Chat log from the meeting on 2024-05-07

From OpenSimulator

Jump to: navigation, search
[11:01 AM PT]  Andrew Hellershanks: Hello, everyone.
[11:01 AM PT]  Gavin.Hird Hi Andrew
[11:01 AM PT]  Ubit Umarov: hi
[11:02 AM PT]  Jagga Meredith: I worked on IBM 1130 Mainframe.  Took punched card input
[11:02 AM PT]  Ubit Umarov: so..  about last week code changes.. not much
[11:03 AM PT]  Gavin.Hird punch cards never exceeded 80 columns, but tabulated print could be 128 cols or more
[11:03 AM PT]  Ubit Umarov: i did a little change to mysql connector 8.3.0 and seems to be working
[11:03 AM PT]  Gavin.Hird does the new version require 8.3.0 or can it use it if you want to?
[11:04 AM PT]  Ubit Umarov: seems 8.0.31 had same issue and i did same fix back then...
[11:04 AM PT]  Ubit Umarov: ours 8.0.31 was working
[11:05 AM PT]  Vincent.Sylvester That connector is going to be a pain going forward, but works for now. Changing it is pretty trivial at least, less so some other new things we supposed to use :/
[11:05 AM PT]  Ubit Umarov: i did place the changed code on opensim-libs
[11:05 AM PT]  Ubit Umarov: i respect to opensource rules :)
[11:05 AM PT]  Ubit Umarov: in..
[11:05 AM PT]  Ubit Umarov: its a huge 2 lines change..  in fact could had benn one
[11:06 AM PT]  Vincent.Sylvester Small change, huge impact
[11:06 AM PT]  Gavin.Hird does the new version require 8.3.0 or can it use it if you want to?
[11:07 AM PT]  Ubit Umarov: you cna use 8.0.31
[11:07 AM PT]  Gavin.Hird good
[11:07 AM PT]  Ubit Umarov: but foe maria you need my 2 lines change
[11:07 AM PT]  Gavin.Hird migration to 8.3.0 might require some more planning...
[11:07 AM PT]  Ubit Umarov: diference is that 8.3.0   was changed to use dotnet sugar and things
[11:07 AM PT]  Gavin.Hird I use the Oracle version
[11:07 AM PT]  Ubit Umarov: so should be a tiny bit faster
[11:08 AM PT]  Ubit Umarov: im talking abotu oracle...
[11:08 AM PT]  Ubit Umarov: it may also use cached buffers that this way are shared on all the app
[11:08 AM PT]  Ubit Umarov: minor things..
[11:09 AM PT]  Ubit Umarov: dotnet shared buffers i mean
[11:09 AM PT]  Ubit Umarov: wel as i was telling added a new midday setting that new viewers want
[11:10 AM PT]  Ubit Umarov: they have the uuid for it hardwired..
[11:10 AM PT]  Ubit Umarov: so we need to have it
[11:11 AM PT]  Ubit Umarov: ( ofc if one goes sl first, get is on viewer cache :) )
[11:11 AM PT]  Ubit Umarov: reason why took me a while to notice we where missing it
[11:11 AM PT]  Ubit Umarov: i also changed the version of Yengine compiler to force compilation
[11:12 AM PT]  Gavin.Hird when abouts was that added to the viewer code?
[11:12 AM PT]  Ubit Umarov: as reported on mantis 9125, scripts where failing on tps
[11:13 AM PT]  Ubit Umarov: the new mididay...  recently.. seen on on fs 7.1.7
[11:13 AM PT]  Ubit Umarov: they just changed the UUID
[11:13 AM PT]  Vincent.Sylvester What about Postgres now. Should we just drop that completely going forward? Eventually people will upgrade to newer versions and it will fail to start. The connector needs fixing or a migration to move that one guid out back to varchar. Though I don't know what else is broken beyond that.
[11:13 AM PT]  Ubit Umarov: fun pstgres has high ratings on some plaves
[11:14 AM PT]  Ubit Umarov: a ot more than maria
[11:14 AM PT]  Joe Magarac: Postgres is generally well thought of.
[11:14 AM PT]  Vincent.Sylvester It is slightly faster than mysql in some metrics, but that also depends on the connector used
[11:14 AM PT]  Joe Magarac: And fewer licensing games.
[11:15 AM PT]  Cuga.Rajal could say that the next release is the last version to support it
[11:15 AM PT]  Ubit Umarov: ( and those where the last week changes on code and things )
[11:16 AM PT]  Vincent.Sylvester If you take current release already and try to get that to run with Postgres latest on ubuntu 22 it blows up
[11:16 AM PT]  Vincent.Sylvester It's already that broken
[11:16 AM PT]  Vincent.Sylvester So I'd vote for removing that OR redoing it and upgrading the connector, which is going to be a pain with the dependencies that requires
[11:17 AM PT]  Vincent.Sylvester I documented on that mantis ticket about this how much pain that is with the unsafe crap that got added and all the other weird dlls it only really uses one function of
[11:17 AM PT]  Vincent.Sylvester We'd have to ship all that along with it
[11:18 AM PT]  Vincent.Sylvester And probably still migrate that guid to varchar, which not sure that's gonna go smoothly
[11:18 AM PT]  Cuga.Rajal are there any metrics of active PG OS users?
[11:19 AM PT]  Vincent.Sylvester I always thought of Gavin as the last holdout, but there might be more still using it, especially the foss fanatics
[11:19 AM PT]  Cuga.Rajal are they even around and if so are they even using trunk
[11:20 AM PT]  Gavin.Hird I would gladly revert to Postgres, but the work to get it up to speed is IMO relatively significant
[11:21 AM PT]  Cuga.Rajal I suppose you could send a message to the listserv and ask
[11:21 AM PT]  Gavin.Hird I suppose you could leave the code in there, just not build it in the project
[11:21 AM PT]  Cuga.Rajal not sure if that group is a good measure
[11:22 AM PT]  Vincent.Sylvester It's probably what like ten hours of work at least to figure out what to change in the connector to have it now blow up
[11:22 AM PT]  Cuga.Rajal it's good practice not to leave broken code lying around that implies it's usable
[11:23 AM PT]  Gavin.Hird leaving it means a bigger chance of getting it working again
[11:23 AM PT]  Gavin.Hird removed nobody will bother
[11:23 AM PT]  Cuga.Rajal create a poll on Google (kidding)
[11:24 AM PT]  Vincent.Sylvester Or they are compelled to contribute a fixed version when support is dropped
[11:25 AM PT]  Vincent.Sylvester Either way I can't include it in migrations rework because I cannot test it, I'd have to get an old version of Postgres on ubuntu focal or something to get it running to test those
[11:25 AM PT]  Ubit Umarov: i didn't look to a updated connector for it
[11:25 AM PT]  Ubit Umarov: can look, but then can't test :)
[11:26 AM PT]  Ubit Umarov: well not even sure what is now broken on it..
[11:26 AM PT]  Vincent.Sylvester guid
[11:26 AM PT]  Ubit Umarov: the connector itselt?
[11:26 AM PT]  Ubit Umarov: they also changed guid?
[11:26 AM PT]  Gavin.Hird there are some missing casts in libopenmetaverse so without those it will not work regaredless of how new the connector is
[11:26 AM PT]  Ubit Umarov: duhh what missing casts?
[11:26 AM PT]  Ubit Umarov: uuid/guid are there
[11:27 AM PT]  Ubit Umarov: i think..
[11:27 AM PT]  Gavin.Hird I have expalined it many years ago, but you did not want to listen so I just gave ujp ,-)
[11:27 AM PT]  Vincent.Sylvester That's the first thing that blew up after I got the connector play nicely in the first place, the latest version requires dependent dlls, let me find the mantis, it's all in there
[11:27 AM PT]  Vincent.Sylvester
[11:28 AM PT]  Ubit Umarov: well years ago UUID was diferent.. i remade it almost 100% :P
[11:28 AM PT]  Vincent.Sylvester That documents my odyssey with npgsql
[11:28 AM PT]  Vincent.Sylvester Get it working to the point the connector stops complaining about missing dependencies only to have it barf on guid
[11:28 AM PT]  Ubit Umarov: on maria and mysql guid it a total mess
[11:29 AM PT]  Gavin.Hird exactly Vincent
[11:29 AM PT]  Vincent.Sylvester I think Postgres only uses it at one spot even
[11:29 AM PT]  Ubit Umarov: then made it binary.. then no deia why moved to 32 chars.. then no idea why to 36 chars
[11:29 AM PT]  Ubit Umarov: sound very bad moves..  but they did them
[11:29 AM PT]  Vincent.Sylvester Could change it, but if it doesn't load at all the migrations won't either, so that means at least putting all those dlls in
[11:30 AM PT]  Ubit Umarov: so we jsut have them as varchar 36 no native type
[11:30 AM PT]  Ubit Umarov: wel and uuid is not guid
[11:30 AM PT]  Ubit Umarov: diferent rules.. etc
[11:30 AM PT]  Ubit Umarov: guid is ms interpretation..
[11:31 AM PT]  Ubit Umarov: but we do have casts.. somewhere :p
[11:32 AM PT]  Gavin.Hird the other issue is that most of the SQL needs to be brushed up to use upsert, which Postgres did not support at all when the code was written
[11:33 AM PT]  Gavin.Hird can probably collapse a lot of code tool in the simple table handler
[11:34 AM PT]  Vincent.Sylvester OpenSim.Data is a total mess
[11:34 AM PT]  Gavin.Hird maybe, but this has nothing to do with that
[11:34 AM PT]  Vincent.Sylvester The other day I added new fields to some tables to record changes via current_timestamp...
[11:34 AM PT]  Vincent.Sylvester 3 tables blew up. One failed as the connector tried to enter a timestamp in imperial formatting while the db only accepts metric one
[11:35 AM PT]  Vincent.Sylvester Another failed because the field count differed
[11:35 AM PT]  Gavin.Hird there is loads of code to handle when to do update and when to do insert, but with upsert it can be collapsed to one code path
[11:35 AM PT]  Vincent.Sylvester Another because it said it couldn't find the field even though it's not meant to look
[11:36 AM PT]  Vincent.Sylvester The various modules parts need to be unified and simplified, some better query design to speed up the returns and reduce the flood of data required for some calls
[11:36 AM PT]  Vincent.Sylvester Groups alone just fetching everything from tables one by one instead of joining things and only grabbing what it needs
[11:36 AM PT]  Gavin.Hird region loading does an awful lot of writes
[11:37 AM PT]  Vincent.Sylvester Yeah so I then fixed the issue with the timestamp with a trigger, guess what, they trigger on insert for two tables and one does trigger update
[11:37 AM PT]  Ubit Umarov: it does? hmm
[11:37 AM PT]  Ubit Umarov: didn't notice...
[11:37 AM PT]  Vincent.Sylvester So some of them instead of changing a record do a full rewrite
[11:37 AM PT]  Gavin.Hird ouch
[11:38 AM PT]  Gavin.Hird well update in postgres is always a full rewrite anyway
[11:39 AM PT]  Vincent.Sylvester My plan for now is to finish the new migration system so at least the base is solid to start fixing the rest of the mess
[11:39 AM PT]  Gavin.Hird ... which is required to back out the transaction
[11:40 AM PT]  Vincent.Sylvester The fact some modules pull data based on index number rather than building a kvp though, that's concerning
[11:40 AM PT]  Ubit Umarov: no ideia why you think region load does a lot of db writes..
[11:40 AM PT]  Gavin.Hird timestamps
[11:40 AM PT]  Gavin.Hird lots
[11:41 AM PT]  Ubit Umarov: hmm and why do we have those? :)
[11:41 AM PT]  Vincent.Sylvester Needs a good dba with a chainsaw to clean up that mess, anyone volunteering? ;)
[11:42 AM PT]  Vincent.Sylvester *crickets*
[11:42 AM PT]  Andrew Hellershanks: :)
[11:42 AM PT]  Jagga Meredith: been awhile since I touched a database and I'm primarily an ORACLE guy
[11:42 AM PT]  Ubit Umarov: we do need version on prims..  but that is now done by that pseudocrc thingy
[11:43 AM PT]  Ubit Umarov: load region should be only reads..
[11:43 AM PT]  Ubit Umarov: but well DB code is a strange thing..
[11:44 AM PT]  Gavin.Hird I have not looked at it since ...3 years ago or so
[11:44 AM PT]  Ubit Umarov: well its mostly the same
[11:44 AM PT]  Andrew Hellershanks: If the DB system was noting when a record was last read that might seem like an update. Other than that I wouldn't expect records to change on reads.
[11:46 AM PT]  Andrew Hellershanks: We are finished with the DB related discussions for today? If so, what is the next topic for today? We are 15 minutes before the top of the hour.
[11:48 AM PT]  Cuga.Rajal nothing new here, except a desire to learn how to improve the Bullet connector and mesher
[11:50 AM PT]  Jagga Meredith: gotta run RL
[11:51 AM PT]  Vincent.Sylvester I'm gonna continue work under the assumption we'll no longer support postgres and just let my code fall into the void on it
[11:51 AM PT]  Andrew Hellershanks: ok, Jagga. Thanks for dropping by.
[11:51 AM PT]  Vincent.Sylvester Maybe someone reading the logs will step up eventually
[11:51 AM PT]  Gavin.Hird it is a tall challenge in many ways
[11:52 AM PT]  Joe Magarac: Still plugging away on Sharpview. Currently: 1) annoying WGPU devs by showing them they have a locking conflict that's killing frame rate, 2) working on first phase of animation control.
[11:53 AM PT]  Vincent.Sylvester Neat
[11:53 AM PT]  Gavin.Hird I have slwoly started to update my viewer code again
[11:53 AM PT]  Vincent.Sylvester Doubly neat :D
[11:53 AM PT]  Gavin.Hird major catchup to get current
[11:54 AM PT]  Joe Magarac: Yes, LL is adding so much stuff that viewers need a lot of attention now.
[11:55 AM PT]  Gavin.Hird it is crazy the roll they are on and it is almost impossible to follow all the changes and background for them unless you have TPV access
[11:55 AM PT]  Ubit Umarov: like inventory code..
[11:55 AM PT]  Ubit Umarov: Beq did recover most of what we need, that ll had removed
[11:55 AM PT]  Joe Magarac: What's "TPV access"?
[11:56 AM PT]  Gavin.Hird TPV access gives better access to their Jira and new equivalent, in addition to Havok and other code
[11:56 AM PT]  Andrew Hellershanks: TPV-> Third Party Viewer
[11:56 AM PT]  Ubit Umarov: they have no jira now :)
[11:57 AM PT]  Joe Magarac: Ah.
[11:57 AM PT]  Ubit Umarov: well have replavements..
[11:57 AM PT]  Joe Magarac: I never asked for that. I just look at LL's github.
[11:57 AM PT]  Gavin.Hird most of the TPV have bought expensive KDU and other licenses used by the LL viewer
[11:58 AM PT]  Andrew Hellershanks: Good to hear programs on Gavin's viewer and on Sharpview.
[11:59 AM PT]  Joe Magarac: Ah. Havok doesn't do much, really. Just a format for displaying the pathfinding mesh and a crappy convex hull decomposition.
[11:59 AM PT]  Joe Magarac: Viewer side, that is.
[11:59 AM PT]  Gavin.Hird In addition after Broadcom buight VMware they are making a fuckery of some of the sw I have used
[11:59 AM PT]  Gavin.Hird not sure how that will end
[11:59 AM PT]  Joe Magarac: And Vivox is going away.
[11:59 AM PT]  Ubit Umarov: viewers for opensim use hacd or hadc one of those :)
[12:00 PM PT]  Ubit Umarov: should be good.. used by many,,  soemhow not that good on viewers
[12:00 PM PT]  Joe Magarac: Next meeting, have to go. Bye!
[12:00 PM PT]  Andrew Hellershanks: Joe, we know. Still ongoing as to what is going to happen about that.
[12:00 PM PT]  Andrew Hellershanks: Bye, Joe. See you next week.
[12:01 PM PT]  Ubit Umarov: they moved to webrtc was we spoke about in prev meetings
[12:01 PM PT]  Ubit Umarov: wel or are moving..
[12:01 PM PT]  Gavin.Hird BugSplat is also paid for lisence
[12:02 PM PT]  Gavin.Hird Tracy might be too
[12:02 PM PT]  Ubit Umarov: didn't they changed to tracy?
[12:02 PM PT]  Ubit Umarov: or its fs?
[12:02 PM PT]  Vincent.Sylvester Don't they all have a free plan to get you addicted to it?
[12:03 PM PT]  Gavin.Hird Bugsplat does not have a useful free
[12:03 PM PT]  Gavin.Hird Tracy probably does
[12:04 PM PT]  Ubit Umarov: well rl calls
[12:04 PM PT]  Ubit Umarov: any other issue?
[12:04 PM PT]  Vincent.Sylvester The free plan doesn't work for what viewers need?
[12:05 PM PT]  Gavin.Hird I suppose you can get it to work, but there is no jira integration in the free, so then you have to juggle that
[12:05 PM PT]  Cuga.Rajal I need to head out RL... Take care all
[12:05 PM PT]  Vincent.Sylvester Ouch yeah that might be a bit annoying hehe
[12:05 PM PT]  Gavin.Hird cheers
[12:05 PM PT]  Ubit Umarov: yeha me 2
[12:06 PM PT]  Ubit Umarov: see you next week if not before
[12:06 PM PT]  Ubit Umarov: bye...
[12:06 PM PT]  Andrew Hellershanks: We are at the top of the hour and most of you are heading out. I will call this meeting to a close. Thank you all for coming. See you again next week.
[12:06 PM PT]  Gavin.Hird all right take care all
Personal tools
About This Wiki