Chat log from the meeting on 2020-02-18

From OpenSimulator

Jump to: navigation, search

[11:00] Gavin.Hird Hi Andrew
[11:00] Telehub: Welcome back to, Dev Outreach, Andrew Hellershanks.
**During meethings, all chat is logged.**
[11:01] Kayaker Magic: Hi Andrew
[11:02] Selby.Evans hi everone
[11:02] Gavin.Hird Hi Selby
[11:02] Kayaker Magic: Not a big draw yet today...
[11:02] Gavin.Hird nope
[11:03] Andrew Hellershanks: I'll wait a bit and see who else shows up.
[11:03] Andrew Hellershanks: Hello, everyone.  :)
[11:03] Gavin.Hird How did the Kitely upgrade go, Selby?
[11:04] Kayaker Magic: Kitely seems to be working well on
[11:04] Gavin.Hird great
[11:05] Andrew Hellershanks: Hello, George
[11:05] Gavin.Hird I hopped over there Saturday morning my time, but was not sure if it was quite done yet at the time
[11:05] Selby.Evans Kitely did an update last night to fix a few problems -- but the upgrade is working fine
[11:05] Gavin.Hird Hi George
[11:05] George Equus: Evening everyone
[11:05] Kayaker Magic: Hi George
[11:05] Selby.Evans Hi George
[11:05] Gavin.Hird I'll check it out again
[11:05] Ubit Umarov: hi all
[11:06] George Equus: Was held up by a full sim mesh landscape refusing to be solid... lol
[11:06] Gavin.Hird Evening Ubit
[11:06] Kayaker Magic: They are having minor issues like teleport hubs not landing you in the right place and group messages not going out
[11:06] Andrew Hellershanks: Hello, Ubit.
[11:06] Selby.Evans The original update was done last Wed. Hi Ubit
[11:06] George Equus: Hi Ubit
[11:06] Ubit Umarov: talking about kitely?
[11:06] Kayaker Magic: Yes
[11:06] Selby.Evans Yes, Ubit
[11:07] Gavin.Hird I was busy building viewer at that time so did not pay attention :-)
[11:07] Ubit Umarov: well had no word from them
[11:07] Gavin.Hird New macOS and Windows versions posted for download yesterday
[11:08] Selby.Evans They are still doing minor fixes, Ubit
[11:08] Ubit Umarov: sure, nice to have news from you folks, but would be nicer to have from them
[11:08] Gavin.Hird yeah
[11:08] Ubit Umarov: im not that expensive ;)
[11:09] Kayaker Magic: I have a question about llGiveInventoryList, which is FUBAR in SL, but I see a note to myself that it could deliver to absent avatars in 0.9.0 but testing in tells me that it cannot deliver to absent avatars any more. Could it ever?
[11:09] Selby.Evans Probably not much talk from them until after March 1. Big customer starting up then
[11:09] Ubit Umarov: i made some changed on that of close
[11:09] Ubit Umarov: but not usre
[11:09] Ubit Umarov: sure
[11:10] Ubit Umarov: how it handles offline at thing point
[11:10] Ubit Umarov: still not happy with that inventory sharing code.. thing we do have a lot wrong
[11:10] Ubit Umarov: at least one big dif
[11:11] Kayaker Magic: Just tried llGiveInventoryList last night on 0.9.2 and and in both systems it gives an ANGRY RED DEBUG WINDOW when trying to deliver to an avatar not present.
[11:11] Ubit Umarov: as soon a item is given, it is placed on target inventory
[11:11] Ubit Umarov: and it is seen
[11:11] Ubit Umarov: a refuse to accept, moves it to trash folder
[11:12] Ubit Umarov: last part is as sl
[11:12] Ubit Umarov: but at SL we do not see it placed on inventory, directly
[11:12] Ubit Umarov: guess they place it on a temporay hidden folder
[11:12] Gavin.Hird I think that was changed in SL say 4-6 months ago
[11:12] Gavin.Hird there was a bit of discussion about it in a TPV meeting
[11:13] Ubit Umarov: temporay, or special
[11:13] Ubit Umarov: oh?
[11:13] Gavin.Hird so there was viewer changes because of changed server behavior in SL
[11:13] Ubit Umarov: yeack
[11:13] Ubit Umarov: good all this is udp still
[11:13] Gavin.Hird I'll see if I can find it somewhere
[11:13] Ubit Umarov: i hate the aisv3 crap
[11:14] Gavin.Hird I think Ansariel mada bunch of comments on the changes
[11:14] Ubit Umarov: well our way is not bad
[11:14] Ubit Umarov: just strange to see it there before pressing accept
[11:14] Gavin.Hird the point was to spare the user from seeing spam inventory
[11:14] Gavin.Hird thayt did not want
[11:14] Ubit Umarov: well from what i see, guess they have a special hidden forlder
[11:15] Ubit Umarov: folder
[11:15] Ubit Umarov: because the fact on deny the thing shows up on trash, means it was on inventory
[11:15] Gavin.Hird could be
[11:16] Ubit Umarov: and also safer for offline
[11:16] Ubit Umarov: just confusing.. always accepted :)
[11:17] Ubit Umarov: well in some cases FS is issue accept message automaticly anyway
[11:17] Gavin.Hird they might have some incoming items queue that is processed on login
[11:17] Ubit Umarov: yeah more breaks for us
[11:18] Ubit Umarov: well a few days ago i notice objects items give was not good
[11:18] Ubit Umarov: and did a few changes
[11:18] Gavin.Hird I'll see if I can track down the viewer changes later
[11:18] Ubit Umarov: we had a mess on messages processing
[11:18] Ubit Umarov: and LL does odd things
[11:18] Ubit Umarov: accept message also means deny
[11:18] Ubit Umarov: in object items case
[11:19] Ubit Umarov: because the message is basicly a folder move
[11:20] Ubit Umarov: accept is moving to correct folder ( we have it there already)
[11:20] Ubit Umarov: deny is move to trash
[11:20] Ubit Umarov: on people inventory to ppl, they do it other way
[11:20] Ubit Umarov: confusing
[11:21] Ubit Umarov: u need to test 0.92 kayaker
[11:21] Ubit Umarov: to see if i improved or made it worse
[11:21] Ubit Umarov: :)
[11:21] Kayaker Magic: Test what? I tested llGiveInventoryList on 0.9.2 and it FAILS NOISILY when giving to an avatar who is not present.
[11:22] Ubit Umarov: well not sure that is suposed to work on that case
[11:22] Gavin.Hird there is this change that involves it: SL-11454 Discarding should accept items to trash
[11:23] Kayaker Magic: Yeah, it is supposed to fail in SL, my memory is that it failed in 0.8 but started working here in, now it is failing again.
[11:23] Ubit Umarov: Gives inventory items to target, creating a new folder to put them in.

• key

group, avatar or prim UUID that is in the same region
[11:23] Ubit Umarov: it is not suposed to do offline
[11:23] Ubit Umarov: yeack group ?
[11:23] Ubit Umarov: Depending upon the type of target this function works differently.
Must be in or able to see into the same Region as the sending object, or recently have been so
[11:24] Kayaker Magic: Yeah, that is the known SL faulure. I thought we started doing better at one point.
[11:24] Gavin.Hird
[11:25] Ubit Umarov: it is spec
[11:25] Gavin.Hird here is the old HG link to it
[11:25] Ubit Umarov: not fail
[11:25] Gavin.Hird searches in the new LL git is a complete mess
[11:25] Ubit Umarov: guess our restriction is to be on same region
[11:25] Ubit Umarov: ( at most on a region on same instance =
[11:28] Kayaker Magic: The workaround is to call llKey2Name with the avatar's UUID. If they are nearby that will work, and then it is safe to call llGiveInventoryList.
[11:28] Ubit Umarov: well seems we do try to send to inventory even if not present
[11:29] Object: Script running
[11:31] Kayaker Magic: Interesting, I just ran my give inventory list test, and it did not generate a debug message here!
[11:32] Kayaker Magic: I could swear it did on my 0.9.2 regions in OSGrid, and I just upgraded yesterday!
[11:32] Andrew Hellershanks: :)
[11:32] Ubit Umarov: did u update yr 0.92. regions ?
[11:32] Ubit Umarov: from code i see no limitations about where the target avatar is
[11:34] Ubit Umarov: ( remember that after a commit, you need to clean and runprebuild)
[11:34] Ubit Umarov: wel or clone into a new folder
[11:34] Ubit Umarov: there where a change that needs that
[11:34] Ubit Umarov: ( and clean before runprebuild )
[11:35] Gavin.Hird You might want to check this commit too
[11:35] Andrew Hellershanks: The LSL wiki page says the target avatar must be in or able to see into the same Region as the sending object.
[11:35] Ubit Umarov: yeap it does
[11:35] Ubit Umarov: we are not checking that
[11:35] Ubit Umarov: ( and will not check by that =
[11:35] Kayaker Magic: LL claimed that the architecture of the server prevented them from ever accepting inventory lists.
[11:36] Ubit Umarov: well ok.. that means there is a child presence on region
[11:36] Ubit Umarov: we can check that :)
[11:36] Gavin.Hird SL-11430: Transaction ID is passed as aux_id from offline message. No special parsing required to get type from binary_bucket.
[11:36] Kayaker Magic: My test worked here when the avatar was not logged on, she logged on and found the inventory.
[11:36] Ubit Umarov: well we do not have aux:id outch :p
[11:37] Kayaker Magic: Then I sent another copy, she got notification and another folder. It is working perfectly!
[11:37] Ubit Umarov: cool
[11:38] Gavin.Hird Hi Bill!
[11:38] George Equus: Hi Bill
[11:38] Bill Blight: Howdy
[11:39] Selby.Evans Hi Bill
[11:39] Bill Blight: Hiya Selby
[11:40] Ubit Umarov: hi bill
[11:40] Ubit Umarov: welcome to your region
[11:40] Ubit Umarov: lol
[11:40] Bill Blight: Haha
[11:40] Andrew Hellershanks: Hello, Bill
[11:41] Ubit Umarov: yeack always wrong stream :p
[11:41] Gavin.Hird So how is the old ticker doing?
[11:42] Ubit Umarov: your region My stream!!! bill
[11:42] Ubit Umarov: well while i have one :)
[11:42] Bill Blight: lol
[11:43] Ubit Umarov: so more news andrew ?
[11:44] Ubit Umarov: yesterady or so i change mysql region store DB
[11:44] Ubit Umarov: replaced storing of floats on doubles, to float
[11:44] Andrew Hellershanks: We have already covered the news of Kitely's upgrade to 0.9
[11:45] Ubit Umarov: saving some useless space on the rows
[11:45] Ubit Umarov: seems the prim tables is already too large for the engine used
[11:45] Andrew Hellershanks: You also added a defaults file for the osslEnable settings.
[11:45] Ubit Umarov: unless you set it to dynamic
[11:45] Ubit Umarov: not nice
[11:46] Andrew Hellershanks: I still don't like the filename chosen but there it is. :)
[11:46] Ubit Umarov: yeah dan asked me that
[11:46] Gavin.Hird how is it too large, Ubit?
[11:46] Ubit Umarov: max row size is 8KB on default
[11:47] Gavin.Hird ok
[11:47] Kayaker Magic: Oh, when I updated my opensim yesterday, I noticed that the osslEnable.ini has been changed, but there was still a copy of OpenSim.ini that referenced it by the old name.
[11:47] Gavin.Hird does it not create additional storage?
[11:47] Ubit Umarov: and we have tons of columns on that
[11:47] Ubit Umarov: 2 or so full text
[11:48] Ubit Umarov: a ton of useless 36chars UUIDS
[11:48] Ubit Umarov: etc
[11:48] Kayaker Magic: OpenSim.ini.example was correct, but there was a copy of OpenSim.ini in the bin folder that still referenced osslEnable.ini directly.
[11:48] Andrew Hellershanks: IIRC, that would still be correct. The new file is osslEnable.ini.example
[11:48] Ubit Umarov: yes i can't change that
[11:49] Ubit Umarov: but on old instalations, there will also be a osslEnable with all entries
[11:49] Ubit Umarov: so minor
[11:49] Ubit Umarov: for now, at least
[11:50] Andrew Hellershanks: Right. Upgrades won't clobber the existing ini file.
[11:50] Ubit Umarov: that thing will always ne a pain no matter what :(
[11:51] Ubit Umarov: about mysql ( and others) no idea why ppl did decide to add columns to all the crap
[11:51] Ubit Umarov: it is a total wast
[11:52] Ubit Umarov: prims are only retrieved as full object group (linkset) and stored as such
[11:52] Ubit Umarov: all those columns are just waste.. it should be a simple blob
[11:52] Ubit Umarov: with a few index ones with uuids
[11:53] Ubit Umarov: no ?
[11:53] Ubit Umarov: ( and on all dbs )
[11:53] Ubit Umarov: in fact SQLITE should be enough for that table
[11:54] Ubit Umarov: fear our code is making sqlite worse than it is
[11:54] Ubit Umarov: well managed/unmaged does make it worse..
[11:55] Ubit Umarov: but the thing should handle all the prims and all region data
[11:55] Ubit Umarov: with no need for mysql/pg on regions
[11:55] Gavin.Hird probably a separate project is needed to g throuhg data storage
[11:55] Andrew Hellershanks: 96 fields in the prims table does seem a bit much.
[11:55] Gavin.Hird I don't have that on regions at all
[11:55] Ubit Umarov: 100
[11:56] Andrew Hellershanks: oh. My copy is out of date.
[11:56] Ubit Umarov: plus the primshapes
[11:56] Andrew Hellershanks: RIght.
[11:56] Ubit Umarov: primshapes are other table.. but all that is part of a prim
[11:56] Gavin.Hird no db is running on the regions
[11:56] Ubit Umarov: and unique to a prim
[11:56] Ubit Umarov: regions need a db
[11:56] Ubit Umarov: for region storage
[11:57] Gavin.Hird not in grid mode. they need a grid db
[11:57] Ubit Umarov: that is silly
[11:57] Ubit Umarov: regions should have local db
[11:57] Gavin.Hird for what?
[11:57] Ubit Umarov: well central one works
[11:57] Gavin.Hird exactly
[11:57] Ubit Umarov: but slower
[11:58] Ubit Umarov: for all region specific data
[11:58] Gavin.Hird having region dbs makes for a cf and too much complexity
[11:58] Ubit Umarov: prims, terrain, etc etc
[11:58] Ubit Umarov: all that info is local to a region
[11:59] Gavin.Hird sure, store it on a shared db instance
[11:59] Ubit Umarov: sure central does easy backups etc
[11:59] Gavin.Hird exactly
[11:59] Gavin.Hird only one db to manage and update
[11:59] Ubit Umarov: otherwise sqlite should be a better option
[11:59] Gavin.Hird as in software update
[12:00] Ubit Umarov: htey do claim more performance than we see
[12:00] Ubit Umarov: and of course.. standalones
[12:00] Ubit Umarov: and regions on osgrid and metro
[12:00] Gavin.Hird for open conect grids it is a bit different
[12:01] Ubit Umarov: only "closed" can use central regions databases
[12:01] Gavin.Hird but even there, if you have many regions (simulators) you can store it in a shared db
[12:01] Ubit Umarov: assuming you ever do backup it :p
[12:01] Andrew Hellershanks: :)
[12:01] Gavin.Hird that is a always a good thing :-))
[12:01] Ubit Umarov: backup is the only reason to use a shared db
[12:02] Ubit Umarov: well that i can remember :)
[12:03] Ubit Umarov: my linux regions use local maria
[12:03] Ubit Umarov: win use local mysql
[12:03] Andrew Hellershanks: An alternative is replication of the region dbs to the central server to keep a single place to do backups. Perhaps not the best way to handle it.
[12:03] Gavin.Hird sounds like over-engineering
[12:03] Ubit Umarov: dunno
[12:03] Ubit Umarov: but remote connections do cost a lot
[12:04] Gavin.Hird invest a little bit more in the shared db instance - cpu, memory, storage, backup, network
[12:04] Gavin.Hird then forget about rgion databases in grid mode
[12:04] Ubit Umarov: but our bad dbs is something ppl mentioned a lot of times
[12:05] Ubit Umarov: inwordz even moved things to cassandra or other engines
[12:05] Gavin.Hird they don't get better by being replicated for every region
[12:05] Ubit Umarov: they are not replicated
[12:05] Andrew Hellershanks: Most of the DB tables don't seem that bad until you look at groups. I've always felt that one was rather bad.
[12:06] Ubit Umarov: what part of information local to a region you missed? :)
[12:06] Gavin.Hird by having one instance for each region - did that make it more understndable? ;-)
[12:06] Ubit Umarov: well prims and primshapes yeack
[12:06] Ubit Umarov: look to that
[12:06] Ubit Umarov: robust tables are another thing
[12:07] Andrew Hellershanks: yea, it had been a while since I looked at prims. I didn't recall it have anywhere near the number of fields as it has now.
[12:07] Ubit Umarov: bc the tradicion of telling people to read the db directly
[12:07] Ubit Umarov: with php etc
[12:07] Gavin.Hird prims are usually cached on the region and for most regions don't change a lot over time
[12:07] Ubit Umarov: all grid tools do direct db access
[12:07] Gavin.Hird so after the initial fetch, it has to go to the central db relatively infrequent
[12:08] Ubit Umarov: physics prims write back a lot
[12:08] Andrew Hellershanks: I just noticed the prims table has three blob type fields.
[12:08] Andrew Hellershanks: That doesn't seem good.
[12:09] Ubit Umarov: good they are bloobs :)
[12:09] Andrew Hellershanks: yes, but you were talking about the limit to row size data earlier. Won't that make it more likely to exceed the default row size limit?
[12:10] Gavin.Hird it is probabaly stored separately
[12:10] Gavin.Hird not sure how mysql does it, but postgres packs the blob data into the default 8k block if it can fit
[12:11] Gavin.Hird if not, it goes to a separate table space for blobs
[12:11] Andrew Hellershanks: What happens if it doesn't fit?
[12:11] Andrew Hellershanks: ok
[12:11] Andrew Hellershanks: We are now past the top of the hour? Any other questions/comments for today before we wrap up?
[12:12] Andrew Hellershanks: I don't see anyone typing so that will do it for today. Thank you all for coming. See you again next week.
[12:13] Gavin.Hird As I said there are new viewer builds at
[12:13] Gavin.Hird the mesh uploader can handle much more complex models without choking
[12:13] Ubit Umarov: bloods are stored elsewhere
[12:13] Gavin.Hird and you can seletively autoplay media for a location - not just on off
[12:14] Ubit Umarov: ( thats why some say they are bad )
[12:14] Ubit Umarov: yeahh guess mysql packs them on the row, if fits, else, elsewhere
[12:15] George Equus: Thanks Andrew. I am off. bye all.
[12:15] Ubit Umarov: but they do not count for the 8k limit

Personal tools
About This Wiki