Bug triage/2007-07-16/Transcript

From Second Life Wiki
Jump to navigation Jump to search

Agenda

[15:01] Stevex Janus: I'm still rying yo catch up on eht 400+ emails from the list
[15:01] Soft Linden: Ping!
[15:01] otakup0pe Neumann: ICMP_ECHO_REPLY
[15:01] Daedalus Young: hi Rob
[15:01] Saijanai Kuhn: stop floating
[15:01] Soft Linden: Hee
[15:01] Stevex Janus: hi rob
[15:01] Soft Linden: Yeah, the list gets a bit noisy on and off :>
[15:01] Rob Linden: hi folks....y'all wanna try out my fancy new three prim meeting area down the hill?
[15:01] Soft Linden: Sure :>
[15:02] Soft Linden: Bring the bug box?
[15:02] Stevex Janus: oku
[15:03] Rob Linden: hopefully, we're close enough that people can still find us
[15:03] otakup0pe Neumann: hard to miss the mass of green dots :o
[15:03] Daedalus Young: follow the green dots
[15:03] Soft Linden shouts: Dale, Saijanai, down the hill!
[15:04] Squirrel Wood: /ao off
[15:04] Gigs Taggart: eek, wet
[15:04] Squirrel Wood: /ao on
[15:04] Dale Glass: interesting
[15:05] Gigs Taggart looks for a copy of the multi-sit-target script :)
[15:05] Soft Linden: What's that do - set a new target whenever someone sitss?
[15:06] Rob Linden: yeah, I'm not exactly the builder with the skilz
[15:06] Gigs Taggart: yep
[15:06] Soft Linden: Nice!
[15:06] Gigs Taggart: it was pretty ingenious :)
[15:06] Gigs Taggart: I don't have a copy on me right now
[15:06] Gigs Taggart: I'll have to drop one on rob later
[15:06] Soft Linden shouts: Aric, Dzonatas - down the hill today!
[15:06] Squirrel Wood gave you AutoSizing Ring Seat 20+.
[15:07] Boroondas Gupte: I think it can be rezzed by MultiGadget
[15:07] Saijanai Kuhn: lotsa lindens here today...
[15:07] Squirrel Wood: Mmm. Lindens ^^
[15:07] Soft Linden: Now we just need this to go floating down the river once triages start
[15:07] Rob Linden: alright....let's get started
[15:07] Rob Linden: VWR-779
[15:08] Rob Linden: https://wiki.secondlife.com/wiki/Open_Source_Meeting/Agenda
[15:08] Daedalus Young: 779 looks good to me
[15:08] Dzonatas Sol tried to sit...
[15:08] Soft Linden: Looks good to me
[15:08] Squirrel Wood: Isn't it http://wiki.secondlife.com/wiki/Bug_triage/Agenda ?
[15:09] Rob Linden: both should work
[15:09] Soft Linden: Squirrel: Yep!
[15:09] Daedalus Young: or even https://wiki.secondlife.com/wiki/Bug_triage/Monday_Agenda
[15:09] Rob Linden: (oh, whoops, you're right
[15:09] Soft Linden: But it was a trick: Now you have to run it, Squirrel.
[15:09] Squirrel Wood: the open sauce one shows me an empty list
[15:10] Soft Linden: VWR-779 import?
[15:10] otakup0pe Neumann: i can't notice a difference in the pictures, but go for it.
[15:10] otakup0pe Neumann: oh. looks liek the text on the top is less crunched
[15:10] Daedalus Young: Nicholaz' version is a bit cleaner
[15:11] Soft Linden: Overrunning that background area, which can make it unreadable with some backgrounds
[15:11] Rob Linden: yeah, ok. we should at least give Ben a heads up. it's debug stuff, so he'll be less concerned, but still
[15:11] Rob Linden: ...but I'll mark it for import
[15:11] Rob Linden: next up: SVC-337
[15:13] Soft Linden: Looks low-priority, but a valid bug if it's right. Might import, but not assign yet
[15:13] Dzonatas Sol: I haven't applied this patch, but I wonder if it conflicts with Dale's Mute Visibility.
[15:13] Dale Glass: that's Able's, not mine
[15:13] Dzonatas Sol: Able's.. my bad
[15:13] Soft Linden: The patch is just for easily reproducing a bug, not for fixing it.
[15:14] Squirrel Wood: depends on whether this is indeed a bug or intended behavior
[15:14] Boroondas Gupte: is there any documentation on what is is supposed to do?
[15:14] Dzonatas Sol: i see that now... "demonstrates".
[15:15] Rob Linden: this hasn't been discussed on sldev yet, has it?
[15:15] Soft Linden: I don't recall it
[15:16] Soft Linden: If it was brought up, it wasn't with the JIRA in it. No hits in search.
[15:16] Rob Linden: does anyone feel confident they know what the behavior should be here?
[15:17] Aric Linden: Rob, if you'd like you can import this and assign it to me
[15:17] Squirrel Wood: that's the point that needs clarification?
[15:17] Aric Linden: I'll have someone on my team repro it and provide feedback on correct behavior
[15:17] Soft Linden: At the very least, we should see if updateMutFlags is supposed to do something with that field. Yeah.
[15:18] Rob Linden: ok...that sounds great, Aric, thanks!
[15:18] Aric Linden: sure
[15:18] Rob Linden: next up: VWR-824
[15:18] Daedalus Young: SL isn't the only realtime 3D app with the transparent problem
[15:18] Squirrel Wood: 824.. isn't that something that is caused by opengl limitations of some sort or another?
[15:19] Daedalus Young: I'm sure I've seen it in Blender's Game Engine too
[15:19] Daedalus Young: also see the linked issues there
[15:20] Soft Linden: That kind of alpha sorting is an unsolved problem for poly-based engines. I'm sure that's duplicated internally...
[15:20] Dzonatas Sol: I'm sure 824 is related to another jira entry... somewhere
[15:20] Daedalus Young: VWR-27
[15:20] Rob Linden: we've discussed this general problem before
[15:20] Daedalus Young: for example
[15:20] Rob Linden: that's probably the one I'm thinking of
[15:20] Soft Linden: I'll close it as a dupe of that
[15:21] Squirrel Wood: I assume SL uses some form of z-buffer to render stuff... which should technically eliminate such z-order / alpha problems ?
[15:21] Daedalus Young: you'll need to have objects subdivided by the renderer
[15:21] Rob Linden: has anyone explained this general problem well, though?
[15:21] Rob Linden: VWR-27 is still kinda hanging out there as a probelm, too
[15:21] Daedalus Young: my comment on VWR-27:
[15:21] Daedalus Young: This happens because the Z-buffer (which you can see if you go to Snapshot > Capture > Depth) doesn't allow transparent texturing. You can make it work in a lot of situations, but it breaks when several objects are both behind -and- in front of eachother (like a ring around a pole, the ring is behind the pole, but the pole is behind the ring). As Thygrrr said, it requires geometry subdivisions, divide the ring in 2 pieces, one piece behind the pole and one piece in front of it.
[15:22] Dzonatas Sol: I like that idea... subdivisions
[15:22] Rob Linden: VWR-203 is the discussion I'm remembering, but it looks like they're all related
[15:22] Soft Linden: But it gets even more complicated than that, when you have 3-4 objects overlapping, the number of subdivisions required is explosive.
[15:23] Daedalus Young: yeah, I've linked some up yesterday
[15:23] Rob Linden: ok, so resolved/duplicate?
[15:23] Dzonatas Sol: yes.
[15:23] Daedalus Young: there are many more issues concerning transparency
[15:23] Squirrel Wood: it happens with objects that have different 3d positions inworld and aren't intersecting at all.
[15:23] Dzonatas Sol: real-time ray-tracers have a way to subdivide fast
[15:23] Boroondas Gupte: classic z-buffer trick only works for opaque objects, for transparent ones you would need a list/array/whatever of several z values per pixel which isn't performant anymore
[15:23] Daedalus Young: also sorting in HUDs wrong
[15:23] Daedalus Young: I believe it's basically all the same problem
[15:23] Rob Linden: next up: VWR-914
[15:24] Gigs Taggart: HUD alpha is newly broken daedalus
[15:24] Gigs Taggart: at least in terms of picking and certain rendering problems
[15:24] Daedalus Young: ok gigs
[15:25] Soft Linden: VWR-914 will likely become more of a problem with Vista as well. Memory needs to be sized based on available texture memory, not the whole of video memory.
[15:25] Gigs Taggart: yeah
[15:25] Soft Linden: Vista uses quite a bit more memory for UI compositing as well.
[15:25] Gigs Taggart: same with linux and Beryl/etc
[15:25] Dzonatas Sol: It has votes. Import without patch?
[15:25] Gigs Taggart: 3D desktop is only going to get more common
[15:25] Squirrel Wood: If you run the transparent glass ui on vista it demands a 256mb gfx card methinks
[15:26] Soft Linden: Yeah. Or I can make a note to ask for suggestions on how to choose the size on-list
[15:26] Aric Linden: It coincides nicely with an internal uptick in vista diagnostics too
[15:26] Gigs Taggart: yes import
[15:26] Rob Linden: k...marked for import
[15:26] Rob Linden: next VWR-1448
[15:27] Squirrel Wood: seen that happen on one of me products
[15:27] Squirrel Wood: quite annoying actually
[15:27] Gigs Taggart: I haven't tried the repro but it looks detailed
[15:27] Gigs Taggart: Looks good to me for import
[15:28] Rob Linden: I'll import
[15:28] Rob Linden: next
[15:28] Rob Linden: VWR-747
[15:28] AusFox Walcott: Oh noes, its teh feds!
[15:28] Gigs Taggart: 747 is straightforward
[15:28] AusFox Walcott hides contraband* <_< >_>
[15:28] Gigs Taggart: a "misfeature" that was waiting to get enough votes to triage again
[15:28] Rob Linden waits for VWR-747 to load
[15:29] Gigs Taggart: I almost made a patch for 747, I expect it to be one liner :)
[15:29] Squirrel Wood: humm.. when uploading a 1022x510 texture, SL will size it down to 512x256 when it would be better of sizing it to 1024x512... not related but I think its worth mentioning ^^
[15:29] Gigs Taggart: squirrl you might want to report that in Jira too :)
[15:30] otakup0pe Neumann: heh
[15:30] Soft Linden: The bigger problem is the number of builders using 1024 when 128 would do ;)
[15:31] otakup0pe Neumann: True.
[15:31] Saijanai Kuhn: There's a solution for that with sculpties at least
[15:31] Saijanai Kuhn: allow lossless compression only for 128 or less
[15:31] Squirrel Wood: I'd use 10240² if there was a need for it (high quality animation :p)
[15:31] Soft Linden: Saijanai - that's brilliant
[15:31] Gigs Taggart: 1024 is useful for "whole object" maps
[15:32] Saijanai Kuhn: not my idea. pretty much the concensus of everyone
[15:32] Daedalus Young: texture atlases
[15:32] Daedalus Young: make one object with different textures, but all textures mapped onto one large texture
[15:32] Gigs Taggart: a 128 lossless is probably bigger than a 1024 compressed :)
[15:32] Squirrel Wood: mipmaps... let people choose a maximum mipmap level to load in clients ?
[15:32] Soft Linden: Even if not lossless, a higher quality level, enough to kill visible macroblock borders
[15:33] Saijanai Kuhn hopes that Gigs is wrong...
[15:33] Gigs Taggart: hehe
[15:33] Gigs Taggart: well I exaggerate
[15:33] Gigs Taggart: to be honest I would love a lossless option
[15:33] Soft Linden: But this one - import?
[15:33] Gigs Taggart: yes 747 import
[15:33] Saijanai Kuhn: 64 with lossless is probably enouygh, from what Qarl says, but we shall see
[15:33] Gigs Taggart: it was waiting for votes
[15:33] Squirrel Wood: I'm all for it
[15:33] Gigs Taggart: it has them now
[15:33] Gigs Taggart: it's going to be like one line of code I bet
[15:34] Rob Linden: ok...I'll import, though I'm still a little confused, to be honest
[15:34] Gigs Taggart: right now upload previews constrain aspect ratio to the powers of two the actual texture is
[15:34] Rob Linden: could someone maybe clear up the description a little bit?
[15:34] Bridie Linden: yes, please
[15:34] Gigs Taggart: ok I'll do it
[15:35] Aric Linden: Thanks. I'm worried that we might be making things worse if we change this.
[15:35] Aric Linden: A better description of what's expected will help avoid that
[15:35] Rob Linden: I think it's worth a discussion from whoever made the change. can't promise a fix, thoguh
[15:36] Dzonatas Sol: How about an extra L$5 for lossless?
[15:36] Rob Linden: next up: VWR-857
[15:36] valebi Aabye: i m looking for a job
[15:36] Gigs Taggart: I'd pay 500L for a 1024 lossless :)
[15:37] Boroondas Gupte: lossless is to be considered when uploading, not when displaying
[15:37] Saijanai Kuhn notes: aren;t we all
[15:37] Dzonatas Sol: blackened avatars on macs... dupe
[15:37] Daedalus Young: yes, seen a dupe of it too
[15:37] valebi Aabye: where is teleport?
[15:37] Rob Linden: oh...yeah, it's linked to the dup, just not marked
[15:37] Aric Linden: I think it would be nice to have a repro for 857
[15:37] valebi Aabye: hi
[15:38] Squirrel Wood: repro? Get a Mac? ^^
[15:38] Daedalus Young: no, not that easy Squirrel
[15:38] Soft Linden: Valebi: Try clicking 'search' at the bottom of the screen and finding a location to teleport to in there
[15:38] Squirrel Wood: aww
[15:38] Daedalus Young: I'm on a Mac and it doesn't always happen
[15:38] Aric Linden: Same here Daedalus.
[15:38] Daedalus Young: have seen it on other avatars though
[15:38] Rob Linden: dup of VWR-771
[15:38] Daedalus Young: usually when I log into a busy area, a highly-prim-attached avatar turns up black
[15:39] Rob Linden: next up: VWR-400
[15:39] Gigs Taggart: I have updated 747
[15:39] Bridie Linden: thx Gigs
[15:39] Soft Linden: ty Gigs :)
[15:39] Rob Linden: ditto
[15:39] Squirrel Wood: packet loss related "bug" ?
[15:39] Daedalus Young: I think 400 is due to packet loss indeed
[15:40] valebi Aabye: soy nueva
[15:40] Gigs Taggart: oh I have news on that front
[15:40] Gigs Taggart: I have installed the latest ubuntu beta which has the traffic shaping module I need to do full latency and packet loss testing
[15:40] Dzonatas Sol: sounds like packet loss
[15:40] Saijanai Kuhn: valebi try going to the second life library and asking for help from the librarians
[15:40] Gigs Taggart: that was holding me up previously
[15:40] valebi Aabye: i need telepot
[15:41] Gigs Taggart: I will be hopefully trying to repro most latency and packet loss related bugs real soon now and posting how to do so to the sldev list :)
[15:41] Saijanai Kuhn: use that to teleplort to the library
[15:41] Bridie Linden: / valeba, you can use the Map to find a location and teleport there.
[15:41] Saijanai Kuhn: there are often people there to help you
[15:41] Bridie Linden: / velebi, sorry
[15:41] valebi Aabye: i must push teleport
[15:42] Dzonatas Sol: yes
[15:42] Saijanai Kuhn: 5-10 people there ATM she should be fine
[15:43] Rob Linden: it doesn't look ats though there's a reliable repro for this, short of what Gigs is talking about
[15:43] Dzonatas Sol: resolve/no repro =/
[15:44] Gigs Taggart: well it's pretty straightforward I think, the client "moves" the prim visually but the server is authoritative in the end
[15:44] Dzonatas Sol: or need more info.. for network info
[15:44] Squirrel Wood: yep. poster needs to provide packet loss detail info for cases when it "happens"
[15:44] Gigs Taggart: hopefully that command will move to TCP :)
[15:45] Rob Linden: k...will resolve/cannot repro for now
[15:45] Aric Linden: for 1043 please import it and assign it to me
[15:45] Rob Linden: next up: VWR-1043 which sounds like we've got a resolution for
[15:45] Saijanai Kuhn: talk about synchornicity. Someone in teh scripters group just described a situation where they could stretch a build to cover 1/2 a sim without unlinking
[15:46] Dzonatas Sol: close.. old
[15:46] Rob Linden: thx aric
[15:46] Dzonatas Sol: not 1043... but next
[15:46] Dzonatas Sol: 63
[15:46] Rob Linden: VWR-63
[15:46] Squirrel Wood: Sai... build it small scale, link, then scale it up
[15:47] Saijanai Kuhn: ah, well, new to me...
[15:47] Dzonatas Sol: 63 has been fixed
[15:47] Dzonatas Sol: i makred is reolved/fixed
[15:48] Dzonatas Sol: *it as
[15:48] Soft Linden: I'd be curious to see the shop though. It sure sounds like it could even just be someone with two objects in the same location. Or too close for decisive Z-ordering.
[15:48] Rob Linden: when was it fixed?
[15:48] Gigs Taggart: soft I remember talking with someone about this bug and they said it wasn't z-ordering
[15:48] Gigs Taggart: soft it almost sounds like video ram corruption
[15:48] Dzonatas Sol: This was a bug up until the 1.13 and 1.14
[15:49] Dzonatas Sol: I used to get it also
[15:49] Dzonatas Sol: it doesn't happen anymore
[15:49] Rob Linden: ok...good enough
[15:49] Rob Linden: next up: SVC-29
[15:49] Aric Linden: I'd be curious to know if this is still an issue
[15:50] Gigs Taggart: yeah, he would know if it was hehe
[15:50] Aric Linden: could we ask him?
[15:50] Gigs Taggart: he's surely the largest user of XML-RPC in the world
[15:50] Aric Linden: really?
[15:50] Rob Linden: which issue are we talking about? SVC-29?
[15:50] Gigs Taggart: I would think so, all those thousands of SLX boxes
[15:50] Aric Linden: yes, the xml-rpc issue
[15:51] Dzonatas Sol: there was a dns issue with xml-rpc... that may be fixed now
[15:51] Aric Linden: it's hard for us to test xml-rpc timing issues internally
[15:52] Gigs Taggart: I think Kelly said xml-rpc is a bottleneck system where it all goes through a single server and that was causing trouble.
[15:52] Squirrel Wood: mayhaps it'd be a good idea to set up a bunch of xmlrpc boxes in different sims and have them monitor "round trip" times of sorts?
[15:52] Gigs Taggart: I don't know if that is still the case.
[15:52] Aric Linden: If it's still a problem, I'd love to know about it and work with him to repro it and analyze the specifics
[15:52] Gigs Taggart: yeah he'd be the guy to go to, I'm sure he gets emails every time the system slows down :)
[15:53] Aric Linden: Rob, can I propose that we ask Nynthan if he's still seeing problems and if so, put him in touch with me?
[15:53] Saijanai Kuhn wonders if there is a debug assistance kit for landowners and lindens. wander into a sim and click option #3 in the HUD...
[15:53] Squirrel Wood: Where's my moose.. err.. money? ^^
[15:53] Gigs Taggart: noo I mean Apotheus
[15:54] Aric Linden: Ahh. thanks Gigs. Apotheus it is. Rob, is that ok with you?
[15:54] Rob Linden: that's fine
[15:54] Boroondas Gupte: SVC-29 is linked to an issue that can't be viewed by public (SVC-1)
[15:55] Rob Linden: yeah, I'm not sure what's up with that.
[15:55] Boroondas Gupte: I saw that with another issue that's been mooved to the security project, or something
[15:56] Rob Linden: next up MISC-23
[15:56] Rob Linden: (for what it's worth, it says I don't have perms to see SVC-1, which is really weird)
[15:56] Daedalus Young: description is a bit unclear, but apparently with very large builds, "edit linked parts" and editing a prim unlinks it
[15:56] Gigs Taggart: MISC-23 I saw a recent complaint about from Lewis Nerd on SLDEV group.
[15:56] Gigs Taggart: spontaneous unlinking
[15:57] Saijanai Kuhn cues Twilight Zone music
[15:57] Squirrel Wood: I have not had that happen to me as of yet
[15:57] Gigs Taggart: it is a little hard to repro, might be packet loss related again
[15:57] Gigs Taggart: Isn't Lewis in UK?
[15:57] Saijanai Kuhn: might be related to the wandering child prims thing
[15:57] Gigs Taggart: high latency/packet loss might be a trigger for this
[15:58] Daedalus Young: I don't knw how edit linked parts works internally, if that unlinks and relinks a linkset, it could be related to packet loss
[15:58] Dzonatas Sol: linksets usually move as one vector
[15:59] Dzonatas Sol: so packet loss is doubted
[15:59] Rob Linden: hrm...ok. what's the next step?
[15:59] Gigs Taggart: I don't know at this point.
[15:59] Daedalus Young: trying to get a solid repro? :S
[15:59] Aric Linden: I'd suggest either a solid repro
[15:59] Rob Linden: do we need some way of marking issues as possible packet loss related?
[15:59] Gigs Taggart: Meta-issue for potentially latency/packet loss related bugs seems called for.
[15:59] Gigs Taggart: yes
[15:59] Dzonatas Sol: IM Lestat
[16:00] Gigs Taggart: it would make testing easier once I develop some test scenarios for high latency/packet loss
[16:00] Dzonatas Sol: OK... I'll IM Lestat and post the reply to the jira
[16:01] Rob Linden: k....looks like we're out of time
[16:01] Daedalus Young: aww, already
[16:01] Squirrel Wood: That was a quite productive hour again ^^
[16:01] Rob Linden: thanks everyone for showing up!
[16:01] Soft Linden: Thanks all :)
[16:02] Squirrel Wood: Thanks for listening to our opinions and suggestions ^^
[16:02] Dzonatas Sol: thank you for your time 'all
[16:02] Daedalus Young: Lindens thanks for letting us help :)
[16:02] Squirrel Wood: /ao on
[16:02] Soft Linden: Who's meeping?
[16:02] Daedalus Young: not me
[16:02] Bridie Linden: bye !
[16:03] Daedalus Young: bye