Chat log from the meeting on 2022-04-26

From OpenSimulator

Revision as of 10:44, 18 May 2022 by Tampa (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
[11:02 AM PT]  Andrew Hellershanks: Hello, everyone.
[11:03 AM PT]  Ubit Umarov: wel so what news do you have?
[11:04 AM PT]  Vincent.Sylvester Trying to fix up the opensim library and how viewers support it. Some fixes to movement behavior and its stability given the issues we had last time. Putting more caps in that viewers can support.
[11:04 AM PT]  Kayaker Magic: I heard a rumor that LL is adding PBR (physically based rendering) features to the viewer. What will have to change in OpenSim for that?
[11:05 AM PT]  Ubit Umarov: as you did notice,  Gavin left last meeting apparently upset  becasue i said the word "firestorm" a bit too many times for his like
[11:06 AM PT]  Ubit Umarov: sad is that while we where here, is did block my access to his machine,  where i was running a test sim.
[11:06 AM PT]  Andrew Hellershanks: I don't think it was the many references to firestorm. I think it was the comment that made it sound like firestorm is the official viewer for OpenSim.
[11:06 AM PT]  Ubit Umarov: after the meeting i could not even take out a oar of the test region.
[11:06 AM PT]  Vincent.Sylvester Which it isn't to clarify
[11:06 AM PT]  Ubit Umarov: sadly he just blocked the use of the machine, not even telling
[11:07 AM PT]  Andrew Hellershanks: Vincent, I know that but that wasn't how it came across to Gavin.
[11:07 AM PT]  Ubit Umarov: So he did terminate that help wa was giving to opensim development
[11:08 AM PT]  Andrew Hellershanks: I'll be right back. I have someone picking up a package shortly. I need to tape up the box.
[11:08 AM PT]  Ubit Umarov: So.. lets just remember a tahnk him for the support he gave for and seems to had been 3 years
[11:08 AM PT]  Vincent.Sylvester Apparently it wasn't, he posted on his forum and I replied to make doubly sure that the intention of the OpenSim project is to support the protocol and handle that as best as possible, working with any viewer dev to work out problems, just as we did this week trying to figure out what to do about the library
[11:08 AM PT]  Ubit Umarov: remember and Thank him.
[11:09 AM PT]  Michael.Christopher hello :)
[11:09 AM PT]  Ubit Umarov: He also did provide some help, for exmple compiling ubOde for macs, and also testing ahd patchin pgslq code
[11:09 AM PT]  Ubit Umarov: apparently he also stopped that
[11:09 AM PT]  Ubit Umarov: time will tell
[11:10 AM PT]  Vincent.Sylvester I had high hopes for Dayturn as a viewer with a focus more on OpenSim than SL, but it's not clear what the goal is now
[11:10 AM PT]  Ubit Umarov: We will not stop talking abotu any viewer, much less the one most people currently use
[11:10 AM PT]  Ubit Umarov: well there is also SceneGate also 100% for opensim
[11:11 AM PT]  Ubit Umarov: we even did refer it less than dayturn here
[11:11 AM PT]  Vincent.Sylvester FS usage is actually decreasing according to my logs anyways so time will tell. FS currently has a few issues that need resolving and lacks support for some things as well, grid status rss comes to mind, that button is still not working
[11:12 AM PT]  Ubit Umarov: wel we never sent viewers needed information for that to work
[11:12 AM PT]  Ubit Umarov: we do now..
[11:12 AM PT]  Ubit Umarov: well  during last week  i had some colaboratuion with Rye from alchemy
[11:13 AM PT]  Ubit Umarov: in past weeks he is beeing here testing and improving alchemy for opensim
[11:13 AM PT]  Ubit Umarov: He did ask me abotu 2 Caps i had tried to add back in 2018, and did gave up, because viewers did ignore them
[11:14 AM PT]  Ubit Umarov: he did modify alch to stop ignoring, and so asked me to add them
[11:14 AM PT]  Ubit Umarov: and i did
[11:14 AM PT]  Andrew Hellershanks: I'm back.
[11:14 AM PT]  Michael.Christopher how did it go?
[11:15 AM PT]  Michael.Christopher wb
[11:15 AM PT]  Ubit Umarov: FetchLIb2 and FetchLibDescendents2 are now up
[11:15 AM PT]  Ubit Umarov: for now only the Alchemy test viewer does use them
[11:15 AM PT]  Michael.Christopher ok
[11:15 AM PT]  Ubit Umarov: in time others can use its code changes and use them also
[11:16 AM PT]  Andrew Hellershanks: Hello, Michael
[11:16 AM PT]  Ubit Umarov: they are not that important, specially viewer side
[11:16 AM PT]  Ubit Umarov: Libray can be fetched from main inventory as before
[11:16 AM PT]  Andrew Hellershanks: Vincent, if FS use is declining according based on your log files, what is taking it place?
[11:16 AM PT]  Ubit Umarov: but region side, can simplify things a bit
[11:17 AM PT]  Vincent.Sylvester Singularity is in there, Cool VL and Alchemy for the most part
[11:17 AM PT]  Ubit Umarov: so last week i did add those 2 caps :)
[11:17 AM PT]  Ubit Umarov: well singu project is dead
[11:17 AM PT]  Ubit Umarov: the last release is very outdated
[11:17 AM PT]  Ubit Umarov: alchemy is supposed to be its replacement
[11:18 AM PT]  Vincent.Sylvester I keep telling people that, but they flip when I block it
[11:18 AM PT]  Andrew Hellershanks: Vincent, ok. I'm a little surprised to hear Sing is in there although I still use it now and then.
[11:18 AM PT]  Ubit Umarov: well Rye told be that, and he should know since it is the same team :)
[11:19 AM PT]  Ubit Umarov: singu was kept longer with its own UI, but having same internal code as Alchemy
[11:19 AM PT]  Ubit Umarov: seems no more..
[11:20 AM PT]  Ubit Umarov: so part of last week was alse testing Alch on those caps, and making some region test code
[11:20 AM PT]  Ubit Umarov: bc sadly alch had a little bug that only was very visible with the new Caps
[11:20 AM PT]  Ubit Umarov: so took a bit to debug, and then get fixed by rye
[11:21 AM PT]  Ubit Umarov: He made some optimixed SIMD code even to convert UUIDs to hexa and back
[11:22 AM PT]  Ubit Umarov: faster hash, etc etc
[11:22 AM PT]  Vincent.Sylvester I also ended up testing viewerasset and getasset caps, which are meant to tell the viewer to get the assets directly from the asset server rather than going through the simulator, but so far that has not worked. It tries to fetch the assets, but somehow the downloads fail. Not quite sure why yet, will have to test some more later.
[11:22 AM PT]  Ubit Umarov: and a tiny bug was hidding somewhere on that "easy to read" SIMD code :)
[11:22 AM PT]  Ubit Umarov: well viewerAsset as no issues
[11:23 AM PT]  Michael.Christopher any chance its as simple as access permissions limited to the simulator so user viewers are blocked as wrong ip?
[11:23 AM PT]  Ubit Umarov: using other url to service them, is outside core :p
[11:24 AM PT]  Ubit Umarov: Region on registration may provide a diferent url for any cap, acording to ini settings
[11:24 AM PT]  Vincent.Sylvester Problem is either the server not providing correct format or viewer not understanding it somehow, will need to test that more. Gavin actually told me quite some time ago that this was kinda broken, but I do not recall as to why that is
[11:24 AM PT]  Ubit Umarov: if that happens,  region takes no more part on cap
[11:24 AM PT]  Vincent.Sylvester Viewers don't tell as to why the download ultimately fails making debugging a bit harder
[11:25 AM PT]  Vincent.Sylvester Will have to idk try and intercept and examine the data or something
[11:25 AM PT]  Ubit Umarov: wel at SL  assets are serviced by CDN
[11:25 AM PT]  Vincent.Sylvester Not quite sure how to approach that yet
[11:25 AM PT]  Ubit Umarov: so must be a simple thing
[11:25 AM PT]  Ubit Umarov: in fact some opensim forks also to it
[11:25 AM PT]  Andrew Hellershanks: I just looked at the Alchemy viewer website and found it is for Windows only. I looked at the Dayturn website and it is for Mac and Raspberry Pi. I run Linux so I can't try out either of those two viewers.
[11:25 AM PT]  Vincent.Sylvester It should be yes, but my guess is the simple 301 redirect I setup serves the data in the wrong format, perhaps it needs all xml base64 encoded still
[11:26 AM PT]  Ubit Umarov: the site is obsolete
[11:26 AM PT]  Ubit Umarov:
[11:26 AM PT]  Ubit Umarov: the beta things are there
[11:27 AM PT]  Ubit Umarov:
[11:27 AM PT]  Vincent.Sylvester At the moment I haven't really seen that much network load on my systems so I wonder if there is even a need to serve assets directly given even the largest regions only have maybe 200mb of them and on most modern connections that doesn't take long to download
[11:27 AM PT]  Ubit Umarov: there windows, mac and linux
[11:28 AM PT]  Ubit Umarov: about the caps, not that there is not even a http redir on those cases
[11:28 AM PT]  Andrew Hellershanks: ty. The Alchemy site didn't mention a Linux version on their main page.
[11:28 AM PT]  Ubit Umarov: the viewer is told abotu the provided url directly
[11:29 AM PT]  Ubit Umarov: and region takes no more part
[11:29 AM PT]  Ubit Umarov: as i said the alch site is obsolete
[11:29 AM PT]  Ubit Umarov: the work on on that gitlab still
[11:29 AM PT]  Andrew Hellershanks: Then what is the official site for Alchemy?
[11:29 AM PT]  Ubit Umarov: guess it is that :)
[11:29 AM PT]  Vincent.Sylvester It's testing builds Andrew, no releases yet
[11:30 AM PT]  Ubit Umarov: just no new releases.. just the beta work
[11:30 AM PT]  Andrew Hellershanks: Vincent, I realize that. I just expected to see a website about the project and not just a code repository.
[11:30 AM PT]  Vincent.Sylvester They do have a website, but it only really gets updates when releases happen
[11:30 AM PT]  Ubit Umarov: well more abotu code changes
[11:31 AM PT]  Vincent.Sylvester Plus the testing builds are just that, for testing, so releasing them out in the wild to users, known for breaking everything not up a tree by three, probably isn't a good idea
[11:31 AM PT]  Vincent.Sylvester Get that stuff tested and working first
[11:31 AM PT]  Ubit Umarov: i fix a small bug i did made that have nasty impact on this region, using firestorm
[11:31 AM PT]  Ubit Umarov: looked ok with other viewers
[11:31 AM PT]  Ubit Umarov: and five on any other regions..
[11:32 AM PT]  Ubit Umarov: This region is running at full 47fps
[11:32 AM PT]  Michael.Christopher right vincent? thats why i havent released anything, showing something too early is usually counter-productive
[11:32 AM PT]  Ubit Umarov: or more correclty its heartbeat rate is 47fps approximatly
[11:32 AM PT]  Ubit Umarov: that did expose my bug :p
[11:33 AM PT]  Ubit Umarov: not clear why other viewers like old suingu and alch had no visible problems
[11:33 AM PT]  Ubit Umarov: we FS we just could not walk
[11:33 AM PT]  Ubit Umarov: with..
[11:33 AM PT]  Ubit Umarov: wel almost...
[11:34 AM PT]  Ubit Umarov: so that is fixed
[11:34 AM PT]  Vincent.Sylvester Also weird how that got worse when trying to walk and turn the avatar
[11:35 AM PT]  Vincent.Sylvester Only time I ever seen something like that before was back when there was issues with lludp allowing too many packets in at once
[11:35 AM PT]  Ubit Umarov: made the sun texture show up on library :)
[11:35 AM PT]  Ubit Umarov: a few changes to library code, of top of supporting the new caps
[11:36 AM PT]  Ubit Umarov: a few more changes on avatar movement, trying to reduce cpu use
[11:36 AM PT]  Ubit Umarov: as i said that is a heavy thing
[11:37 AM PT]  Ubit Umarov: viewers to throttle,  but can send like 90 updates per second
[11:37 AM PT]  Ubit Umarov: ...
[11:37 AM PT]  Michael.Christopher did anyone else hear mention of a working viewer for oculus other than the work in progress mention on austin tate's blog?
[11:38 AM PT]  Ubit Umarov: if that is actual unthrottle movement that means 90* number of avatars,  packets send
[11:38 AM PT]  Ubit Umarov: like 9000 packets with 100 avatars in a region :)
[11:38 AM PT]  Ubit Umarov: per second..
[11:39 AM PT]  Ubit Umarov: so some heavy throttling and needed
[11:39 AM PT]  Ubit Umarov: is needed
[11:40 AM PT]  Ubit Umarov: it all be a code area to visit and revisit again :(
[11:40 AM PT]  Vincent.Sylvester Also for sake of scene update queue, which it seems is the primary target of some scripters these days. Seen quite a few regions "stop responding" due to scripts sending so many scene updates that the queue just breaks
[11:41 AM PT]  Vincent.Sylvester It doesn't automatically recover like heartbeat, once dead you have to restart the simulator
[11:41 AM PT]  Ubit Umarov: well a mantis on that, could help
[11:41 AM PT]  Vincent.Sylvester Nothing to really fix I think, it's just overload
[11:41 AM PT]  Ubit Umarov: with details about how to repo, ofc
[11:42 AM PT]  Ubit Umarov: opensim had no idea about the need of throtlling on many areas
[11:42 AM PT]  Ubit Umarov: so sure some need it
[11:42 AM PT]  Ubit Umarov: see the simple llSay case..
[11:42 AM PT]  Ubit Umarov: it had no throotle
[11:43 AM PT]  Ubit Umarov: a very basic script could tunr a region useless just with it
[11:43 AM PT]  Ubit Umarov: llemail had no throttling..
[11:44 AM PT]  Andrew Hellershanks: VIncent, when I looked at one of the event queues it appeared as if packets would be dropped if more than a specific amount were already in the queue. It could kill a script that would need a response to something that got dropped from the queue. That doesn't mean the queue system is dead but it could very well cause a script to wait on an event that will never happen.
[11:44 AM PT]  Ubit Umarov: he was not talking abotu event queues
[11:44 AM PT]  Andrew Hellershanks: No, but I had no reason to think that other queues would be handled any differently.
[11:45 AM PT]  Ubit Umarov: ofc any other is handled diferent
[11:45 AM PT]  Ubit Umarov: even queues have own rules..
[11:45 AM PT]  Andrew Hellershanks: So other queues don't impose a limit on their size?
[11:45 AM PT]  Ubit Umarov: remember the "opensim has no limits" ?
[11:46 AM PT]  Andrew Hellershanks: If there is no limit why do some queues fall over under a given load?
[11:46 AM PT]  Vincent.Sylvester Once the scene update queue overloads no scripts run at all, avatars can get stuck or animations stop. It's a hard failure once overloaded the entire system stops running anything. I suspect it does produce some internal error, but that's masked. Haven't found anything that says "scene update queue full" or something like that. Usually this is caused by scripts doing too much stuff at once or, well, the suspect scripts are usually of a similar kind so it is fairly easy to just remove everything by those authors and things work fine.
[11:47 AM PT]  Ubit Umarov: bc there are no limits where LIMITS ARE NEEDED
[11:47 AM PT]  Ubit Umarov: lol
[11:47 AM PT]  Ubit Umarov: :p
[11:47 AM PT]  Andrew Hellershanks: That implies the queues do have limits. If there is no limit I would just expect major lag while the code catches up on the data in a queue.
[11:47 AM PT]  Ubit Umarov: that "rule" was just of the nonsense opensim still has all over the code
[11:47 AM PT]  Vincent.Sylvester I haven't bothered looking into this as it's only specific regions that overload and once those scripts and objects are removed everything works fine going forward so I file this under misuse rather than bug
[11:48 AM PT]  Ubit Umarov: been fixed, slowly and case by case
[11:48 AM PT]  Andrew Hellershanks nods
[11:48 AM PT]  Vincent.Sylvester I suspect it is some sort of array or list and once you try to add something above integer max size the entire construct just falls apart and fails in subsequent code
[11:49 AM PT]  Vincent.Sylvester Checking if list longer than 0 when it is rolled over int -2 billion
[11:49 AM PT]  Ubit Umarov: remember how some ossl was even added to by pass NEEDED mechanings
[11:49 AM PT]  Ubit Umarov: lke the stupid OSSL notecards code
[11:49 AM PT]  Ubit Umarov: that ignored that notecards are assets
[11:49 AM PT]  Ubit Umarov: across the web, many times
[11:50 AM PT]  Ubit Umarov: so to bypass the DataServer mechanics, was..  well stupid is the simple current word
[11:50 AM PT]  Ubit Umarov: correct also
[11:50 AM PT]  Vincent.Sylvester The problem with notecards is the information on when and who created them is part of their data and gets hashed so each notecard, even if empty, has a different hash and thus full asset information along with it totaling a few kb each time, that adds up quickly
[11:51 AM PT]  Ubit Umarov: now to fix it, a hidden and more complex dataserv like mechanics is needed
[11:51 AM PT]  Vincent.Sylvester The only way to solve that is treating notecard contents as asset only and saving the creation time and other description on the asset table, where they, as empty notecards, would refer to the same asset
[11:51 AM PT]  Vincent.Sylvester But re-wiring assets to that degree is not easy and might not convert either
[11:51 AM PT]  Kayaker Magic: Make notecards mutable?
[11:52 AM PT]  Ubit Umarov: wel the materials assets wher fun
[11:52 AM PT]  Ubit Umarov: a new asset on every click of the mouse wheel changin a numeric vlaue like mask
[11:52 AM PT]  Vincent.Sylvester No, creating notecards themselves is fine, if done responsibly osMakeNotecard is a fine addition. People just don't seem to understand that you don't use them to store running data every 5 seconds.
[11:52 AM PT]  Ubit Umarov: like at least 10 new assets per second
[11:53 AM PT]  Vincent.Sylvester There is a perfectly reasonable way to store data, Prim Description
[11:53 AM PT]  Ubit Umarov: well as we can,  we try to improve things
[11:54 AM PT]  Ubit Umarov: in some cases a mantis with repo steps, can ofc, help
[11:54 AM PT]  Ubit Umarov: but as i said, opensim was broken by design, on that.. so not easy fixes
[11:55 AM PT]  Vincent.Sylvester Fix one thing, something else breaks, rinse and repeat
[11:55 AM PT]  Andrew Hellershanks: I have often wondered about the handling of NCs and scripts.
[11:57 AM PT]  Ubit Umarov: wel think i made a few more code changes, but minor
[11:57 AM PT]  Ubit Umarov: so what news do you have?
[11:57 AM PT]  Ubit Umarov: :)
[11:58 AM PT]  Vincent.Sylvester What might make sense is a on_derez event triggered on simulator shutdown that could let scripts know to write their data to a notecard so the runtime data isn't lost between clearing the scriptengine data, on a binary change for example. Some have kept the same data across binary versions, not sure that is a good idea
[11:59 AM PT]  Ubit Umarov: jezz and an never stopping reagion looping aroud all that
[11:59 AM PT]  Vincent.Sylvester lol yeah that's the downside of that
[11:59 AM PT]  Andrew Hellershanks: Back on the situation with Gavin and the Dayturn viewer, I looked at the Dayturn website and found a forum thread from him that resulted from last weeks meeting. It also contains comments from Tampa. It is long so I haven't read all of it yet. If you want to read it have a look at the web page at
[11:59 AM PT]  Ubit Umarov: ppl already to kill the process  while the poor thing is doign normal backup
[11:59 AM PT]  Ubit Umarov: go kill...
[12:00 PM PT]  Andrew Hellershanks: We are at the top of the hour. Any other last minute items for today?
[12:00 PM PT]  Michael.Christopher not here
[12:00 PM PT]  Vincent.Sylvester It would ultimately need a timeout to give it a grace period of say 10 seconds or so to run things and then it would just proceed with the shutdown, not ideal
[12:01 PM PT]  Andrew Hellershanks: Heading out, Kayaker?
[12:01 PM PT]  Kayaker Magic: I need to run. See you all next week!
[12:01 PM PT]  Vincent.Sylvester Might look into that, but more concerned with the whole prim deletion thing from the mantis I filed the other day
[12:01 PM PT]  Andrew Hellershanks: Kayaker, See you again next week.
[12:01 PM PT]  Michael.Christopher cya kayaker!
[12:01 PM PT]  Ubit Umarov: well on the prim delete is more confusing to not see them on region
[12:02 PM PT]  Ubit Umarov: but still colliding with them
[12:02 PM PT]  Ubit Umarov: because the objects are not really deleted until the all take to trash is done
[12:02 PM PT]  Ubit Umarov: viewers are just told to not show htem
[12:03 PM PT]  Selby.Evans Must go -- Bye all
[12:03 PM PT]  Michael.Christopher bye!
[12:03 PM PT]  Ubit Umarov: cya
[12:03 PM PT]  Vincent.Sylvester Even just a dump of all uuids "to be deleted" on shutdown would help, something to even let me know stuff is still there
[12:04 PM PT]  Vincent.Sylvester Or even a "show delete queue" I can delay shutdown manually if found to have items
[12:04 PM PT]  Vincent.Sylvester Just something to say "hey stuff is still processing"
[12:04 PM PT]  Ubit Umarov: :)
[12:05 PM PT]  Vincent.Sylvester Like with shutdown since I wrap all commands I do a show users, if that says people are still there it sends an alert first and then shuts down later
[12:06 PM PT]  Vincent.Sylvester So a "show delete queue" saying "Items to be deleted: 5" I can easily crawl from the log and tell shutdown to halt until that says 0
[12:06 PM PT]  Vincent.Sylvester Not really elegant, but better than objects coming back from the dead
[12:06 PM PT]  Ubit Umarov: take to trash is irritating
[12:07 PM PT]  Andrew Hellershanks: My first thought about that is to delay the shutdown until after the delete queue is empty.
[12:07 PM PT]  Ubit Umarov: i just set things to temporary
[12:07 PM PT]  Ubit Umarov: so the are just really gone
[12:07 PM PT]  Vincent.Sylvester Problem is delete is a keyboard hotkey, set temporary isn't
[12:07 PM PT]  Vincent.Sylvester Literally delete button
[12:08 PM PT]  Ubit Umarov: yeah
[12:08 PM PT]  Ubit Umarov: is nice to have rececl bin
[12:09 PM PT]  Vincent.Sylvester Delaying shutdown is problematic, it can hang when take to inventory fails, better would be to just dump them to disk or mark them so the deletion can proceed after the restart, perhaps a flag in the prims table?
[12:09 PM PT]  Ubit Umarov: just the the on region obejects are particular heavy to do
[12:09 PM PT]  Vincent.Sylvester I'd be happy just to be able to check if there is still a queue
[12:11 PM PT]  Ubit Umarov: ofc some of those may be 1000 prims objects each with 1000 items on contents :p
[12:11 PM PT]  Ubit Umarov: er even more, since "opensim has no limits"
[12:12 PM PT]  Vincent.Sylvester Well having a way to check would open up avenues to write force delete function as well for anyone that desires that
[12:13 PM PT]  Vincent.Sylvester Adding commands is fairly easy, the big issue is interfacing with that queue and figuring out what's in it, that I haven't quite figured out
[12:13 PM PT]  Ubit Umarov: "Complete movement of Prim itive into Lbsa Plaza"
[12:13 PM PT]  Ubit Umarov: nice nick :)
[12:15 PM PT]  Andrew Hellershanks: :)
[12:15 PM PT]  Andrew Hellershanks: If there isn't anything more for today I'll wrap up the meeting.
[12:16 PM PT]  Andrew Hellershanks: Thank you all for coming. See you again next week.
Personal tools
About This Wiki