 slower ll lists kay ?
[11:54] Arielle Popstar: i did mention someone saying slow descent is an issue in .9
[11:55] Ubit Umarov: what slow descent ?
[11:55] Arielle Popstar: but they dont seem comfortable mantising it
[11:55] Kayaker Magic: No, I'm switching to mono 4.12 (I think...) which has some important fixes but is not as slow as 5.x
[11:55] Ubit Umarov: what slow descent ?
[11:56] Kayaker Magic: I tried to write some c# programs to show the same decrease in array fetching, but could not get it to happen.
[11:56] Andrew Hellershanks: Linking prims takes a lot longer in 0.9 than it did in 0.8.2
[11:56] Ubit Umarov: it is a array of objects
[11:57] Arielle Popstar: meaning?
[11:57] Arielle Popstar: its not fixable?
[11:57] Kayaker Magic: So an array of floats in c# would not show the same timing as an array of objects?
[11:57] Ubit Umarov: kay... ll list is a array of objects..
[11:58] Ubit Umarov: no bc of encapsulation
[11:58] Ubit Umarov: ( ie putting the float inside a object and vice versa )
[11:59] Ubit Umarov: but what descent ????
[12:00] Jak Daniels: do you mean an av falls to the ground more slowly?
[12:00] Arielle Popstar: he had a platform that when the avatar lands on it, it starts to descend downwards
[12:00] Arielle Popstar: in .9 it goes fast
[12:00] Arielle Popstar: same with someone else and parachutes
[12:01] Ubit Umarov: that depends on how the platform is done
[12:01] Bill.Blight Also depends on the physics engine
[12:01] Ubit Umarov: yeap
[12:02] Arielle Popstar: the platform is on ubode but the parachute i suspect on bullet
[12:02] Arielle Popstar: i would have to confoirm thAT
[12:02] Ubit Umarov: well that all depends on the scripts
[12:02] Ubit Umarov: hard to say from descrition
[12:02] Arielle Popstar: well assume the same scripts both used on .8.2.1
[12:02] Arielle Popstar: i know :(
[12:03] Ubit Umarov: well some functions where fixed etc
[12:03] Bill.Blight I have never seen that type of tunneling on prims in .9 unless we have intentionally increased the meters per second to insane levels ..
[12:03] Ubit Umarov: that means 0.8 values may give dif response.. can't tell
[12:04] Arielle Popstar: i will ask them both this week and let you know
[12:04] Kayaker Magic: Gota run, I'll try that rez/delete/time(second rez) idea to test MySQL optimization. Report back next week.
[12:04] Ubit Umarov: THAT IS NOT easy
[12:04] Ubit Umarov: to benchmark
[12:04] Ubit Umarov: i already told andrew
[12:05] Ubit Umarov: and benchmark that is not that relevant
[12:05] Arielle Popstar: well anything to help speed up the rezzing of mesh would be great
[12:05] Andrew Hellershanks: Don't need a special benchmark to tell when something is taking 10 to 100 times longer.
[12:06] Bill.Blight mesh rezzes fine if it is optimized mesh
[12:06] Andrew Hellershanks: Code changes that only have a minor impact on performance are harder to benchmark.
[12:06] Ubit Umarov: well to me rez of 128 prims takes lot less than 1s, if 73 takes 17s to you.. well
[12:07] Bill.Blight ^^
[12:07] Andrew Hellershanks: I don't have a problem with rez times.
[12:07] Arielle Popstar: talking avatars and the meshes they wearing
[12:07] Andrew Hellershanks: Linking of prims in 0.8.2 vs in 0.9 takes substantially different amounts of time.
[12:07] Ubit Umarov: i meant link
[12:07] Bill.Blight worn mesh should be optimized as well, and if you mean HG avatars, that is also dependent on their remote asset server sending the information
[12:08] Andrew Hellershanks: Almost instant to link 73 prims in 0.8.2 vs 17 seconds in 0.9 on the same computer with the same mono.
[12:08] Sheera.Khan wow
[12:08] Ubit Umarov: what prims ?
[12:08] Bill.Blight What viewer?
[12:08] Andrew Hellershanks: Ordinary default cubes.
[12:08] Sheera.Khan linked some 10000 prims once^^
[12:09] Ubit Umarov: ( suicidal cat )
[12:09] Bill.Blight I did over 500 prims in one shot last night and took about 2 seconds
[12:09] Andrew Hellershanks: Bill, Firestorm but IIRC, I've also seen the problem with Sing.
[12:09] Arielle Popstar: i remember when everyone was testing that
[12:09] Arielle Popstar: Neb did a huge amount too
[12:09] Arielle Popstar: sing was the best
[12:10] Arielle Popstar: fs choked on 500 if i rememebr right
[12:10] Andrew Hellershanks: Interestingly (to me), using osBreakLink to unlink the 73 prims is faster in 0.9 than in 0.8.2
[12:10] Ubit Umarov: opos bad place to do this :P
[12:10] Andrew Hellershanks: That looks like more than 73 prims.
[12:11] Bill.Blight maybe you should have put that on the roof
[12:11] Bill.Blight LOL
[12:11] Andrew Hellershanks: At least Ubit didn't put me in the middle of all those prims. :)
[12:11] Arielle Popstar: going to link the whole structure
[12:11] Bill.Blight He's not done yet andrew, give him a minute
[12:11] Andrew Hellershanks: Bill, :)
[12:12] Ubit Umarov: bahh select only my objs selected all
[12:12] Ubit Umarov: :)
[12:13] Ubit Umarov: ok.. lets see
[12:13] Andrew Hellershanks: Hm... 384 prims to link
[12:14] George Equus: Now, make a working Rubics Cube of it  :)
[12:14] Andrew Hellershanks: Are you going to way when you start the link process and when it is done?
[12:14] Andrew Hellershanks: hehe... I bet someone has done that. It wouldn't be hard to do.
[12:15] Andrew Hellershanks: The tricky part is adding in the action of throwing it against the far wall when you can't solve it. ;)
[12:15] George Equus: But how about solving it??
[12:15] George Equus: Right..
[12:16] Andrew Hellershanks: Oops. Ubit tried to link all those prims and knocked himself out of the grid.
[12:16] Ubit Umarov: my fs did not liked the way i did selec them :P
[12:16] Sheera.Khan hihihi
[12:16] Ubit Umarov: well they are linked :)
[12:17] Andrew Hellershanks: How long did it take?
[12:17] Ubit Umarov: unlinked
[12:18] Andrew Hellershanks: Some of those prims moved.
[12:19] Jak Daniels: a few are missing now
[12:19] Ubit Umarov: uff no idea what version is here
[12:19] George Equus: Not missing I think, but unalligned
[12:20] George Equus: had that problems some time ago when had to unlink an old link set to correct it, it scatterd all prims all around
[12:20] Jak Daniels: some big holes.... still the same number of prims?
[12:20] Bill.Blight yes this is not a good test, as it is not core exactly
[12:20] Ubit Umarov: yeap no idea what his running here
[12:21] Ubit Umarov: but on link, issue is not much performance, but correctness
[12:21] George Equus: If unlinked it scatters
[12:21] George Equus: 52 prims
[12:22] George Equus: well... it did way back, had hell with it
[12:22] Andrew Hellershanks: Taking a long time to link when building things gets to be a pain.
[12:22] Andrew Hellershanks: I just wonder why there is such a noticable difference between 082 and 09
[12:23] Ubit Umarov: link suffers from same as select
[12:24] Andrew Hellershanks: The other (slightly more important) issue that still needs to be resolved. There are still a number of discrepancies in the OSSL enable file vs what the code states for permissions.
[12:24] Ubit Umarov: protocol was never planed for more than 256 prims
[12:24] Bill.Blight **
[12:25] Bill.Blight that too
[12:25] Andrew Hellershanks: Right. I have someone telling me they had issues but they were linking over 256 prims together. They started with 98 prims and adding 198 more.
[12:25] Sheera.Khan I just tested it in 0.8.3 and 1000 Prims take about 2 mins, where the first 850 are done in abaut 40 seconds...
[12:25] Bill.Blight just like a lot of things we try to do in opensimulator, we are exceeding what we are supposed to be able to do, and yet we complain that it does not work perfect .. that is just my .02
[12:25] Andrew Hellershanks nods
[12:25] Arielle Popstar: I know I have done over 500 with sing several years ago
[12:26] Andrew Hellershanks: I don't know if the linking issue I'm seeing is due to changes in the messaging between the grid and viewer or something else in the SOG/SOP code.
[12:27] Andrew Hellershanks: I'll run more checks to verify if the time difference I see occurs in more than one viewer.
[12:27] Andrew Hellershanks: ok. It is almost a half hour past the top of the hour. If there isn't any other final comments it is time to wrap things up for today.
[12:28] Andrew Hellershanks: Ok, I don't see anyone typing anything so that will do it for this week. Thank you for dropping in.
[12:28] Andrew Hellershanks: See you again next week.
[12:29] Nebs Metal Bar Stool v1.6 (w/sit & launch): Goodbye..
[12:29] Sheera.Khan Thank you Andrew for starting the framework for the mute list :-)
[12:29] George Equus: Thanks Andrew
[12:29] Andrew Hellershanks: It is saying the OnRemoveMuteListEntry doesn't exist. I need to track down the proper name (or need to add some other code to make the hook available).

Personal tools
About This Wiki