User:Andrew Linden/Office Hours/2009 04 07
< User:Andrew Linden | Office Hours
Jump to navigation
Jump to search
Revision as of 15:59, 9 April 2009 by Andrew Linden (talk | contribs) (formatted with Zero's perl script)
Transcript of Andrew Linden's office hours:
[10:57] | Second Life: | Your object 'Object' has been returned to your inventory lost and found folder by Simon Linden from parcel 'Linden Village - Simon Linden' at Denby 216, 60 due to parcel auto return. |
[10:57] | Moon Metty: | hi Andrew :) |
[10:57] | Roberto Salubrius: | Physics and Simulatior issues... I have not worked with physics for some time now |
[10:57] | Uni Ninetails: | hihi Andrew |
[10:57] | Roberto Salubrius: | hi Andrew |
[10:57] | Andrew Linden: | Hello |
[10:58] | Moon Metty: | it's quiet hehe |
[10:58] | Moon Metty: | ahh, there's Kitto :) |
[10:58] | Uni Ninetails: | Hmm wondering where the rest are |
[10:58] | Andrew Linden: | Simon won't be here today, he's on vacation. |
[10:59] | Moon Metty: | mmm, nice |
[10:59] | Kitto Flora: | Hello |
[10:59] | Andrew Linden: | I'm at home sick, but I can sit here and type. |
[10:59] | Uni Ninetails: | ack |
[10:59] | Moon Metty: | oh dear |
[10:59] | Moon Metty: | a flu? |
[10:59] | Kitto Flora: | You'll be even sicker soon :) |
[10:59] | Andrew Linden: | It is just a cold. |
[11:00] | Uni Ninetails: | Hey Squirrel |
[11:00] | Squirrel Wood: | Yellow |
[11:00] | Andrew Linden: | But I'm tired and cranky. Dunno how useful these office hours will be. |
[11:00] | Moon Metty: | nuts! |
[11:00] | Andrew Linden: | I also don't have much news. |
[11:00] | Kitto Flora: | San Francisco is as full of viruses as a Microsoft operating system. |
[11:00] | Moon Metty: | we'll see :) |
[11:00] | Roberto Salubrius: | hi kitto and squirrel |
[11:01] | Kitto Flora: | Hello, hello |
[11:01] | Roberto Salubrius: | I gotta say that win 7 looks promising |
[11:01] | Squirrel Wood: | yep |
[11:01] | Squirrel Wood: | win7 is what vista should have been |
[11:01] | Uni Ninetails: | mmm been playing with it for a while on the test bench |
[11:01] | Roberto Salubrius: | same |
[11:02] | Roberto Salubrius: | well I got build 7068 now |
[11:02] | Squirrel Wood: | I run it on a laptop with 1GB ram... running a private WoW shard there... just fine ^^ |
[11:02] | Moon Metty: | Andrew, this came by on the bug triage: https://jira.secondlife.com/browse/SVC-2741 |
[11:02] | Moon Metty: | i see it has been imported now |
[11:03] | Andrew Linden: | Yeah, I think Alexa Linden was asking me about that earlier this morning. |
[11:03] | Moon Metty: | hehe ok |
[11:03] | Moon Metty: | then it's ok |
[11:04] | Roberto Salubrius: | hi anto |
[11:04] | Andrew Linden: | It turns out we don't actually allow small prims to have very small collision shapes in the physics engine. |
[11:04] | Kitto Flora: | does bounding box data get generated by Havok4? |
[11:05] | Andrew Linden: | We actually clamp them to be about 0.2 or 0.1 meters on a side, and if they are very small we replace them with 0.2m^3 box or something |
[11:05] | Moon Metty: | hmm ok |
[11:05] | Andrew Linden: | the physics engine doesn't like really small objects -- they don't collide well with large objects, and Havok has a 0.1m "collision tolerance" anyway. |
[11:05] | Moon Metty: | yes, that's what i thought |
[11:05] | Andrew Linden: | Yeah, the bounding box reported by the script engine is using the collision shape's |
[11:06] | Roberto Salubrius: | well that can even be seen just by rendering BB |
[11:06] | Andrew Linden: | so we could report a more accurate bounding box... but would it be useful? |
[11:06] | Andrew Linden: | why would people use the smaller more accurate bounding box value? |
[11:06] | Moon Metty: | then i wonder why the jira was imported |
[11:06] | Squirrel Wood: | It was imported for clarification mathinks |
[11:07] | Squirrel Wood: | methnks even.. |
[11:07] | Kitto Flora: | because the result is not a 0.1M cube? |
[11:07] | Squirrel Wood: | and to explain that this is not a bug but expected functionality |
[11:07] | Andrew Linden: | I suggested that perhaps we could report a more accurate bounding box for the apparent shape if necessary. It shouldn't be too hard. |
[11:07] | Moon Metty: | alright :) |
[11:07] | Roberto Salubrius: | well that would do wonders for sculpts |
[11:07] | Squirrel Wood: | what about just reporting a bounding box that fits the shape ? |
[11:07] | Eata Kitty: | I can't see what you could do with it |
[11:08] | Andrew Linden: | I guess I'm wondering... what are people using llBoundingBox() for? |
[11:08] | Squirrel Wood: | and ommit the 0.1 boundary |
[11:08] | Squirrel Wood: | bullet collisions and such |
[11:08] | Roberto Salubrius: | Andrew personally I have used it for weapons |
[11:08] | Roberto Salubrius: | that's why the small BB is somewhat important for small bullets |
[11:09] | Andrew Linden: | would it be more useful to have a BB that approximated the visible shape (even if it were tiny)? |
[11:09] | Andrew Linden: | Or should it be a BB for the collision shape? |
[11:09] | Squirrel Wood: | I would prefer both actually |
[11:10] | Eata Kitty: | I don't follow why you would need to know the BB |
[11:10] | Andrew Linden: | Squirrel, are you saying you want very small collision shapes? (< 0.2m) |
[11:10] | Roberto Salubrius: | personally if I was asked to choose i would go for number 1 to aproximate it for the visible shape... the collision well we can live with the ones we have now... just a few errors now and then |
[11:11] | Roberto Salubrius: | however the other would be good to get better info about the sculpts large or small |
[11:11] | Roberto Salubrius: | but that's just my opinion |
[11:11] | Squirrel Wood: | What I mean is return both. the physical bounding box and the prims actual bounding box (which may be smaller) |
[11:12] | Andrew Linden: | I seem to recall having a conversation about the lower limits to llGetBoundingBox()... but I can't remember if it were in these office hours or not. |
[11:12] | Kitto Flora: | Why dont you simply use the prim's dimentions? |
[11:12] | Andrew Linden: | I think someone had a legit use for wanting to know a more accurate BB. |
[11:12] | Andrew Linden: | I think they were using the call to query the size of the object. |
[11:12] | Eata Kitty: | I imagine thats much easier for multiprim objects than trying to work it out via scale |
[11:13] | Andrew Linden: | Is there a llGetScale(). I can't remember. |
[11:13] | Eata Kitty: | Yes |
[11:13] | Eata Kitty: | But that only tells you one prim |
[11:13] | Andrew Linden: | ok. But llGetBoundingBox() works for the entire object, I think, not the individual prim. |
[11:13] | Moon Metty: | yeah |
[11:14] | Eata Kitty: | So better for overall size |
[11:14] | Squirrel Wood: | llGetBoundingBox() and llGetPhysicalBoundingBox() ? |
[11:14] | Andrew Linden: | Hrm... I don't have many utilities for computing the BB of an object independent of its true collision shape. |
[11:14] | Kitto Flora: | Then you should say you want the bounding box of an object, not a prim. |
[11:14] | Andrew Linden: | That is, we rely on the physics engine for that llGetBoundingBox() call. |
[11:15] | Andrew Linden: | Ok well it has been imported. I'll be seeing it soon on the internal triage list |
[11:15] | Andrew Linden: | which I usually process right after these Tuesday meetings, but not today. |
[11:16] | Kitto Flora: | Those that want to use this function: Are you trying to get the bounding box for a object that does not contain the quering script? |
[11:16] | Moon Metty: | nope, you should take it easy, and have a hot cup of tea :) |
[11:16] | Kitto Flora: | What next, the bad news or the worse news? |
[11:17] | Andrew Linden: | I think server-1.26 is deploying today... the initial experimental deployment. |
[11:17] | Andrew Linden: | On 500 or so regions I think. |
[11:17] | Uni Ninetails: | Isnt it wednesday? |
[11:17] | Moon Metty: | we tortured aditi yesterday |
[11:17] | Moon Metty: | all sims stayed up |
[11:17] | Uni Ninetails: | oop nm |
[11:17] | Andrew Linden: | Oh is it? Perhaps they are only deploying the new "central" code today. |
[11:18] | Uni Ninetails: | centrals today |
[11:18] | Andrew Linden: | I know that our internal #deploy IRC channel was busy this morning deploying something. |
[11:19] | Kitto Flora: | Andrew: In the 'colors' sims, where GSLR runs there are some regions with huge lag on border crossing. > 1 minute. |
[11:19] | Moon Metty: | one minute ??? |
[11:19] | Kitto Flora: | Yes |
[11:19] | Uni Ninetails: | yikes |
[11:19] | Kitto Flora: | like 60 seconds + |
[11:20] | Andrew Linden: | does the simulator totally block for that long? or is that just how long it takes to cross the boundary? |
[11:20] | Roberto Salubrius: | sorry what does GSLR means ? |
[11:20] | Kitto Flora: | I think just how long to cross |
[11:20] | Squirrel Wood: | worry not. I had a TP attempt take 5+ minutes yesterday... the bar stopped at some point while it was still trying to negotiate the tp.... after about five minutes it just decided to go back to normal without timeout or other error messages. |
[11:20] | Kitto Flora: | Greater Second Life Railway |
[11:20] | Roberto Salubrius: | thx Kitto |
[11:21] | Kitto Flora: | Squirrel: Thats an older problem |
[11:21] | Squirrel Wood: | and it was a tp from one point in the sim to another point in the same sim |
[11:21] | Kitto Flora: | Periwinkle to Purple is a problem |
[11:21] | Kitto Flora: | The other direction is ok |
[11:21] | Moon Metty: | Kitto, does the whole sim freeze during that time, or only your connection? |
[11:22] | Kitto Flora: | Squirrel: Thats a different problem |
[11:22] | Andrew Linden: | The 60+ only has to do with the border crossing? |
[11:22] | Kitto Flora: | Yes |
[11:22] | Roberto Salubrius: | I had an odd thing happen to me today I was on 1 sim then I had my home on the same sim |
[11:22] | Kitto Flora: | The sims are running well, have n mS of 'spare time' |
[11:22] | Andrew Linden: | Ok so the simulator continues to simulate everything else while that crossing is taking forever. |
[11:22] | Squirrel Wood: | does it still take that long when you wear nothing? no prims, no scripts... ? |
[11:22] | Roberto Salubrius: | so I did a control-alt-h to go to home |
[11:22] | Roberto Salubrius: | and I ended on another sim |
[11:22] | Kitto Flora: | What one sees depends on if one is riding the vehicle or watching it cross |
[11:23] | Kitto Flora: | Andrew: Yes, it seems so |
[11:23] | Kitto Flora: | Squirrel: Yes. Mostly I have no attachments |
[11:23] | Andrew Linden: | Ok, well I don't know what is wrong offhand. It is probably a server or network misconfig. |
[11:23] | Kitto Flora: | ANdrew: My thoughts too |
[11:24] | Andrew Linden: | Periwinkle to Purple lags you say? |
[11:24] | Kitto Flora: | As same sim code, other regions, theres only the usal ~5 - 10 second lag |
[11:24] | Kitto Flora: | Andrew : Yes |
[11:24] | Kitto Flora: | Periwinkle to either esat or west |
[11:24] | Kitto Flora: | east |
[11:24] | Andrew Linden: | Ok, I'll take note and suggest to the current "GridMonkey" that they look into it. |
[11:25] | Kitto Flora: | And theres other sims along the track that have the same problem |
[11:25] | Eata Kitty: | I was crossing regions on a physical boat, it was about 26 prims of sculpts, very slow to cross. Much slower than flying over on your own |
[11:25] | Kitto Flora: | Watching a railcar cross, from the desitinattion sim.... it appears to stop at the border for 1+ minute |
[11:26] | Kitto Flora: | Riding the car -your Av goes off to strange places for a minute |
[11:26] | Kitto Flora: | then reappears on the track just across the border |
[11:26] | Roberto Salubrius: | to <0,0,0< |
[11:26] | Kitto Flora: | some 20 seconds later the car appears - and carries on moving |
[11:26] | Roberto Salubrius: | s/</> |
[11:27] | Kitto Flora: | Amazingly - the railcar survives all this - it does not get 'lost' - though its rezzed from Inv |
[11:28] | Kitto Flora: | Meanwhile on the Wengen line... cars rezzed from rezzer - wel loose a couple or more cars a day. Vanished. |
[11:29] | Anto Infinity: | is this all happening recently Kitto |
[11:29] | Kitto Flora: | (A minutes silence for the demise of sl?) |
[11:29] | Eata Kitty raises a flag to half mast | |
[11:29] | Kitto Flora: | Well I was obserbing it over the last 24 hours |
[11:30] | Kitto Flora: | How long the Periwinkle problem has been apparent I am not sure - It was not like that say 6 months back. |
[11:30] | Anto Infinity: | as the silent partner in SLFS....I can tell you we have lost more ferries in the last 24 hours than in the last month |
[11:30] | Roberto Salubrius: | SLFS ? |
[11:30] | Kitto Flora: | croosing sim borders? |
[11:30] | Eata Kitty: | ferry system I guess :p |
[11:31] | Anto Infinity: | Susie's Little Ferry Service....runs little boats all over the grid....there is a monitor board and most of the ligths are now out |
[11:31] | Kitto Flora: | Do you know if they get returned to her? |
[11:31] | Andrew Linden: | Huh... this all started in the last 24 hours? |
[11:31] | Anto Infinity: | usually yes...... |
[11:32] | Andrew Linden: | Did it perhaps all start around 07:00 this morning? |
[11:32] | Kitto Flora: | MAybe 'gone off world' |
[11:32] | Kitto Flora: | I presume ferrys dont have border crossing control? |
[11:32] | Anto Infinity: | I couldn't say I only logged on a couple of hours ago.... |
[11:32] | Kitto Flora: | Andrew: Not what I am seeing |
[11:32] | Kitto Flora: | Andrew: I was seeing the problem yesterday |
[11:32] | Andrew Linden: | ok |
[11:33] | Kitto Flora: | Well Yesterday evening |
[11:33] | Kitto Flora: | say 5pm on |
[11:33] | Kitto Flora: | Before that : Cannot say, was not working on it there |
[11:34] | Andrew Linden: | ok well I'll ask around about it later |
[11:34] | Uni Ninetails: | can i throw a possible theory; Somthing not happy with migration to the new network? seeing as that was fairly recent? |
[11:34] | Andrew Linden: | maybe even go test it after this meeting (no triage meeting at noon) |
[11:34] | Anto Infinity: | Periwinkle to Purple is included in one of the routes..I will go take a look after the meeting |
[11:34] | Kitto Flora: | We can ride cars Andrew - its trippy :) |
[11:34] | Andrew Linden: | That is possible Uni, however that transition was complete a few weeks ago. |
[11:35] | Uni Ninetails: | yeah |
[11:35] | Kitto Flora: | Andrew: Theres a lotta messages in the Console when this happens, but I dont understand them all |
[11:35] | Uni Ninetails: | as mah old boss told me check dns first well chacke the cables before that but thats a given XD |
[11:36] | Kitto Flora: | Also a mass of 'packet out of sequence' when one recovers |
[11:37] | Andrew Linden: | We've got an interesting server imaging system. We use "systemimager" to update almost every server every day. |
[11:37] | Kitto Flora: | what gets 'updated'? |
[11:37] | Andrew Linden: | So what can happen is that someone makes an erroneous config change that gets pushed out with the new image |
[11:38] | Andrew Linden: | Kitto, almost everything gets updated. There are only a few directories that are left alone. |
[11:38] | Eata Kitty: | Is there a changelist for 1.26 |
[11:38] | Andrew Linden: | Well, it does a sync and updates any differences |
[11:38] | Andrew Linden: | and removes anything that is not on the image |
[11:38] | Kitto Flora: | Is this OS stuff? Or Simulater and support programs? |
[11:38] | Roberto Salubrius: | like Rsyncing a disk image ? |
[11:38] | Uni Ninetails: | http://wiki.secondlife.com/wiki/Release_Notes/Second_Life_Server/1.26 |
[11:39] | Andrew Linden: | This is OS stuff. yes, rsyncing a disk image. |
[11:39] | Eata Kitty: | Aha |
[11:39] | Anto Infinity: | and you do this update every 24 hours...seems like a lit of work |
[11:39] | Eata Kitty: | Thanks |
[11:39] | Andrew Linden: | I mention it as a vector for problems in SL. |
[11:39] | Kitto Flora: | Any way to see ther version of the OS update? |
[11:39] | Andrew Linden: | More than once a bad config has been pushed out in the daily update |
[11:39] | Roberto Salubrius: | well if it works like that a simple line could bring the grid down |
[11:40] | Andrew Linden: | Indeed Roberto, but we try to double-check the changes that go into the system image. |
[11:41] | Roberto Salubrius: | good that should be a human redundancy thing team 1 checks team 2 and viceversa |
[11:41] | Andrew Linden: | As a matter of fact, all servers get the same config files too, but they will execute their specific configs by pivoting on their MAC address. |
[11:42] | Andrew Linden: | No Kitto, there is no version update on the system image that I know of. |
[11:42] | Andrew Linden: | Yes, the update is run daily by a cron-job (automated script). |
[11:42] | Kitto Flora: | Then as residents we cannot correlate 'sl' problems with whats on the servers :( |
[11:43] | Roberto Salubrius: | well we can see unexpected behavior |
[11:43] | Roberto Salubrius: | then call it upon a linden ( If you are premium that is ) |
[11:43] | Andrew Linden: | Indeed, but the Periwinkle --> Purple problem could be a network config problem, or might be a server config problem. |
[11:43] | Andrew Linden: | Or it might be some other problem. Dunno. |
[11:43] | Uni Ninetails: | duff netwrk card, port switch rattycable etc etc |
[11:44] | Roberto Salubrius: | maybe the route... ( just guessing ideas ) |
[11:44] | Kitto Flora: | Is there any news on the 'lag spike' problem? Turns out that Moon Meytty has been chasing the same thing. |
[11:44] | Andrew Linden: | No news on that Kitto. |
[11:44] | Moon Metty: | yes, for a while now |
[11:45] | Moon Metty: | Neptune had long frezzes today ... |
[11:45] | Roberto Salubrius: | lag-spike random or on avi entry or exit ? |
[11:45] | Moon Metty: | 6 seconds with 2 agents on the region |
[11:45] | Kitto Flora: | Seems I will have to pack another notecard with products: More SL bugs I cannot do anything about' |
[11:46] | Moon Metty: | i use a script that checks the realtime timestamp |
[11:46] | Andrew Linden: | I will try to investigate that a bit right after this meeting Kitto. I still haven't gone to that region and looked at the in-world stats (CTRL+SHIFT+1) |
[11:46] | Moon Metty: | it the script is inactive for longer that 0.5 seconds it gives a warning |
[11:47] | Moon Metty: | the statistics window is tricky, as it doesn't get updates during a freeze |
[11:47] | Kitto Flora: | The Periwinkle problem is worse... in that is severe. |
[11:47] | Andrew Linden: | Moon, think in Kitto's case the server isn't blocking... but the script is not getting runtime. |
[11:47] | Kitto Flora: | Also may be easier to fix - if its a config problem |
[11:47] | Moon Metty: | yes, Kitto's case is different |
[11:48] | Moon Metty: | i'm not typing :) |
[11:48] | Kitto Flora: | I have been looking only at script (not) running time |
[11:48] | Andrew Linden: | Moon, what region are you noticing the spiky lag? |
[11:48] | Moon Metty: | well, a lot of sims |
[11:48] | Andrew Linden: | My guess would be some scripted content bursting some action. |
[11:49] | Moon Metty: | only very empty sims suffer less |
[11:49] | Uni Ninetails: | everytime i hear spiky lag i think someones running temp ressers |
[11:49] | Andrew Linden: | Oh, perhaps you're noticing the lag spikes from login/TP. |
[11:49] | Andrew Linden: | Yes, temp rezzers could do it. |
[11:49] | Moon Metty: | login, tp and border crossings are always good for a short freeze |
[11:49] | Roberto Salubrius: | we monitor neptune mostly... and login/tp is not related to the long ones |
[11:50] | Roberto Salubrius: | I know I have been there watching also :) |
[11:50] | Moon Metty: | yes, the long freezes have a different source |
[11:50] | Roberto Salubrius: | and well we have worked on neptune for what moon 2 years now ? |
[11:50] | Roberto Salubrius: | we know that sim |
[11:50] | Moon Metty: | yes |
[11:51] | Kitto Flora: | More busy sims do have more lag events |
[11:51] | Roberto Salubrius: | and the timings are random way random... |
[11:51] | Kitto Flora: | several per minute is bad |
[11:51] | Kitto Flora: | Light load sims its down at a few a day |
[11:52] | Moon Metty: | default |
float old_time = (float)llGetSubString(llGetTimestamp(), 17, 26); | ||
{ /52041 calculate the time since the last loop | ||
float now = (float)llGetSubString(llGetTimestamp(), 17, 26); | ||
float elapsed = now - old_time; | ||
if (elapsed < 0) elapsed += 60.0; | ||
old_time = now; | ||
// warn if it was more than 0.5 seconds ago | ||
if (elapsed > 0.5) llOwnerSay(llGetSubString((string)elapsed, 0 , 3) + " second lagspike"); | ||
llSleep(0.001); // enabling this line makes this script less accurate, but more server-friendly | ||
[11:52] | Roberto Salubrius: | well I guess moon if you send the data to a PHP offworld we could try to coorelate it and see if we can make any sort of statistical anaysis on the time to see if it even has a fractal distribution ( worst case scenario ) |
[11:52] | Moon Metty: | a very simple script |
[11:53] | Roberto Salubrius: | andrew before you wrap can I ask you something in private ? |
[11:53] | Andrew Linden: | Sure Roberto. |
[11:53] | Roberto Salubrius: | thx |
[11:54] | Andrew Linden: | I'm going to be on vacation next week, Tues and Thurs both |
[11:54] | Moon Metty: | hrmmm, a gesture interfered with the script |
[11:54] | Andrew Linden: | I think Simon will be attending these office hours |
[11:54] | Moon Metty: | lol |
[11:54] | Andrew Linden: | I'll try to verify that. |
[11:54] | Uni Ninetails: | shift swap ^^ |
[11:55] | Kitto Flora: | Simon will have to talk :) |
[11:55] | Moon Metty: | yup |
[11:57] | Uni Ninetails: | Andrew just a quickie... cause my Estate owner keeps asking and i want to know personally as a sim owner. Im assuming Windlight control on the simside, for esate wners and managers via Region/estate is on hold for the forseeable future? |
[11:59] | Andrew Linden: | Uni I think that idea has been kicked around the lab |
[11:59] | Andrew Linden: | dunno what the state of it is I'll have to ask some of the other devs |
[11:59] | Uni Ninetails: | Just seems a bit of a waste.... |
[11:59] | Andrew Linden: | I think that it is something that we want to do. |
[11:59] | Uni Ninetails: | Its a big thing for me, but i know its not high on the list of prioritues atm |
[12:00] | Uni Ninetails: | *priorities |
[12:00] | Andrew Linden: | one thing about that uni... it requires both simulator and viewer work to function properly |
[12:00] | Andrew Linden: | which means it can take a long time to reach true fruition... the viewer is taking a long time to get updated versions |
[12:01] | Uni Ninetails: | aye i had expected some harmony between both sides of the fence to be required. Hopefully in the future :) dont leave it too long though XD |
[12:02] | Andrew Linden: | I'll make a note to ask so I can answer that question next time I see it. |
[12:02] | Uni Ninetails: | Thanks i appreciate it |
[12:02] | Andrew Linden: | Ok, I'm going to go look at Periwinkle, and then a few other regions. |
[12:02] | Andrew Linden: | Thanks for coming. |
[12:02] | Kitto Flora: | TY :) |
[12:02] | Moon Metty: | thank you you Andrew :) |
[12:02] | Uni Ninetails: | Ok dont work to hard :) |
[12:02] | Moon Metty: | get better soon! |
[12:02] | Uni Ninetails: | yus! |
[12:03] | Uni Ninetails: | Kay time to go Have a good day all! |