Simulator User Group/Transcripts/2012.10.23

From Second Life Wiki
Jump to: navigation, search

Simulator_User_Group

Prev 2012.10.19 Next 2012.10.26

List of Speakers

Baker Linden Davido Chrome
Kallista Destiny Kelly Linden
Lucia Nightfire Nalates Urriah
Qie Niangao Rex Cronon
Simon Linden Tankmaster Finesmith
Theresa Tennyson Whirly Fizzle
Yuzuru Jewell

Transcript

[12:02] Simon Linden: Let's see ... release news ... today there was an update to the main channel, as planned

[12:02] Baker Linden: I'm still investigating object rez lag, and how to go about fixing it.

[12:03] Simon Linden: The code that went out should have minimal impact -- it's a change that should make simulators run better on our new hardware

[12:04] Simon Linden: Tomorrow the RC channels will all get updated

[12:04] Tankmaster Finesmith: thats a good thing

[12:04] Simon Linden: and ...

[12:04] Simon Linden starts drum roll

[12:04] Kallista Destiny: Uh oh

[12:04] Tankmaster Finesmith: group list loading?

[12:04] Simon Linden: Magnum should get bug fixes, including Baker's work on large group queries

[12:04] Baker Linden: awwwwwwwwwwwwwwwwwwww yeahhhhhhhhhhhhhhhh

[12:04] Tankmaster Finesmith: \o/

[12:04] Kallista Destiny: YEAH!!!!!!

[12:05] Qie Niangao: yay!

[12:05] Rex Cronon: hello everybody

[12:05] Nalates Urriah: That will make many happy

[12:06] Simon Linden: There is a pretty good number of little fixes in that update, so we'll keep fingers crossed that it's solid and nothing acts badly

[12:07] Simon Linden: The release notes are here: http://community.secondlife.com/t5/Second-Life-Server/Deploys-for-the-week-of-2012-10-22/td-p/1709151

[12:07] Simon Linden: but there isn't any detail on that update

[12:07] Simon Linden: I'm not sure when that will get more information

[12:07] Kallista Destiny: Usually that gets updates Wed AM

[12:08] Simon Linden: Right, that's the regular schedule for putting updates into the RC channels

[12:08] Kallista Destiny: no the web page

[12:08] Simon Linden: ah, I get it

[12:08] Kallista Destiny chuckles

[12:09] Simon Linden: That's it for server news ... there's a project viewer some of you might be interested in at http://community.secondlife.com/t5/Featured-News/The-Communications-Hub-User-Interface-CHUI-Launches-Today/ba-p/1708805

[12:09] Simon Linden: Let's not dig into viewer design issues today, but I thought I'd point it out

[12:10] Kallista Destiny: I've seen an unusual number of avatars waking across the sky after crossing a Region boundry (two in an hour, where it's usually 1 a month)

[12:10] Simon Linden: So the invisible table is open for topics, questions or comments

[12:10] Tankmaster Finesmith: oh, do you by chance know where the sorce code is fo that viewer?

[12:10] Kallista Destiny: I thought I was back in 2006 got s moment

[12:10] Kallista Destiny: for a*

[12:11] Rex Cronon: why if the sim i try to login is down than i can't login? why don't i get sent to any running sim instead of saying that i can't login?

[12:11] Simon Linden: No, I don't know that Tankmaster.

[12:11] Tankmaster Finesmith: ok, thx

[12:12] Rex Cronon: ?

[12:12] Kallista Destiny chuckles, "Nothing like going to an Adult info hub."

[12:13] Rex Cronon: i assume there r checkes that don't allow everybody into an adult sim

[12:13] Lucia Nightfire: ha, lol

[12:13] Simon Linden: I forget when Oz's office hours are, but I think he'd know what the status is

[12:13] Kallista Destiny: I think it uses your highest preference

[12:13] Tankmaster Finesmith: mon @ 7:00, thurs @ 15:00

[12:14] Kallista Destiny: g,m,a

[12:14] Baker Linden: Tank, are the changes for large group management in a release build of Firestorm?

[12:14] Theresa Tennyson: Because you might have logged off NEKKID.

[12:14] Simon Linden: yeah, you really have to set your prefs for content level or you might get passed into some interesting areas

[12:14] Tankmaster Finesmith: no, baker

[12:14] Baker Linden: ok.

[12:14] Tankmaster Finesmith: just in our development repo

[12:14] Rex Cronon: if u don't happen so know a few sim names u r stuck outside unable to login

[12:14] Tankmaster Finesmith: our beta testers have a build of it

[12:14] Tankmaster Finesmith: but nothing more public

[12:15] Baker Linden: ok

[12:15] Simon Linden: Are those changes in Linden's beta viewer now?

[12:15] Tankmaster Finesmith: no

[12:15] Baker Linden: I'm unsure -- probably not

[12:15] Tankmaster Finesmith: everything on the LL dev is blocked untill they get the crash rate down

[12:15] Tankmaster Finesmith: its at 14% atm

[12:15] Kallista Destiny: Which is a good thing

[12:15] Tankmaster Finesmith: last release was 10%

[12:16] Kallista Destiny: not the crash rate, the block

[12:16] Simon Linden: Right, it jumped up.

[12:16] Tankmaster Finesmith: removing tcmalloc exposed a lot of bad code

[12:16] Simon Linden: Do you know if FS uses tcmalloc for memory management?

[12:16] Tankmaster Finesmith: it does, yes

[12:16] Tankmaster Finesmith: inharited from the ll code base

[12:17] Theresa Tennyson: Is Firestorm LAA?

[12:17] Tankmaster Finesmith: we are based on a build that still has it enabled

[12:17] Tankmaster Finesmith: yes, threasa

[12:17] Tankmaster Finesmith: and so is LL dev

[12:17] Tankmaster Finesmith: they pulled in my changeset to enable that

[12:17] Simon Linden: Right, that makes sense ... and yes, removing it seems to expose a bunch of problems. We really do need to track down and fix them, but doing that through releases and fixes isn't a good idea

[12:18] Tankmaster Finesmith: last official release has tcmalloc enabled, its the beta builds that are working to fix issues with it off

[12:19] Davido Chrome: What is tcmalloc?

[12:19] Tankmaster Finesmith: heya kelly

[12:19] Tankmaster Finesmith: tcmalloc is a third party lib that aligns memory to make it more effecent

[12:20] Tankmaster Finesmith: but it adds overhead, so it takes up more memory, and it has issues with non SL releated software like teamviewer and skydrive

[12:20] Kelly Linden: 'aligns memory' is probably the wrong term. It handles memory allocation. In some cases it is faster than the default allocators.

[12:20] Tankmaster Finesmith: which causes the viewer to crash when those other apps are installed

[12:20] Simon Linden: It works faster than the typical system memory allocator -- we use it on the simulators as well to get better performance

[12:20] Davido Chrome: Ah, does that mean that tcmalloc will work better when you have fixed the bad code you found?

[12:20] Tankmaster Finesmith: it aligns it to 16byte

[12:20] Tankmaster Finesmith: or bit?

[12:21] Simon Linden: Possibly, but more importantly finding those bugs will make the whole system work smoother

[12:21] Kallista Destiny is glad that Kelly is not all blinged up

[12:21] Kelly Linden: alignment is not a primary feature of tcmalloc that I'm aware of.

[12:21] Tankmaster Finesmith: no, katt, they are removing tcmalloc do the the crash issues it has with other software

[12:21] Simon Linden: It's a sub-block allocator -- it will grab a big block of memory from the system, than hand out small blocks

[12:21] Tankmaster Finesmith: probably not, kelly, but thats how LL is using it

[12:21] Nalates Urriah: I thought it was mostly for aloocating and freeing memory for use in a program.

[12:22] Nalates Urriah: allocating^

[12:22] Kallista Destiny: tcmalloc is a special case of the standard malloc

[12:22] Davido Chrome: Ah. And if there are other applications that do the same thing they get cranky and start to fight until they crash?

[12:22] Simon Linden: Alignment becomes an issue when dealing with other systems -- Havok, for example, wants a lot of the memory aligned properly for performance

[12:22] Tankmaster Finesmith: it was added becuase if its ability to easly align blocks, making it more effecent with SSE2 code

[12:22] Kelly Linden: I'm pretty sure alignment is not why we are using tcmalloc.

[12:23] Kelly Linden: We were using tcmalloc long before we added SSE2 optimizations.

[12:23] Kelly Linden: (at least on the server)

[12:23] Simon Linden: No, it's there for performance

[12:23] Tankmaster Finesmith: thats what i was told, but idk, obviously, as i wasnt appart of the dicussion to add it :P

[12:23] Tankmaster Finesmith: the viewer got it around the time of mesh code i think

[12:23] Simon Linden: There's been a bunch of work around alignment when removing TC malloc in the viewer

[12:24] Rex Cronon: if u r getting rid of tcmalloc, what r u using instead?

[12:24] Tankmaster Finesmith: doing it properly, hand code

[12:24] Tankmaster Finesmith: not relying on third party functions

[12:25] Kallista Destiny: I thought intel hardware was more or less immune to alignment issues.

[12:25] Simon Linden: The system allocators are a lot better than they used to be, so tcmalloc's advantage isn't as big as it was before

[12:25] Davido Chrome: I am on AMD processors?

[12:25] Kelly Linden: Rex: I think the default allocator. With recent compilers there isn't as much of a performance improvement from tcmalloc, and some of tcmalloc's behaviors can lead to memory bloat which may cause its own issues.

[12:25] Tankmaster Finesmith: while it may work, its more effecent aligned

[12:25] Rex Cronon: hand code for thread mem allocation? that has pottential of getting interesting:)

[12:27] Simon Linden: threads are always interesting

[12:27] Kallista Destiny: There is hardware that will make a program die, horribly, if operands are not properly aligned.

[12:27] Rex Cronon: u guys might be better off switching to either c# or java:)

[12:27] Kallista Destiny: I360 et seq

[12:27] Simon Linden: ah, no thanks, Rex :)

[12:27] Tankmaster Finesmith: java, hell no

[12:28] Kallista Destiny: Re the problem is the C# bloats more and runs slower

[12:28] Tankmaster Finesmith: c isnt as cross OS capible as c++

[12:28] Kelly Linden: On a different topic, LeTigre this last week had some library updates. One update of note was to the curl library that changed its behavior specifically around caching.

[12:28] Tankmaster Finesmith: c#*

[12:29] Kelly Linden: The effect was that all outgoing requests on non-letigre regions have a Pragma: no-cache header in them, because curl helpfully added this to all requests.

[12:29] Rex Cronon: how about ada:)

[12:29] Kelly Linden: The new version on LeTigre fixed this! Yay, except now llHTTPRequest may return cached results if the 3rd party web server has caching enabled.

[12:30] Kallista Destiny: How about PL\1 or 1401 Autocoder :( come-on.

[12:30] Whirly Fizzle: Oh, Im sure I saw a JIRA reporting that recently

[12:30] Tankmaster Finesmith: you did

[12:30] Meeter: Timecheck : User Group is half over

[12:31] Kelly Linden: There is a work around already. You can manually specify Pragma: no-cache by adding [HTTP_CUSTOM_HEADER, "Pragma", "no-cache"] to the llHTTPRequest

[12:31] Nalates Urriah: The 3rd party server, depending on how set up, should handle per request caching. Is there going to be an announcment of the change in the blog or forum?

[12:31] Nalates Urriah: That'll work

[12:31] Simon Linden: I believe the forum release notes are getting details of it

[12:32] Kelly Linden: However we are discussing how this could effect some existing systems that would suddenly start getting cached results, when they wouldn't before.

[12:32] Kelly Linden: Well, more precisely they *could* suddenly start getting cached results, depending on the configuration of the remote server.

[12:33] Davido Chrome: People would have to update their code to include no-cache instruction in their request string?

[12:33] Kelly Linden: In their LSL yes, not in the request string but in the parameters list

[12:34] Kelly Linden: We could modify the server version so that all llHTTPRequests add the "Pragma: no-cache" header but that seems like a step backwards in some ways since it takes the control away from the scripter

[12:35] Davido Chrome: Isn't that sent as string in LSL? I have been looking loosly at Database web interface at work. The one I have been in contact uses commands sent as XML strings.

[12:35] Kelly Linden: Headers are not a part of the request body.

[12:35] Kelly Linden: (I'm confused by your question)

[12:36] Rex Cronon: how old will the cache be?

[12:36] Davido Chrome: Kelly, can't you have it affect only new scripts or scripts edited after the change?

[12:36] Kelly Linden: Rex: depends on the configuration of the external server.

[12:36] Kelly Linden: AlveKatt, I can think of no nice way to do that.

[12:37] Davido Chrome: Kelly, ignore that question, I am an intern, I am in the progress of learning that stuff. So I am probably fumbling a bit...

[12:37] Kelly Linden: I could add handling for another parameter option, or a special "Pragma: cache" header that would allow you to skip the no-cache header.

[12:38] Rex Cronon: if the external web server uses flash wouldn't that fix it?

[12:38] Kelly Linden: That is too broad of a statement and relating to unrelated technologies. Some pages that use flash may be affected, some may not.

[12:38] Rex Cronon: i mean flash the stream

[12:38] Qie Niangao: Yeah, I guess to maintain backward compatibility, the "proper" fix is to add some llHTTPRequest() option for ALLOW_CACHE that defaults to FALSE, but... honestly I don't know how much content depends on the old no-cache behavior.

[12:39] Davido Chrome: A lot of feature additions seems to be hard to implenent due to backwards compatibility. Maybe implementing a system that lets you have old scripts run as usual but new ones use the new standards would be a good idea if it can be done.

[12:40] Kelly Linden: Davido: in cases like this that means a lot of special case branching and tacking on extra data to all scripts.

[12:40] Kelly Linden: it is .... ugly.

[12:40] Kelly Linden: Would we have multiple versions of the llHTTPRequest implementation that branched based on a set of flags attached to the script?

[12:41] Davido Chrome: New functions then, that do the same thing but with the new feature, then just mark the old stuff as deprecated and have the compiler just ignore attempts to implement it the old way. but still let old scripts use the old way?

[12:41] Kelly Linden: Qie: Systems that are most likely to be affected are those that frequently hit the exact same url and expect the data to change. Maybe they are getting a counter or checking on something's status.

[12:42] Kelly Linden: Davido: while that is possible, I think it is a bit extreme for this case.

[12:42] Theresa Tennyson: Is the change currently on LeTigre or is it going to happen tomorrow?

[12:42] Davido Chrome: Ok,

[12:42] Kelly Linden: It is on LeTigre now.

[12:43] Theresa Tennyson: If there were piles of dying breedables I think we would have heard about it by now then.

[12:43] Kelly Linden: Theresa: probably.

[12:43] Simon Linden: The current (new) behavior will not change with tomorrow's update

[12:43] Davido Chrome: Ok, you need to fix that. Those breedables need to die.

[12:44] Kallista Destiny: Or un fix it...

[12:44] Tankmaster Finesmith: lol davido

[12:46] Kelly Linden: If anyone uses or develops systems they think *might* be affected please give them a try on LeTigre this week and let me know. Or if you know others that might, encourage them to test on LeTigre this week. Thanks

[12:47] Davido Chrome: I guess those Breedables bring in a lot of doe for LL... With all those Lindens that get circulated by Breeding...

[12:47] Qie Niangao: Thanks Kelly. I wonder if it would be worth a post to the scripting forum

[12:47] Kelly Linden: As a side note apparently we forgot to document/announce HTTP_CUSTOM_HEADER when it shipped. So, um, yeah. That is out everwhere and now on the llHTTPRequest wiki page.

[12:48] Simon Linden: Are there any other questions or topics for today?

[12:49] Tankmaster Finesmith: guess not?

[12:49] Davido Chrome: Is there any documentation about Andrews interest lists somewhere?

[12:49] Simon Linden listens to the crickets

[12:49] Qie Niangao: Do we know any more about Toysoldier Thor's bug... BUG-355

[12:49] Simon Linden: No Alve, we haven't described that system

[12:49] Baker Linden: He doesn't right now

[12:50] Theresa Tennyson: Looks like Andrew updated things with the interest list yesterday, and I haven't found anything that it's broken yet.

[12:50] Qie Niangao: (aka MAINT-1682, about sim-crippling network performance, with moderate-to-high avatar counts)

[12:50] Davido Chrome: Because it's still a bit problematic when people float around in fast sports. You never really know where they are when they just entered your view, as it takes a second for the viewer to get their correct position.

[12:51] Simon Linden: I don't have any news on BUG-355

[12:51] Qie Niangao: okay. reports are that it's still happening, so... probably will keep getting asked about it.

[12:51] Nalates Urriah: Talk to Nyx, I think we saw that in Barrowdale yesterday.

[12:52] Davido Chrome: I had a motorcycle that wasn't mine spam me 100 sit position menues until I blocked it earlier today.

[12:52] Nalates Urriah: The region had 0.03 TD and 45 FPS... w/free script time and no PF...

[12:53] Tankmaster Finesmith: in relationship to the viewer discussion earlier, just wanted to bring attention to STORM-1914

[12:53] JIRA-helper: http://jira.secondlife.com/browse/STORM-1914

[#STORM-1914] Test Firestorm crash fixes - Second Life Bug Tracker

[12:53] Davido Chrome: Argh, MAINT-1682, permission violation...

[12:54] Tankmaster Finesmith: we are in the process of passing back a list of crash fixes from FS to the main LL viewer

[12:54] Simon Linden: the MAINT bugs are internal to Linden only

[12:54] Tankmaster Finesmith: yes, most MAINT jiras are closed to all but lindens

[12:55] Rex Cronon: top secret:)

[12:55] Lucia Nightfire: Anyone been having regions recently stop granting capabilities to people out of the blue? I've had it several times in the last few days, sometimes right after spam attacks...

[12:55] Meeter: Timecheck : User Group is almost over

[12:56] Lucia Nightfire: I wrote jira on a hopeful deterrant too https://jira.secondlife.com/browse/BUG-498

[12:56] Baker Linden: I've heard of it, Lucy. While I was investigating teleport timeouts, I ran into it

[12:56] Lucia Nightfire: for anyone that can access that

[12:56] Yuzuru Jewell: Permission Violation

[12:56] Whirly Fizzle: Theres been cases of that coming into Firestorm support too Lucia.

[12:57] Davido Chrome: Are we talking about permissions tied to group tag now?

[12:58] Lucia Nightfire: worst part abotu it is if its post spam, I can't return prims in the region nor enter it if I was outside it when caps stopped being granted

[12:58] Baker Linden: right.

[12:58] Tankmaster Finesmith: UG jiras are closed to only the reporter, lindens, code contributors, and a handful of other helpful individuals.

[12:59] Kallista Destiny: No about the region failing to pass the capabilities packet(s) to the viewer

[13:00] Rex Cronon: ok. i g2g. tc everybody and have a nice day:)

[13:00] Kelly Linden: back to work for me. o/

[13:00] Qie Niangao: Thanks Kelly

[13:00] Yuzuru Jewell: Thank you, Kelly.

[13:00] Kallista Destiny: thanks folks

[13:00] Davido Chrome: Thanks for today.

[13:00] Meeter: Thank you for coming to the Server User Group

[13:00] Baker Linden: thanks for coming everyone. See you friday!

[13:01] Nalates Urriah: Thx Lindens

[13:01] Simon Linden: Thanks everyone for coming today

[13:01] Qie Niangao: Thanks Simon.



Simulator_User_Group

Prev 2012.10.19 Next 2012.10.26