Open Development User Group/Archive/2011-06-13-post-meeting

This is the conversation held after the Monday June13, 2011 Open Developers User Group (please forgive the fact that this in the raw log and not wikified at the moment WolfPup Lowenhar 11:34, 13 June 2011 (PDT))

[2011/06/13 08:08] Wolfpup (wolfpup.lowenhar): the Open Source https://bitbucket.org/lindenlab/llconvexdecompositionstub/overview

[2011/06/13 08:08] Wolfpup (wolfpup.lowenhar): the fact is that we need to be able to have it asctualy functional

[2011/06/13 08:08] ArminWeatherHax: libdirectfb has a very different version in opensuse and debian/like

[2011/06/13 08:09] FERRET (lestat.demain): WHAT PHYSICS SYSTEM ARE WE GOING TO USE FOR IT?

[2011/06/13 08:09] FERRET (lestat.demain): OOPS

[2011/06/13 08:09] FERRET (lestat.demain): sorry was commenting code

[2011/06/13 08:09] Wolfpup (wolfpup.lowenhar): as it stands right now that dose not do any thing at all and from what i heard @ the beta meeting mesh will be comeing to the main grind in about a month

[2011/06/13 08:10] Wolfpup (wolfpup.lowenhar): that is the question lestat

[2011/06/13 08:10] Wolfpup (wolfpup.lowenhar): and how are we going to get it merged in

[2011/06/13 08:11] Wolfpup (wolfpup.lowenhar): john's code is no longer to be considered as he is even sugesting to use HACD to replace his

[2011/06/13 08:11] FERRET (lestat.demain): physX or cuda or is there another system?

[2011/06/13 08:11] Wolfpup (wolfpup.lowenhar): so that leave the bullet system and HACD

[2011/06/13 08:12] Wolfpup (wolfpup.lowenhar): cuda and physx are stricly nVida lestat

[2011/06/13 08:12] FERRET (lestat.demain): ugh thats like saying havoc is purely linux

[2011/06/13 08:12] Wolfpup (wolfpup.lowenhar): we need some thing for ALL vid cards and Open Source

[2011/06/13 08:12] Phoenix-FirestormViewer Rocks (tankmaster.finesmith): openCL

[2011/06/13 08:13] Wolfpup (wolfpup.lowenhar): the LL version of convex looks to be mainly dealing with the decomp with just a little physics

[2011/06/13 08:15] Twisted Laws: HACD looks better to use to me as its mostly just the part of whats in bullet system that we need ... http://sourceforge.net/projects/hacd/

[2011/06/13 08:15] বরুনদাশ গুপ্তে (boroondas.gupte): I don't think there's need to have the convex decomposition run on the graphics card. It's a task ran once per upload. Not something running all the time.

[2011/06/13 08:15] Wolfpup (wolfpup.lowenhar): if we can come to a concensus on the source code to use we could then see if LL would import the code to the stub so that we are all useing the same code

[2011/06/13 08:16] FERRET (lestat.demain): heh if u wanna wait 6 months sure...but i go by their kdu example

[2011/06/13 08:16] ArminWeatherHax: does the stub now compile without changes? (does now compile the stub without changes?)

[2011/06/13 08:16] Wolfpup (wolfpup.lowenhar): Lestat KDU is not open source

[2011/06/13 08:16] Twisted Laws: it doesn't seem so critical tho as there is an external method of accomplishing it

[2011/06/13 08:17] Twisted Laws: when i first looked at this i thought it was used more often then just the upload

[2011/06/13 08:17] বরুনদাশ গুপ্তে (boroondas.gupte): Well, we don't need LL, at least in the beginning. Just clone the stub and declare one clone the 'official' repo for open source work on mesh decomposition.

[2011/06/13 08:17] Wolfpup (wolfpup.lowenhar): yes the stub compiles but you do not have and functionality as the calculations are not there

[2011/06/13 08:18] Wolfpup (wolfpup.lowenhar): well i already have a online set up i think

[2011/06/13 08:18] Wolfpup (wolfpup.lowenhar): yes i do

[2011/06/13 08:18] Wolfpup (wolfpup.lowenhar): https://bitbucket.org/WolfpupL/llconvexdecompositionos

[2011/06/13 08:19] ArminWeatherHax: the last time I compiled it it needed some changes https://github.com/ArminW/kokua-llconvexdecompositionstub/commits/master

[2011/06/13 08:19] Wolfpup (wolfpup.lowenhar): but all that has in it is the core from LL right now

[2011/06/13 08:19] বরুনদাশ গুপ্তে (boroondas.gupte): I thought the core was what's missing and we just have the interface?

[2011/06/13 08:19] Wolfpup (wolfpup.lowenhar): there is no extrac code it it as yet

[2011/06/13 08:20] Wolfpup (wolfpup.lowenhar): your right Boroondas

[2011/06/13 08:20] Wolfpup (wolfpup.lowenhar): im just calling the interface the core for now as we have to work off theat

[2011/06/13 08:21] Wolfpup (wolfpup.lowenhar): as that is the 'core' connection

[2011/06/13 08:23] Wolfpup (wolfpup.lowenhar): so now the question is do we want to use the 'stable' bullet system or the 'alpha' HACD

[2011/06/13 08:24] FERRET (lestat.demain): go with the stable

[2011/06/13 08:24] বরুনদাশ গুপ্তে (boroondas.gupte): I'd go for HACD, so we don't have all the physics stuff on bord we won't ever need.

[2011/06/13 08:24] Twisted Laws: i vote for HACD

[2011/06/13 08:25] Wolfpup (wolfpup.lowenhar): the bullet system has some of the HACD incorperadted into the linear math lib that they use

[2011/06/13 08:25] বরুনদাশ গুপ্তে (boroondas.gupte): into the linear math lib?

[2011/06/13 08:25] বরুনদাশ গুপ্তে (boroondas.gupte): that doesn't make any sense

[2011/06/13 08:25] Wolfpup (wolfpup.lowenhar): we do not have to bring all of the bullet just the convex and linear

[2011/06/13 08:27] Wolfpup (wolfpup.lowenhar): well im seeing two people saying to use HACD directly any others?

[2011/06/13 08:29] Twisted Laws: hacd is an active opensource project (i think) and more closely related to what we need (i think)

[2011/06/13 08:30] বরুনদাশ গুপ্তে (boroondas.gupte): has anyone looked at the interfaces of bullet and HACD so far and compared them to the interface in the stub? I guess the closer the match, the easier it'll be to write the glue between them.

[2011/06/13 08:30] Wolfpup (wolfpup.lowenhar): and bullet is useing in think=gs like blender which is what a lot of people use to make sculpties and mesh objects for SL with

[2011/06/13 08:31] Wolfpup (wolfpup.lowenhar): well the bullet system is still useing John's code and the interface seems more consistant withthe the LL interface

[2011/06/13 08:32] Wolfpup (wolfpup.lowenhar): from what i have seen

[2011/06/13 08:32] Wolfpup (wolfpup.lowenhar): but im no expert

[2011/06/13 08:33] Twisted Laws: sry, i must go, rl calls

[2011/06/13 08:33] Wolfpup (wolfpup.lowenhar): it would need some one with more codeing knowlage than me to know for certian

[2011/06/13 08:33] FooBar (data.landau): b srr rl called me away

[2011/06/13 08:34] FooBar (data.landau): reading ^^^^

[2011/06/13 08:35] FERRET (lestat.demain): i still say go with the bullet system...its compatible and also must have fewer bugs :)

[2011/06/13 08:36] Wolfpup (wolfpup.lowenhar): Boroondas what do you think?

[2011/06/13 08:37] FERRET (lestat.demain): besides...1 day someones going to have to write a decent building interface to use within SL itself

[2011/06/13 08:37] FooBar (data.landau): well if its just conv and math from bullet i would agree with the fury and go bullet

[2011/06/13 08:38] বরুনদাশ গুপ্তে (boroondas.gupte): FERRET, do you think a building interface would need viewer-side physics?

[2011/06/13 08:38] FooBar (data.landau): its going to be a mess to start with and so stability will help

[2011/06/13 08:38] FERRET (lestat.demain): to perform with any decent speed in a lagged environment like the sandboxes...definately

[2011/06/13 08:38] Wolfpup (wolfpup.lowenhar): Boroondas it might if the person in question is testing the object in a editing window

[2011/06/13 08:39] FooBar (data.landau): from what i can make out physics is used to define the bounding box of the object so you cant walk through it

[2011/06/13 08:39] বরুনদাশ গুপ্তে (boroondas.gupte): the other way around: physics might use bounding boxes to detect collisions

[2011/06/13 08:40] FooBar (data.landau): ut im no coder my knowlage is very limited

[2011/06/13 08:40] FooBar (data.landau): but*

[2011/06/13 08:40] Wolfpup (wolfpup.lowenhar): that is right data as it is noe sculpties have a bounding box the size of the prim if it was non-sculpted and mesh tightens that bounding box to the mesh its self

[2011/06/13 08:42] বরুনদাশ গুপ্তে (boroondas.gupte): If you repeat that sentence(s?) with punctuation we might be able to understand them ...

[2011/06/13 08:42] FERRET (lestat.demain): buries head laffin

[2011/06/13 08:43] বরুনদাশ গুপ্তে (boroondas.gupte): ?

[2011/06/13 08:43] Wolfpup (wolfpup.lowenhar): ok ill try again

[2011/06/13 08:44] Wolfpup (wolfpup.lowenhar): yes you are right data

[2011/06/13 08:44] FooBar (data.landau): As it is no sculpties have a bounding box. The size of the prim, if it was non-sculpted. And mesh tightens that bounding box to the mesh its self

[2011/06/13 08:45] FooBar (data.landau): .

[2011/06/13 08:45] Wolfpup (wolfpup.lowenhar): the bounding box of a sculpty is determaned by the non sculpted sixe of the object

[2011/06/13 08:45] বরুনদাশ গুপ্তে (boroondas.gupte): ah

[2011/06/13 08:45] Wolfpup (wolfpup.lowenhar): mesh objects pull that bounding box closed to the mesh sculpty

[2011/06/13 08:46] বরুনদাশ গুপ্তে (boroondas.gupte): i.e. sculpt map only affects visual shape, not physical.

[2011/06/13 08:46] বরুনদাশ গুপ্তে (boroondas.gupte): ?

[2011/06/13 08:46] FooBar (data.landau): y (and)

[2011/06/13 08:46] Wolfpup (wolfpup.lowenhar): that is correct Boroondas

[2011/06/13 08:47] Wolfpup (wolfpup.lowenhar): where as mesh objects are a two part sculpty

[2011/06/13 08:47] Wolfpup (wolfpup.lowenhar): the visual and the physical

[2011/06/13 08:47] বরুনদাশ গুপ্তে (boroondas.gupte): kinda

[2011/06/13 08:48] বরুনদাশ গুপ্তে (boroondas.gupte): mesh objects are of course not texture-driven in their (visual of physical) shape

[2011/06/13 08:48] FooBar (data.landau): saying that tho, from what i can remember from the last time i ply with bullet, im not sure how well bullet actualy does that

[2011/06/13 08:49] FooBar (data.landau): but its been a few years , so im sure there might have been improvements since then

[2011/06/13 08:49] বরুনদাশ গুপ্তে (boroondas.gupte): Gotta go, tc all

[2011/06/13 08:50] Wolfpup (wolfpup.lowenhar): data if you get and build the bullet code there is a convex api text that you can 'play' with object in

[2011/06/13 08:50] /me waves, too

[2011/06/13 08:50] বরুনদাশ গুপ্তে (boroondas.gupte): if the comments on John's blog can be believed, bullet might switch to HACD: http://codesuppository.blogspot.com/2011/05/hacd-hierarchical-approximate-convex.html#c5916931305323549215

[2011/06/13 08:51] FooBar (data.landau): yer, i was waiting to see which way to go b4 i did play,,