Chat log from the meeting on 2018-01-23

From OpenSimulator

Jump to: navigation, search

[11:07] Andrew.Hellershanks What's new and exciting?
[11:08] Andrew.Hellershanks Ubit has been busy working on the code this past week but he has been working on a branch.
[11:08] Kayaker Magic: Why a separate branch?
[11:09] Andrew.Hellershanks The changes are potentially disruptive. He is cleaning up how things are done according to .NET standards.
[11:12] Andrew.Hellershanks Ubit is about to arrive. You can ask him for the details in a moment, Kayaker.
[11:13] George Equus: Can't go here directly, nothing rez for ages if someone is already on the region. In sandbox plazas or even direct log-in to my own region all rez in a minute, then you can get here and see stuff.
[11:13] George Equus: Hi Ubit
[11:14] Sheera Khan: I had issues with the direct TP too... "Error connecting to grid" - but the TP via Wright Plaza as Relay worked fine
[11:14] Andrew.Hellershanks Ubit, I was just mentioning that most of the work you have done this past week has been on a branch.
[11:15] George Equus: This region seems to have some sort of problem from the start Is odd..
[11:15] Ubit Umarov: yes, only added a fix for macs on both master and that branch: httptests
[11:16] Ubit Umarov: about that branch, it started as name says some testing on http.. mods on httpserver.dll and main code
[11:16] Ubit Umarov: but... in time i did add a few more tests etc...
[11:16] Ubit Umarov: so now it is more as Next Master...
[11:17] Ubit Umarov: recently i pushed on it the minimal requirements for opensim to be .net framework 4.6
[11:17] Ubit Umarov: that means no more XP  :(
[11:18] Andrew.Hellershanks You decided to just go for the full 4.6 rather than just 4.5?
[11:18] Ubit Umarov: and on linux may mean mono 5.x
[11:18] Andrew.Hellershanks You'll want to save that note about no XP support for the next release notes.
[11:18] Ubit Umarov: .net 4.5 is outdated... does not contain some security patches
[11:19] Ubit Umarov: only present on 4.5.2
[11:19] JayR.Cela @ Ubit / yes the latest versions of Mono is 5.x
[11:19] Ubit Umarov: but 4.5.2 seems to have similar requirements as 4.6
[11:19] Ubit Umarov: ( and 4.5 also does not suport XP)
[11:19] Andrew.Hellershanks ok
[11:20] George Equus: XP is a bit outdated too.. def. no security for it so .NET security is bit academic anyway on that os...
[11:20] Ubit Umarov: some ppl told me that mono 4.8 is to avoid...
[11:20] Andrew.Hellershanks Another on the 4.x series to avoid? :P
[11:20] George Equus: It was hard to leave..
[11:20] Andrew.Hellershanks I won't have a problem with mono. My distro has mono 5.4
[11:21] George Equus: But 7 is fine too.
[11:21] Ubit Umarov: current stable is
[11:22] Ubit Umarov: httptests does include all changes on master
[11:22] Andrew.Hellershanks Did you add a tag marking the point at which httptests is no longer compatible with the older .NET and mono versions?
[11:22] Ubit Umarov: so you are invited to crash on it.. err i mean test it
[11:23] Andrew.Hellershanks I've been good at finding issues in another program lately. I might find some in httptests if there are any to be found.
[11:24] Ubit Umarov: btw httptests is only avaiable on git .. and gitview if you can navigate to it :)
[11:25] Andrew.Hellershanks You can also get it by using git gui if you know how to use that program.
[11:25] Leighton.Marjoram does any one have the git address for master tried to find it on the download and building pages of the OS website
[11:25] Ubit Umarov: err its ode.dll ( all os) may require also a recent linux/windows
[11:26] Ubit Umarov: it is also a bit diferent from master, and i do not have old machines to compile it :(
[11:26] Ubit Umarov: sure andrew has the url :p
[11:26] Leighton.Marjoram found it :)
[11:27] Sheera Khan: ok, you beat me Leighton ;-)
[11:27] Ubit Umarov: of course.. like normal master httptests is unstable :)
[11:27] Leighton.Marjoram hehe took some rummaging to find :)
[11:28] Ubit Umarov: may work today.. and blow up tomorrow ;)
[11:28] Andrew.Hellershanks Leighton, I was about to say the links are in the wiki
[11:29] Andrew.Hellershanks
[11:30] Ubit Umarov: you can navigate btw branchs with git checkout as you know
[11:30] Leighton.Marjoram I was skim reading the pages and missed the section linking to that page.
[11:30] Sheera Khan: the wiki references some elderly versions^^ git checkout -b 0.7.3-post-fixes origin/0.7.3-post-fixes
[11:30] Ubit Umarov: do not forget to do a clean and runprebuild each time
[11:30] Andrew.Hellershanks ok. I don't know what page links to it. I just found it by an Internet search.
[11:30] Ubit Umarov: we have all them on git
[11:30] Ubit Umarov: or most of them
[11:31] Leighton.Marjoram I think its the building page on the wiki
[11:31] Andrew.Hellershanks hm... yes, that page does reference 0.7.3 What's one more page that needs an updated. :P
[11:31] Ubit Umarov: isn't that just a example ??
[11:31] Sheera Khan: it sure is...
[11:32] Andrew.Hellershanks yes, it is. Hope no one misses that part.
[11:32] Ubit Umarov: you can see a list of branch at
[11:33] Ubit Umarov: heads section
[11:34] Ubit Umarov: 0.1 prestable
[11:34] Ubit Umarov: errr opensim has no stable :p
[11:34] Andrew.Hellershanks Yup. No place for livestock. ;)
[11:37] Ubit Umarov: speak andrew
[11:37] Andrew Hellershanks: Is there anything else we have to discuss for this week?
[11:38] Sheera.Khan On a slightly unrelated note: does anyone happen to know if vertices in an avatar mesh, which are switched to transparent, still contribute to rendering weight?
[11:38] Andrew Hellershanks: I've been working on some other projects so I still haven't gotten around to working on the change log summary for 0.9.
[11:39] Ubit Umarov: think vertices and transparence are totaly dif things.. so guess yes
[11:39] Sheera.Khan i.e. would a Ruth2 without those parts I have to switch to transparent put less stress on a client?
[11:39] Ubit Umarov: no
[11:40] Sheera.Khan the server shouldn't be effected I guess
[11:40] Andrew Hellershanks: Sheera, I'm not sure how you would mark a vertex as transparent. I would think that is just changing texture and not affecting the model so it would still count all parts of the mesh.
[11:40] Sheera.Khan it is as you said Andrew
[11:40] Ubit Umarov: mesh is a object, textures are applied to it later
[11:40] Andrew Hellershanks: right
[11:40] Sheera.Khan I tried to keep the question coming before the meeting is over ^^
[11:40] Ubit Umarov: so the gpu must get the all mesh
[11:41] Andrew Hellershanks: Sheera, I keep thinking one of these days we may wind up ending a meeting early. :)
[11:41] Kayaker Magic: HA!
[11:41] Sheera.Khan that's why I shortened my question a bit (too much)
[11:42] Andrew Hellershanks: We have time still if you want to ask the longer version of the question.
[11:42] Sheera.Khan [11:39] Ubit Umarov: no seems to be the short and precise answer ^^
[11:43] Ubit Umarov: actually transp may increase the gpu load.. but don't know :)
[11:43] Andrew Hellershanks: Ubit, that's true. The viewer would have to work out what can be seen behind the part of the model that was marked transparent.
[11:43] Kayaker Magic: Transparency has never been well supported in OpenGL, it is a bit of a kluge to do it in the viewer.
[11:44] Sheera.Khan My idea was to take Ruth2 to Blender and delete the parts I have to blank out permanently like head and legs
[11:44] Ubit Umarov: well that is another thing you will be changing the mesh :)
[11:44] Andrew Hellershanks: Sheera, if you are manipulating the object in Blender you can just delete what you don't need instead of using a texture to hide parts.
[11:44] Sheera.Khan thus saving quite some vertices which shouldn't contribute to render load anymore then
[11:44] Ubit Umarov: way better...
[11:45] Andrew Hellershanks: yes
[11:45] Sheera.Khan right, that was my proposal: not using the transparency via HUD but really deleting the parts in Blender
[11:46] Ubit Umarov: don't forget to fix uv maps etc.. hmm why im i talkig about things i really do not know?? lol
[11:46] Lucy Afarensis: It can be fun
[11:46] Andrew Hellershanks: :)
[11:47] Ubit Umarov: avatars are tricky .. bones.. vetice weights etc etc
[11:48] Ubit Umarov: well guess objects for animesh will have all those also
[11:49] Kayaker Magic: Is there hope that we will support armatures in OpenSim mesh objects one day?
[11:49] Sheera.Khan before tackling all that (given my very limited Blender skills) I try to find out if it makes sense at all
[11:49] Ubit Umarov: no point without viewers supporting them also
[11:50] Kayaker Magic: Oh dang, I'm looking into Ruth2 in Blender and I was hoping you (Sheera) was an expert I could hit up with questions!
[11:50] Sheera.Khan sorry Kay, I'm definitly not :-(
[11:50] Andrew Hellershanks: The blind leading the blind? ;)
[11:51] Andrew Hellershanks: I've seen the link for Ruth 2 but I haven't tried downloading it yet.
[11:51] Sheera.Khan I'm just trying to reduce the load I put on servers and viewers...
[11:51] Ubit Umarov: and whats the thing about it? seems just another mesh avatar.. or not ?
[11:52] Sheera.Khan it's one avatar with a clear copyright ...
[11:52] Ubit Umarov: what is special about it?
[11:52] Ubit Umarov: ahh ok
[11:52] Kayaker Magic: Open source, good quality
[11:52] Lucy Afarensis: Mesh avatars load so much more slowly than system avies
[11:52] Kayaker Magic: backwards compatible UVmap
[11:52] Sheera.Khan most other mesh-avatars have some obscure origins
[11:52] Ubit Umarov: system ones don't load
[11:52] Kayaker Magic: low poly
[11:53] Ubit Umarov: they are on your disk
[11:53] Ubit Umarov: and yeap low poly etc
[11:53] Lucy Afarensis: and mesh is not ?
[11:53] Andrew Hellershanks: Mesh for avatars would not be low poly
[11:53] Ubit Umarov: no mesh is on the region, you need to download it
[11:54] Ubit Umarov: to the viewer cache.. then ok is faster
[11:54] Kayaker Magic: Ruth is low poly and high quality compared to some. Also open source and compatible with SL clothing.
[11:54] Andrew Hellershanks: I wonder how mesh avatars impacts avatar complexity values. Pushes it up no doubt, but by how much?
[11:54] Ubit Umarov: ( the one on disk is also called ruth )
[11:54] Leighton.Marjoram you can download the model and other things from git
[11:55] Sheera.Khan not sooo much as you might think Andrew
[11:55] Andrew Hellershanks: Sheera, that's good. Some avatars have too high a complexity level as it is.
[11:55] Leighton.Marjoram Ruth2 has a complexity of about 90000 which is extremely low compared to the SL ones
[11:55] Sheera.Khan my avatar as system body and as Ruth2 differ by about 1% in complexity
[11:55] Andrew Hellershanks: BTW, AFAICT, osgrid is online again.
[11:56] Ubit Umarov: really andrew
[11:56] Ubit Umarov: we didn't had noticed that :p
[11:56] Andrew Hellershanks: Um... oh. yea. It was this one. :) hehe
[11:56] Andrew Hellershanks: I don't know when they finished their migration work.
[11:57] Andrew Hellershanks: The website mentions new ini files are available for download.
[11:57] Sheera.Khan but I try to reduce that further
[11:57] Ubit Umarov: they now use core profiles
[11:58] Andrew Hellershanks: The code wants triangles so using tris to quads in Blender won't really help much.
[11:58] Sheera.Khan ohhh, and I don't have that pesky message anymore: "No profiles could be found" :-)
[11:58] Andrew Hellershanks: Ubit, they also mention the mutelist module. Not sure which one they are using.
[11:58] Ubit Umarov: no but others will lol
[11:58] Ubit Umarov: depends what grids are using
[11:59] Andrew Hellershanks: I've not had a problem with profiles in here. I just get one message about a texture that doesn't load every time I log in here.
[12:00] Sheera.Khan a big "Thank you" to the grid admins for taking that step forward :-)
[12:01] Andrew Hellershanks: We managed to get through another hour. :)
[12:01] Ubit Umarov: well i hope not many grids use the opensimprofiles module ( ie the not core one )
[12:01] Sheera.Khan I've encountered it in OSGrid only ^^
[12:02] Andrew Hellershanks: Ubit, You hope that because it doesn't have HG support?
[12:02] Ubit Umarov: seems the migration is easy.. same db
[12:02] Ubit Umarov: both have hg
[12:02] Ubit Umarov: just don't talk btw them
[12:02] Ubit Umarov: ( if i remember.. )
[12:03] Andrew Hellershanks: When I looked at the core profiles code I saw some HG related stuff in the code that isn't in the non core one.
[12:03] Ubit Umarov: ahh on httptests my new mutesmodule is the oficial core one
[12:03] Ubit Umarov: old one and xmutes removed
[12:03] Ubit Umarov: and it does not do HG, intencionaly
[12:04] Ubit Umarov: a HG visitor will have mutes on viewer cache
[12:04] Ubit Umarov: but they are deleted on next relog on home grid
[12:05] Ubit Umarov: ie can add mutes relative to the visited grid into viewer and its cache
[12:05] Sheera.Khan but then they get reloaded from the grid database if the grid supports them, right?
[12:05] Ubit Umarov: those are not sent to any grid db
[12:06] Ubit Umarov: see ?
[12:06] Sheera.Khan ah, so the mutes in the home-grid stay permanent whereas mutes of HG-visitors are temporary?
[12:07] Andrew Hellershanks: um... what if you were in your home grid and muted an HG avatar. Is that made persistent?
[12:07] Ubit Umarov: they will work.. etc.. but no storage.. and will be deleted from viewer on next relog on home grid
[12:07] Ubit Umarov: hmm yes, if i mute you, it will be stored
[12:07] Ubit Umarov: well if osgrid was using core module :)
[12:08] Andrew Hellershanks: :)
[12:08] Sheera.Khan hmmm, but I'm from a different grid and thus I'M a HG-visitor
[12:08] Andrew Hellershanks: I haven't tested my module in an HG enabled environment.
[12:08] Ubit Umarov: no idea on what they are using
[12:08] Ubit Umarov: just doing HG on that is not that simple
[12:08] George Equus: I muted an ava on another grid, still in my Blocked list
[12:08] Ubit Umarov: well not as simple as it seems :)
[12:09] Ubit Umarov: so i made that decision..
[12:09] Ubit Umarov: ( also questionable if we really wanted to store those )
[12:09] Andrew Hellershanks: The latest osgrid ini file uses OpenSimMuteList for the mutelist module.
[12:10] Ubit Umarov: George im talking about the core module
[12:10] Andrew Hellershanks wonders if the L is supposed to be capitalized.
[12:10] Ubit Umarov: named MuteListModuleTst on master, MuteListModule on httptests
[12:11] George Equus: OK Ubit
[12:11] Andrew Hellershanks: Um... no. That L isn't supposed to be capitalized. I'll have to talk with nebadon or dan about that.
[12:11] Ubit Umarov: well actually httptests as a more recent version :)
[12:11] Ubit Umarov: what i said is relative to httptests
[12:12] Ubit Umarov: i can't make it master
[12:12] Ubit Umarov: our jenkins etc uses mono 2.x.. whould smoke
[12:13] Andrew Hellershanks: Jenkins is using mono 2?? eek. I would have expected at least 3.x
[12:13] Ubit Umarov: that does not even suport all .net4.0
[12:14] Ubit Umarov: but bf making httptests master, a release will be done
[12:14] Ubit Umarov: with updated 0.9.0 code ( ie master )
[12:15] Ubit Umarov: then master will be 0.9.1 ( ie httptests )
[12:15] Ubit Umarov: and bummm happy and funny crashs for all :p
[12:15] Andrew Hellershanks: :)
[12:15] Ubit Umarov: like is normal on master :)
[12:16] Sheera.Khan Welcome to my laboratory where safty is number one priority... booom
[12:16] Ubit Umarov: :)
[12:16] Ubit Umarov: my code is so good.. SEGMENTION FAULT
[12:16] Andrew Hellershanks: heh
[12:17] Ubit Umarov: i did learn long ago to not say "my code is good" :)
[12:17] Andrew Hellershanks: Good code just takes longer to crash. ;)
[12:17] Kayaker Magic: lol
[12:18] Ubit Umarov: ( of course we can always blame andrew )
[12:18] Andrew Hellershanks: You can but I probably didn't do the code for what crashed.
[12:18] Ubit Umarov: the code sees you and crashed :p
[12:18] Sheera.Khan back again to .NET 4.6 vs. 5.0 for a second...
[12:19] Sheera.Khan might it be a good idea to set minimum to 5.0 even? Are there features not covered by 4.6+ which might be needed?
[12:20] Ubit Umarov: i don't understand mono .net comp.. they have no mapping :)
[12:20] Kayaker Magic: There is my mantis about llList fetches in scripts running 50 times slower in Mono 5.x...
[12:21] Sheera.Khan I mean: if we have to change the minimum requirement, should we do it so we don't have to change it again for the next release?
[12:21] Sheera.Khan ok, that would be a valid reason Kay ...
[12:21] Ubit Umarov: no ill not go to .netcore2.0 :p
[12:22] Ubit Umarov: or even .net4.7.1
[12:22] Kayaker Magic: I'd like to learn enough about C# to build a standalone test program to demonstrate the problem without OpenSim and submit it to the mono mantis system...
[12:23] Andrew Hellershanks: Kayaker, do you know C?
[12:23] Kayaker Magic: Oh yeah, I call it C-- (not being a fan of C++)
[12:24] Ubit Umarov: you did not forgot appdomains false right?
[12:24] Kayaker Magic: Yes, on or off that isn't effecting the speed of list fetches in mono 5
[12:24] Ubit Umarov: i think c# is a lost easier oop that c++
[12:24] Andrew Hellershanks: Kayaker, ok. If you know C you can look at the OS C# code and could put something together.
[12:25] Ubit Umarov: well it must affect
[12:25] Andrew Hellershanks: Knowing C and the basics of OOP I've been able to do some useful things in C#.
[12:25] Kayaker Magic: I tried writing something, building an array, timing tests (requires knowing the C# API) and did not see the problem.
[12:25] Ubit Umarov: appdomains cost a lot no matter the .net version
[12:26] Kayaker Magic: Need to build an array of objects, not just atomic types, to see the problem. I am told.
[12:26] Andrew Hellershanks: Ubit, that always makes me wonder why that would be the case.
[12:26] Ubit Umarov: objects cost a bit more than basic types
[12:27] Ubit Umarov: ( they wrap basic types etc )
[12:27] Ubit Umarov: but that is done all over .. so should not be a issue
[12:27] Ubit Umarov: what andrew  ??
[12:28] Andrew Hellershanks: I wouldn't expect app domains should really have some major impact on execution speed. Of course, I don'tk now that much about how the scripting engine works.
[12:29] Ubit Umarov: well read about them
[12:29] Andrew Hellershanks: Is it a C# thing or a .NET thing?
[12:29] Ubit Umarov: they are just a bit simplified interprocess comunication
[12:29] Ubit Umarov: tons of code on each crossing
[12:30] Ubit Umarov: and in many cases full serialization of parameters and return values
[12:33] Ubit Umarov: its odd.. our lists currently are arrays
[12:33] Ubit Umarov: and arrays are used every where
[12:33] Ubit Umarov: Queues are arrays Dictionaries are arrays.. etc
[12:35] Andrew Hellershanks: its now half past. Time to wrap it up.
[12:36] Ubit Umarov: ohh did you know we can't login on a grid that has robust and regions down?
[12:36] Ubit Umarov: ufff im so silly...
[12:36] Andrew Hellershanks: If both those things are down that would be a problem.
[12:37] Andrew Hellershanks: If Robust is up but regions are down you get stuck when it comes to trying to connect to a region.
[12:38] Andrew Hellershanks: Let's call it a day. See you all next week.

Personal tools
About This Wiki