User:Ama Omega/archive/Office Hours/2010-08-02

From Second Life Wiki
< User:Ama Omega‎ | archive‎ | Office Hours
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Transcript

[09:00] Ardy Lay: Hello Kelly
[09:01] Kelly Linden: sweet, is Techwolf leading the discussion today?
[09:01] Latif Khalifa: hello :)
[09:01] tx Oh: 'lo
[09:01] Talarus Luan: Sit on him
[09:01] Kelly Linden: I'll just hide back here.
[09:01] Ardy Lay: Hehe
[09:02] Liisa Runo: you should buid your own office, and script your seat to allow only you to sit on it
[09:02] Latif Khalifa: wan't this seat? :)
[09:02] Kelly Linden: Ok, I need a couple of minutes to call my dentist, sorry, brb.
[09:02] Latif Khalifa: ok
[09:03] Kelly Linden: ok, I'm back but I'm gonna have to make this a short office hours, I need to leave in half an hour.
[09:03] Kelly Linden: I swallowed my temp crown on Sat night
[09:03] Talarus Luan: O.o
[09:03] Kelly Linden: ANYWAY. scripting.
[09:03] Latif Khalifa: must've been a good party :)
[09:04] Kelly Linden: hah. I wish.
[09:04] tx Oh: oh dear
[09:04] Latif Khalifa: Kelly, any ETA on mono 2.6?
[09:04] Kelly Linden: We are working on it right now. Our best hope is 1.44 but we need to be a bit on the lucky side to hit that.
[09:05] Latif Khalifa: nice, with acellerated release cycle that is near :)
[09:05] Kelly Linden: Well 1.44 is going to be the first acellerated release, I think, though I'm not as up on how 1.42 is going
[09:05] tx Oh: whats new?
[09:06] Latif Khalifa: at the Beta Server office hours, Oskar was struggling to descibe 1.42 and tcmalloc hanges
[09:06] Latif Khalifa: changes
[09:06] Latif Khalifa: he said Kelly would be the person to ask
[09:06] Kelly Linden: tcmalloc is a 4 line change that has the potential to effect everything
[09:06] Kelly Linden: so it scares the hell out of oskar
[09:06] Latif Khalifa: hehehe
[09:07] Kelly Linden: But really, what we need is some real world testing with good monitoring of memory usage
[09:07] tx Oh: anything else new?
[09:07] Kelly Linden: Anyway the key benefit of moving to tcmalloc is that when a sim bloats we can poke it and find out where the memory is being used.
[09:07] Kelly Linden: Which is way better than 'hunh, its huge, lets log in to see what is going on? maybe look at logs?'
[09:08] Latif Khalifa: Will it enable you to say "Region XXX process takes 2gb ram and is ballooning, go to agni sim and see what's up with it"?
[09:08] Kelly Linden: We can already see what regions are balooning or using too much memory
[09:08] Kelly Linden: but we can only kind of guess by looking at logs what is happening
[09:08] Kelly Linden: With tcmalloc we can get it to print a map of where all the memory is.
[09:09] Latif Khalifa: oh cool, so it will make fixing "lag due to resource leaks swap hell" possible
[09:09] Kelly Linden: Yeah. Really we are going to need to get the script memory limits project back in gear to really address the swapping problem
[09:10] Kelly Linden: Combined with a smarter region conductor to better arrange regions on hosts
[09:10] Talarus Luan: Is that still on-target for this year?
[09:10] tx Oh: how transparentis the mem usage to the land owner?
[09:10] Kelly Linden: Talarus - it is kind of in the air because it is important but we are not yet working on it. And there are some new people debating the design
[09:10] Kelly Linden: tx: it is very transparent.
[09:10] Liisa Runo: combined with new and more efficient ways for us to improve our scripts
[09:11] Kelly Linden: land owners can see a list of objects and how much memory they use right now.
[09:11] Techwolf Lupindo: Whater limits are disided, please don't kill/crash scitps. That won't save memeory, only break content.
[09:11] tx Oh: estate owners can, but land owners?
[09:11] Talarus Luan: OK, so if we say "later this year or early next", it won't be too far out of the ballpark? ;)
[09:11] Latif Khalifa: i don't think that's the full story. there are resource leaks. "sim freezing" gets worse and worse over uptime
[09:11] Kelly Linden: tx yes land owners can
[09:12] Morgaine Dinova: That's very good Kelly, if tcmalloc gives you better instrumentation, otherwise operating almost blind.
[09:12] Kelly Linden: talarus - that is the goal but at this point who knows
[09:12] Talarus Luan: kk
[09:13] Kelly Linden: Latif: right and tcmalloc is going to help us see what is going on
[09:13] Rex Cronon: hello everybody
[09:13] Kelly Linden: which is why we need it
[09:13] Morgaine Dinova: It's like in the viewer, we've diagnosed so many issues by looking at the instrumentation, like Fast Timers output. It's key.
[09:14] Kelly Linden: llCastRay is still being discussed / designed by Falcon
[09:15] Kelly Linden: I confirmed that it is going to truncate rays to the region edge (rather than just failing as it does now)
[09:15] Kaluura Boa: Good news!
[09:15] Kelly Linden: Falcon is looking into better obj->obj coms instead of a specialized event when an object is hit.
[09:16] Ardy Lay: Uhmm... FOund it again: http://jira.secondlife.com/browse/SVC-5598
[09:16] Kelly Linden: One suggestion was llSayToObject(key id, string msg) that would trigger a listen event only in the object
[09:16] Morgaine Dinova: Yep, that's good. Should do the same with megaprims --- just truncate to sim edge as an encrouchment strategy, and be done with it, get on to something else.
[09:16] Talarus Luan: Be still my wishful heart. :P
[09:16] Kelly Linden: I'm not quite sure where he is with that though since I haven't talked to him since early last week.
[09:16] Rex Cronon: lots of peope have been waiting for such a funtion, kelly:)
[09:17] Talarus Luan: Next, you're going to tell us that llTeleportAgent/llTeleportObject is being worked on. :D
[09:17] Kelly Linden: oh Ardy, I'll assign that to me but I know what is wrong, and why it is taking so long to fix.
[09:17] Kelly Linden: Ardy: short version is it is a viewer problem.
[09:18] Kelly Linden: And the better news is I think I'm allowed to muck with the viewer these days.
[09:18] Kelly Linden: The server actually reports all the information, if I'm remembering right. The viewer just screws it up.
[09:19] Liisa Runo: wanna be a hero? add all the new stuff to 1.4 client
[09:19] Kelly Linden: SG 1.4?
[09:19] Ardy Lay: Hehe, thanks Kelly. I originally filed that in VWR, I think.
[09:19] Liisa Runo: no SG, the mainstream client
[09:19] Morgaine Dinova: wanna be a hero? Revert 2.o UI to 1.4 :P
[09:19] Rex Cronon: when should we expect llInputBox(key user, string value, string text) return text:)
[09:19] Talarus Luan: There's a 1.4 mainstream client? O.o
[09:19] Kelly Linden: it sure looks like a server bug, but our automated and non-automated server tests that pull the data directly get all the parcel details
[09:20] Liisa Runo: 1.23 is the mainstream
[09:20] Kelly Linden: 1.4 was about 6 years ago
[09:20] Techwolf Lupindo: Kelly, if you do managed to fix it in the client, I hope there is a small commit that is notialbe in viewer-externial. Some TPV are adding script reporting to 1.4/.5 code base.
[09:20] tx Oh: i have a 2 1/2 year old feature request on llGiveMoney on SVC-1190
[09:20] Talarus Luan: Yeah.. :P
[09:20] Kelly Linden: Techwolf - maybe they are getting it right!
[09:20] Kelly Linden: GiveMoney returning a status code I assume?
[09:20] Talarus Luan: 1.40, then. :p
[09:21] tx Oh: give a reason argument to the function
[09:21] Kelly Linden: Ah.
[09:21] Liisa Runo: yea, my bad, not 1.4..... meant 1.24 or what ever, the client most people use
[09:21] Kelly Linden: We should roll that into however we do a version that reports success
[09:21] Techwolf Lupindo: 1.23.5/1.4/1.5
[09:21] Kelly Linden: As far as I'm aware we are still not working on 1.23 based code.
[09:21] Liisa Runo: yea, and that is FAIL
[09:21] tx Oh: it's also good for the transactions
[09:22] Kelly Linden: <shrug> it is and it isn't.
[09:22] Liisa Runo: you could be the hero
[09:22] Techwolf Lupindo: 1.4/1.5 is still active by the snowglobe community.
[09:22] Kelly Linden: Should I spend the next 8 minutes (wow, short) on software development theory and practice to discuss 1.23 vs developing 2.0 or is there more scripting topics?
[09:22] Ardy Lay: Liisa, try Snowglobe 1.x?
[09:22] Morgaine Dinova: It's not the 2.x base code that is a Fail, it's the 2.x UI. Put the 1.23 UI on the 2.x base code and you have Success.
[09:22] Latif Khalifa: make a SNOW- jira request
[09:23] Latif Khalifa: it has a good chance of getting looked at
[09:23] tx Oh: 2.0
[09:23] Talarus Luan: Well, I mentioned llTeleportAgent/Object. :P
[09:23] tx Oh: the innerts are not open sourced
[09:24] Latif Khalifa: llTeleportAgent() agent would solve a lot of ugly workarounds
[09:24] Latif Khalifa: with "sit" teleporters
[09:24] Talarus Luan: Andrew promised it years ago. :P
[09:24] Kelly Linden: Would it be ok to only work within a single sim?
[09:24] Liisa Runo: no
[09:24] Talarus Luan: Sure.. that's what most of the workarounds are used for anyway.
[09:24] Latif Khalifa: it would be an improvement, yes
[09:24] Rex Cronon: it is kind of useless to work within only one sim:(
[09:24] Kaluura Boa: Yeah...
[09:24] Kelly Linden: Is it ok to ask the person being teleported for permission first?
[09:25] Latif Khalifa: Kelly, no
[09:25] tx Oh: would be nice to be opensim compatible, with script tp's
[09:25] Liisa Runo: in one sim we can alreary warp the whole prim to the destination
[09:25] Talarus Luan: Sure, and autogrant if attachment or sitting
[09:25] Latif Khalifa: yeah
[09:25] Kelly Linden: Am I allowed to put anyone who walks by my parcel into an indefinite TP loop?
[09:25] Latif Khalifa: i'd make it like animation permission.
[09:25] Ardy Lay: I have been chatting an app/URI to them.
[09:25] Latif Khalifa: if i sit on the object it's auto granted
[09:25] Latif Khalifa: otherwise ask
[09:25] Techwolf Lupindo: Kelly, I think he was refering to autogrant on sit.
[09:26] Latif Khalifa: just like anim permission
[09:26] Kelly Linden: Ah. Then how is it better than the sit hack?
[09:26] Techwolf Lupindo: On sit, no popup needed.
[09:26] Liisa Runo: not allowd to put on TP loop, when the first TP happens, the agent is not sitting on the prim anymore, and the prim dont have perms to TP the agent anymore
[09:26] Rex Cronon: no way , autogrant
[09:26] Talarus Luan: You can get off having to support the hacks like PosJump and WarpPos indefinitely into the future.
[09:26] Latif Khalifa: sit hack moves object with warp pos, has trouble if you pass across the parcles with no object entry
[09:27] Kelly Linden: posjump and warppos are a different issue
[09:27] Latif Khalifa: that's what teleporters use
[09:27] Talarus Luan: I'm aware, but they are both ugly hacks.
[09:27] Kaluura Boa: TAlking about perms, having a way to revoke permissions would be good too
[09:27] Liisa Runo: i vote for autogrant, i can already log off people who sit on my prims, TP is not that dangerous
[09:27] tx Oh: i made a multi region sit tp (with physics on the borders)
[09:27] Rex Cronon: how can u log off somebody?
[09:27] Talarus Luan: ..and they will be used more and more, and you'll never be able to "fix" them without gnashing of teeth and massive bellyaching.
[09:27] Liisa Runo: not telling
[09:27] Latif Khalifa: so llTeleportAgent() even if it only worked within a sim would be a great improvement. and I would handle permissions in teh same way as anim permission does
[09:27] Techwolf Lupindo: llteleport would only need to check destation if allowed. If allowed, just go there, don't try to move across parcels.
[09:27] Rex Cronon: than it is a hack, lisa:)
[09:28] Morgaine Dinova: If you can log somebody off, you're a griefer.
[09:28] Kelly Linden: talarus we past that point on warppos and similar years ago
[09:28] Latif Khalifa: i think majority of warppos use is for intra sim teleporters
[09:28] Liisa Runo: im not a griefer if i know stuff, im a griefer if i use the stuff to grief
[09:28] Kelly Linden: in fact, twice we have made specific changes to the sim to unbreak those hacks
[09:28] Talarus Luan: Well, you can more easily sunset hacks than you can sunset canonical features.
[09:28] Rex Cronon: i didn't say u r a griefer. i said is a hack:)
[09:28] Talarus Luan: At some point, you can justify no longer supporting them.
[09:29] Talarus Luan: But not if there is no "official" replacement.
[09:29] Latif Khalifa: Talarus, if you killed warppos you'd kill a ton of stuff out there, but that's besides the point ;) What we need is llTeleportAgent()
[09:29] Liisa Runo: anyway, we could maybe have some kind of vote after deep discussion about the autogrant
[09:30] Kelly Linden: All right I have to head out now. Sorry for the short time.
[09:30] Morgaine Dinova: Liisa: if you know of a way to log people off, and have not told Lindens about it, then even if you;'re not a griefer yourself, you're supporting the griefer community.
[09:30] Talarus Luan: That or uncap llSetPos et al
[09:30] Liisa Runo: i have told LL about it
[09:30] tx Oh: byebye
[09:30] Latif Khalifa: Thanks for your time Kelly, and good luck at the dentist's :)
[09:30] Kelly Linden: Morgaine and Liisa, this isn't the place for that discussion.
[09:30] Kelly Linden: bye everyone.

Generated with SLog Wikifier