Time Nick Message 00:58 rubenwardy Thanks 00:58 rubenwardy And good night 13:11 Krock >.< 13:47 Krock What to do if Lua's garbage collector is not fast enough? GC makes it RAII compliant but deletion is delayed way too much 13:56 Krock will merge #9150 in 15 minutes 13:56 ShadowBot https://github.com/minetest/minetest/issues/9150 -- Attachments: Fix interpolation from (0,0,0) after detach by SmallJoker 13:59 Krock branch squashed & function renamed 14:11 Krock merging 14:14 Krock #8910 close as Can't Fix upstream issue? 14:14 ShadowBot https://github.com/minetest/minetest/issues/8910 -- Add PNG file name in PNG profile warnings 14:17 p_gimeno Krock: run collectgarbage manually? 14:18 Krock well that's among a :close() function a possibility 14:18 Krock but that's not instant RAII 14:22 ANAND Whoa, #9150 fixes all the camera offset bugs? 14:22 ShadowBot https://github.com/minetest/minetest/issues/9150 -- Attachments: Fix interpolation from (0,0,0) after detach by SmallJoker 14:28 Krock not sure 14:32 ANAND Seems to fix most of them, going by the closed issues :) 14:32 ANAND Awesome work, thanks! 14:32 Krock np 14:33 Krock however, objects attached to LocalPlayer's CAO still don't update instantly (for some reason not handled by Irrlicht) 15:30 hecks How do I get the magical, consecrated libstdc DLL that the latest win64 build requires 15:30 hecks stealing it from sfan no longer works 15:30 sfan5 "latest win64 build" from where? 15:31 hecks from the buildbot, from git itself 15:31 hecks it doesn't pack three DLLs, libstdc, pthread and uh something 15:31 sfan5 git is a version control program 15:31 sfan5 do you mean gitlab? 15:31 hecks from the git REPO 15:31 hecks I usually steal the missing DLLs from your builds 15:31 sfan5 oh you've built it yourself 15:32 sfan5 the DLLs you need come with the mingw toolchain installed on your system 15:32 hecks yeah, to test joker's merge from like an hour ago 15:33 sfan5 in /usr/x86_64-w64-mingw32/lib or similar 15:33 hecks oh they're there 15:33 hecks why thank you 16:08 Krock > Then maybe disable client-side predictions for drops altogether? 16:08 Krock ANAND: results in double-dropped items on laggy servers 16:08 Krock players expect that the item is dropped instantly 16:09 Krock same with nodes. you don't wait until the server confirms that it's really dug 16:09 Krock you'd end up waiting for the server confirmation that might arrive a second late 19:14 Krock Well so... anyone here? 19:17 p_gimeno do I count? 19:17 Krock yeah, although I also hoped for some core developers 19:17 sfan5 hi 19:17 Krock tbh I wonder why something like #8665 is still open 19:17 ShadowBot https://github.com/minetest/minetest/issues/8665 -- Formspec: change the appearance of the cursor on fields and co. by DS-Minetest 19:17 Krock o/ sfan5 19:18 Krock the meeting points on the wiki could be way longer but adding more wouldn't help much at all 19:18 Krock concept discussions would probably fit best, rather than reviewing which happened in 0.4.x times 19:21 Krock nore, sofar: would you possibly have some free time to spend on a short dev meeting? 19:25 Krock back in 2014 there were Zeno, RBA, BlockMen and hmmmm helping to process the PRs and issues. in exchange there are now two new members, including me. Question is whether it would help to actively look for potential devs 19:27 Krock I actually also talked with rubenwardy and nore about the situation but we couldn't really recall someone who'd be interested in joining the team 19:28 Krock numberZero was a candidate, but he didn't seem to be interested or just disappeared 19:30 Krock p_gimeno: too bad you no longer have a GitHub account :/ 19:30 p_gimeno easy fix, just migrate to GitLab ;) 19:31 p_gimeno couldn't resist, sorry 19:34 Krock fair enough. they didn't kill the golden goose yet, though. 19:35 p_gimeno I always follow this chat, and look at some issues from time to time and comment when I think it's appropriate 19:38 Krock that's also a great way of contributing, even though the messages are lost if it's not a larger discussion 19:43 p_gimeno I'm out of the loop for the most part these days though, so it isn't very frequent. Last I remember is noting that the Irrlicht particles PR would bring back #8363. I could get motivated again though, but contributing this way is somewhat limiting. 19:43 ShadowBot https://github.com/minetest/minetest/issues/8363 -- Attached particle spawners no longer rotate with object 19:44 p_gimeno I tried for a long time playing with the Irrlicht particles PR to see if I could make attached particles work, which would automatically fix this, but I was unable. 19:45 Krock yet another camera shift issue? 19:53 p_gimeno I thought so, but then I tried at 0,0,0 and couldn't make it work there either 20:04 rubenwardy p_gimeno: I've migrated to gitlab 20:06 p_gimeno rubenwardy: migrated what? can we now handle issues and PRs there? 20:06 Krock just him personally 20:06 p_gimeno ah 20:08 p_gimeno so, like the personal repos I guess 20:10 p_gimeno is that camera offset thing really necessary? can't a different approach work better than that? I don't think any other 3D project handles it that way 20:10 sfan5 the offset approach is quite prone to bugs too 20:10 sfan5 as seen multiple ... times ... recently 20:11 p_gimeno gtg for a while sorry, will read the backlog later 20:11 Krock p_gimeno: you can try to disable it in client/camera.cpp (or set the offset values to > 33k) 20:11 Krock the issue is that the fastRaceRows vertices will have gaps in between 20:12 Krock it gets more noticeable the further you go from spawn 20:12 Krock *makeFastFace rows 20:13 Krock that's mainly because f32 are used to calculate the vertex positions 20:14 sfan5 hmm if it's so easy to disable maybe we could ask the issue reports to test with offsets disabled 20:14 sfan5 (regarding recent issues with entities, camera and attachments) 20:15 sfan5 would narrow the problem down by a lot 20:15 Krock well, if it doesn't work within 200m from (0,0,0) it's surely broken anyway 20:15 Krock camera offset only matters when going further than 200m into one direction 20:30 Krock > ./util/travis/script.sh: line 68: 12233 Segmentation fault (core dumped) ${CMD} 20:35 paramat i had to go out but am here from now on 20:37 Krock welcome. there's a long monologue in the logs. 20:38 Krock for now it might be best to check who could be added to the team 20:43 p_gimeno back, yeah I know that it's a precision issue, I just think there must be a better way to handle it 20:44 Krock offset each node? 20:44 Krock it would be more consistent but have the same bugs as 200m so after all it wouldn't matter at all 20:46 p_gimeno I'm not sure exactly how, but the general idea is that at render time, you subtract the camera position from every object somehow... maybe Irrlicht is not prepared to do that, I just know how I'd do it 20:46 p_gimeno think of it as a continuous offset, rather than discrete 20:47 p_gimeno if I want to translate that idea into code, I need to look into a lot of things 20:48 Krock offset = floor(pos) 20:48 Krock pos -= offset 20:49 p_gimeno why not offset = pos; pos = 0 ? 20:50 p_gimeno but I never know when to apply camera offset 20:50 Krock so to say - all SceneNodes would be relative to you 20:50 p_gimeno yes 20:51 Krock you need to apply the camera offset when dealing with Irrlicht SceneNodes. Only there. 20:51 p_gimeno I see 20:51 Krock the camera sets the origin (kinda) 20:51 Krock and all other SceneNodes are placed around it using the offset from the camera 20:52 p_gimeno it makes some sense, in that it's the AbsolutePosition calculation that screws the precision 20:53 paramat yes, i read logs 20:53 p_gimeno addition of big floats plus small floats destroys the precision of the small floats 20:54 Krock after all it's still needed to have the absolute position of entities for speed calculations - at least server-side 20:54 Krock and for client-side prediction as well 21:01 p_gimeno in the interval from 16K to 32K, a single has a precision of 1/512, i.e. you can go from 20000.0 to 20000.001953125 but not anywhere in between, that's ~2 mm but the accumulated error in the matrix computations will make things much worse 21:02 p_gimeno can the camera go far away from the player in any circumstances? 21:03 * Krock grammar-corrects the team discussion post 21:04 Krock p_gimeno: yes. attachments. 21:04 p_gimeno if it can't, player-relative coordinates could work, but for guaranteed results, camera-relative coordinates will always work 21:04 Krock when the player is attached to an entity, the model is in another location than the eyes 21:04 p_gimeno I see 21:04 Krock this is somewhat weird but the mods might depend on this 21:04 p_gimeno I don't think there are models >200 nodes big though :) 21:05 Krock see also: game.cpp ->getEyeOffset 21:05 p_gimeno or, well, things probably aren't expected to work flawlessly in such cases 21:05 Krock no. 200m is a safe distance for likely all use-cases in Minetest 21:13 p_gimeno so what happens when a 200m barrier is crossed? are all scene nodes moved? 21:14 Krock AFAIK detached and re-attached to the scene 21:14 Krock content_cao.cpp GenericCAO::step() when the visuals expired 21:15 p_gimeno would doing that every frame be too much? 21:15 p_gimeno I'll look, thanks 21:15 Krock "AFAIK" because I'm not 100% sure whether that's exactly executed when passing the barrier 21:15 Krock I didn't measure the time but I'd expect more imprecisions due to position movements 21:17 Krock if you detach the SceneNode, the position in the map remains the same but is no longer relative to its parent 21:17 Krock rather - relative to the origin (0,0,0) 21:18 Krock moving the origin could btw mess up the entire mapblock mesh placement. just so that I mentioned it 21:19 p_gimeno I think that the authoritative source for the object position is what's used to re-add the object to the scene, i.e. there's no accumulated error. Otherwise it could be a chaos. 21:20 p_gimeno or, well, I *hope* it's done that way 21:21 Krock *shrug* 21:29 paramat #1119 21:29 ShadowBot https://github.com/minetest/minetest/issues/1119 -- Fix rendering glitches when far from the center of the map by Ekdohibs 21:29 paramat 'every frame' sounds scary 21:44 rubenwardy I'm around now 21:58 p_gimeno https://github.com/minetest/minetest/pull/1119/commits/da63e5863b83e9b8bf18faa6bb6479b69838d382#diff-c2f90db5de8c33a259c27113939c63c5R1417 21:58 p_gimeno I guess that's costly to do it once per frame 22:02 p_gimeno also, it's incremental, i.e. subject to accumulated error 22:38 sfan5 merging #8813 in 5 minutes 22:38 ShadowBot https://github.com/minetest/minetest/issues/8813 -- Clean up craft replacements docs by pauloue 22:39 sfan5 also #9173 22:39 ShadowBot https://github.com/minetest/minetest/issues/9173 -- Run luacheck in travis by rubenwardy 22:50 sfan5 ... 22:50 sfan5 merged