Time Nick Message 01:22 lhofhansl Does the ScopeProfiler report seconds or milliseconds? It says it reports [ms] but the numbers seem to be second (and in the code the ms are divided by 1000). 16:56 Newbyte Are there any plans for Minetest to move away from Irrlicht? 17:00 Krock no but like 100 requests to do it 17:02 Newbyte requests meaning people asking to do it or pull requests? 17:02 Newbyte asking for it to be done* 17:02 MTDiscord >pull requests 17:02 MTDiscord that was a good one ? 17:03 Newbyte yes, that seemed too good to be true 17:04 Newbyte if I ask it this way: is the general consensus that Irrlicht should be kept? or is it just kept because no one has put in the work to replace the Irrlicht parts of the code? 17:04 rubenwardy there's no general consensus 17:05 rubenwardy the current plan is to maintain irrlicht ourselves as needed, and to increase abstraction to reduce how much it's spread in the code base 17:05 Newbyte got it, thanks 17:05 rubenwardy https://github.com/minetest/irrlicht 17:05 rubenwardy replacing it is a lot of work 17:06 Newbyte does your fork support GLES? 17:06 rubenwardy there's also the question of what to 17:06 rubenwardy Yes 17:06 rubenwardy Minetest has full support for GL ES too 17:07 rubenwardy you need to enable a compile flag and link with the correct Irrlicht 17:51 sfan5 the forked repo currently does not include GLES code 18:33 nerzhul replacing irrlicht = rewrite the whole client 18:38 Krock and major parts of the server as wll 18:49 celeron55 it's possible to incorporate irrlicht into minetest, rename the library and make it completely our own as needed 18:50 celeron55 that'll probably happen sooner than later 18:58 nerzhul Krock: it's not needed to rewrite the server, our problem is more the client than the server :p 18:58 nerzhul and making irrlicht inside MT means maintain the GL layer, it's hard 19:00 MTDiscord Using GL is not hard at all. 19:01 MTDiscord Irrlicht could theoretically benefit us if we were fully using it's abstractions, but we aren't; rather are we circumventing them. 19:01 celeron55 i think being able to use opengl directly will be both a benefit and a necessity in the future 19:02 celeron55 and the necessary skills are out there waiting for irrlicht to get out of the way 19:02 MTDiscord Yes 19:02 nerzhul we curerntly have nobody with GL skills 19:03 celeron55 can't really say that when even i have some, but you could point at hecks for example 19:13 pgimeno what are the parts that depend the most on Irrlicht? I'd say scene graph, mesh loading and GUI 19:13 pgimeno I may be missing something major 19:14 rubenwardy it's possible to incorporate irrlicht into minetest, rename the library and make it completely our own as needed 19:14 rubenwardy this is the supertuxkart approach 19:14 MTDiscord TBH we'd be better off using a non-experimental GUI framework :P 19:14 rubenwardy I think it should be a separate repo though 19:14 MTDiscord definitely 19:14 MTDiscord Irrlicht is relatively bloated 19:15 pgimeno the scene graph is one big hinderance, causing unfixable bugs and preventing expansion of the world beyond 64K x 64K x 64K 19:16 celeron55 distributions will try to use their existing irrlicht packages for minetest and it has to be avoided then when we have our own irrlicht modifications 19:16 celeron55 it can be a separate repo but the name and the resulting library name has to change 19:16 MTDiscord Building our own scene graph shouldn't be too hard 19:16 pgimeno is linking statically out of the question? 19:16 celeron55 also header names 19:17 pgimeno @appguru not too easy either, considering it handles animations too 19:17 celeron55 pgimeno: distros absolutely hate anything statically linked and you always have to explain to them if they can't link dynamically 19:18 pgimeno well, "it's not Irrlicht anymore" is a good excuse 19:18 celeron55 i guess it's fine as long as it's a git submodule and the build just works 19:18 celeron55 then they'll get the hint 19:22 pgimeno for mesh loading, Assimp can be used instead; for GUI, I'm not sure there's any standard library that can be used as a replacement 19:23 pgimeno there's the possibility of rendering the GUI from Lua, of course, but that seems somewhat far away in the future 19:35 rubenwardy pgimeno: v-rob has a working branch for that 19:35 pgimeno for what of everything? drawing a GUI from Lua? 19:36 MTDiscord "considering it handles animations too" - yeah, poorly: it utilizes the CPU and reloads the buffers 19:36 rubenwardy latter 19:36 rubenwardy #10524 19:36 ShadowBot https://github.com/minetest/minetest/issues/10524 -- CSM Events and Drawing Testing by v-rob