(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
[11:02 AM] Andrew Hellershanks: Hello, everyone.
[11:02 AM] Anthropic.Claude @hg.zetaworlds.com: hi
[11:03 AM] Gavin.Hird @grid.xmir.org:8002: I see CLaude is here too :-))
[11:04 AM] Jagga.Meridith @hg.zetaworlds.com: Ubit?
[11:04 AM] Andrew Hellershanks: Let me see if he is on IRC
[11:04 AM] Vincent.Sylvester @hg.zetaworlds.com: Might not be feeling well or his internet went poof again
[11:04 AM] Andrew Hellershanks: He isn't on IRC.
[11:05 AM] Gavin.Hird @grid.xmir.org:8002: does it go poof often?
[11:05 AM] Vincent.Sylvester @hg.zetaworlds.com: It was last week
[11:05 AM] Gavin.Hird @grid.xmir.org:8002: oh
[11:06 AM] Andrew Hellershanks: There was a DoS attack against his provider, IIRC.
[11:06 AM] Vincent.Sylvester @hg.zetaworlds.com: Most, well all, this weeks commits are my fault anyways so I might as well take the blame for them
[11:06 AM] Gavin.Hird @grid.xmir.org:8002: git blame
[11:06 AM] Andrew Hellershanks: :)
[11:07 AM] Vincent.Sylvester @hg.zetaworlds.com: A bug LL added had me looking through user picks and I found a bunch of nonsense in there which he fixed for the most part. There is still some stuff I'd change to reduce the complexity of that system further, but considering it's not time critical it's more bother than it's worth
[11:08 AM] Vincent.Sylvester @hg.zetaworlds.com: In debugging why the chat ranges were not showing properly despite setting them I found that the concierge module would be loaded despite not being configured, but because it is a sub module of the chat module it would fall back on any function call it doesn't implement itself
[11:08 AM] Vincent.Sylvester @hg.zetaworlds.com: It doesn't implement RegionLoaded, which is what sets the chat ranges to be part of opensim extras map
[11:09 AM] Vincent.Sylvester @hg.zetaworlds.com: So it calls the chat module again, without init, so there is no config for it, meaning it uses defaults, defaults which then get added because concierge on most startups will run after the chat module has
[11:09 AM] Vincent.Sylvester @hg.zetaworlds.com: Overwriting the already properly set ranges
[11:10 AM] Gavin.Hird @grid.xmir.org:8002: we have found a couple patterns like that in the viewer code too :-))
[11:10 AM] Vincent.Sylvester @hg.zetaworlds.com: His fix to add a dummy RegionLoaded to concierge works in the sense when you only have the chat module enabled, but when concierge is enabled it does mean it no longer sets the correct chat ranges as concierge is setup to replace the chat module completely
[11:11 AM] Vincent.Sylvester @hg.zetaworlds.com: The option I have in local copy is to just check if the config is null and pull it from the scene if so, then set the ranges via that, which is a roundabout way of doing it, but will work
[11:12 AM] Vincent.Sylvester @hg.zetaworlds.com: It prompted me to check the loading process of modules to find that while a lot of them do have their own bool for enabled they still get called just the same, including all the events like RegionLoaded
[11:12 AM] Vincent.Sylvester @hg.zetaworlds.com: Even if they are not actually configured to run
[11:12 AM] Cuga.Rajal @rajal.org:9000: I saw notes about concierge changes in the commits, are those for fixing this?
[11:12 AM] Gavin.Hird @grid.xmir.org:8002: sweet
[11:12 AM] Andrew Hellershanks: Hello, Lyr
[11:12 AM] Lyr.Lobo @cc.opensimulator.org:8002: Hello *grins*
[11:12 AM] Vincent.Sylvester @hg.zetaworlds.com: So I went through all 200+ modules, adding an Enabled bool that the module loader can check for and avoid calling the RegionLoaded and other events if the init actually disabled the module in the first place
[11:12 AM] Cuga.Rajal @rajal.org:9000: Hi Lyr =)
[11:13 AM] Vincent.Sylvester @hg.zetaworlds.com: No point in running those calls when the module is meant to not be used
[11:13 AM] Vincent.Sylvester @hg.zetaworlds.com: This does speed up the startup process measurably as well
[11:13 AM] Gavin.Hird @grid.xmir.org:8002: I can imagine
[11:13 AM] Vincent.Sylvester @hg.zetaworlds.com: Course it's a pain and a half to go through all of that code to make sure the modules aren't doing something that needs to be done despite not actually being configured and so on
[11:13 AM] Lyr.Lobo @cc.opensimulator.org:8002: /me grins and waves
[11:14 AM] Gavin.Hird @grid.xmir.org:8002: there is ai for that
[11:14 AM] Vincent.Sylvester @hg.zetaworlds.com: I still have to review all that to see if some of that can't be done a little less complicated now with that bool added to not do so much internal logic
[11:14 AM] Vincent.Sylvester @hg.zetaworlds.com: Doesn't have much impact on runtime so not sure it's worth going through all that
[11:14 AM] Gavin.Hird @grid.xmir.org:8002: that is actually one of the things it is pretty darn good at
[11:15 AM] Vincent.Sylvester @hg.zetaworlds.com: In adding that bool I did find a bunch of modules that had no internal logic at all as well, making me wonder what they are there for
[11:16 AM] Vincent.Sylvester @hg.zetaworlds.com: A number of modules acting as base for others, which is a construct that is just ripe for problems
[11:16 AM] Gavin.Hird @grid.xmir.org:8002: so a complier would just ignore them?
[11:17 AM] Vincent.Sylvester @hg.zetaworlds.com: No but they are like base modules that provide some basic functions and then you have sub modules that are actually what's loaded and they implement actual logic
[11:17 AM] Gavin.Hird @grid.xmir.org:8002: ok
[11:17 AM] Vincent.Sylvester @hg.zetaworlds.com: I suppose it's so you can just specify things like assetservice and then whatever is actually enabled takes over
[11:17 AM] Gavin.Hird @grid.xmir.org:8002: makes sense
[11:18 AM] Vincent.Sylvester @hg.zetaworlds.com: Why the concierge module is like that idk, I wasn't the evidently braindead person that wrote the thing. I don't really get what that is for and I haven't seen it be used so I was inclined to just nuke it, since most of what it does do could be integrated into chat module by just adding the options there
[11:19 AM] Vincent.Sylvester @hg.zetaworlds.com: Since that is designed to effectively replace the chat module anyways when enabled it's a bit of a double whammy that seems unnecessary
[11:19 AM] Vincent.Sylvester @hg.zetaworlds.com: Only thing that it does is mean less logic on chat I suppose
[11:20 AM] Lyr.Lobo @cc.opensimulator.org:8002: sessions are only 20 minutes. we use voice in Discord most often as Skype went offline, and we pipe it into YouTube to match with the video stream. You can see examples of past conference sessions on the AvaconInc channel
[11:20 AM] Gavin.Hird @grid.xmir.org:8002: concierge sounds like something that could run on first login to guide a user?
[11:21 AM] Vincent.Sylvester @hg.zetaworlds.com: It seems to be some type of chat welcome bot or something
[11:21 AM] Gavin.Hird @grid.xmir.org:8002: right
[11:21 AM] Andrew Hellershanks: Seems a reasonable guess as to its purpose.
[11:22 AM] Gavin.Hird @grid.xmir.org:8002: could actually be useful, say at the opensim conference
[11:22 AM] Vincent.Sylvester @hg.zetaworlds.com: Spent today helping TJ debug the (???) issue as we found region that reliably reproduces the problem. It seems to be a viewer side thing that assumes the url to retrieve the names from before it actually has the correct one
[11:22 AM] Lyr.Lobo @cc.opensimulator.org:8002: /me smiles
[11:23 AM] Vincent.Sylvester @hg.zetaworlds.com: There is possibly some server side thing in there to make that less problematic, but we do eventually get a request through caps for the names
[11:23 AM] Cuga.Rajal @rajal.org:9000: SL has same issue with ????
[11:23 AM] Vincent.Sylvester @hg.zetaworlds.com: Unfortunately when replying to that caps call the viewer then never updates all the places it needs to to properly display the names
[11:24 AM] Vincent.Sylvester @hg.zetaworlds.com: They sort of do. They have had some trouble with their live testing when urls change between simulator versions so they found a bug in that a couple days ago
[11:24 AM] Gavin.Hird @grid.xmir.org:8002: what is the caps call?
[11:24 AM] Vincent.Sylvester @hg.zetaworlds.com: But that's only sort of related to this
[11:24 AM] Vincent.Sylvester @hg.zetaworlds.com: GetDisplayNames
[11:24 AM] Gavin.Hird @grid.xmir.org:8002: aha
[11:24 AM] Vincent.Sylvester @hg.zetaworlds.com: Before that comes a call to uuidnameresult or something that is handled elsewhere, but the viewer posts this to the region url of the departing region, not the one it goes to
[11:25 AM] Andrew Hellershanks: Here comes Ubit.
[11:25 AM] Vincent.Sylvester @hg.zetaworlds.com: setCapabiltiesReceived is set allowing it to ask for that despite not actually having changed the url yet
[11:25 AM] Andrew Hellershanks: Hello, Ubit.
[11:25 AM] Gavin.Hird @grid.xmir.org:8002: I might have claude to look at what it does. the good thing it looks at both the server and viewer code at the same time to figure out what is supposed to go on
[11:25 AM] Ubit Umarov: (hi)
[11:25 AM] Gavin.Hird @grid.xmir.org:8002: hi
[11:26 AM] Vincent.Sylvester @hg.zetaworlds.com: TJ is looking into ways to add something in there to make sure it only asks once it actually has the correct url
[11:26 AM] Vincent.Sylvester @hg.zetaworlds.com: Fun thing to debug since we couldn't find a region to reliably repro the problem until today
[11:26 AM] Cuga.Rajal @rajal.org:9000: That delay should still be pretty short
[11:27 AM] Vincent.Sylvester @hg.zetaworlds.com: It sends the caps one about a minute after arriving on the region so it does get the correct names, just that routine doesn't touch the nearby floater or the radar so only the floating name signs update properly
[11:27 AM] Vincent.Sylvester @hg.zetaworlds.com: And yeah that part needs fixing too, just in case
[11:28 AM] Ubit Umarov: guess one minute with some lag
[11:29 AM] Vincent.Sylvester @hg.zetaworlds.com: The funny thing is if the uuidnameresult thing hits throttle or cannot respond it tells the viewer to retry after 30 seconds, so while the teleport is going it queues up a coro to ask the old region again if it has already lost the circuit
[11:29 AM] Ubit Umarov: ( generating them is a tiny bit heavy )
[11:30 AM] Vincent.Sylvester @hg.zetaworlds.com: Then the region sends it into the void
[11:30 AM] Vincent.Sylvester @hg.zetaworlds.com: The joys of async
[11:31 AM] Andrew Hellershanks: brb
[11:32 AM] Gavin.Hird @grid.xmir.org:8002: we found a few code paths in the viewer where HG teleports in particular would send the viewer into timeout void so things never got marked as unavailable for the rest of the login session when it fact it was till being fetched by the grid server and therefor was temporary unavailable
[11:32 AM] Gavin.Hird @grid.xmir.org:8002: things go*
[11:32 AM] Gavin.Hird @grid.xmir.org:8002: got even
[11:33 AM] Vincent.Sylvester @hg.zetaworlds.com: There seem to be a lot of assumptions of order in those things when the underlying design is totally async and random depending on what gets delayed
[11:33 AM] Gavin.Hird @grid.xmir.org:8002: the code is written for a single unified grid. LL code don't know anything about HG so there a big holes in behavior because of that
[11:34 AM] Vincent.Sylvester @hg.zetaworlds.com: I found some similar things on the server end that still lack checks for whether the client is even still there
[11:34 AM] Ubit Umarov: like at sl it does matter what region sends a answer to some things, since they are global
[11:34 AM] Vincent.Sylvester @hg.zetaworlds.com: I mean I'd argue it is just good design to not rely on a server always behaving the same way either
[11:34 AM] Gavin.Hird @grid.xmir.org:8002: yes
[11:34 AM] Ubit Umarov: that ofc fails even on a sigle os grid
[11:34 AM] Vincent.Sylvester @hg.zetaworlds.com: And like I said they have found that to bite them in their tests too hence some fixes they are adding
[11:35 AM] Vincent.Sylvester @hg.zetaworlds.com: As even there the urls can differ
[11:35 AM] Vincent.Sylvester @hg.zetaworlds.com: Just less likely
[11:35 AM] Vincent.Sylvester @hg.zetaworlds.com: Never assume, the random sun eruption of particles will ruin your timings
[11:37 AM] Gavin.Hird @grid.xmir.org:8002: for instance we found that if a sound was unavailable when arriving at the grid with a hg transfer, it could result in all sounds not being fetched because of a 7 minute timeout, and only then started coming in
[11:37 AM] Vincent.Sylvester @hg.zetaworlds.com: No doubt some of those timeouts are from times when attempts were made to support dial up
[11:38 AM] Gavin.Hird @grid.xmir.org:8002: we parked them in some sort time retry queue, and fetched those available immediately
[11:38 AM] Gavin.Hird @grid.xmir.org:8002: short time*
[11:39 AM] Vincent.Sylvester @hg.zetaworlds.com: LL has been adding some multi threading to things now too, which is going to be fun to debug. No doubt as reports of region loading neighbors is causing considerable lockups in the viewer while it does whatever it does to fetch the initial data
[11:40 AM] Vincent.Sylvester @hg.zetaworlds.com: I have seen that as bad as 7 or 8 seconds frame time
[11:40 AM] Andrew Hellershanks: I'm back
[11:40 AM] Gavin.Hird @grid.xmir.org:8002: there are what I call fetch stalls involving neighboring regions that are not easy to resolve
[11:41 AM] Vincent.Sylvester @hg.zetaworlds.com: Especially on large var regions it can lock up a long time making me think it's terrain related, but I have not debugged what is going on
[11:41 AM] Vincent.Sylvester @hg.zetaworlds.com: I just noticed it got worse over the last year or so
[11:42 AM] Gavin.Hird @grid.xmir.org:8002: no it it not terrain related, it is a combination of a burst of textures and stuff coming in and decoding it fast enough
[11:42 AM] Vincent.Sylvester @hg.zetaworlds.com: Viewer requesting more than it can process I suppose
[11:43 AM] Vincent.Sylvester @hg.zetaworlds.com: It's weird that it doesn't defer any of that or throttle, it just overloads itself
[11:43 AM] Vincent.Sylvester @hg.zetaworlds.com: And I thought I had bad eating habits
[11:43 AM] Gavin.Hird @grid.xmir.org:8002: so we have added two queues that both use all cores available for both texture and sound decoding which sped up thing a lot
[11:44 AM] Vincent.Sylvester @hg.zetaworlds.com: Only testing I did on this was to see if different bandwidth setting helped it and I couldn't confirm that
[11:44 AM] Andrew Hellershanks: Hello, Sarah. Didn't see you arrive.
[11:44 AM] Gavin.Hird @grid.xmir.org:8002: we found the same Vincent
[11:45 AM] Vincent.Sylvester @hg.zetaworlds.com: It's good to see that's getting attention, cause it has been annoying me more and more the worse it got
[11:46 AM] Vincent.Sylvester @hg.zetaworlds.com: If it's annoying me then others must be furious
[11:46 AM] Gavin.Hird @grid.xmir.org:8002: on each frame the viewer has in essence only 2 ms to process new stuff before a frame is computed, so there are limitations to howmuch can be done and particularly if done on the main thread
[11:47 AM] Andrew Hellershanks: Just to interrupt the discussion for a moment. The welcome page of this grid indicates an important(?) meeting this Sunday (the 27th at 11am grid time in Hurliman Plaza. They are encourage people to attend.
[11:47 AM] Andrew Hellershanks: /me notes two minor typos in that announcement
[11:47 AM] Cuga.Rajal @rajal.org:9000: Did they say what about?
[11:47 AM] Andrew Hellershanks: /me frowns
[11:48 AM] Ubit Umarov: guess future of osgrid
[11:48 AM] Cuga.Rajal @rajal.org:9000: k thx
[11:48 AM] Gavin.Hird @grid.xmir.org:8002: any hints?
[11:48 AM] Lyr.Lobo @cc.opensimulator.org:8002: yes, and while they say not shutting down, probably decision time
[11:48 AM] Ubit Umarov: they also ask for ppl to use local accounts
[11:48 AM] Andrew Hellershanks: No, there wasn't anything clear about it. The one thing they did say is that it isn't about the grid closing.
[11:49 AM] Lyr.Lobo @cc.opensimulator.org:8002: server test *grins*
[11:49 AM] Andrew Hellershanks: Right. Forgot to mention that. Local accounts rather than attendance via HG.
[11:49 AM] Jagga.Meridith @hg.zetaworlds.com: I have church. I'm the organist so they like if I show up
[11:49 AM] Lyr.Lobo @cc.opensimulator.org:8002: on Hurlimann
[11:50 AM] Lyr.Lobo @cc.opensimulator.org:8002: *grins and nods*
[11:50 AM] Andrew Hellershanks: Jagga, :)
[11:50 AM] Cuga.Rajal @rajal.org:9000: "Our database is too big"
[11:50 AM] Gavin.Hird @grid.xmir.org:8002: Have they not heard about streaming Jagga?
[11:51 AM] Andrew Hellershanks: Cuga, That has been an issue for some time. Even with deduping.
[11:51 AM] Cuga.Rajal @rajal.org:9000: yeah
[11:52 AM] Cuga.Rajal @rajal.org:9000: IMHO, they should retire old accounts and delete assets unique to those accounts
[11:52 AM] Andrew Hellershanks: That's all I have to say about the upcoming event. Just wanted to mention it before anyone needs to leave.
[11:52 AM] Lyr.Lobo @cc.opensimulator.org:8002: Thanks, Andrew
[11:52 AM] Andrew Hellershanks: yw, Lyr.
[11:53 AM] Andrew Hellershanks: Cuga, Easier said than done. Hard to know whether the assets are referenced by objects held by other avatars
[11:53 AM] Cuga.Rajal @rajal.org:9000: agree
[11:53 AM] Vincent.Sylvester @hg.zetaworlds.com: I forgot to mention last week regarding libgdiplus where that code lives. The specific fix for the font rendering is this: https://gitlab.winehq.org/mono/libgdiplus/-/commit/bc1f6dc08a659661ee4c56e67c241f998a9a2743 You can also find the current working version in there if you want to build from source to fix a local install
[11:54 AM] Lyr.Lobo @cc.opensimulator.org:8002: /me steps away for a moment
[11:54 AM] Cuga.Rajal @rajal.org:9000: So we can install the fixed update by hand before it gets pushed into the repos
[11:54 AM] Andrew Hellershanks: Thanks for the reminder, Vincent. I wanted to check that out. I was busy getting ready for an industry trade show that was held last week.
[11:56 AM] Cuga.Rajal @rajal.org:9000: vincent do you know if that will be merged into an upcoming release?
[11:56 AM] Cuga.Rajal @rajal.org:9000: of libgdiplus
[11:57 AM] Vincent.Sylvester @hg.zetaworlds.com: I would hope so. It is on their main branch now, which is what the distros should build from and I did yell at canonical to do so, but of course that's a bit like screaming into a pillow as of late
[11:57 AM] Vincent.Sylvester @hg.zetaworlds.com: It takes some time for the distros to take note and build if it isn't critical security stuff
[11:57 AM] Cuga.Rajal @rajal.org:9000: 👍
[12:01 PM] Gavin.Hird @grid.xmir.org:8002: I posted an update of the Mac viewer on Wednesday with a fix that made it run on macOS 27 which was released last Monday.
[12:01 PM] Andrew Hellershanks: I'm building the latest from git now.
[12:01 PM] Vincent.Sylvester @hg.zetaworlds.com: Oh I should update my mini then
[12:01 PM] Cuga.Rajal @rajal.org:9000: Yay Gavin
[12:01 PM] Cuga.Rajal @rajal.org:9000: you said it would never hapen LOL
[12:02 PM] Andrew Hellershanks: uh oh... incoming. Dino coming this way.
[12:02 PM] Gavin.Hird @grid.xmir.org:8002: did I? It still runs under Rosetta 2. macOS 28 will never happen without a major rewrite (owrking on)
[12:03 PM] Cuga.Rajal @rajal.org:9000: Oh, i thought Rosetta was gone in 27
[12:03 PM] Anthropic.Claude @hg.zetaworlds.com: ow
[12:03 PM] Gavin.Hird @grid.xmir.org:8002: nope, it is the last release that supports it
[12:03 PM] Cuga.Rajal @rajal.org:9000: ah, ok thanks
[12:03 PM] Lyr.Lobo @cc.opensimulator.org:8002: Many thanks for the great meeting! I must dash. Have a great week! The OpenSimulator Call for Proposals opens 30 September 2026 for the conference, which is 5-6 December.
[12:03 PM] Lyr.Lobo @cc.opensimulator.org:8002: See you next week
[12:04 PM] Andrew Hellershanks: tc, Lyr. See you next week.
[12:04 PM] Gavin.Hird @grid.xmir.org:8002: bye Lyr
[12:04 PM] Ubit Umarov: have fun Lyr :)
[12:04 PM] Cuga.Rajal @rajal.org:9000: tc Lyr
[12:04 PM] Lyr.Lobo @cc.opensimulator.org:8002: you too *grins*
[12:04 PM] Andrew Hellershanks: We are at the top of the hour. Any last minute item(s) for today?
[12:05 PM] Cuga.Rajal @rajal.org:9000: I must go too. Take care everybody
[12:05 PM] Andrew Hellershanks: ok, Cuga.
[12:05 PM] Andrew Hellershanks: That will do it for another week. Thank you all for coming. See you again next week.
