Chat log from the meeting on 2017-10-03

From OpenSimulator

Jump to: navigation, search

[10:57] Arielle.Popstar just looked at the git changes
[10:57] Arielle.Popstar or lack of them]\
[10:58] Sheera Khan: I#d say it's rather quiet at the moment
[10:58] Arielle.Popstar nod
[10:59] Sheera Khan: In a way I take it as a positive sign...
[11:08] Andrew.Hellershanks I realized i missed last weeks meeting. There was some stuff happening at my house that needed my attention. It wasn't until well after the meeting was over that I remembered it was that time of week.
[11:10] Simulator Version v0.5 ruft: Rev Paela OpenSim 4.0.1 Dev (Unix/Mono)
[11:12] Andrew.Hellershanks What new and exciting things have you been up to in the world of OpenSim this week?
[11:12] Leighton.Marjoram trying to get SSL encrypted logins enabled have had varying degrees of success.
[11:13] Leighton.Marjoram Doesnt seem to play well with Jopensim ...
[11:14] Leighton.Marjoram its keeping me busy and turning me greyer than I already am
[11:14] Andrew.Hellershanks Leighton, sounds good. Have you needed to make many changes to the code?
[11:15] Andrew.Hellershanks Some code features will do that to a person. I was having some trouble when I was trying to replace some mail related code.
[11:15] Leighton.Marjoram Apart from the network config block nothing code wise. I am using a branch from Bill Blights github.
[11:15] Sheera Khan: Metro has switched to SSL some time ago and we had some issues with it too, but it got sorted out
[11:16] Leighton.Marjoram Nods @ Sheera
[11:16] Andrew.Hellershanks IIRC, Avination had SSL based logins.
[11:16] Sheera Khan: but I guess we don't use Jopensim
[11:17] Leighton.Marjoram
[11:17] Leighton.Marjoram I hope I can figure it out it is the last piece of the puzzle for virtual world counselling
[11:18] Andrew.Hellershanks osCore2 makes me wonder what happened to osCore1.
[11:18] Leighton.Marjoram lol
[11:19] Leighton.Marjoram this is the closest I have been to a complete solution that could be used for professional practice.
[11:20] Andrew.Hellershanks Good luck with that project. I'm sure a lot of grids would switch to SSL based logins if (when) it becomes available.
[11:20] Leighton.Marjoram thank you Andrew
[11:21] Andrew.Hellershanks I haven't been working on anything OpenSim related recently. I've been too busy working on a personal project.
[11:22] Sheera Khan: OS development is a bit slow atm - but maybe that's a good sign ...
[11:22] Leighton.Marjoram It would be interesting to know how many people have upgraded to 0.9
[11:23] Leighton.Marjoram I thought there would be more noise than there had been since its release
[11:23] Sheera Khan: meaning OS0.9 is stable and not to much errors had to get fixed immediately
[11:23] Andrew.Hellershanks yes, it has been slow. I see Gavin recently provided an updated libBulletSim library for use with MacOS.
[11:24] Gavin.Hird yes
[11:24] Gavin.Hird it works ok, but I wish I had some raycast scriupts to test it with
[11:25] George Equus: I still have one script refusing to run (on ubODE) in 0.9 but that is all. as far as I have seen.
[11:25] Andrew.Hellershanks Gavin, talk to Kayaker. He has talked about raycast issues in the past. He may have some scripts you could use to run some tests.
[11:26] Gavin.Hird will do
[11:26] Andrew.Hellershanks George, refuses to run?
[11:27] Andrew.Hellershanks George, Have you filed a mantis about it or talked with Ubit?
[11:28] George Equus: It's a bot patrolling my region, it just sit there now, twitching occasionally, don't do it's round. Thought would simply hand over the script and info on sim version it stopped.
[11:28] George Equus: Mantis is not someting I use, don't have the proper tech knowledge for that
[11:28] Andrew.Hellershanks You should definitely talk to Ubit about it.
[11:29] Andrew.Hellershanks What method do you use to make the bot move about the sim?
[11:30] George Equus: I see now that actually did sent a notecard to Dan with all facts and the script on Sept. 16  :)
[11:31] George Equus: My knowledge of scripting is virtually non existant so can't describe it technically.
[11:32] Andrew.Hellershanks George, I'm going to guess that you didn't write the script. Where did you get it?
[11:34] George Equus: There Andrew...Locked.Semaphore seem to have vanished, been looking for him
[11:35] Andrew.Hellershanks I'll have a more detailed look at it later. I see it uses keyframe motion so I'm surprised if that stopped working.
[11:35] George Equus: I did manage to change co ordinates to fit my environment, as for the code itself, I am clue less.
[11:36] George Equus: Some change in the sim code at a specific version made it stop so should probably be fairly easy to pin point what changes was made and might have triggered this.
[11:37] George Equus: Easy for a programmer to see I imagine
[11:37] Andrew.Hellershanks nods
[11:38] George Equus: osgrid-opensim-07092017.v9.0.1.cc5d6f6
In this and previous version the script work as should.
[11:38] George Equus: next one it stopped
[11:38] George Equus: osgrid-opensim-07232017.v0.9.1.6bf96f8
[11:39] George Equus: and from then on
[11:40] Bill.Blight George the copy of the script you gave me runs fine on my grid, and I am pretty much on the latest master
[11:40] George Equus: Not urgent for me, but should probably be fixed, not sure how many use that script but might be functions in similar that also produce errors
[11:41] Andrew.Hellershanks George, That will make it easy for someone to dig in to the cause.
[11:41] George Equus: Yes Bill, and it was tested here as well, on Project secret
[11:42] Bill.Blight Have you manually cleared your scriptengines cache?
[11:42] Bill.Blight that might help
[11:42] George Equus: How do I do that?
[11:43] George Equus: I run my simulator without any deep knowledge, as is out of the box
[11:43] Ubit Umarov: just delete the folder
[11:43] Bill.Blight If you don't have many scripts that you care if they get reset, just delete the whole scriptengines directory
[11:43] Ubit Umarov: yeap... state is lost doing that
[11:43] George Equus: Turn off sim and then restart after I take it?
[11:44] Bill.Blight shutdown the sim, delete the scriptengines folder, restart, will rebuild all the scripts and their states
[11:44] George Equus: I got loads of scripts lol
[11:44] Andrew.Hellershanks George, I just checked the two hash codes. There is two weeks between those two commits. It will still help someone narrow down where the change occured that stoped the script from working.
[11:44] George Equus: might get me busy...
[11:44] Ubit Umarov: but that as no relation to changes btw bullet and ubOde, only version updates
[11:45] Bill.Blight if you have not recompiled them and or cleared that cache, it very well could be a compilation error, as at some point in there when KFM was being fixed, some constants may have changed.
[11:45] George Equus: OK. thanks for taking interest. I might test deleting 5that cache...
[11:45] Ubit Umarov: its a large script ?
[11:45] George Equus: Have re compiled
[11:46] George Equus: no change
[11:46] George Equus: I'll send it to you Ubit, not that large
[11:46] Ubit Umarov: ok
[11:46] Andrew.Hellershanks You can sort by file name then delete all files that start with the name CommonCompiler. That will preserve the script state files.
[11:46] Ubit Umarov: ill take a look hmm one of this days
[11:46] Ubit Umarov: my test region ubode is half broken now
[11:46] Ubit Umarov: lol
[11:47] Bill.Blight which is fine, as long as the script current state is not what is causing it to not work ..
[11:47] Andrew.Hellershanks Ubit, trying to fix the broken half or break the other half?  ;)
[11:47] Bill.Blight (but a reset of the script should have fixed that)
[11:47] Andrew.Hellershanks True, Bill
[11:47] Ubit Umarov: well decided to clean up the code
[11:48] Ubit Umarov flüstert: of course over cleans :)
[11:48] Ubit Umarov: hmm this script works on just a box george ?
[11:48] Bill.Blight You will find the scriptengines folder by default in your bin folder, unless you moved it for some reason in the configgs
[11:49] Andrew Hellershanks: The other thing I notice about that script is that it appears to restart the KFM when it reaches the end instead of setting KFM to loop.
[11:49] George Equus: Ubit, Yes on a box is fine
[11:50] Bill.Blight Yeah I noticed that too, should still work, but may cause excessive script time
[11:53] Ubit Umarov: KFM_FORWARD] does not loop
[11:54] Gavin.Hird I just got a message on the Robust console. Anyone got a clue what's going on?
[11:54] Gavin.Hird 20:49:33 - [HELO]: hi, HEAD was called
20:49:33 - [BASE HTTP SERVER]: Requested method [link_region] from threw exception: Object reference not set to an instance of an object at OpenSim.Server.Handlers.Hypergrid.HypergridHandlers.LinkRegionRequest (Nwc.XmlRpc.XmlRpcRequest request, System.Net.IPEndPoint remoteClient) [0x00013] in <f25133c2ae4a44559ef22b7770027e79>:0
at OpenSim.Framework.Servers.HttpServer.BaseHttpServer.HandleXmlRpcRequests (OpenSim.Framework.Servers.HttpServer.OSHttpRequest request, OpenSim.Framework.Servers.HttpServer.OSHttpResponse response) [0x002b2] in <45dce7b225674f5e83f5f89a075b6e88>:0
[11:54] Andrew Hellershanks: Gavin, Kayaker is here. He isn't seeing avatar names at the moment. If you IM him he can pass you some scripts to test Raycasting.
[11:54] Ubit Umarov: mb just failed comms
[11:54] Kayaker.Magic You cannot use KFM_LOOP when you change the distance nd time every timer event.
[11:55] Andrew Hellershanks: Right. If that is what is being done in the script I hadn't noticed it.
[11:57] Andrew Hellershanks: I have someone seeing some odd messages in an 0.8.2 based grid. Any ideas about the following?
[11:57] Andrew Hellershanks: OpenSim.Framework.Util Run threadfunc 374 (Queued 6, Running 290)
OpenSim.Framework.Util Exit threadfunc 339 (63 ms)
[11:57] Ubit Umarov: about timers, you know opensim defaults have very low resolution... you may need to change one or two settings to improve it
[11:57] Andrew Hellershanks: Is that something to do with lack of threads on the Linux based machine using mono?
[11:58] Andrew Hellershanks: They see a lot of those types of messages on the console.
[11:58] Ubit Umarov shouts: Andrew we know nothing about 0.8 now
[11:58] Kayaker.Magic The default minimum timer used to be 0.5 seconds. I think I noticed that was decreased to 0.2 in the latest OSGrid distribution.
[12:00] Gavin.Hird Looks like they run out of threads Andrew
[12:00] Andrew Hellershanks: I still see the ini file saying MinTimerInterval default is 0.5
[12:00] Andrew Hellershanks: Ubit,, :)
[12:01] Andrew Hellershanks: Gavin, that is what I'm thinking. I'll get them to dig a bit deeper in to their setup.
[12:01] Ubit Umarov: well at least i don't.. never used/coded 0.8
[12:01] Gavin.Hird the threadpool limits is the first place to look if they run smarthreadpool
[12:02] Andrew Hellershanks: That would be a good place to start the investigation. I'd forgotten about that setting.
[12:03] Andrew Hellershanks: I just noticed we are at the top of the hour. Any last comments/questions/thoughts for today?
[12:03] Ubit Umarov: timers? think there are 2 related to them
[12:04] Andrew Hellershanks: Gavin, I just looked at the OpenSim.ini file and see there is a min/max threads setting. I'll have them check the value they are using and if needed, bump it up.
[12:04] Ubit Umarov: that message may also be a thread timeout... no idea of what was that timeout on 0.8
[12:05] Ubit Umarov: i mean a workjob timeout
[12:05] Andrew Hellershanks nods
[12:05] Bill.Blight You may also have to set the MONO_THREADS_PER_CPU=
[12:05] Ubit Umarov: hm workitem
[12:05] Bill.Blight environment variable
[12:05] Ubit Umarov: nahh don't mess with that
[12:05] Andrew Hellershanks: Bill, AFAIK that is already set but I'll double check.
[12:05] Bill.Blight If it is default , you surely need to mess with it .. LOL
[12:05] Bill.Blight :P
[12:06] Ubit Umarov: no you don't
[12:06] Andrew Hellershanks: When it says that some threads are queued it does sound like it has run out of available threads.
[12:06] Gavin.Hird MONO_THREADS_PER_CPU= has more to do with concurrency of the threads than actual available slots
[12:06] Bill.Blight kicks ubit in the shins
[12:06] Bill.Blight yes I know that
[12:06] Bill.Blight LOL
[12:06] Ubit Umarov: it does not run out of threads
[12:06] Ubit Umarov: 0.8 made a mess on that
[12:06] Andrew Hellershanks: Bill: now, now. Play nice. :)
[12:06] Bill.Blight I am referencing an old arguement on IRC about that
[12:06] Bill.Blight LOL
[12:06] Ubit Umarov: one thing is threads.. other is workitems
[12:07] Ubit Umarov: and the code does a mess btw both
[12:07] Ubit Umarov: you can run with only one thread allowed
[12:07] Ubit Umarov: in theory at least
[12:07] Andrew Hellershanks: ok. I wonder about running out of threads when there are settings in OS.ini that appear to be used to set limits on the number of threads.
[12:07] Ubit Umarov: ( not nice... region will freeze :p )
[12:08] Bill.Blight yes I know, like I said, I was just being cheeky, about an old argument on IRC about that setting
[12:08] Sheera.Khan Those environment variables can be set differently in different linux distributions. Is there a wiki page on how to set them and what values to use?
[12:08] Ubit Umarov: keep them as linux defaults
[12:08] Gavin.Hird when I ran smartpool I definitely had crashes because the simulator ran out of threads
[12:08] Sheera.Khan different distros use different defaults...
[12:09] Gavin.Hird now I run just the thread setting
[12:09] Ubit Umarov: number of threads is relevant because of things like http and scripts
[12:09] Sheera.Khan or at least may do so
[12:09] Ubit Umarov: bc how code is
[12:09] Andrew Hellershanks: Bill, I just checked. MONO_THREADS_PER_CPU is set on the persons machine.
[12:09] Ubit Umarov: both scripts and http connections hold a thread
[12:10] Simulator Version v0.5 ruft: Rev Paela OpenSim 4.0.1 Dev (Unix/Mono)
[12:10] Ubit Umarov: so if there are a lot of them, taking a lot of time, and we can't add more threads, region may stall until they get released
[12:10] Ubit Umarov: but scripts have their own set of threads
[12:11] Ubit Umarov: ( that is the llSleep issue for example.. it holds a thread )
[12:11] Kayaker.Magic And I have seen bad scripts hold down threads (with llSleep) and bog down the rest of the scripts in a region.
[12:11] Andrew Hellershanks: llSleep is best avoided when possible.
[12:11] Gavin.Hird ODE used a large number of threads when generating physics
[12:11] Ubit Umarov: yeap.. that is still a issue
[12:12] Bill.Blight llSleep, the root of all evil
[12:12] Bill.Blight :P
[12:12] Ubit Umarov: ode is single thread
[12:12] Sheera.Khan hmm, per default the variable MONO_THREADS_PER_CPU is not set... In my startup script I have set "export MONO_THREADS_PER_CPU=2048"
[12:12] Ubit Umarov: runs on hearbeat
[12:12] Bill.Blight that sets it for that instance
[12:12] Andrew Hellershanks: That is the value I've been using on my own machine.
[12:12] Ubit Umarov: export MONO_THREADS_PER_CPU=2048" this has hidden costs...
[12:13] Gavin.Hird if you gave it free for all ti allocated about 1900 mono threads and ran mono at 770% CPU on an i7
[12:13] Bill.Blight just depends on how many cpu cores you have lower values are fine for a system with a lot of cores, values that high can rob other processes of threads
[12:13] Andrew Hellershanks: Gavin, ouch
[12:13] Bill.Blight umm rob them of cpu ticks not threads
[12:13] Gavin.Hird it got the job done fast . haha
[12:14] Ubit Umarov: 770% ??
[12:14] Gavin.Hird yes
[12:14] Ubit Umarov: guess someone is beeing fooled :P
[12:14] Andrew Hellershanks: Ubit, what sort of costs? AFAIK, that is still a suggested setting and value for people using mono to run OS.
[12:14] Gavin.Hird 7 cores running 100% mono + one 70%
[12:14] Gavin.Hird the rest was used by the kernel
[12:15] Andrew Hellershanks nods to Ubit
[12:15] Gavin.Hird bye George
[12:15] Andrew Hellershanks: Any other last minutes items before we lose many other people?
[12:15] Ubit Umarov: hidden bc they don't even show up on debug tools in some cases
[12:15] Andrew Hellershanks: See you next time, Bill
[12:16] Ubit Umarov: on win not even in some memory uses
[12:16] Andrew Hellershanks: Ubit: hidden indeed.
[12:16] Ubit Umarov: they are 2 deep for the tools to see
[12:17] Ubit Umarov: but ode does not use threads...
[12:17] Ubit Umarov: as i said, it runs on heartbeat
[12:17] Andrew Hellershanks: right
[12:17] Ubit Umarov: ( it does massive use of stack )
[12:17] Gavin.Hird it sure use threads when generating physics
[12:17] Ubit Umarov: no it does not
[12:18] Ubit Umarov: ubmeshmerizer uses one or two extra threads only
[12:18] Gavin.Hird this was before
[12:18] Gavin.Hird with the old ODE
[12:18] Ubit Umarov: ode was never multithread
[12:18] Ubit Umarov: it did crash
[12:19] Ubit Umarov: new version could be.. but i disable it on compile :p
[12:19] Ubit Umarov: bc i don't like that fine grained multithread on a server
[12:20] Ubit Umarov: and more threads == slower ;p
[12:20] Gavin.Hird depends on how the operating system handle it
[12:20] Ubit Umarov: no
[12:20] Ubit Umarov: depends on the application structure and data
[12:20] Gavin.Hird
[12:21] Gavin.Hird yes it does Ubit
[12:21] Ubit Umarov: forgot the name of the law again
[12:21] Ubit Umarov: sure it also depends on the OS...
[12:22] Ubit Umarov: in some cases .net can take ms to create a new thread
[12:22] Gavin.Hird if you set mono to run with Threads rather than Smartthreadpool on macOS, mono use GCD to handle and schedule them
[12:23] Gavin.Hird and it is very efficient
[12:23] Ubit Umarov: yeap but threads is not nice
[12:23] Ubit Umarov: doesn't not have the flexibility smartpool has
[12:24] Ubit Umarov: i mean the .net main Thread pool
[12:25] Ubit Umarov: well opensim threading is still a issue
[12:25] Ubit Umarov: i did reduce a bit of the abuse on 0.9.. but still a lot
[12:25] Ubit Umarov: for example
[12:26] Ubit Umarov: selecting 300 prims on 0.8 issued 300 threads ( if allowed )
[12:26] Andrew Hellershanks: eek
[12:26] Ubit Umarov: that is clear a abuse
[12:27] Andrew Hellershanks: Almost half past now. I need to get going. Good meeting. Thank you all for coming. See you again next week.

Personal tools
About This Wiki