Chat log from the meeting on 2021-11-02

From OpenSimulator

Jump to: navigation, search
[11:03] Andrew Hellershanks: Hm... the logging cube isn't working today.
[11:03] Andrew Hellershanks: Hello, everyone.
[11:03] Ubit Umarov: opos di di kill it?
[11:03] Andrew Hellershanks: Ubit, I don't know. What did you do to it?
[11:03] Gavin.Hird poor thing, hehe
[11:04] Ubit Umarov: strange the notecard is empty
[11:04] Andrew Hellershanks: The script in it is just called "New script"??
[11:04] Ubit Umarov: ofc
[11:05] Andrew Hellershanks: :)
[11:05] Ubit Umarov: is there other name for scripts?
[11:06] Ubit Umarov: oops seems i did
[11:06] Andrew Hellershanks: Not much to report in the way of news of OpenSim changes this past week.
[11:06] Ubit Umarov: some issue on timers grrr
[11:07] Gavin.Hird did you fix the timers?
[11:07] Andrew Hellershanks: Timer issues?
[11:07] Ubit Umarov: just changed a bit
[11:07] Ubit Umarov: well guess you need to do the log the hard way this time :(
[11:08] Andrew Hellershanks: Not a problem as long as the region doesn't crash. Even if it does I will save the chat log before I terminate my viewer.
[11:08] Ubit Umarov: ok
[11:09] Ubit Umarov: the logs machine is also dead
[11:09] Ubit Umarov: hmm or not
[11:10] Vincent.Sylvester I managed to clear out 370 mantis issues overall, with a good 100 just in the past few weeks
[11:10] Vincent.Sylvester Confirmed a few of as well
[11:11] Ubit Umarov: eckk the code seems fine..  dang
[11:12] Andrew Hellershanks: Vincent, sounds good. A lot of "not a bug", or obsolete ones, or ... ?
[11:12] Andrew Hellershanks: Ubit, we may have to wait for Bill to have a look at it.
[11:13] Vincent.Sylvester Been closing ODE related ones since those mostly don't apply anymore, few were bugs that were fixed or patches merged and no one noticed, rest mostly cannot reproduce or things not really fixable or reasonable to implement
[11:13] Andrew Hellershanks: ok, ty.
[11:13] Vincent.Sylvester There are some old ones now confirmed that surprised me, example scripts not saving properly when not running
[11:14] Vincent.Sylvester I thought that had been fixed, but apparently not
[11:15] Vincent.Sylvester I did have a look at one thing after the fact, osGetLinkParams which had a bit sloppy code... well code I would write so has to be bad, currently testing an improvement on that
[11:16] Vincent.Sylvester Something to reduce array access amount on that make it a tiny bit faster
[11:16] Gavin.Hird speaking of ODE and physics, there is a change to the next version of my viewer where items set to 100% transparent will not be rendered, which should make it less costly to use transparent prims as physics
[11:17] Vincent.Sylvester You mentioned that before I thought all viewers did that already
[11:17] Andrew Hellershanks: Surprising that it wasn't already a feature of current viewers, Gavin.
[11:17] Gavin.Hird they render it fully
[11:17] Vincent.Sylvester Makes sense for those that don't know how to re-import their meshes with proper physics
[11:17] Gavin.Hird as in goes through the entire redering pipeline
[11:18] Vincent.Sylvester Maybe while in there can fix the water shaders :x
[11:19] Andrew Hellershanks: Gavin, will it still handle the cases where only part of the object is set at 100% transparency?
[11:19] Gavin.Hird LL does a lot of optimization work on the renderer right now, but it only works on Windows
[11:19] Gavin.Hird on the Mac it reners at half the FPS before the optimizations
[11:19] Gavin.Hird yes
[11:20] Ubit Umarov: if uses the main transparency only no ?
[11:20] Andrew Hellershanks: Good. I can imagine some cases where one side of a prim is transparent but you can still see parts of other sides at the same time.
[11:20] Ubit Umarov: it..
[11:21] Gavin.Hird if the face is set to trans it will not be rendered
[11:21] Ubit Umarov: yeah a simple opt they could had done ages ago
[11:22] Gavin.Hird of course but the cam usually don't see through the front and back at the same time?
[11:22] Ubit Umarov: dang seems my co2 saving commit does have issue
[11:22] Ubit Umarov: code looks fine..  but just got other report of error
[11:23] Gavin.Hird so it is the direction of view that counts for want is selected for rendering
[11:23] Gavin.Hird what even
[11:24] Ubit Umarov: sl ignores back side of prims
[11:24] Ubit Umarov: on render
[11:24] Vincent.Sylvester It would be really cool if we have more nunit tests checking various aspects as it stands they don't really catch very many errors
[11:24] Andrew Hellershanks: I forget about the unit tests in OS. I haven't run them in ages.
[11:25] Vincent.Sylvester They all go through yet I have had some cataclysmic failures in the past when making minor changes
[11:26] Ubit Umarov: unit can only catch a few things
[11:27] Ubit Umarov: very hard to make such tests also
[11:27] Ubit Umarov: and should never be done by same coder
[11:28] Andrew Hellershanks: Indeed. I have written about 100 tests for some code I have written and I've barely scratched the surface of what needs to be tested. I've mostly just done testing for some standalone utility subroutines. Other parts of the system are harder to test.
[11:28] Ubit Umarov: many of our tests are just bad
[11:28] Ubit Umarov: i did remove a few even :P
[11:29] Vincent.Sylvester Realistically probably need ten times what is currently done or even more than that
[11:29] Andrew Hellershanks: I wouldn't be surprised by that count.
[11:29] Vincent.Sylvester Project needs another intern to do the honors lol
[11:29] Ubit Umarov: some just had wrong assumptiosn about what should be done
[11:30] Vincent.Sylvester The funniest is the versioninfo test, because I put build numbers into it and I keep breaking that each time another digit is added
[11:30] Andrew Hellershanks: It shouldn't be too hard to write some tests to check code behaviour for a lot of SOG/SOP operations.
[11:31] Vincent.Sylvester Some presence and teleport operations, few more on linkset stuff, quite a few areas that are really in need
[11:32] Vincent.Sylvester Script engine, but poking around in that is easier said than done
[11:33] Andrew Hellershanks: Tests for database operations, including items such as inventory management.
[11:34] Andrew Hellershanks: Testing DB operations can get tricky. You have DB table creation and migration for several different database engines.
[11:34] Vincent.Sylvester Inventory still has a bug allowing entry of bad data for parentFolderID which breaks inventory, stumbled upon that ticket as well
[11:35] Andrew Hellershanks: Sounds like something a unit test could catch.
[11:36] Vincent.Sylvester Among other things, quite a few of the confirmed tickets I guess translate to a unit test once the underlying issue is resolved
[11:36] Andrew Hellershanks: The number of unit tests goes up quickly when you start writing tests to check for normal/expected behaviour then add tests to see what happens when you pass bad or invalid data.
[11:37] Vincent.Sylvester There are a few of these and I'm sure way more that I haven't found yet
[11:37] Andrew Hellershanks: For the routines I test in my code I sometimes have as many as half a dozen tests being run against a single function.
[11:38] Vincent.Sylvester 8903 is the ticket id of the inventory thing by the way
[11:38] Andrew Hellershanks pulls up the mantis report
[11:38] Andrew Hellershanks: ty, Vincent. I'm monitoring that one now so I can look at it more closely later.
[11:39] Vincent.Sylvester It has a patch, huh never changed the status on it, will do that later, but yeah I did add a little check to prevent the worst part about the parent folder getting broken
[11:39] Vincent.Sylvester Rest I guess is down to forcing lowercase on the keys somewhere
[11:40] Vincent.Sylvester Though that doesn't seem to have a negative impact
[11:40] Gavin.Hird Kirsten viewer.- is that still a thing?
[11:40] Vincent.Sylvester Yes apparently there was an update to it this year
[11:40] Vincent.Sylvester I even got it logged in, kinda
[11:41] Vincent.Sylvester After that inventory fiasco I didn't check further yet, though did leave at note for Kirsten to not forget us
[11:42] Vincent.Sylvester Used to work fine with OS in the past I primarily used it before Firestorm because of the pretty rendering
[11:42] Andrew Hellershanks: I haven't heard of Kirsten viewer before.
[11:42] Gavin.Hird even if it was updated a year ago, it is probably massively out of date
[11:43] Gavin.Hird it primarily focused on addition rendering options as far as I can remember
[11:43] Gavin.Hird it was a thing up to 2014-ish
[11:44] Vincent.Sylvester I have seen some rather pretty regions hence getting more out of the rendering for some nice pictures seemed worth the effort to me
[11:44] Gavin.Hird not if it distroys your inventory
[11:44] Vincent.Sylvester It did login, but crashed on something if I recall, then it broke my inventory, took me a while to find that
[11:45] Vincent.Sylvester Well the patch on OS end should prevent bad data going in
[11:45] Andrew Hellershanks: Should be simple enough to force UUID strings to lower case before it gets sent to the DB code for saving.
[11:45] Vincent.Sylvester It's just the main inventory folder the most parent one, should have nullkey as parent not its own id as parent
[11:47] Vincent.Sylvester Should be a simple first for the viewer too, but I wonder where that came from initially, can you imagine if all SL viewers do this any of them will break inventory
[11:47] Vincent.Sylvester Users trying different viewers isn't unheard of
[11:47] Gavin.Hird all the uuids should be stored with the uuid datatype in the databases
[11:47] Gavin.Hird now it is usually varchar
[11:48] Vincent.Sylvester At least forces 36 chars on all of them now I think
[11:48] Gavin.Hird but of course that is a massive upheaval bopth code side and to migrate existing databases
[11:49] Vincent.Sylvester The case didn't seem to matter much in terms of inventory though it did double up folders because they are not unique, probably need to look into connectors for inventory and force lowercase somewhere in there
[11:49] Andrew Hellershanks: If DB queries are done case insensitive there would be no need to migrate existing DBs to deal with the issue. Hosever, there may be other scripts used by a grid that could be affected by the issue.
[11:49] Gavin.Hird postgres db use uuid as type for a number of tables, but not all
[11:49] Gavin.Hird such as assets, but nit inventoryitems
[11:50] Vincent.Sylvester Adding .ToLower() in the right place is trivial, just need to find the right place heh
[11:50] Gavin.Hird uuid datatyep enforces more than just case
[11:50] Andrew Hellershanks: Gavin, by "uuid as type" do you mean used as an index?
[11:50] Gavin.Hird it also knows about the 3 different uuid formats that exist
[11:51] Gavin.Hird no it is a native datatype of postgres and modern MySL
[11:51] Gavin.Hird MYSQL
[11:51] Ubit Umarov: well that patch has not much relation to the issue
[11:51] Ubit Umarov: may fix some strange case,..
[11:52] Ubit Umarov: and has no relation at all with tolower
[11:52] Vincent.Sylvester It just prevents viewers setting parent and folder id to the same
[11:52] Vincent.Sylvester Like I said, it's just that, the rest I haven't looked into yet
[11:53] Vincent.Sylvester tolower needs to go into inventory somewhere... no idea where yet haven't looked
[11:53] Ubit Umarov: ( fix the timer issues
[11:53] Ubit Umarov: fixed
[11:53] Ubit Umarov: can't commit now :)
[11:53] Gavin.Hird of course you can
[11:54] Gavin.Hird just do it ;-)
[11:54] Andrew Hellershanks: :)
[11:54] Ubit Umarov: bad boy :p
[11:54] Vincent.Sylvester I think I have that patch locally so I'll look into inventory see if I can find the correct place to add the tolower
[11:54] Ubit Umarov: wants to see us all killed :)
[11:54] Vincent.Sylvester See what Kirsten viewer breaks next from there lol
[11:54] Gavin.Hird I suggest you don't use it
[11:54] Andrew Hellershanks: Ubit, you will just be testing the protection against restarting this place during meeting hours. ;)
[11:54] Ubit Umarov: :)
[11:55] Gavin.Hird there is very little point in chasing bugs from unsupported viewers
[11:55] Vincent.Sylvester Wiki think still mentions how you can get SL viewer connected to OpenSim and if Kirsten is based on that others could break things as well
[11:55] Andrew Hellershanks: Gavin, That is true if those bugs are only in the one viewer.
[11:55] Ubit Umarov: see? one tries to reduce CO2 and gets into worse trouble..
[11:55] Vincent.Sylvester Not to mention we should protect against bad data no matter from where
[11:56] Gavin.Hird sure, but a better methiod is to convert to using native uuid data type for such data at the backend
[11:56] Vincent.Sylvester I'm looking at my mariadb install and it doesn't seem to have that type yet
[11:57] Gavin.Hird a lot of the code is already in there, only it is only used to the pgsql section of the code
[11:58] Gavin.Hird
[11:58] Andrew Hellershanks: Ubit, I saw the reference to CO2 in the changelog and meant to ask you what it meant. I don't think it means carbon dioxide. :)
[11:59] Selby.Evans
[11:59] Vincent.Sylvester Ah I am still on 10.6 so that's why
[12:00] Selby.Evans Musr go -- bye all
[12:01] Ubit Umarov: thx selby.Evans
[12:01] Andrew Hellershanks: ok, Selby. See you next week.
[12:01] Andrew Hellershanks: We are at the top of the hour. Any final comments/questions/thoughts for today?
[12:02] Jagga Meredith: CO2?
[12:02] Andrew Hellershanks: That's what I was also asking.
[12:02] Vincent.Sylvester whispers: It
[12:02] Gavin.Hird who cares about CO2
[12:02] Gavin.Hird it is atrace gas of 0.04% of the atmopshere
[12:03] Vincent.Sylvester CO2 is what plants crave
[12:03] Gavin.Hird exactly
[12:04] Gavin.Hird uder 280 ppm plant growth goes down dramatically.  at 150 ppm it stops completely
[12:04] Gavin.Hird they want us down to around 280 again
[12:05] Ubit Umarov: yes co2
[12:05] Ubit Umarov: reducing cpu work saves power, so CO2 :p
[12:05] Vincent.Sylvester Just go kill a cow if the timer thing don't work, CO2 impact likely the same rofl
[12:05] Andrew Hellershanks: If that is what it was really about and not something else its time to wrap up todays meeting.
[12:06] Ubit Umarov: well less opensim cows may also do it yes
[12:06] Gavin.Hird :-)
[12:06] Andrew Hellershanks: Thank you all for coming. See you again next week.
Personal tools
About This Wiki