Chat log from the meeting on 2016-07-12

From OpenSimulator

Revision as of 13:10, 12 July 2016 by Sheera Khan (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

[11:03] Andrew Hellershanks: Hello, everyone.
[11:05] Andrew Hellershanks: I've got my computer back up and running after swapping out the bad drive in the RAID array. So far all is well but I'm not 100% sure the array was properly rebuilt.
[11:05] Andrew Hellershanks: At least I didn't lose anything when one of the drives died on me.
[11:05] Gavin.Hird glad to hear
[11:08] Andrew Hellershanks: Development work on 0.9 continued this past week but there are still a few outstanding issues on the ToDo list for an 0.9 release.
[11:09] Andrew Hellershanks: Its nice to see that the development activity has picked up in the last few weeks.
[11:11] Andrew Hellershanks: Has anyone had any run across any new problems in OpenSimulator during this past week?
[11:12] Gavin.Hird it's been pretty stable. I am on a build that is 5 hours old
[11:13] Andrew Hellershanks: ok.
[11:13] Gavin.Hird But I get a bit nervous over commit messages like this one " Change the user verification call back to use the host name instead of the
IP address. Fixes NGINX on OSGrid getting confuzzled." as it sounds like there is special code for OSgrid placed there
[11:14] Andrew Hellershanks: Two changes this past week worth noting are some changes to partially (mostly?) address mantis #7904. That issue has to do with prims that have a large number of items in their prim contents.
[11:15] Andrew Hellershanks: Gavin, I think that is related the other change of note I was going to mention.
[11:15] Gavin.Hird ok
[11:18] Andrew Hellershanks: Another change this week was to deal with HG TP issues between different versions of OpenSim. It should allow some TP's between some grids where previously you might have had differences in the HG support between the grids.
[11:20] Andrew Hellershanks: Gavin: I just found the comment about that change in my IRC history.
[11:21] Gavin.Hird there is nothing wrong with the change, it was only the commit message that sounded like it was OSG specific
[11:21] Gavin.Hird which would not be so good if that was the case
[11:22] Andrew Hellershanks: Melanie stated that HttpWebRequest, when fed with a URL that has an IP as the host portion, will not produce a host: header in recent mono. An earlier mono produced a host: header with the ip address as host. nginx refuses to work without a host header present so the change makes regions outside of osgrid, in particular standalones, work with nginx again.
[11:22] Andrew Hellershanks nods to Gavin.
[11:23] Andrew Hellershanks: The commit messages are often quite terse and don't always give you any insight as to why the change was needed.
[11:24] Andrew Hellershanks: That's all I have in the way of some news items regarding recent development for this week.
[11:28] Andrew Hellershanks: Looking again at the comments from Melanie, I don't think that change affects HG TP's between grids.
[11:29] Andrew Hellershanks: Any other topics of discussion for today?
[11:29] Gavin.Hird not really
[11:30] Andrew Hellershanks: This meeting might be a very short one for a change. :)
[11:30] Gavin.Hird busy with viewer stuff this week
[11:30] Andrew Hellershanks: Gavin, routine bugs or is there something new and interesting being worked on?
[11:31] Gavin.Hird I suppose it is mostly of interest for Mac users, as it has to do with how / if the viewer runs on the upcoming macOS 10.12
[11:31] Andrew Hellershanks: ok
[11:32] Gavin.Hird they are removing support for some deprecated technologies that still are in the viewers
[11:32] Gavin.Hird wich makes at least one of the viewers tested unable to run at all
[11:32] Simulator Version v0.5 ruft: OpenSim Dev f87d5d0: 2016-07-11 22:32:27 +0100 (Unix/Mono)
[11:33] Andrew Hellershanks: That doesnt sound good.
[11:33] Gavin.Hird it was announced as deprecated in macOS 10.6 and we are now on 10.12
[11:33] Gavin.Hird so that is about 6 years ago
[11:33] Andrew Hellershanks: ah, ok.
[11:33] vegaslon plutonian: will kokua support lightshare in opensim?
[11:34] Gavin.Hird I have to check that
[11:34] Gavin.Hird but I think it does
[11:34] Gavin.Hird because I have an old Lightshare control panel that is ancient
[11:34] Gavin.Hird and I believe it worked the last time I tested it
[11:34] Gavin.Hird but
[11:35] Gavin.Hird I will go over the code and see if there are issues
[11:35] vegaslon plutonian: if it does work. should update opensim wiki for lightshare to state that it does work now
[11:35] Gavin.Hird In the Grids tab of preferences there is a LightShare for Opensim button to enable
[11:35] Andrew Hellershanks: There is an avatar I haven't seen in a while. :)
[11:36] Andrew Hellershanks: avatar outfit, I should say.
[11:37] Gavin.Hird there might be some issues with Windlight too in Kokua that I need to check
[11:37] Nebadon Izumi: I saw that Ubit was just looking at windlight stuff
[11:37] Nebadon Izumi: guess there is a mantis or something
[11:37] Gavin.Hird yep
[11:37] Gavin.Hird there is
[11:38] Gavin.Hird 7961
[11:38] Andrew Hellershanks: Hm... I was thinking Windlight when Lightshare was mentioned. I don't recall hearing about Lightshare.
[11:38] Gavin.Hird I don't think very many use Lightshare
[11:38] Gavin.Hird but I believe it works
[11:39] vegaslon plutonian: lightshare is opensims windlight modules but we allow a few more options like sending windlight to only one user
[11:39] Andrew Hellershanks: ah, ok. I was just looking at a page about it.
[11:40] Andrew Hellershanks: I haven't tried using the Lightshare/Windlight settings.
[11:40] vegaslon plutonian: it is the closest thing you can get to parcel windlight outside of firestorm viewer
[11:40] George Equus: Sorry about that..
[11:41] Gavin.Hird It is one of those things you need to know pretty good to be able to test it correctly
[11:41] Andrew Hellershanks: Particles are a bit like that.
[11:41] Gavin.Hird Is the opensim behavior supposed to be indentical to LL behavior?
[11:42] Ubit Umarov: you ask hard questions :p
[11:42] Gavin.Hird hehe
[11:42] Gavin.Hird it is hard to test a moving target :-)
[11:42] Ubit Umarov: from what i seen checking that mantis
[11:43] Ubit Umarov: lightshare module adds scripting to windlight
[11:43] vegaslon plutonian: that would be preferable I believe. Seen weird lighting effects though like lighting suddenly changing to noon when enter another parcel
[11:43] Nebadon Izumi: yea I have a pretty cool controller somewhere
[11:43] Ubit Umarov: without any script region stores and gives back what viewer tells about enviroment
[11:43] LightShare: Oh no...[ loading Default settings from NC, be patient.. ]
[11:43] Gavin.Hird I have a very old controller from 2008-ish
[11:44] Gavin.Hird a panel
[11:44] Gavin.Hird that's the one
[11:44] vegaslon plutonian: yep that worked, windlight just changed for me
[11:44] Nebadon Izumi: grab a copy :)
[11:44] Nebadon Izumi: if you want
[11:44] Ubit Umarov: but i never played with that here or sl
[11:44] George Equus: Never tried using it, got one too
[11:44] Nebadon Izumi: seems to work here
[11:45] George Equus: I think... recognize it anyway
[11:45] Gavin.Hird I don't think every function works any more
[11:45] Nebadon Izumi: ya would not surprise me
[11:45] vegaslon plutonian: created a system once that was changed windlight for people based on their elevation
[11:46] Gavin.Hird in this one you can set the sun height from a lattitude
[11:47] Gavin.Hird or the sun to a geo location
[11:47] Ubit Umarov: ( why do i see @hg.osg in my name? odd)
[11:48] Gavin.Hird did you tp infrom somewhere else?
[11:48] Nebadon Izumi: cache
[11:48] Nebadon Izumi: it happens sometimes Ubit
[11:48] Nebadon Izumi: if you log into another grid that has it cached as @hg.osg
[11:48] Ubit Umarov: maybe because i killed the viewer on an external hg region
[11:48] Ubit Umarov: and relogged back here
[11:49] Nebadon Izumi: i have it happen if i log into Avacon grid
[11:49] Nebadon Izumi: before i log into OSgrid
[11:49] Nebadon Izumi: then after that I always see the @hg.osgrid
[11:49] Nebadon Izumi: i have a ton of my stuff on that grid
[11:49] Nebadon Izumi: same with oscc grid
[11:50] Ubit Umarov: well i still click on you and get my profile :)
[11:50] Gavin.Hird looks like MOSES is giving up on OpenSim
[11:50] Nebadon Izumi: that is because im wearing mesh
[11:50] Gavin.Hird they talk about MOSES-Halcyon
[11:50] Andrew Hellershanks: Ubit, Can you comment on mantis #7812? Is that only an issue in a particular physics engine?
[11:51] Nebadon Izumi: my actual avatar is totally enclosed inside a mesh
[11:51] Ubit Umarov: oh they are ?
[11:51] Nebadon Izumi: i hate that bug
[11:51] Ubit Umarov: yeap me 2
[11:51] Nebadon Izumi: you can right click my name above head
[11:51] Nebadon Izumi: and get my profile
[11:51] Ubit Umarov: but for some reason mel told me not to look into it
[11:51] Andrew Hellershanks: Gavin, could that just be their name for their fork(?) of OS?
[11:51] Gavin.Hird I am actaully usure
[11:52] Gavin.Hird Unfortunately, this means that Opensim-based MOSES is on its last legs. We will, of course, keep MOSES-Opensim up as long as it is useful, but new development is already occurring on MOSES-Halcyon
[11:52] Nebadon Izumi: I am 100% ok with that
[11:53] Nebadon Izumi: i worked pretty close with those guys and they always made opensim run really bad
[11:53] Nebadon Izumi: and then blamed it all on us
[11:53] Nebadon Izumi: i experienced things on moses that I was absolutely unable to recreate on stock opensim
[11:53] Nebadon Izumi: dozens of bugs
[11:54] Nebadon Izumi: their problem is they want to use Simiangrid back end
[11:54] Gavin.Hird they did run simian-grid?
[11:54] Nebadon Izumi: and its absolutely wretched
[11:54] Nebadon Izumi: yea
[11:54] Nebadon Izumi: its sooo buggy
[11:54] Nebadon Izumi: inventory is a complete disaster on simiangrid
[11:54] Ubit Umarov: Halcyon is a fork of OpenSimulator that has been used by the Inworldz grid for some time. It forked from Opensim several years ago, and because of that, has the following differences from an opensim-based grid:
[11:54] Gavin.Hird ok, thx Ubit
[11:55] Gavin.Hird makes sense now that you say it
[11:55] Ubit Umarov: route to use phyX
[11:55] Nebadon Izumi: the real reason doug is going that route is because he wants PhysX
[11:55] Nebadon Izumi: which in my opinion without fixing 1000 other things is a big waste of time
[11:55] Ubit Umarov: yeap
[11:55] Nebadon Izumi: he wants 1000s of physical objects
[11:56] Andrew Hellershanks: nebadon, Ubit: You both said you hated a bug. Where you commenting on mantis #7812 or the other issue you were talking about?
[11:56] Nebadon Izumi: but physics engine isnt the problem
[11:56] Nebadon Izumi: its the ability to update the viewer with 1000s of updates that is the problem
[11:56] Gavin.Hird those guys always want to blow up things, hehe
[11:56] Andrew Hellershanks: Gavin, I wonder why? ;)
[11:56] Nebadon Izumi: i've shown hands down that bullet and ubode are just as good as PhysX
[11:56] Nebadon Izumi: to do what they want to do
[11:56] Nebadon Izumi: they think they can do Hardware accelerated physics
[11:57] Nebadon Izumi: on server side
[11:57] Nebadon Izumi: they are going to have the same exact problem
[11:57] Ubit Umarov: hmm mantis 7812?
[11:57] Nebadon Izumi: without fixing all the other stuff first
[11:57] George Equus: PhysX is real old isn't it? Seem to remeber games sometimes used it 10 Y back
[11:57] Nebadon Izumi: honestly I am not sure we can ever fix the SL viewer enough
[11:57] Gavin.Hird plus you have to stuff the server with graphics cards
[11:57] Nebadon Izumi: to do what they want to do
[11:57] Nebadon Izumi: it would really need new protocols and viewer
[11:57] Ubit Umarov: physX is up to date
[11:57] Nebadon Izumi: offload some of that physics to the viewer itself
[11:58] Ubit Umarov: now owned by nvidia
[11:58] Nebadon Izumi: well thats one thing they do have
[11:58] Ubit Umarov: and provided with all gpu drivers i think
[11:58] Nebadon Izumi: MOSES has servers with 1tb ram in them
[11:58] George Equus: So still used today then I take it
[11:58] Nebadon Izumi: they could put 8 NVIDIA cards in their servers
[11:58] Nebadon Izumi: lol
[11:58] Gavin.Hird the viewer should be able to do a lot of the physics that only is local to the avatar for instance
[11:58] Nebadon Izumi: still all that hardware and it runs like poo
[11:58] Nebadon Izumi: no better than stock opensim
[11:59] Nebadon Izumi: infact in some cases much worse
[11:59] Andrew Hellershanks: Ubit:
[11:59] Nebadon Izumi: its certainly worth pursuing, but OpenSimulator isnt going to get them what they want
[11:59] Ubit Umarov: seems a raycastv3 bug.. i hate raycastv3 :p
[12:00] Ubit Umarov: didn't remove it because bullet doesn't do raycasts :)
[12:00] Nebadon Izumi: ya Misterblue says he wants to implement bulletsim raycast eventually
[12:01] Andrew Hellershanks: Ubit, I have some grid owners who think that 7812 would need to be fixed before a switch to 0.9 as it would be a big problem for one of their regions.
[12:01] Ubit Umarov: actually bullet may also see the inside of a sphere.. not sure
[12:01] Gavin.Hird I am sure there are more pressing issues than raycast
[12:02] Nebadon Izumi: ya hes been looking into increase in CPU levels
[12:02] Gavin.Hird good
[12:02] Nebadon Izumi: he has a big to-do list :)
[12:02] Andrew Hellershanks: Gavin, I don't see it in the list of 0.9 bugs to be fixed for the upcoming release.
[12:02] Ubit Umarov: err what list ?
[12:02] Nebadon Izumi: Raycast definitely wont be in 0.9 release
[12:02] Nebadon Izumi: improved raycast that is
[12:02] Andrew Hellershanks: Ubit:
[12:03] Ubit Umarov: oops :)
[12:03] Andrew Hellershanks: Nebadon, do you recall if there is a mantis # re: the change in CPU usage?
[12:03] Nebadon Izumi: hmmm
[12:03] Nebadon Izumi: i am not sure actually
[12:03] Gavin.Hird I think I made one
[12:04] Andrew Hellershanks: ok. It is important enough it should be noted on the 0.9 bug list
[12:04] Ubit Umarov: ohh one got resolved it seems lol
[12:04] Ubit Umarov: let me see
[12:04] Gavin.Hird 7925
[12:04] Andrew Hellershanks: ok, thanks Gavin. I'll get that added to the bug list page.
[12:04] Ubit Umarov: damm 7959 i already knew its fixed
[12:05] Gavin.Hird great
[12:06] Ubit Umarov: This comes in addition to an about 50% increase in CPU on 0.9dev over release version.  ??
[12:06] Gavin.Hird it does
[12:07] Gavin.Hird a simulator that was idling at about 12% now is 95% in 0.9 + latest mono
[12:08] Andrew Hellershanks: That's going to be a show stopper for me. It would max out the servers I'm running.
[12:08] Ubit Umarov: 0.8 also compiled with same mono ?
[12:08] Gavin.Hird no
[12:09] Ubit Umarov: well that something to check before anything.. since mono.. is.. well its mono :)
[12:09] Gavin.Hird mono 4.4 says deep locking is much more expensive now
[12:09] Gavin.Hird don't know how much of that is in the server code
[12:10] Ubit Umarov: we do have tons of locks
[12:12] Nebadon Izumi: we are like that Bridge in france :P
[12:12] Nebadon Izumi: Paris-Love-locks.jpg
[12:12] Nebadon Izumi: haha
[12:12] Andrew Hellershanks: Are we saying there is two different sources of increased CPU loads in 0.9?
[12:12] Gavin.Hird wow
[12:12] vegaslon plutonian: all your locks are belong to mono
[12:13] Andrew Hellershanks: hehe
[12:13] Nebadon Izumi: they actually had to cut all those locks off
[12:13] Nebadon Izumi: the bridge was about to collapse it got so heavy
[12:13] Nebadon Izumi: haha
[12:13] Ubit Umarov: lol they had to remove those locks neb .. the bridge was about to fall :)
[12:13] Andrew Hellershanks: Wow. I didn't know it had gotten that bad.
[12:13] Ubit Umarov: yeap
[12:14] Gavin.Hird from the mono release notes:
[12:14] Gavin.Hird Unmanaged Thin Locks
This avoids allocation of new MonoThreadsSync structures for the typical cases of shallow nesting of uncontended locks. The necessary information (owner, nest and status bits) is embedded in the synchronisation field of a MonoObject. The lock is inflated when the nest overflows or when there is contention on the lock.

This greatly improves performance (over 10 times) for the case of locking a lock only a few times, due to the inflation avoidance. It also improves performance up to 25% for the case of locking a thin lock. A decrease in performance (up to 3 times slower) can happen for the case of repetitive nesting of a lock, due to the need of a CAS at each entry/exit.
[12:14] Nebadon Izumi: is there a URL for this?
[12:14] Gavin.Hird
[12:14] Nebadon Izumi: thanks
[12:14] Ubit Umarov: the locks issue i seen it also :)
[12:14] Gavin.Hird yw
[12:15] Ubit Umarov: but i still have no clue for such a difference
[12:15] Nebadon Izumi: cooperative GC mode sounds interesting
[12:15] Nebadon Izumi: i may have to test that
[12:15] Ubit Umarov: ( also don't notice it in windows all my regions including a oar of Sisyphus... show 0 to 1% load :) )
[12:16] Andrew Hellershanks: Ubit, I find it surprising that a number of locks on the bridge would get it close to collapse. Can their combined weight be more than a large number of vehicles on the bridge? Sounds like a design flaw.
[12:16] Gavin.Hird I suppose the railing was not dimensioned for the load
[12:16] Ubit Umarov: look to the number of locks there :)
[12:17] Andrew Hellershanks: Ubit, yea, but still...
[12:17] Nebadon Izumi: Plugh > The bridge carries more than 700,000 locks with an estimated combined weight roughly the same as 20 elephants.
[12:17] Ubit Umarov: and andrew bridges aren't made all that strong
[12:17] Nebadon Izumi: ya its an ancient bridge too
[12:17] Gavin.Hird don't see many of those on bridges around here
[12:17] Ubit Umarov: there are issues if you fill one with people
[12:18] Andrew Hellershanks: If it is an ancient bridge that might explain it. It wouldn't be designed to handle the types of vehicle loads of modern day.
[12:18] Ubit Umarov: even modern ones
[12:19] Ubit Umarov: you need to check before sending a crowd into one
[12:19] Gavin.Hird there is a bridge in Oslo that has a sign "Can carry 200 soldiers if they don't march."
[12:19] Ubit Umarov: yeap soldiers can't march on any bridge
[12:19] Ubit Umarov: ( but thats to avoid ressonance )
[12:21] Andrew Hellershanks: That's why columns of soldiers break stride when crossing bridges.
[12:21] Ubit Umarov: about the load.. someone needs to test 0.9 with identical conditions as 0.8 including mono
[12:21] Ubit Umarov: compilation and execution mono..
[12:21] Nebadon Izumi: well I need to run, i have a big vehicle simulation im working on, have to place about another 200 waypoints
[12:22] Andrew Hellershanks: I have mono 4.3.x, 4.4.1, and 4.5.x
[12:23] Ubit Umarov: hmm nested locks.. hope we don't have many of those
[12:23] Ubit Umarov: well i don't have a working linux now :(
[12:24] Andrew Hellershanks: I do but I need to complete tests to make sure my drive mirror is a proper mirror after my drive swap.
[12:25] Andrew Hellershanks: Its almost half past. Time to draw this meeting to an end for this week.
[12:26] Andrew Hellershanks: Thanks for dropping in. See you next week.

Personal tools
About This Wiki