Simulator User Group/Transcripts/2012.07.13

From Second Life Wiki
Jump to navigation Jump to search


Prev 2012.07.10 Next 2012.07.17

List of Speakers

Andrew Linden Ardy Lay
Ashiri Sands Baxter Aubin
Draconis Neurocam Kallista Destiny
Kelly Linden Nalates Urriah
Rex Cronon Tankmaster Finesmith


[16:00] Meeter: Welcome to the Server User Group

[16:01] Nalates Urriah: Seems a little deserted. Did I miss a cancelation notice?

[16:01] Ardy Lay: Good question

[16:02] Nalates Urriah: I'd say... no cancelation on the wiki

[16:02] Ardy Lay: I hate hard disks

[16:02] Draconis Neurocam: i hate phones

[16:02] Ardy Lay: Just had a new 3TB drive fail after I got it loaded up.

[16:02] Rex Cronon: hello everybody

[16:02] Nalates Urriah: That would suck Ardy

[16:02] Ashiri Sands: Total fail?

[16:02] Andrew Linden: Hello.

[16:03] Nalates Urriah: Hey Andrew

[16:03] Ashiri Sands: hi Andrew

[16:03] Andrew Linden: Simon won't be able to make it today.

[16:03] Rex Cronon: hi andrew

[16:03] Rex Cronon: AFK...............................

[16:03] Rex Cronon: i need to go afk for a few minutes

[16:03] Ardy Lay: Is he chained to a computer with a busted hard drive? That would be my day.

[16:04] Ardy Lay: I live on thanks to a 120GB SSD

[16:04] Nalates Urriah: Which brand drive failed?

[16:04] Ardy Lay: Guess

[16:05] Andrew Linden: odd, my chat history doesn't show everybody's name in chat history

[16:05] Nalates Urriah: Maxtor

[16:05] Ardy Lay: Maxtor still exists?

[16:05] Kallista Destiny: Really

[16:05] Nalates Urriah: I haven't checked lately

[16:05] Andrew Linden: but the names show up when I close the chat history floater

[16:06] Ardy Lay: That's interesting

[16:06] Ardy Lay: Sometimes the names show up late and get displayed twice on a line when they do.

[16:06] Ashiri Sands: latest release viewer or beta?

[16:06] Andrew Linden: this isn't the latest...

[16:07] Andrew Linden: Second Life 3.3.3 (260300)

[16:07] Andrew Linden: on linux

[16:07] Ardy Lay: THey change from "Loading ..." to "Andrew LindenAndrew Linden:", for example.

[16:07] Andrew Linden: oh well, I'll ask someone for a copy of the transcripts if they never show up by the end of the hour

[16:08] Andrew Linden: Oh yeah, and Baker won't be coming. He's doing touristy stuff around San Francisco, I think.

[16:08] Kallista Destiny: well Friday the 13th is on a Friday this month

[16:09] Andrew Linden: as for news...

[16:09] Andrew Linden: I've just been looking into the intrestlist stuff still, trying to get more info

[16:10] Andrew Linden: in preparation for the big meeting where we'll try to decide on strategy and architecture design

[16:10] Andrew Linden: the only other thing I did of interest was look into some inventory corruption issues and how to help with that

[16:11] Kallista Destiny: Is that the inventory and folders ourside of inventory?

[16:11] Andrew Linden: one of the tools that used to be used for fixing inventory has some bugs in it

[16:11] Kallista Destiny: /e see Telly arrive and puts on her sunglasses

[16:12] Kelly Linden: Who is telly?

[16:12] Ardy Lay: I know a coupl of people that have some inventory items that have become the wrong type. That sort of thing? I know another whose inventory is only partially loading, suddenly.

[16:12] Rex Cronon: back

[16:12] Andrew Linden: not the perl tool that was used by some of our internal web pages

[16:12] Andrew Linden: so I'm going to try to fix that in my copious spare time

[16:12] Kallista Destiny: Ahhhhhh....

[16:13] Ardy Lay: THey would LOVE to get their inventories fixed up so they can use them again.

[16:13] Draconis Neurocam: you can't just replace the usage of perl with python, rather than redo the perl one?

[16:13] Andrew Linden: Unfortunately not Draconis, at least not without more overhaul than I can do on the side.

[16:13] Draconis Neurocam: ah

[16:14] Andrew Linden: In any case, I'll be looking for some examples of corrupted inventory that I can look at and test my fixes against

[16:14] Andrew Linden: of course, I'll be copying such inventories to the beta grid (aditi) first for test

[16:14] Ardy Lay: How yould you like those examples delivered?

[16:14] Andrew Linden: I just need to an account name and some sort of consent

[16:14] Andrew Linden: that is, an IM from the individual

[16:14] Ardy Lay: Have them IM you?

[16:15] Andrew Linden: yes

[16:15] Ardy Lay: Okay

[16:15] Andrew Linden: I've got a command line interface tool that actually works

[16:15] Rex Cronon: how can u find out if any of your inventory items r corrupted or not? u rezz them, one by one?

[16:15] Nalates Urriah: You are probably going to be picky and require consent come form the inventory owner... OTHERWISE I'll be happy to give my concent for THEIR inventory.:)

[16:16] Andrew Linden: so if anyone needs their inventory fixed immediately they can send me an IM (for the next week or so)

[16:16] Ardy Lay: Thanks Andrew.

[16:16] Andrew Linden: since I'm interested in looking at the types of inventory problems that have occured

[16:16] Andrew Linden: and would like to find out if my CLI script solves all of them or not

[16:17] Andrew Linden: er... not "inventory problems" in general but things that apper to be inventory corruption

[16:17] Ardy Lay: Right

[16:17] Andrew Linden: that is, large inventories will usually not load correctly in a 1.23 viewer

[16:17] Andrew Linden: so I can't fix that problem

[16:17] Ardy Lay: When you get that automatic inventory organizer ready let us know. ;-)

[16:17] Andrew Linden: but things like undeletable folders that show up at the topmost level

[16:17] Kallista Destiny: For what value of large?

[16:18] Kallista Destiny: 50K 100K 200K 1M?

[16:18] Andrew Linden: or maybe even folders that went missing and you're pretty sure you didn't delete

[16:18] Andrew Linden: I know that an inventory with 25k folders will have trouble showing up in 1.23

[16:19] Kallista Destiny: Andrew you might take a look at SVC-7653

[16:19] JIRA-helper:

[ "LL entory repair tools broken; avatars can be permanently disabled by previously fixable inventory folder glitches." CHOP -839 *AngusGraham Ceawlin

[16:19] Andrew Linden: really what happens with 1.23 is...

[16:19] Ardy Lay: How about wearables that the viewer says are objects but cannot be rezzed?

[16:19] Andrew Linden: the more items/folders you have there the less likely you're going to see them all show up

[16:19] Andrew Linden: some will show, but some may appear to be missing, although they are not lost

[16:20] Andrew Linden: and the more packet loss you're experiencing the more things will probably go missing

[16:20] Nalates Urriah: IS the packet loss thng only for 1.23?

[16:20] Andrew Linden: Hrm... wearables that can't be rezzed...

[16:20] Andrew Linden: sure, I'll take a look, although that could be an asset problem

[16:21] Ardy Lay: I think somebody told me one of their shapes is now labeled as a shirt and it doesn't work either way.

[16:21] Kallista Destiny: This is the inventory corruption that gets fixed by those scripts, I think

[16:21] Andrew Linden: Ok thanks for that pointer at SVC-7653

[16:22] Kallista Destiny: nods nods Our Whirlly Fizz has been dealing with that a lot, as you can see

[16:22] Andrew Linden: Pathfinding went into RC this week, on Magnum

[16:22] Ardy Lay: THere are probably orphaned items in there owned by me. PRetty sure I managed to nullify some object parent UUID fields some time ago.

[16:22] Andrew Linden: So far it looks pretty good, but a few bugs have been reported

[16:23] Andrew Linden: the worst one that I know about is that some ground vehicles are broken

[16:23] Andrew Linden: Falcon and Maestro looked into some of those reports

[16:23] Andrew Linden: they had trouble reproducing but eventually did so on two vehicles I think

[16:24] Andrew Linden: upon closer examination they found some... design problems with the vehicles

[16:25] Andrew Linden: one had all the child prims set to "NO SHAPE" and the root prim was a hollow box, collapsed very thin in the Z axis

[16:25] Andrew Linden: so the collision shape was a concave mesh

[16:25] Ardy Lay: Empirical design yields dubious results when no sound engineering practices are available.

[16:25] Andrew Linden: I think that one also had some extra "gravity" that was pushing the vehicle very hard into the ground

[16:26] Andrew Linden: the result was that the vehicle was partially falling through the terrain

[16:26] Andrew Linden: which uses a "mesh shape" instead of a "heightfield" for its collision representation in the physics engine

[16:26] Andrew Linden: which makes for cheaper collisions and much better performance when computing the navigation mesh (needed for the pathfinding features)

[16:27] Andrew Linden: but can cause snagging and strange collision problems from partial penetrations

[16:27] Andrew Linden: so anyway, this one vehicle was partially penetrating because it was so thin and was being pushed hard into the ground

[16:28] Andrew Linden: another vehicle was very sensitive to the number of collision points, and also had a very thin box for its collision shape

[16:28] Ardy Lay: Can we see physics shapes in the viewer metadata displays?

[16:28] Ashiri Sands: Sound like the designer was trying to make up for the initial design flaw

[16:28] Andrew Linden: it is not clear if we'll be able to recover these vehicles as they are designed

[16:28] Rex Cronon: i think we need a automatic vehicle tester. u put vehicle on it and it tells u if something wrong with it

[16:29] Ardy Lay: I see a menu item but it immediately crashes viewer.

[16:29] Andrew Linden: Ardy, I'm not sure... I think that viewer option is only for the "static" objects in the world

[16:29] Andrew Linden: that is a new navmesh feature, right?

[16:29] Ardy Lay: Not in this viewer

[16:29] Andrew Linden: er... part of the pathfinding viewer

[16:30] Ardy Lay: I am not using PF viewer, I hav viewer-development

[16:30] Andrew Linden: oh. huh...

[16:30] Andrew Linden: no I'm not familiar with that option

[16:30] Ardy Lay: Develop > Render Metadata > Physics Shapes - crash to desktop

[16:30] Meeter: Timecheck : User Group is half over

[16:31] Ardy Lay: Heh, maybe it just divides by zero. ;-)

[16:31] Andrew Linden: but I think there is some object collision information that may be visible for static objects in the pathfinding viewer

[16:31] Andrew Linden: if so, you could try setting a dynamic object static, and make it part of the navmesh, and then trigger a navmesh rebuld...

[16:32] Andrew Linden: and then you could download the navmesh and collision shape data and maybe be able to see that object's collision shape

[16:32] Ardy Lay: Heh, last time I did that the region crashed. I'll assume that is fixed now.

[16:32] Kallista Destiny: Oh, Andrew there was a report that there is an issue wihere some % of mesh objects simply fail to cross out of PF regions let me see if I can find it.

[16:32] Andrew Linden: Yeah, we think we did a pretty good job of fixing the server crash bugs in the pathfinding servers.

[16:33] Tankmaster Finesmith: famous last words :D

[16:33] Andrew Linden: Actually, I think that failure of mesh region crossings is a know (expected) issue

[16:34] Andrew Linden: the reason being that we use some Havok black-magic (optimized data I/O) for transmitting mesh collision info across region boundaries

[16:34] Ardy Lay: I guess I am gonna have to give up on my personal custom viewer. I can't afford to keep buying licenses for the libraries that require it. I want to use PF and upload mesh. KDU was expensive enough.

[16:34] Andrew Linden: and there is sometimes a version incompatibility when the physics engine gets updated

[16:35] Andrew Linden: as was the case for the pathfinding project

[16:35] Andrew Linden: such mesh objects can cross from old physics engines to new, but not new to old

[16:35] Kallista Destiny: Yesh but the % was rather low (15% I believe))

[16:36] Andrew Linden: So, some mesh objects would cross from new to old, but not all?

[16:37] Andrew Linden: I guess the table is open.

[16:38] Andrew Linden: Of course Kelly, if you have any news go ahead.

[16:38] Kallista Destiny: yes here is the notice Seems that Mesh Arrows have under some special Circumstances Problems to cross Regionboundaries. We tested it with Mesh and Sculpted Arrows, in 1 of 15 cases the Mesh Arrows failed, the Sculpted didnt. Please inform your people to switch to sculpted Arrows when they fight over Regionboundaries.


[16:38] Kelly Linden: I just finished up some work so llGetInventoryCount and llGetInventoryName were significantly faster, especially with large object inventories.

[16:39] Kelly Linden: At the same time I'm looking at a cap on the number of items that can go in an object's inventory - so if you have feedback on that I'd appreciate it.

[16:40] Rex Cronon: there r objects with over 2k in their inventory. what will happen with them if u restrict max inv size?

[16:40] Kelly Linden: Legacy objects will retain their inventories but no new inventory can be added to them.

[16:41] Rex Cronon: we need at least 2048 items per object inventory:)

[16:41] Kelly Linden: If you have MAX + 100 then you can rez, take, copy etc but you can't add inventory to it until it is under MAX

[16:41] Kelly Linden: Okay Rex.

[16:41] Kelly Linden: (why?)

[16:42] Ardy Lay: Can you make the limit for script different from the limit for other task inventory types?

[16:42] Andrew Linden: If the object is no-modify can the inv items be removed? I can't remember if that is allowed or not.

[16:42] Kelly Linden: Ardy: already done. The limit for scripts already exists and is 1k per entire linked set

[16:42] Ardy Lay: Cool

[16:42] Kelly Linden: Andrew: no, I don't think so.

[16:42] Ardy Lay: I was thinking about all the texture viewers out there with large numbers of textures in them.

[16:42] Rex Cronon: look. a linked object can be of max 256 objects, each one needs at least one script, one texture, one sound, possibly one sound file

[16:43] Andrew Linden: if you can't take an overfull no-modify object... can you delete it?

[16:43] Kelly Linden: You CAN take an overfull object

[16:43] Tankmaster Finesmith: why do you want to put a cap into the amount of items stored inside a prim?

[16:43] Kelly Linden: You CAN take, you CAN rez, you CAN copy overful objects.

[16:43] Rex Cronon: if u add them up u get to around 2048 things per prim

[16:44] Kelly Linden: you just can't add more inventory to them.

[16:44] Kelly Linden: Rex: that seems backwards - you mean 2048 per object because one of each at least on each prim?

[16:44] Andrew Linden: Ah I see... I read back. You can't copy or add inventory *items* but you can copy or take the object.

[16:44] Kelly Linden: Andre correct.

[16:44] Rex Cronon: i mean at least the root prim needs that many

[16:45] Kelly Linden: Ah, okay.

[16:45] Rex Cronon: otherwise u have to put things in each one of the prims and is pain to edit them

[16:46] Rex Cronon: and i do mean each prim in the linked object

[16:46] Ardy Lay: A thousand scripts per linkset still sounds like a hell of a lot to me.

[16:46] Ardy Lay: Probably won't hit many legitimate products

[16:46] Rex Cronon: ardy. usually 90% of them r sleeping

[16:46] Kelly Linden: I'm looking at a limit for each prim, since that is how they are stored in the sim.

[16:46] Ardy Lay: Rex, sleeping doesn't unload the host

[16:46] Kelly Linden: 1k scripts is rare.

[16:46] Andrew Linden: well scripts are already specifically capped on a per-object basis (during rez of that object)

[16:47] Tankmaster Finesmith: you dont need a script in each prim to do things like changing color or size

[16:47] Ardy Lay: Right

[16:47] Rex Cronon: in the root u need at least on script per prim

[16:47] Rex Cronon: if u try to animate the object

[16:48] Ardy Lay: Demonstrate this, Rex>

[16:48] Ardy Lay: You have such an animated object to show us?

[16:49] Kelly Linden: was it rex who had the animating skeleton?

[16:49] Ardy Lay: The animated object I am familiar with use a single script but I don't own any of them.

[16:49] Ardy Lay: -object +objects

[16:49] Kelly Linden: someone had a very fluid animating stick figure skeleton

[16:49] Rex Cronon: i did some custom thing, and i need one script per prim

[16:50] Ardy Lay: Kelly, what was the update traffic from that like?

[16:50] Rex Cronon: and sometimes u want to change texture, sound, and particles for each prim in the linked set

[16:50] Kelly Linden: Ardy: I don't recall.

[16:50] Draconis Neurocam: it was fadeout razorfen

[16:50] Ashiri Sands: I'm still wondering what the limits for textures would be

[16:50] Kelly Linden: probably significant

[16:50] Draconis Neurocam: who made the skeleton

[16:50] Kelly Linden: I don't recall.

[16:50] Andrew Linden: I saw an animated medieval guard in armor that looked fantastic... dunno how many textures it was using...

[16:50] Andrew Linden: but it was a big UDP object update storm

[16:51] Andrew Linden: lots of packets being sent to move each little piece

[16:51] Rex Cronon: sadly there are also "bullets" that have thounsands of scripts used for spamming:(

[16:52] Ashiri Sands: not just spamming

[16:52] Rex Cronon: self-replicators too

[16:52] Kelly Linden: bullets don't have thousands of scripts if they are a single object any more. The rez will fail.

[16:52] Ardy Lay: I have seen one animated tail generate enough update traffic to cause distant clients to time out. I don't understand that behavior but when I ejected the tail's wearer the problem ended and update traffic coming to me dropped from over 3megabytes per second to less than 120 kilobytes per second.

[16:52] Kelly Linden: (been this way for months I believe) they may have hundreds.

[16:53] Rex Cronon: u right kelly

[16:53] Kallista Destiny: Like thi one?

[16:54] Rex Cronon: i haven't see them in months

[16:54] Kallista Destiny: this

[16:54] Ardy Lay: Oh, somebody dropped an object called Diablo that downed a region. Looked like ti had a lot of scripts in it.

[16:54] Ardy Lay: Maybe this 1K script limit will stop that.

[16:55] Kelly Linden: The 1k script limit is old.

[16:55] Ardy Lay: I say it downed a region, maybe it just tied up I/O long enough for clients to disconnect.

[16:55] Meeter: Timecheck : User Group is almost over

[16:55] Ardy Lay: Ah, I don't know then.

[16:56] Ardy Lay: This was about 2 weeks ago.

[16:56] Ardy Lay: I think I was IMing Andrew about chat lag when it happened.

[16:57] Kelly Linden: As always the Maint Server branches include some new anti-griefing fixes. So, maybe that will help. :)

[16:57] Ardy Lay: Okay

[16:57] Tankmaster Finesmith: good to hear

[16:57] Ashiri Sands: Always playing catch-up it seems ≪⁕⁔⁕≫

[16:57] Ardy Lay: You need to red-pill those guys.

[16:57] Kelly Linden: On BlueSteel now. Got another in internal QA and writing code for a third. Trying to keep that pipe flowing.

[16:57] Kelly Linden: :)

[16:58] Ardy Lay: Glad to hear that, Kelly

[16:58] Kelly Linden: Have a good weekend.

[16:58] Andrew Linden: I need some help with the transcripts today

[16:58] Rex Cronon: tc kelly

[16:59] Andrew Linden: usernames are still not showing up in my chat history

[16:59] Tankmaster Finesmith: tc kelly, andrew, simon

[16:59] Ardy Lay: OKay Andrew. Want me to put it in a notecard or email?

[16:59] Ashiri Sands: Just display names showing?

[16:59] Andrew Linden: so I would appreciate it if someone could save the transcripts and send them to me

[16:59] Andrew Linden: via notecard or email

[16:59] Ardy Lay: I will do that, Andrew.

[16:59] Andrew Linden: Yeah, so everyone would be called "Loading..."

[17:00] Andrew Linden: Thanks Ardy

[17:00] Rex Cronon: loading... rex

[17:00] Rex Cronon: lol

[17:00] Rex Cronon: now i am downloadable:)

[17:00] Andrew Linden: That would be a very hard transcript to make sense of...

[17:00] Meeter: Thank you for coming to the Server User Group

[17:00] Kallista Destiny: Andrew jut a moment I'll have a transcript fo you


Prev 2012.07.10 Next 2012.07.17