Chat log from the meeting on 2017-05-09

From OpenSimulator

Jump to: navigation, search

[11:03] Basil Sosides: what is the matter with the Maptiles in OSgrid ????
[11:04] Basil Sosides: they are gone
[11:04] Arielle Popstar: oh thought that was just me Basil
[11:04] Basil Sosides: no maptiles at all, nowhere !!!
[11:04] Arielle Popstar: anyone on IRC Opensim-dev?
[11:05] Selby.Evans See Gavin -- you lost your skin and didn't even notice.
[11:05] Leighton.Marjoram is that the telegraph system of the internet (1 up from carrier pidgeon)
[11:05] Gavin.Hird I just received a texture rebake
[11:05] Arielle Popstar: i see skin and clothes now
[11:05] Leighton.Marjoram you have appeared gavin
[11:05] Selby.Evans I see you normal now
[11:05] Arielle Popstar: 2 tin cans and string :)
[11:05] Sheera Khan: Bongo drums, I hear them ^^
[11:05] Kayaker Magic: +1 Leighton, IRC is not my favorite form of communication
[11:06] Leighton.Marjoram I wish they would use Slack or Discord bring the technology into the modern age.
[11:06] Gavin.Hird I installed the http server update of yesterday and wierd things are happening...
[11:06] Gavin.Hird so will revert later
[11:06] George Equus: Gavin look like a bust standing on the seat, no lover body
[11:07] Leighton.Marjoram you have legs
[11:07] Leighton.Marjoram @ Gavin
[11:07] Kayaker Magic: Now Gavin is all here!
[11:07] George Equus: Yep
[11:07] Leighton.Marjoram I thought it was a new top half only avatar lol
[11:07] Gavin.Hird If I make a selction of sit, then I go into hiding
[11:08] Gavin.Hird like wearing an alpha lower part, haha
[11:08] George Equus: Radar is very slow on getting names... Only Selby and Gavin so far
[11:08] Leighton.Marjoram thats what I thought Gavin and I kinda liked it.
[11:11] Basil Sosides: Andrew, are the maptiles from OSgrid out of order ??
[11:11] James.atLLOUD HI all - wow, I arrived right back where I sat last week.
[11:11] Leighton.Marjoram that reminds me does any one know a viewer to use for xml exports?
[11:11] Arielle Popstar: last location?
[11:11] Leighton.Marjoram Tried Imprudence but no meshes show in the viewer?
[11:11] Andrew.Hellershanks Sorry for being late. It has been a busy day for me.
[11:12] Gavin.Hird Imprudence does not support mesh
[11:12] Arielle Popstar: FS and Sing have different xml formats
[11:12] Selby.Evans I used Firestorm for XML exports from SL
[11:12] Gavin.Hird Kokua is the follow on with mesh support
[11:12] Kayaker Magic: I thought NONE of the viewers would save builds that had mesh in them.
[11:12] Leighton.Marjoram Thanks Gavin
[11:12] James.atLLOUD yeah, I've firestorm for xml export - just avatar shapes.
[11:12] Ubit Umarov: ahh nice white bento avies :)
[11:41] Sheera.Khan the UV maps for the Bento-avatars here are the same as the UV maps for system avatars - the standard skins we use here fit :-) (Except for the wings and the tail of course)
[11:12] Arielle Popstar: they do Kayaker
[11:12] Leighton.Marjoram I am using the oxp backup in FS
[11:12] Gavin.Hird Kokua can save DAE - which is XML sortoff
[11:13] Selby.Evans My exports were prims
[11:13] Ubit Umarov: ohh even the angel that fails to upload on FS
[11:13] Basil Sosides: Andrew, are the maptiles from OSgrid out of order ??
[11:13] Leighton.Marjoram Or collada
[11:13] Arielle Popstar: i wear them ok
[11:13] Ubit Umarov: dan will fix them when he returns from work ( the maps )
[11:14] Andrew.Hellershanks Basil, Looks like the map tiles aren't working right. I only see one tile for this region
[11:14] Kayaker Magic: Basil: where do you not see maptiles? I see them in the map here.
[11:14] Basil Sosides: ok-
[11:14] Leighton.Marjoram Oh nice Arielle
[11:14] Kayaker Magic: I search for my regions in the map and I see maptiles there.
[11:14] Arielle Popstar: needs some texturing  :)
[11:14] Leighton.Marjoram is that the extended avastar 2.0 avatar
[11:14] Basil Sosides: no Maptiles at all
[11:15] Andrew.Hellershanks Leighton, I use Singularity for exports. IIRC, Firestorm will let you do XML exports but the two viewers don't use the same file format so you can't export in on and import using the other.
[11:15] Basil Sosides: Singu or Firestorm or Maptiles
[11:15] Leighton.Marjoram Thank you Andrew thats good to know
[11:15] Arielle Popstar: can copy the ones at the front Leighton
[11:15] Arielle Popstar: Buy didnt work for me
[11:16] Andrew.Hellershanks bah. Can't export in one (not "in on"). Darn typo :)
[11:16] Basil Sosides: oh Map is back i think
[11:16] Arielle Popstar: Sheera brought them
[11:16] Nebadon Izumi: yes I restarted the grid services
[11:16] Andrew.Hellershanks Sheera is not only the meeting secretary but also grid cartographer? ;)
[11:17] Arielle Popstar: Jill of all trades
[11:17] Andrew Hellershanks: indeed
[11:18] George Equus: Map is fine for me in FS
[11:18] Arielle Popstar: any testing started on the Permissions?
[11:18] Ubit Umarov: i gave a prim to my alt
[11:18] Ubit Umarov: :p
[11:18] Leighton.Marjoram Map is ok for me on FS
[11:18] Arielle Popstar: how did that go? :)
[11:18] Ubit Umarov: and he lost it.. damm
[11:19] Arielle Popstar: lost but not found
[11:19] Ubit Umarov: such a pretty box.. and he lost it
[11:19] Ubit Umarov: ;)
[11:19] Arielle Popstar: hate when that happens
[11:20] Andrew Hellershanks: I don't see any other map tiles yet in Sing
[11:20] Ubit Umarov: my alt is worse than me misplaces everything
[11:20] Arielle Popstar: luckily lot more pretty boxes out there
[11:20] Gavin.Hird I hate it more when half the prims of a build gets randomly auto-retruned.
[11:20] Gavin.Hird way to ruin a scene
[11:21] Nebadon Izumi: I have never seen that happen
[11:21] Andrew Hellershanks: I have a grid where one prim of mine gets auto-returned at odd times once in a while. No idea why it happens.
[11:21] Gavin.Hird It can happen in SL if you convert a parcel from group land to regular
[11:21] Gavin.Hird probably here too
[11:21] Andrew Hellershanks: It is usually the same one from the same parcel every time when it does happen.
[11:21] Basil Sosides: lol, i clicked the Tower Elevator
[11:22] Ubit Umarov: better not test that here on a relevant building :)
[11:22] Gavin.Hird if the parcel gets full because you loose the 10% bonus, very interesting things happen
[11:23] Basil Sosides: i have an Elevator on my Land too, runs up to 2000m, Physical Vehicle...
[11:23] Andrew Hellershanks: Let's see... what news to pass on since last week...
[11:23] Andrew Hellershanks thinks
[11:23] Basil Sosides: ok.
[11:23] Ubit Umarov: ( hide.. andrew is thinking.. brain may explode )
[11:24] Ubit Umarov: ;)
[11:24] Arielle Popstar: :p
[11:24] Andrew Hellershanks: I still have some updates to complete for the PHP part of the OpenSimProfile module. This is due to some differences in return values between mysql_* calls and PDO calls when making SQL queries.
[11:24] Kayaker Magic: Anyone looking at Mantis #8127? Which I would like to rename "Scripts on Microsoft servers run up to 10,000 times slower"
[11:24] Ubit Umarov: grrr just move to core profiles
[11:25] Andrew Hellershanks: diva has made some more progress on testing object related perms and setting up automated testing.
[11:25] Andrew Hellershanks: diva and Ubit have been working on object related perms and additional fixes have been applied to git master in the last week.
[11:26] Andrew Hellershanks: Ubit, You can stay on the add on module unless you are using a system running PHP 7.
[11:27] Andrew Hellershanks: I've worked up a list of commands needed to migrate from OSP to core. The only part that isn't easy to migrate is attachments to group notices. The two profile modules store the attachment data in slightly different ways.
[11:28] Andrew Hellershanks looks at mantis 8127
[11:28] Arielle Popstar:
[11:28] Andrew Hellershanks: Oh, that is the list to vector issue
[11:29] Ubit Umarov: actually its timing
[11:29] Ubit Umarov: and im not sure the way time was measured wasn't the issue
[11:30] Kayaker Magic: 8127 is not just list to vector, it is reading anything from a list on a SIM hosted on a Microsoft server.
[11:30] Ubit Umarov: Typical values are between 10 and 20ms was the typical resolution of llGetTime(); on Windows
[11:30] Andrew Hellershanks: I have no way to test that as I don't have any installs of OpenSim running under Windows
[11:31] Ubit Umarov: master now has 1ms resolution
[11:31] Kayaker Magic: I was testing different things on different servers and didn't see the correlation with MS .NET at first.
[11:31] Kayaker Magic: Unfortunately lots of OSGrid regions are hosted on PCs.
[11:32] Basil Sosides: true Kay
[11:32] Ubit Umarov: Kayaker on windows normal time functions have a resolution of 10 to 20ms
[11:32] Andrew Hellershanks: Timing on standard PCs used to be based on timer ticks of 18 (or was it 18.2) per second.
[11:33] Kayaker Magic: Not time functions: llList2Vector(for example) takes 10ms on a Windows server!
[11:33] Basil Sosides: should i change my OS ? to Linux ?
[11:33] Ubit Umarov: hmm and some functions have time penalty also
[11:33] Andrew Hellershanks: That's up to you, Basil.
[11:33] Basil Sosides: yes
[11:34] Kayaker Magic: I have scripts that read a lot of lists, run them on a region hosted on Linux, they run fine. run them on a Microsoft hosted region, they take HOURS.
[11:34] Basil Sosides: or i call Bill Gates and tell him whats wrong !!!!
[11:34] Ubit Umarov: hm is this doint anything??
[11:34] Ubit Umarov: lol
[11:34] Andrew Hellershanks: Ubit, some do. The LSL wiki sites will list any time penalties.
[11:34] Arielle Popstar: every ms os Kay?
[11:36] Andrew Hellershanks: Kayaker, if you are using the same version of OpenSim and the only difference is the operating system being used then the problem is something in Windows.
[11:36] Kayaker Magic: Well, I have not tested every MS system in the world yet....
[11:36] Ubit Umarov: kay you can check my box script
[11:36] Ubit Umarov: well this is linux i guess lol
[11:36] Andrew Hellershanks: I don't know if there would be a way to change anything in OS to make the scripts run faster in Windows.
[11:36] Arielle Popstar: i am meaning 5, 8.1, 10 server editions
[11:37] Arielle Popstar: sorry 7^
[11:37] Kayaker Magic: Fetching values from a list is a pretty basic function. Without usable memory, can you call LSL a language any more?
[11:37] Andrew Hellershanks: Kayaker, #8127 lists the operating system in question as Linux and not Windows (in the summary area)
[11:38] Ubit Umarov: 11:38] Ubit Umarov: well box is doing the same on my win box
[11:38] Ubit Umarov: shows 0 time
[11:38] Andrew Hellershanks: Kayaker, which version of Windows have you been using to run the tests?
[11:38] Kayaker Magic: When I first submitted the mantis I had not noticed the correlation, which system it was failing on.
[11:38] Andrew Hellershanks: Kayaker, np.
[11:39] Ubit Umarov: i say again
[11:39] Ubit Umarov: that script has flaws as a time measument one
[11:39] Kayaker Magic: Several different versions, I just had a friend install SOAS on a Windows 10 and it had the same problem.
[11:40] Andrew Hellershanks: ok.
[11:40] Kayaker Magic: The script in the mantis has flawed time measurement?
[11:40] Ubit Umarov: yes
[11:40] Andrew Hellershanks: Kayaker, The bug report won't let me change the OS entry. Perhaps you can do that when you get a moment.
[11:40] Kayaker Magic: because llGetTime has lower resolution on some systems?
[11:41] Ubit Umarov: that is one yes
[11:41] Ubit Umarov: test on master
[11:41] Andrew Hellershanks: Kayaker, btw, you can also use lLGetAndResetTime()
[11:41] Ubit Umarov: llgetime now has 1ms on all platforms
[11:42] Kayaker Magic: But the test is measuring the llGetTime before and after a list fetch, the differences I am getting are still orders of magnitude.
[11:43] Ubit Umarov: well i just repeated the test on win7 and display was 0
[11:43] Kayaker Magic: May I have your test Ubit?
[11:43] Andrew Hellershanks: I don't know if some of the difference is the delay between when the script resets the time and when it gets it.
[11:44] Ubit Umarov: its c&p of your script :)
[11:44] Ubit Umarov: but sure
[11:45] Kayaker Magic: I also have scripts that do not do timing, they just do lots of list fetches, and I do timing with my head. On some systems, done in a second, on another system, done in 2 HOURS. Is that not a good timing methodology? Is the timer in my head wrong also?
[11:45] Andrew Hellershanks: I haven't used the timing functions in a long time but when I did need them to determine the timing on something I used the get and reset function to minimize errors in the timing measurement.
[11:46] Ubit Umarov: its not the issue on that script
[11:46] Ubit Umarov: but timers have low resulution on default settings
[11:46] Ubit Umarov: there are 2 config options related to those
[11:47] Ubit Umarov: they should be 50ms but whatever
[11:47] Andrew Hellershanks: The LSL wiki's say it gets time in seconds. One says it has sub-second precision without saying what level of precision.
[11:47] Andrew Hellershanks: Ubit, what are the settings that affect the timer precision?
[11:48] Kayaker Magic: I do not care about the llSetTimer, I use it for 0.5 seconds that should work with default settings.
[11:48] Kayaker Magic: I only use llSetTimerEvent for a statistically long number of list fetches.
[11:49] Kayaker Magic: Tell you what, I will make Ubit's changes to my test program, try it out on a few systems come back next week
[11:49] Kayaker Magic: I predict there will be a problem, but I may not be communicating the correct problem now,.
[11:49] Ubit Umarov: i did not changed it
[11:49] Ubit Umarov: hmm i think
[11:50] Ubit Umarov: i DID change llGetTime months ago on master
[11:51] Andrew Hellershanks: The tests need to be run enough times to eliminate any impact there may on the timing due to other processes running on the machine.
[11:51] Andrew Hellershanks: Other scrips running in the region could also have an impact on the timing.
[11:51] Kayaker Magic: Any other timing methodologies I should change? Longer llSetTimerEvent to get away from resolution of that timer?
[11:51] Kayaker Magic: Is llGetAndResetTime more accurate than what I did?
[11:52] Ubit Umarov: same thing
[11:52] Kayaker Magic: I am seeing FOUR ORDERS OF MAGNITUDE, I don't see how these little things can help.
[11:52] Ubit Umarov: its the same timer
[11:52] Ubit Umarov: well then its something else .. very strange
[11:53] Ubit Umarov: bc seems to work ok on win7 and linux
[11:53] Andrew Hellershanks: Kayaker, you could print when it enters and exits the timer handler to make sure there isn't any overlap. I wouldn't expect there to be. A half second between tests should be plenty.
[11:54] Ubit Umarov: also don't forget threads don't run all the time
[11:54] Ubit Umarov: OS does take cpu from them... its the basis of multitasking
[11:55] Ubit Umarov: and that on windows is again around 18ms min
[11:56] Andrew Hellershanks: You could set the timer interval to a much longer amount and do many more tests in the timer handler.
[11:56] Kayaker Magic: But threads don't time slice, do they? If I call llGetTime, fetch from a list, and call llgetTime, the thread was running for that whole time.
[11:56] Ubit Umarov: of course they do
[11:56] Ubit Umarov: you only have 4 or 6 cores
[11:56] Ubit Umarov: and 100's of threads
[11:57] Kayaker Magic: So threads timeslice but scripts do not, as much as we would like them to to prevent llSleep(10) from locking up a thread?
[11:57] Ubit Umarov: rules on the timeslicing.. well OS dependent etc
[11:57] Ubit Umarov: a script event runs on a thread
[11:58] Ubit Umarov: llSleep problem is that it blocks a thread
[11:58] Ubit Umarov: and there is a limit on number of threads for scripts
[11:58] Kayaker Magic: I see, a blocked thread still gives up its core.
[11:58] Ubit Umarov: so 2 many on llsleep.. no more threads to do other scripts
[11:59] Ubit Umarov: the sleeping thread is not using cpu
[11:59] Ubit Umarov: diferent issue see?
[11:59] Kayaker Magic: I see
[12:00] Ubit Umarov: ( actually very small sleep times can put a thread in spinlock.. just burning cpu for the requested time.. but details ;) )
[12:01] Kayaker Magic: I don't have anything new to test, are there no timing methodology changes to the test script? What can I test? More different systems?
[12:01] Ubit Umarov: you don't need precise timing i guess
[12:01] Andrew Hellershanks: Kayaker You said there are four orders of magnitude different in the times. What are typical values of the times you get under Linux vs under Windows?
[12:02] Andrew Hellershanks: More accurate timing may be the issue rather than more precision.
[12:02] Ubit Umarov: Ahh and there is GC.. it can stop all threads
[12:02] Ubit Umarov: grr i hate GC
[12:02] Kayaker Magic: llResetTimer; llList2Vector; llGetTimer returns 10ms on some systems, 0.00001 on other systems.
[12:03] Andrew Hellershanks: Auto GC in C# (or other OOPL) is supposed to be a big help to the programmer but it can also be a pain.
[12:03] Sheera.Khan but the gc won't run for hours...
[12:03] Ubit Umarov: GC is a huge %%%/%/$$$$%&&/&
[12:03] Andrew Hellershanks: 10mS va 10uS? That is a big difference.
[12:03] Sheera.Khan Kay talks about a difference in runtime between secs on Linux vs. hours on Windows
[12:03] Ubit Umarov: guys...
[12:03] Ubit Umarov: that is OLD news
[12:04] Ubit Umarov: Windows time is low resolution time
[12:04] Andrew Hellershanks: I would suggest doing at least 10 if not 100 times the number of tests in the loop to try and even out some of the times that may be affect by other things happening in the system.
[12:04] Ubit Umarov: to have better one needs to use performance counters
[12:04] Kayaker Magic: Andrew: I changed the Environment on that Mantis (Didn't know I could do that!)
[12:04] Ubit Umarov: and that is what i use now on master
[12:04] Andrew Hellershanks: Kayaker, ok. ty
[12:05] Kayaker Magic: I could get academic and calculate mean and std....
[12:05] Andrew Hellershanks: It still mentions Linux mono but it may update later. Chrome may have cached the page.
[12:05] Ubit Umarov: mean and std only have meaning on random events
[12:05] Ubit Umarov: that is not the case :p
[12:06] Ubit Umarov: well unless on a very large time scale
[12:06] Andrew Hellershanks: Kayaker you are the original bug reporter so you should be able to change a number of the items in the summary area of the report.
[12:06] Ubit Umarov: but a task switch may happen reading time..
[12:06] Ubit Umarov: and that is a bias added to avg
[12:07] Kayaker Magic: Oh, that is a problem I have with the Mantis page: When a form asks me what version of OS I am using, does that mean OpenSim or Operating System? Collisions in acronym space.
[12:07] Ubit Umarov: well whatever
[12:07] Ubit Umarov: Windows and desktop linux are not RTOS ( real time os )
[12:07] Andrew Hellershanks: OS in that case meant operating system.
[12:08] Andrew Hellershanks: yea, OS reference can be confusing.
[12:08] Ubit Umarov: had lots of fun seeing a guy controling a fpga timing with linux
[12:08] Ubit Umarov: so so funny :)
[12:09] Ubit Umarov: until i made him see he HAD A FPGA he could do the timing on it :)
[12:09] Andrew Hellershanks: I wonder if the text can be changed in mantis to clarify what is being asked for (other than by trying to determine that from any available choices offered if you have to pick from a list)
[12:09] Andrew Hellershanks: Ubit, :D
[12:09] Kayaker Magic: Yeah, long time ago I tried to count frames in a video, Microsoft Windows would go away for 100ms sometimes and mess up my count.
[12:09] Andrew Hellershanks: That is funny, Ubit.
[12:09] Ubit Umarov: yeap audio sync was a nasty issue also
[12:10] Ubit Umarov: on audio and video digit
[12:10] Ubit Umarov: using the sound card was a fail most the time
[12:10] Ubit Umarov: one had to use hardware that did both things
[12:11] Andrew Hellershanks: I worked on a project where I had to count pulses on a flow meter using interrupts and work out pulses per second. I had the system tell me time between readings when I asked for the count so I could take in to account the actual time that passed since the previous count was received.
[12:11] Ubit Umarov: you can't use microcontrolers programming style on linux
[12:12] Ubit Umarov: unless timings are large.. seconds scale
[12:12] Andrew Hellershanks: Ubit, that was a Linux based system where I was dealing with the flow meter.
[12:12] Ubit Umarov: but on linux one can change settings
[12:13] Ubit Umarov: in fact one can suspend multitask
[12:13] Andrew Hellershanks: Ubit, true but I just used the defaults. I didn't do any special tweaking. I just knew that other things were going on in the system so I coudn't guarantee that my readings would always be exactly 1 second apart.
[12:14] Andrew Hellershanks: I forgot to ask if someone wanted to coordinate some of the testing efforts. I will have to remember to ask next week.
[12:14] Sheera.Khan IMA wants to...
[12:15] Sheera.Khan so Selby would be the one to ask for it
[12:15] Andrew Hellershanks: yes, but we need a specific contact person.
[12:15] Sheera.Khan and the devs for sure ;-)
[12:15] Andrew Hellershanks: Right. Selby was the person I meant to ask about that.
[12:15] Roland.Francis I promised to hand it over to IMA - and Selby Evans has (so he told me) discussed an approach with "you guys (whoever)"
[12:16] Roland.Francis there will be a IMA meeting on Friday to introduce and discuss the topic
[12:16] Sheera.Khan is it at 6 PM again?
[12:16] Andrew Hellershanks: The people who are available and willing to help with testing need to pass their names on to the person who can coordinate the testing effort.
[12:16] Roland.Francis sorry, don't know by heart, Sheera
[12:17] Roland.Francis 1:30 PST as I recall
[12:17] Sheera.Khan the friday meeting is scheduled for 6:30 PM SLT... that translates to about 3:30 AM for me...
[12:17] Roland.Francis as promised, I'll report back "one time" only, as I'm not engaging with IMA
[12:18] Roland.Francis but make sure that the testing will be taken on and is managed
[12:18] Roland.Francis Selby told me he is "on this"

Personal tools
About This Wiki