Chat log from the meeting on 2018-05-08

From OpenSimulator

Jump to: navigation, search

[11:02] Andrew.Hellershanks Sitting on an invisible chair made it appear. :)
[11:04] Andrew.Hellershanks I'm only about half-way through reading last weeks meeting notes.
[11:08] Sheera Khan: So, what's new in OpenSim this week? Is there a Kitten-Module yet?
[11:09] Sheera Khan: *giggles*
[11:09] Ubit Umarov: :=
[11:09] Andrew.Hellershanks I updated one of the PHP components of the addon search module. It now populates the popularplaces table. There may still be an issue with it and how it reports region maturity. I'm looking in to that as it was ok on my own Standalone.
[11:10] Andrew.Hellershanks Ubit has been doing a lot of work on Warp3D in the past week or so. One thing I've noticed is no more white line down the right edge of maptiles.
[11:12] Sheera Khan: That's a great thing :-)
[11:12] Andrew.Hellershanks Ubit, is Warp3D still known to (or suspected of) leaking memory?
[11:13] Ubit Umarov: you tell me andrew  :p
[11:13] Ubit Umarov: a few leaks where removed already sometime ago
[11:13] Andrew.Hellershanks :) I know it was stated as leaking memory in the past. I haven't checked recently whether it leaks or not.
[11:13] Ubit Umarov: ( we are talking about true leaks not lasy gc )
[11:14] Ubit Umarov: there was a major one on the lib used... no more
[11:14] Bill.Blight It does not appear to be as bad as it was, I'm not sure they were actual leaks or it just used a lot of memory if you had render meshes on and you had a crap ton of mesh on the sim for it to render
[11:14] Andrew.Hellershanks I never noticed leaks caused by Warp3D but I was never testing for that.
[11:14] Kayaker Magic: The grid I've seen the Leaky Robust on, it does have the map tiles set to update often.
[11:14] Ubit Umarov: the damm thing did not released a large buffer pinned
[11:15] Andrew.Hellershanks Kayaker, I have maptiles set to generate only on region start. I'm not sure if they really do get generated on start. That's something else I need to check.
[11:17] Andrew.Hellershanks Last week someone said Dan suspected Presence as the cause for memory leakage. That would make sense. Memory use by Robust seems to go up faster when there is more people logging in and out.
[11:17] Andrew.Hellershanks It is physical ram that gets leaked. A server will eventually crash with "out of memory" errors if not caught in time.
[11:19] Andrew.Hellershanks What other topic(s) would you like to talk about today?
[11:19] Ubit Umarov: ( just a reminder httptests is the dev branck now.. new things go to it not master )
[11:19] Ubit Umarov: and branch :)
[11:20] Sheera Khan: is there any time line if and when the old Jenkins gets an Update?
[11:20] Kayaker Magic: I had a family emergency this week and didn't have time to break anything in httptests yet.
[11:20] Ubit Umarov: still no idea when it is updated
[11:21] Andrew.Hellershanks I switched over to httptests for my Standalone this week. I haven't been running OS locally for a while due to work on other projects.
[11:21] Bill.Blight I assume it will get updated when the people with the keys to the kingdom have time to do it .. but that is just a guess ..
[11:21] Arielle.Popstar who hs the keys?
[11:21] Arielle.Popstar Diva/
[11:21] Arielle.Popstar ?
[11:21] Sheera Khan: *giggles*
[11:21] Andrew.Hellershanks I have the keys but I haven't looked at Jenkins to know what is involved with updating it.
[11:22] Ubit Umarov: no not diva in particular
[11:22] Bill.Blight it is not just jenkins the whole os needs updated
[11:22] Bill.Blight it is like ubuntu 12 or something like that
[11:22] Arielle.Popstar who does the releases now?
[11:22] Sheera Khan: apt-get dist-upgrade ^^
[11:22] Andrew.Hellershanks yea, the server is quite out of date.
[11:22] Sheera Khan: *giggles*
[11:23] Bill.Blight yeah Sheera, that should have been happening on a regular basis, but guessing it didn't
[11:23] Andrew.Hellershanks Sheera, in theory that would work. The problem is that it is so far behind you would have to do that multiple times, advancing one release at a time, to minimize breakage.
[11:23] Arielle.Popstar so its broken for linux?
[11:23] Ubit Umarov: possible losing the box on the way :p
[11:23] Sheera Khan: notice the important word: minimize ^^
[11:24] Andrew.Hellershanks When you are multiple releases behind the quickest thing to do is a fresh install. Not easy to do remotely.
[11:25] JayR Cela: i'm not noticing any problems with 16.06 ubuntu fork
[11:25] JayR Cela: 16.04
[11:25] Bill.Blight that is already 2 versions behind JayR
[11:25] Sheera Khan: depends on the provider... I can reinstall my server via provider Web-Frontend in about 5 minutes
[11:25] Bill.Blight 18 is the current mainstream
[11:25] Andrew.Hellershanks Sheera, yea. Some distros have always recommended updates be done as a fresh install. Ubuntu(?) used to be one that updated well. Still it is a risk.
[11:26] JayR Cela: @ Bill / i prefer to wait for at least a point release
[11:26] Andrew.Hellershanks The server isn't running with a GUI desktop so that reduces one area of potential system failure.
[11:26] Bill.Blight Ubuntu 18.04 LTS

Bionic Beaver

Release Notes

April 26, 2018
[11:27] Bill.Blight that point enough for you
[11:27] Andrew.Hellershanks Sheera, a reinstall of the OS via the providers tools often involves wiping the drive. That's one step we definitely don't want done.
[11:27] JayR Cela: @Bill I'll update when it's ready
[11:27] Bill.Blight the LTS was released two weeks ago,
[11:28] Sheera Khan: It most certainly does involve wiping the disk ...
[11:28] Andrew.Hellershanks Bill, That must be a Canadian focused release. ;)
[11:28] Bill.Blight haha
[11:28] Arielle.Popstar i dont understand the problem if people can build it fine on their own machines. Why is this jenkins thing an issue?
[11:29] Andrew.Hellershanks One of the upgrade issues was about latest mono only working with a newer kernel than is running on the server.
[11:29] Ubit Umarov: its a QA thing
[11:29] Andrew.Hellershanks If there isn't a kernel issue some tools/packages could be updated without doing a full system upgrade.
[11:29] Bill.Blight not being able to run automated builds and testing before putting out code is a Q/A issue
[11:29] Sheera Khan: well, Ubit is the QA thingie for httptests ^^
[11:29] Ubit Umarov: on each commit code is compiled and some tests are done
[11:29] Andrew.Hellershanks nods to bill.
[11:30] Ubit Umarov: anhh bill is one of the suicidal testers
[11:30] Ubit Umarov: hope you all soon :P
[11:30] Bill.Blight yes
[11:30] Bill.Blight I build and test httptests on my jenkins
[11:30] TG.Lucan Hosted hg safari last week
[11:30] Arielle.Popstar so only jenkins machine can run the tests?
[11:31] Ubit Umarov: it is a organization thing
[11:31] Ubit Umarov: that box is the oficial core thing...
[11:31] Arielle.Popstar make bill's the official core thing then
[11:31] Ubit Umarov: it should do it for all commits
[11:32] Bill.Blight hahaha, not gonna happen
[11:32] Bill.Blight although I did offer at one time
[11:32] Arielle.Popstar sounds like an easy solution
[11:32] Bill.Blight I get tired of offering help and being slapped in the face for it
[11:32] Arielle.Popstar no comment
[11:32] Andrew.Hellershanks I've never slapped you, Bill.
[11:32] TG.Lucan @Arielle jenkins is a popular QA system
[11:33] Ubit Umarov: i did noticed you do have a kinda flat face Bill possible from that :p
[11:33] Bill.Blight yep
[11:33] Arielle.Popstar you're not the only one Bill
[11:33] Andrew.Hellershanks :)
[11:34] Ubit Umarov: but iworse is not jenkis, its mono
[11:34] Ubit Umarov: that box is in 2.10.x
[11:34] Andrew.Hellershanks mono can easily be updated.
[11:34] Arielle.Popstar Thx TG. Just questioning why only one "official" machine is allowed to update master
[11:34] Bill.Blight It is a box operating sysem / Mono issue, simply the box has needed an upgrade for a long time
[11:35] TG.Lucan mono runtime has been converted to use coop
[11:35] Bill.Blight I can answer that question easily, with the exception of Ubit, and "sometimes" Andrew, hehe, if the idea does not come from inside of core it automatically sucks ...
[11:36] Arielle.Popstar agrees
[11:36] Arielle.Popstar so have to convince core the idea came from Ubit then it is a shoo in
[11:36] Bill.Blight Not saying there are not a lot of sucky ideas floating around, but this really does seem to prevent people willing to help , from helping, or even trying to help these days ..
[11:37] TG.Lucan Only known problem with httptests on Linux are exceptions during shutdown
[11:37] Andrew.Hellershanks Apparent rejection of ideas coming from people outside that of the main developers on projects is an all too common problem.
[11:37] Arielle.Popstar lapsing into micro management
[11:37] Andrew.Hellershanks TG.Lucan, I noticed that myself the other day when I was testing httptests.
[11:37] Bill.Blight absentee micro-mangement in this case
[11:38] Leighton.Marjoram it is getting to point where it is getting in the way of progress. It also paints a picture that people dont care that the machine is out of date. Both not good.
[11:39] Andrew.Hellershanks isn't sure how to read Bill's "sometimes" Andrew comment. :)
[11:40] Andrew.Hellershanks If the mono on the server is only at 2.10 it should be possible to update it to something more recent even if the kernel prevents use of the most recent version of mono.
[11:40] Ubit Umarov: ( yeach bill andrew does accept anyones ideas, specially the ones that suck :p )
[11:40] Bill.Blight if Ubit was not taking this Bull by the horns and just driving forward with trying to improve it, would ANYTHING be getting done, would anybody even realize the box is so out of date ... It is like a ship with nobody at the wheel, but the engineer keeps throwing fuel in the engine
[11:40] Ubit Umarov hides
[11:40] Arielle.Popstar you all do realize that if a core dev has not contributed to master in 6 months, tyhey no longer have voting rights?
[11:41] Arielle.Popstar does that leave any core dev other then andrew, ubit and mister blue with a say so?
[11:41] Ubit Umarov: hmm bill that image made me remember titanic, no idea why...
[11:42] Arielle.Popstar oh and Mew to i think
[11:42] Bill.Blight the analogy fits even better with the Titanic
[11:43] Arielle.Popstar so vote on updating jenkins and use your keys andrew
[11:43] Arielle.Popstar no need to wait on absentee devs
[11:43] Bill.Blight updating jenkins is not really the issue, saving all the past build info is, one small slip up and you loose all build history
[11:43] Arielle.Popstar cant back up first?
[11:44] Bill.Blight you don't loose commits, because you can pull those from a git that is a mirror of it
[11:44] Bill.Blight but the git is also on that box , so a good mirror or several would need to be made
[11:44] Ubit Umarov: git web site.. lots of stuff :)
[11:45] Bill.Blight yes, the box is not just a one trick pony, so while I agree it needs to be done, it needs to be done correctly,
[11:45] Andrew.Hellershanks I didn't find Jenkins installed on the server via a package of that name.
[11:45] Arielle.Popstar didnt Neb used to do it?
[11:45] Bill.Blight most likely it is a standalone .war
[11:46] Ubit Umarov: many ppl install things by hand
[11:46] Ubit Umarov: not from packages
[11:46] Bill.Blight I doubt there was a package when those repositories on that version were made
[11:46] Andrew.Hellershanks It is using RPM package managment. I used to roll my own custom RPM packages.
[11:46] Andrew.Hellershanks Been a while since I last did that.
[11:47] Bill.Blight but updating jenkins is not the hard part ... LOL
[11:47] Bill.Blight updating the whole box is
[11:48] Andrew Hellershanks: Updating the box is also likely to break running programs on the machine that were not installed via packages.
[11:48] TG.Lucan someone needs to host another box, switch dns bring up new stuff, update old box switch back
[11:49] Andrew Hellershanks: Include making a backup of the current server before switching DNS.
[11:49] TG.Lucan of course
[11:50] TG.Lucan but the original is still in tact... the whole point
[11:50] Arielle Popstar: sounds like we need someone willing to be the scapegoat when it breaks to update the box
[11:50] Andrew Hellershanks: :)
[11:50] JayR Cela: are there any further idea's or plans to stop using UDP file tranfer protocol ?
[11:50] TG.Lucan why?
[11:50] Bill.Blight UDP will never die in opensim I am willing to bet
[11:51] Arielle Popstar: if newer viewers no longer support it for s/l....
[11:51] Ubit Umarov: even recent sl viewers still use udp filetransport bahh
[11:51] JayR Cela: well Linden Lab is finally dumping it along with the FireStorm viewer
[11:52] JayR Cela: so just thought I would ask
[11:52] Ubit Umarov: yeah but they mb forgetting some parts :p
[11:52] TG.Lucan udp and tcp share the same IP layer, only tcp provide a control stack that would otherwise be part of the application
[11:52] Bill.Blight ok this is where I have an issue, people tend to say "Why do we have to follow SL, why can't we do our own thing" , then when something like this comes up , "OMG FOLLOW SL"
[11:52] Arielle Popstar: lol yes
[11:53] Arielle Popstar: but i dont want to see a specific opensim viewer
[11:53] Arielle Popstar: s/l still drives and motivates development here
[11:53] Bill.Blight Then talk to the viewer devs, they are the ones making those decisions
[11:53] JayR Cela: I'm using Singularity Viewer (64 bit) 1.8.7 (6994) Apr 9 2018 03:13:50 (Singularity Test)
[11:53] JayR Cela: works great for me
[11:54] Bill.Blight I still use FS 5.0.7 and various versions of Singularity, work fine
[11:54] Arielle Popstar: sing is too crashy for me
[11:54] Andrew Hellershanks: I mostly use Sing and sometimes FS.
[11:54] JayR Cela: @Arielle it depends on the grid
[11:55] Ubit Umarov: alch seems working also
[11:55] JayR Cela: some people are running poorly set up grid configurations
[11:55] Bill.Blight I like Alchemy, just don't like the 512m draw distance
[11:56] Arielle Popstar: Alchemy doesnt have RLV does it?
[11:56] Bill.Blight don't know, never had a reason to check
[11:57] TG.Lucan @ubit During shutdown..1 or 2 exceptions depending on prior usage, no exceptions if no logins.
[WATCHDOG]: Removing thread
System.ArgumentNullException: Value cannot be null.
[11:57] Bill.Blight does not appear it does according tot he chatter on thsite
[11:58] Bill.Blight wasn't that the same error you got Andrew?
[11:58] Andrew Hellershanks: TG.Lucan, I have been wondering if it was related to use of scripts to generate dynamic texture but it happens.
[11:58] Andrew Hellershanks: Bill, yes that looks like the same issue I see.
[11:59] Andrew Hellershanks: oops... but it happens even if I don't tell a device to update a dynamic texture.
[11:59] TG.Lucan Just noted, no logins to the sim, no shutdown exceptions
[11:59] mayLee Jun: remove the linden client beacon
[11:59] Ubit Umarov: well those ones do help region to shutdown faster :p
[11:59] mayLee Jun: in browser
[12:00] TG.Lucan All info I have is from the mono stack :)
[12:00] TG.Lucan *other
[12:00] Andrew Hellershanks: Which version of mono are you using, TG.Lucan?
[12:00] TG.Lucan 5.15
[12:00] Andrew Hellershanks: ok. I've been testing with mono 5.10
[12:01] TG.Lucan are you running debug?
[12:01] Andrew Hellershanks: Yes
[12:02] TG.Lucan I'll switch to an earlier mono to check
[12:02] Ubit Umarov: keep also in mind that debug mode does have performance impact
[12:02] TG.Lucan sure,
[12:02] TG.Lucan compile with debug also :)
[12:03] Andrew Hellershanks: Ubit, Worth noting. Any idea how much of a performance hit one has running Debug?
[12:03] Ubit Umarov: there a performance diference now
[12:03] Bill.Blight I'd call the hit "very noticeable"
[12:03] Ubit Umarov: new JIT does optimize more.. but not on debug
[12:03] TG.Lucan WE have no users to satisfy, we are a test grid
[12:04] Leighton.Marjoram I am going to head of as RL is calling. Bye for now and see you next week.
[12:04] Ubit Umarov: not dif so large as on C, but getting closer
[12:06] Andrew Hellershanks nods
[12:06] Bill.Blight Just going to quote Ubit from earlier so it makes it in to the logs one more time ...
[12:06] Bill.Blight [11:19] Ubit.Umarov ( just a reminder httptests is the dev branch now.. new things go to it not master )
[12:06] Andrew Hellershanks: We are just past the hour mark. Any other last topics?
[12:07] TG.Lucan We have been testing reduced timeouts, especially on HG connections.. sometime people can be stuck for several minutes
[12:07] Arielle Popstar: yes have seen that
[12:08] Bill.Blight I actually increased the time before the sim considers you "Not there"
[12:08] Arielle Popstar: FS can be a bugger to shutdown incases like that
[12:08] Bill.Blight or should say, "Not arrived"
[12:08] Bill.Blight gives it time to establish the connection before retrying
[12:09] TG.Lucan Things like HG friends, when no longer available, hang around for a while.. kind of hits performance
[12:09] Bill.Blight my change was only for arrival on the sim, had nothing to do with persistence
[12:10] TG.Lucan we placed them in a seperate config, so easier to work with
[12:10] Dragon Moorlord: ive been having issues with people access their sim
[12:10] Dragon Moorlord: not able to find presence even after restart
[12:11] Bill.Blight - if(m_updateAgentReceivedAfterTransferEvent.WaitOne(10000))
+ if(m_updateAgentReceivedAfterTransferEvent.WaitOne(20000))
[12:12] Dragon Moorlord: ok, add that to
[12:12] Dragon Moorlord: or what
[12:12] Dragon Moorlord: lol
[12:12] Bill.Blight diff --git a/OpenSim/Region/Framework/Scenes/ScenePresence.cs b/OpenSim/Region/Framework/Scenes/ScenePresence.cs
index 17603b6a19..60e1e108ef 100644
--- a/OpenSim/Region/Framework/Scenes/ScenePresence.cs
+++ b/OpenSim/Region/Framework/Scenes/ScenePresence.cs
@@ -1976,7 +1976,7 @@ namespace OpenSim.Region.Framework.Scenes

- if(m_updateAgentReceivedAfterTransferEvent.WaitOne(10000))
+ if(m_updateAgentReceivedAfterTransferEvent.WaitOne(20000))
UUID originID = UUID.Zero;
[12:13] Dragon Moorlord: hmm
[12:13] TG.Lucan this is a cool one UpdateAgent(destination, (IAgentData)data, ctx, 200000)
[12:13] TG.Lucan 200 seconds
[12:14] Bill.Blight yeah but wouldn't that fire on ever TP?
[12:14] Bill.Blight every
[12:14] Bill.Blight weird
[12:14] TG.Lucan its a timeout
[12:14] TG.Lucan so cancel if successful
[12:15] Bill.Blight yeah the one I changed is a cancel if successful but only applies to transfers from other regions
[12:15] TG.Lucan open source .. tinker tinker :)
[12:15] Arielle Popstar: i go make a coffee during one of those timeouts
[12:16] Bill.Blight I don't remember how many retries are in that loop, would have to go look at the full code, may be only 1
[12:16] TG.Lucan and a sandwich
[12:17] TG.Lucan maybe dinner for two
[12:17] Bill.Blight no there are none, I remember now, changed that to fix the "Agent never arrived" issue
[12:17] Bill.Blight for people on slow networks , or if several were landing at once, some would "never arrive"
[12:18] Bill.Blight after changing that I don't see it anymore

Personal tools
About This Wiki