User:Andrew Linden/Office Hours/2009 01 27

Transcript of Andrew Linden's office hours:

[11:02]	Qie Niangao:	greetings, Andrew
[11:02]	Andrew Linden:	Hi Qie
[11:02]	Kitto Flora:	Hello Andrew, Qie
[11:03]	Qie Niangao:	hi, Kitto
[11:04]	Qie Niangao:	Hello, Simon
[11:04]	Rex Cronon:	hello everybody
[11:04]	Kitto Flora:	Hi Simon, Rex
[11:04]	Qie Niangao:	Hi, Rex
[11:04]	Rex Cronon:	hiii
[11:05]	Simon Linden:	Hi everyone
[11:05]	Andrew Linden:	ok, so let's start with announcements.
[11:05]	TeaCup whispers:	Tea Time!!
[11:06]	Andrew Linden:	server-1.25.4 has been fully deployed last week. I think there is a 1.25.5 patch planned for the not too distant future.
[11:06]	Andrew Linden:	There are a few MONO/script bug fixes and also some "save the database" changes that will be in that patch.
[11:07]	Andrew Linden:	We managed to reduce our database query rate by 20% over the weekend, and also made some other optimization changes to the db that should give us some breathing room for the next few months.
[11:08]	Andrew Linden:	There is an internal team being put together to work on helping that part of the system scale so we hopefully don't have another fire at the end of those months.
[11:09]	Andrew Linden:	server-1.26 has branched, which means any bug fixes I happen to make (as well as most of the ones recently fixed but not yet passed QA) will not be in 1.26, but will go into server-1.27
[11:09]	Andrew Linden:	ok, that's all I had
[11:09]	Andrew Linden:	you got anything Simon?
[11:09]	Simon Linden:	No news here
[11:10]	Andrew Linden:	Ok, anybody got a topic to talk about?
[11:10]	Kitto Flora:	Good-ish news:
[11:10]	Kitto Flora:	Since 1.25 arrived Railccars have not gone missing
[11:10]	Kitto Flora:	But its short period
[11:10]	Andrew Linden:	Kitto, what was the disappearance rate?
[11:10]	Arawn Spitteler thinks that sounds promising.
[11:10]	Andrew Linden:	two or three a week?
[11:10]	Kitto Flora:	Yes
[11:10]	Simon Linden:	That's good to hear ... definitely ping us again if you continue to see it happening
[11:11]	Kitto Flora:	At least 1 a week average
[11:11]	Kitto Flora:	I ran 2 cars overnight
[11:11]	Kitto Flora:	They are both still running
[11:11]	Andrew Linden:	Ok, so it might be statistically significant. We'll know more next week I expect.
[11:11]	Kitto Flora:	Yes
[11:11]	Kitto Flora:	BAd news:
[11:11]	Kitto Flora:	The texture rez last frame problem is still there
[11:11]	Kitto Flora:	Maybe last 2 frames even
[11:12]	Andrew Linden:	? I think I'm unfamiliar with that problem.
[11:12]	Simon Linden:	yes, I've been trying forever to get that into the main branch of our code
[11:12]	Andrew Linden:	Got a jira number?
[11:12]	Kitto Flora:	Simon knows about it
[11:12]	Andrew Linden:	oh ok
[11:12]	Sunman Loring:	hi
[11:12]	Andrew Linden:	What branch is it fixed in Simon?
[11:12]	Simon Linden:	The sim does not send enough data for a given texture level
[11:12]	Andrew Linden:	Oh that. I saw that somewhere in the logs
[11:12]	Hatch: Hello, Avatar!
[11:12]	Kitto Flora:	Is it supposed to be fixxed in 1.25 + tthe current RC viewer?
[11:13]	Simon Linden:	sim-texture-pipeline ... it's been stuck in our merge process forever
[11:13]	Andrew Linden:	oh right, that is where I saw it. What is the holdup Simon?
[11:13]	Simon Linden:	Our #$@ merge process :)
[11:14]	Simon Linden:	(some of you may guess this is something that i get frustrated with)
[11:14]	Kitto Flora:	So its it supposed to be fixed in current? OR still waiting?
[11:14]	Andrew Linden:	it hasn't received the lock yet, bumped by other priorities? or other problems?
[11:14]	Arawn Spitteler:	Mono compile is really slow.
[11:14]	Hatch: Hello, Avatar!
[11:14]	Andrew Linden:	I'm wondering mostly to see if it something that I can help with.
[11:14]	Simon Linden:	Andrew - we should take this offline, but it's stuck with QA and our merge process
[11:14]	Simon Linden:	well, if you can push that one fix into 1.26, you'll make a lot of people happy
[11:15]	Simon Linden:	I fixed it back in early October and am really frustrated about how hard it is to get our code tested and integrated
[11:15]	Simon Linden:	Solving that's an internal issue, however
[11:15]	Andrew Linden:	Hrm... that would be tough. server-1.26 has branched and the #deploy group is now guarding that branch with barred teeth.
[11:15]	Kitto Flora:	YEah - but you can tell the bosses that the residents are pissed about it :)
[11:16]	Andrew Linden:	on the other hand... this is a bug that has been broken for , like, forever, right?
[11:16]	Kitto Flora:	Sunday Avs on line peaked at 82,578
[11:16]	Kitto Flora:	A new record
[11:16]	Simon Linden:	Yes, it's been around a long time
[11:16]	Andrew Linden:	Yeah Kitto, a large number of us were online watching that.
[11:16]	Sunman Loring:	hi ella
[11:17]	Andrew Linden:	The concern was that the database would fall over on Sunday.
[11:17]	Simon Linden:	BTW, check out the latest nightly viewer ... the texture download fix for viewer-side is in there, and we saw great speed increases
[11:17]	Andrew Linden:	We were wondering how our emergency fixes were going to work.
[11:17]	Sunman Loring:	after the last fix of database it works much better
[11:18]	Andrew Linden:	Sunman, I'm curious: what aspects of the system are you talking about when you say it works much better?
[11:18]	Sunman Loring:	oh
[11:18]	Sunman Loring:	just in resident view
[11:19]	Sunman Loring:	at least from me
[11:19]	Andrew Linden:	That is, I'm curious about what metrics you've got your thumb on that tell you that the database is working better.
[11:19]	Sunman Loring:	usually i saw that with a refresh of inventory
[11:19]	Andrew Linden:	ah ok, so inventory updates when taking a copy to inventory?
[11:19]	Andrew Linden:	or rezzing objects from inventory?
[11:20]	Sunman Loring:	rezzing objects
[11:20]	Sunman Loring:	saving objects
[11:20]	Sunman Loring:	attaching objects
[11:20]	Andrew Linden:	ok cool
[11:20]	Sunman Loring:	detaching
[11:20]	Sunman Loring:	creating
[11:20]	Sunman Loring:	deleting
[11:21]	Andrew Linden:	Those usually operate against the inventory databases, which are separate
[11:21]	Sunman Loring:	and fetching inventory too
[11:21]	Andrew Linden:	but there must be some query connections that cause them to slow down when the main database has problems.
[11:21]	Arawn Spitteler should probably have jiraed this floating text issue, some time ago.
[11:21]	Andrew Linden:	that is, those operations slow down with the main database, not the inventory databases themselves
[11:22]	Sunman Loring:	ok
[11:22]	Andrew Linden:	Arawn, what floating text issue are you referring to?
[11:22]	Arawn Spitteler:	My pathcut green block, near the railing, in the north. Where do you see its floating text?
[11:23]	Andrew Linden:	I see it floating... but right on top of the object.
[11:23]	Arawn Spitteler:	I suppose it's a viewer issue, but can't be sure.
[11:23]	Arawn Spitteler:	I think it's supposed to be above.
[11:24]	Arawn Spitteler:	I've seen under, when a floor was sabotaged, but this seems improper.
[11:24]	Rex Cronon:	it depens on the prim rotation
[11:25]	Sunman Loring:	i use that for make screens
[11:25]	Sunman Loring:	with the prim and floating text
[11:25]	Andrew Linden:	Hrm... the llSetText() call does not have a position offset as an argument
[11:25]	Andrew Linden:	Arwan, I'm still a little unclear on the problem.
[11:26]	Andrew Linden:	You expect the floating text to be at a different spot?
[11:26]	Andrew Linden:	Since there is no floating text offset... yes I think it is a viewer-side bug.
[11:27]	Andrew Linden:	It is up to the viewer to figure out where to place the floating text.
[11:27]	Arawn Spitteler:	But it's reproducible on others' viewers.
[11:27]	Sunman Loring:	you can do the offset in the script
[11:27]	Kitto Flora:	The vertical offset of the text from center depends on the local z-sizze of the prim, no matter what the prim rotation?
[11:27]	Andrew Linden:	Ah that sounds likely Kitto.
[11:28]	Kitto Flora:	The green parth cut prim is rotated
[11:28]	Arawn Spitteler:	I've only found it curious, when doing a path cut trap door.
[11:28]	Object: Hello, Avatar!
[11:28]	Kitto Flora:	So where the text appears depends on the prim 'length'
[11:28]	Arawn Spitteler:	I'd expect it to be according to prim bounding box, but that explains why my one prim bridge had the text so high.
[11:29]	Qie Niangao:	I kinda think it's a "feature" at this point. I seem to remember dimpling boxes to adjust where the text would appear, or something like that.
[11:29]	Arawn Spitteler:	Unqualified Feature/
[11:29]	Arawn Spitteler:	?
[11:29]	Andrew Linden:	I think it should be fixed.
[11:29]	Kitto Flora:	Yo ucan make text appear higher by addin \n.\n.\n.
[11:29]	Kitto Flora:	Cant make it lower tho
[11:30]	Andrew Linden:	the viewer should figure out the world-bounding box of the object and put the text above it
[11:30]	Kitto Flora:	That would be nice
[11:30]	Kitto Flora:	Better would be llSetTextOffset(...)
[11:30]	Kitto Flora:	that included vector for location
[11:31]	Andrew Linden:	Here is a little bit of trivia...
[11:31]	Arawn Spitteler:	Also, TextScale could be set, but that's requiring new data being sent to the viewer.
[11:31]	Andrew Linden:	the floating text is an object setting that is independent of the script that created it
[11:32]	Andrew Linden:	so you can add a script to set it, then delete the script, and the floating text will remain
[11:32]	Kitto Flora:	Sure. Its a property, like sit target and such
[11:32]	Andrew Linden:	wouldn't it be nice if there were a way to set the floating text through the UI?
[11:32]	Kitto Flora:	Yes
[11:32]	Kitto Flora:	Many have noted that, Andrew
[11:33]	Arawn Spitteler set his block temp: Oh, speaking of properties, The goddess Bastet promised dire consequences, if anyone fixes the Co-sitting bug.
[11:33]	Andrew Linden:	I'm not sure what the best way to do that. In the past we would have added a new message template to the various UDP messages that get sent to the server
[11:33]	Andrew Linden:	however today we can no longer add new UDP messages
[11:33]	Arawn Spitteler:	That's right, UI is where Floating Text could be translated...
[11:33]	Kitto Flora:	Good
[11:33]	Simon Linden:	what's the co-sitting bug, and how powerful is goddess Bastet?
[11:33]	Kitto Flora:	death to the UDP
[11:33]	Kitto Flora:	(those that survive :)
[11:33]	Andrew Linden:	What is the co-sitting bug?
[11:34]	Andrew Linden:	Is that why you sent me that picture Arawn?
[11:34]	Kitto Flora:	angels on head o f a pin?
[11:34]	Arawn Spitteler:	As to the last, it depends if you want children, or not. I sent Andrew a picture, of the co-sitting bug, and will have to look for the Jira, but two avatars sit in the same location. Also called Lap Sitting.
[11:35]	Andrew Linden does not want to revisit the "find a place to sit" code anytime soon.
[11:36]	Arawn Spitteler:	http://jira.secondlife.com/browse/SVC-3724
[11:36]	Simon Linden:	Is it something that really bothers people? It sounds like it has some interesting side effects that people want, so if it's not causing pain, we probably won't look at it
[11:36]	Arawn Spitteler:	Was that you, that fixed it?
[11:36]	Simon Linden:	We have enough painful things to look at
[11:36]	Taunter Goodnight:	Taunter Goodnight
[11:37]	Morgaine Alter:	but griefers do use that lap sit to harrass others with :(
[11:37]	Arawn Spitteler:	It's sometimes embarrassing, I first co-sat the groom at a weding reception, but it's probably as much fun as momentary embarrasment
[11:37]	Andrew Linden:	I didn't fix sitting... in the Havok4 code I changed how the "sit spot" is calculated.
[11:37]	Arawn Spitteler:	For a while, it was hard to sit on a prim.
[11:37]	Kitto Flora:	That fixed some stuf, broke others Andrew
[11:38]	Simon Linden:	Hmm, interesting ... looks like it's just not detecting the other close avatar
[11:38]	Andrew Linden:	You know you've been in SL too long when you start enjoying the embarrassing bugs.
[11:38]	Simon Linden:	Maybe it should be optional : "So-and-so would like to sit on your lap. Is that OK? <Yes> <OMG No>"
[11:38]	Arawn Spitteler:	It happens on seats, so I don't know if it's sit targets, but when I've occupied the sit target of my Morph, others have bene able to sit on the prim.
[11:39]	Andrew Linden:	The "find a place to sit" code uses ray-tracing to find the spot.
[11:39]	Andrew Linden:	The ray trace tests should hit the avatar.
[11:39]	Morgaine Alter:	laughs
[11:39]	Kitto Flora:	Pile on Andrew!
[11:39]	Andrew Linden:	If they are not, then there is something wrong with the ray tracing.
[11:39]	Arawn Spitteler doesn't know how or why to trace Ray: I'm still in the dark, as to why vehicles don't have a datum that they're being ridden.
[11:40]	Andrew Linden:	I think Kelly or maybe Simon added some collision group checking to the ray trace tests.
[11:40]	Andrew Linden:	I wonder if the collision group logic is failing to hit the avatar prim.
[11:40]	Simon Linden:	the ray trace must not be hitting the AV, or we don't check to see if there's already an AV on the object with some X distance
[11:41]	Andrew Linden:	hrm... what feature there would you like Arawn? You want vehicles to be able to figure out if there is a driver?
[11:41]	Sunman Loring:	but sometimes this is fun
[11:41]	Sunman Loring:	lol
[11:41]	Kitto Flora:	In the scripted sit case it is really a failure to script properly
[11:41]	Arawn Spitteler:	It's a matter of where it hgappens, like the need to mute improper animations in church. Eventually, it should be a sim variable.
[11:42]	Kitto Flora:	In the default sit target case, sometimes its .. useful? Sometimes its an annoyance
[11:42]	Andrew Linden:	Oy. The problem with per-region or per-estate settings is that they clutter the code logic.
[11:43]	Andrew Linden:	if (special setting) { do this; } else { do the normal thing; }
[11:43]	Arawn Spitteler:	I picture, eventually, that Open Grid will offer choices of Physics Engine, so owners will have an intebger of Grid Sim Owner Flags, and might have another integer of Sim Engine Flags.
[11:44]	Andrew Linden:	but the useful bit can be achieved using various llSitTarget()'s and animation offsets, right?
[11:44]	Kitto Flora:	YEah - but the prim owner has to do the scripting
[11:44]	Arawn Spitteler:	Only one sit target to a prim, but llSetLink Rotation can be used to reset any sitting avatar.
[11:44]	Kitto Flora:	A lot of places , the prim is just at default
[11:45]	Arawn Spitteler:	If Prim Rotation weren't so borked, it would be easier for me, to have an adjustable scale couch prim.
[11:46]	Arawn Spitteler:	Physics Engine should strive for Newtonian Compliancy.
[11:46]	Kitto Flora suspects that it does
[11:46]	Arawn Spitteler:	A lot of High School Physics Tests should be modeleed in world
[11:46]	Kitto Flora:	And whast that got to do with sitting?
[11:47]	Arawn Spitteler:	Prim Rotation, in llSetLinkPrimitiveParams, is another issue.
[11:47]	Andrew Linden:	oh right. That bug.
[11:48]	Andrew Linden makes a note, but has no idea when he'll be able to get to that one.
[11:49]	Andrew Linden:	As I recall that bug requires a new param type added to the script engine.
[11:49]	Arawn Spitteler:	I think the current prefered suggestion, is to create new integers, PRIM_ROT_LOCAL and PRIM_ROT_GLOBAL, with PRIM_ROT_BORKED as legacy
[11:49]	Andrew Linden:	right, that is what I meant. New type integers.
[11:50]	Andrew Linden:	er... I don't think you could "fix" the legacy PRIM_ROT_* method.
[11:50]	Andrew Linden:	There is already lots of content that has worked around the bug (annoying as it is).
[11:51]	Arawn Spitteler:	To much content depends on the error
[11:51]	Arawn Spitteler:	I understand the code is a mess also
[11:51]	Andrew Linden:	Anybody got the jira for that bug for reference?
[11:51]	Qie Niangao:	hmmm... but the constants would just be for llSetLinkPrimitiveParams()... same problem with llSetRot on a child prim
[11:52]	Arawn Spitteler:	Have child prims got the same problem? That's awful. It's basically, a way of keeping new scripters confused.
[11:53]	Andrew Linden:	Those bugs really shouldn't be hard to fix. Adding a new integet to the script syntax is the hardest part.
[11:53]	Andrew Linden:	I'll make a note and see if I can get to those SetRot bugs
[11:53]	Qie Niangao:	SVC-93
[11:54]	Andrew Linden:	Yeah, that's the one. Thanks Qie.
[11:54]	Andrew Linden:	http://jira.secondlife.com/browse/SVC-93
[11:54]	Arawn Spitteler:	http://jira.secondlife.com/browse/SVC-93
[11:55]	Andrew Linden:	Ok this hour is almost up, and I've got another meeting right after.
[11:55]	Arawn Spitteler:	How many integers are available? Is it integers or flaghs?
[11:55]	Andrew Linden:	If anyone has questions or issues they've been saving for the end, now is the time to raise them.
[11:56]	Andrew Linden:	Arawn, they are interegers I believe, not bit flags.
[11:56]	Arawn Spitteler:	Gi es alot more leeway
[11:57]	Andrew Linden:	Alright then. Let's adjourn a little early today.
[11:57]	Andrew Linden:	Thanks everyone for coming.
[11:57]	Qie Niangao:	Thanks for your time, Andrew.
[11:57]	Kitto Flora:	Ty all
[11:57]	Kitto Flora:	Byebye
[11:57]	Sunman Loring:	Thanks Andrew
[11:58]	Arawn Spitteler:	Take yopur next meeting on the road; Lag Orbiting is loads of fun.