Time Nick Message 00:00 RealBadAngel VanessaE have some problems with it too, but in her case even a benchmark runs way too slow 00:01 hmmmm skipped to 43fcfbfe05578d7471d40c8c087fd04e24b264b5, which is the first bad one i saw so far 00:02 Sokomine that is very difficult to tell. fps on my low-grade hardware depend a lot on circumstances. when loading a large area, fps still drop drasticly. staring at the floor while moving forward gives considerably better speed than looking at the horizon. most worlds vary too much to really tell speed changes 00:02 hmmmm but not quite as bad as HEAD 00:02 hmmmm at 43fcfbf the same scene is at like 17 FPS 00:03 RealBadAngel main factor to slow down rendering when keeping same distance are leaves 00:03 RealBadAngel fancy leaves are proffesional fps killer 00:03 est31 heh 00:05 RealBadAngel https://github.com/RealBadAngel/default_normals/blob/master/screenshot1.png 00:05 RealBadAngel btw ^^ 00:05 RealBadAngel steady 60fps even with minimap on and relief mapping 00:06 RealBadAngel which is most expensive feature in our shaders 00:06 hmmmm yup, as I suspected... 00:06 hmmmm it's 43fcfbf 00:07 RealBadAngel whats that? 00:07 hmmmm git show 43fcfbf 00:07 hmmmm RealBadAngel, have you ever tried this http://aras-p.info/blog/2010/09/29/glsl-optimizer/ ? 00:08 RealBadAngel shaders? 00:08 RealBadAngel do you actually use parallax mapping? 00:08 hmmmm nope 00:08 hmmmm I have all fancy graphics settings disabled 00:08 RealBadAngel so how can parallax mapping can be a regression for you? 00:09 hmmmm because maybe other things modified in the shaders that don't have to do with parallax mapping slowed down rendering 00:10 RealBadAngel what is your GPU and drivers 00:10 RealBadAngel this time i want straight answer 00:10 RealBadAngel im still fighting with that the same code works differently (or doesnt work at all) on different gpus 00:12 RealBadAngel hmmm, also i have fixed some issues in shaders for AMD gpus in last commit 00:12 hmmmm why do my drivers matter 00:12 hmmmm it clearly has nothing at all to do with drivers 00:12 RealBadAngel can you check if that has any effect for you? 00:12 hmmmm it works on one commit, it doesn't work on another commit 00:13 hmmmm my drivers didn't magically change while bisecting 00:13 RealBadAngel do you understand when im sayin some thing can work on nvidia and doesnt work on amd for example? 00:13 RealBadAngel why the fuck you cant answer a question what gpu do you have? 00:14 hmmmm i'm not using your fancy graphics at all 00:14 RealBadAngel damn it 00:14 hmmmm you're just trying to use drivers as an excuse 00:14 hmmmm that's why i'm not answering 00:14 RealBadAngel GPU brand 00:14 RealBadAngel open gl version 00:14 RealBadAngel etc 00:15 RealBadAngel how can i fix the issue for you if youre stubborn 00:15 hmmmm I don't really need you to fix the issue 00:15 hmmmm I'm working on it 00:15 hmmmm this sounds like another shitty excuse like the whole "oh this is only a problem for threads under AMD cpus" - wtf? 00:15 RealBadAngel when Krock had issue with amd and opengl 2.1 i spent a few hours with him online to fix it 00:15 RealBadAngel and i fixed that 00:16 RealBadAngel but you cant even answer a simple question 00:16 RealBadAngel im not a houdini to ask what can be wrong on your side :P 00:16 RealBadAngel *to know 00:18 RealBadAngel also as i said, the code you suspect to be a reason for a slowdown has been already changed 00:18 est31 but the regression is still there 00:19 RealBadAngel did hmmm checked today's commits? 00:20 RealBadAngel its really hard to develop shaders that will work for all gpus around 00:20 RealBadAngel and keeping mouth shut but complaing doesnt help that at all 00:22 RealBadAngel also that really might be possible ive already eliminated that regression 00:23 RealBadAngel amd compilers treat ifdefs differently than nvidia 00:24 RealBadAngel thats why im not any longer checking if something is defined but a variable's value 00:24 est31 what the fuck 00:24 RealBadAngel that could be possible that hmmm had autogeneration enabled thx to that 00:25 RealBadAngel and that feature is costly 00:26 RealBadAngel est31, what? 00:27 est31 compilers treating ifdefs differently 00:27 est31 stupid compilers 00:27 est31 isnt this a well defined standard 00:27 RealBadAngel amd vs nvidia 00:27 RealBadAngel its not 00:28 RealBadAngel thats why for example Krock had bumpmapping enabled 00:28 RealBadAngel while there were no def for that in shaders code 00:29 RealBadAngel amd happily compiled the code pissing on ifdef 00:29 hmmmm (44, 23, -29, yaw=93) | d105bf2: noshaders: 30/13 FPS, shaders: 20/9 FPS | 43fcfbf: noshaders: 23/8, shaders: 15/6 FPS 00:30 hmmmm alright, so it's not necessarily a shader problem 00:31 RealBadAngel in any case shaders should be faster than no shaders version 00:31 RealBadAngel hmmm please try my todays commit if it change anything for you 00:31 hmmmm RealBadAngel: the regression is even worse as of current HEAD 00:32 hmmmm RealBadAngel: why do you keep doing this: https://github.com/minetest/minetest/commit/43fcfbfe05578d7471d40c8c087fd04e24b264b5#diff-c2f90db5de8c33a259c27113939c63c5R1440 00:32 hmmmm what's with the tabs in place of spaces 00:32 hmmmm please fix your editor settings 00:34 RealBadAngel which line? 00:34 hmmmm the one i highlighted on github.... 00:35 est31 yes I've noticed you doing it too 00:36 RealBadAngel ooops 00:36 hmmmm also who keeps doing postincrement on std::map iterators 00:37 hmmmm https://github.com/minetest/minetest/blob/43fcfbfe05578d7471d40c8c087fd04e24b264b5/src/mapblock_mesh.cpp#L1218 <---- wtf?? 00:37 hmmmm in any case, I'm beginning to suspect the slowdown is caused by changing S3DVertex to S3DVertexTangents 00:37 hmmmm I trusted you when you said they were the same.. 00:40 RealBadAngel idk what happens on your side, i dont even know your env 00:40 RealBadAngel so i cant help you at all 00:41 RealBadAngel what "wtf" ? 00:42 hmmmm the indentation is all messed up 00:42 hmmmm how could you look at that code and think everything is just fine and dandy 00:42 hmmmm sometimes I wonder what your code would look like if we didn't have code style guidelines... 00:43 RealBadAngel damn, its a bit too long 00:43 RealBadAngel its a few pages 00:43 RealBadAngel havent noticed that 00:45 RealBadAngel hmmmm, if tangents were the reason you will get worse fps without shaders 00:46 RealBadAngel strip nodes shaders to only final color blend and check 00:46 hmmmm there are literally no other functional changes to rendering with shaders disabled in that commit 00:46 hmmmm my common sense is telling me that has to be the reason 00:50 hmmmm hmm 00:50 hmmmm so for now I guess all I can really do is just disable shaders and suck up the difference in performance 00:51 hmmmm can't change it to use the old vertex format without messing up all the recent shader work 00:51 RealBadAngel what you should do is to check if you dont have special effects enabled thx to compilers 00:51 hmmmm ...what? 00:51 RealBadAngel you should remove their code from shaders and check clean shader 00:52 RealBadAngel you heard 00:52 est31 shaders are completely disabled 00:52 RealBadAngel noshaders: 30/13 FPS, shaders: 20/9 FPS 00:52 RealBadAngel i can see that you have better performance without shaders 00:53 RealBadAngel while it should be in different order 00:53 RealBadAngel game IS faster with shaders 00:53 VanessaE_mobile Shaders are always slower for me as well 00:53 hmmmm that has literally never been the case for me 00:54 hmmmm and i'm using a pretty generic graphics setup 00:54 hmmmm nvidia proprietary, a powerful, modern kepler card 00:56 RealBadAngel so you shouldnt be affected by amd issues, thats clear 00:58 hmmmm besides, I've mentioned several times before that my card is a GTX660 00:58 hmmmm ctrl+F in the logs 00:58 hmmmm i don't like repeating myself 00:59 RealBadAngel you could copy paste that not repeat ;) 01:00 RealBadAngel are you pretty sure that commit with tangents is the one from which you get regression? 01:00 VanessaE_mobile Hmmm like me you have a good setup that just insists on under performing in minetest 01:04 hmmmm buy a $180 video card, get 8 FPS on minetest 01:04 hmmmm sounds about right 01:04 hmmmm fwiw I totally understand why Kenney abandoned minetest and I would do the same 01:06 VanessaE_mobile Hmmmm did you see the screen shot I posted yesterday ? 01:06 hmmmm no, what 01:07 VanessaE_mobile http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/11-client-screenshot---2015-07-02.png 01:08 VanessaE_mobile This. Note the total fps across all 11 instances running there 01:08 VanessaE_mobile Around 300 (est) 01:09 RealBadAngel hmmmm, kenney wanted to use skyboxes for example. the way c55 implemented it wasnt good for him 01:10 RealBadAngel if you ask for the reasons 01:10 est31 also he didnt like formspecs 01:10 est31 the engine was too restrictive to him 01:11 RealBadAngel "fuck this, it has formspecs! im leavin!" indeed a good reason 01:11 RealBadAngel maybe we shall all leave then :P 01:11 est31 heh 01:12 VanessaE_mobile bbl 01:13 VanessaE_mobile (I'll see the rest on my pc later) 01:15 RealBadAngel hmmmm, ive compared your gpu to mine, yours is slightly faster 01:15 RealBadAngel you should be getting at least the same results as i do 01:16 RealBadAngel btw, what CPU do you have? 01:16 hmmmm irrelevant 01:16 hmmmm you don't even know what scene I was using to benchmark 01:17 RealBadAngel im trying to find the differences in setups 01:17 RealBadAngel hmmm i dont remember seeing such low framerates even on dreambuilder world 01:18 RealBadAngel so your scene doesnt matter 01:18 RealBadAngel you dont have propably scenes to stress tests the gpu as im using 01:19 hmmmm what does matter though, is that I was as consistent as possible throughout my own benchmarking 01:19 RealBadAngel ok 01:19 hmmmm the relative difference in framerates is what matters 01:21 RealBadAngel can you give me seed for the world you are testing? 01:21 RealBadAngel so i can check the same scen 01:23 RealBadAngel btw, what compile options are you using? 01:23 Niebieski What's an open source .x .b3d file editor/creator ? 01:23 RealBadAngel blender 01:24 Niebieski for some reason it's site is not opening for me. 01:24 RealBadAngel it can import/export in those formats 01:24 RealBadAngel sudo apt-get install blender 01:24 Niebieski Oh. 01:25 Niebieski is there an other variant that is less in size ? 01:25 RealBadAngel idk 01:26 Niebieski Alright thanks. 01:26 RealBadAngel and blender.org doesnt seems to be down 01:26 Niebieski Trust me it's always down for me from a very long time. 01:27 RealBadAngel i just checked it, works 01:27 Niebieski It's very weird, it keeps connecting to the site but nothing shows up. 01:27 RealBadAngel hmmmm, can you give me the seed? 01:29 est31 RealBadAngel, if hmmmm sais thre regression shows up when he's looking at A WALL then the seed doesnt matter at all 01:29 est31 the* 01:29 RealBadAngel then what for he gave us coords and yaw? 01:30 RealBadAngel so we should build a wall at given coords? 01:31 RealBadAngel if something works perfectly on one box and fails on the other imho we shall find what differs those boxes 01:31 RealBadAngel that should find locating the reason 01:32 est31 RealBadAngel, hmmmm is very competent, and working right now on a fix 01:43 hmmmm :\ 01:43 hmmmm I don't know how I can possibly fix this without screwing other things up too in the process 01:43 hmmmm at least I see the problem 01:43 RealBadAngel hmmmm, yes? 01:44 RealBadAngel whats that? 02:26 hmmmm est31: I am not able to replicate it looking at the *ground*, which is a better indicator of clipping almost all blocks 02:26 hmmmm looking at a wall and having the FPS sag is an indication that the occlusion culling algorithm blows 02:30 est31 ok 02:36 hmmmm I don't think that minetest is ready for a release in a month 02:36 hmmmm does anybody else feel the same way? 02:36 hmmmm we should fix all the issues on our plate that we know about and can actually do something about, before we think about releasing 02:42 hmmmm at current HEAD, that scene is 13/5 with shaders 02:45 est31 hmmmm, define those issues 02:45 est31 there is already a list here https://github.com/minetest/minetest/milestones/0.4.13 02:45 est31 you can fill in the further issues 02:46 hmmmm ok sure 02:46 hmmmm the static object stored in block bug 02:46 hmmmm the multiplayer avatar texture wrap bug 02:46 hmmmm the massive performance regression 02:47 est31 ? 02:47 hmmmm the more recent performance regression too 02:47 est31 lets hope thats fixed today 02:47 hmmmm hahahahahaha. 02:47 hmmmm not a chance. 02:48 est31 and what other massive perf regression do you speak about 02:48 est31 also with "static object stored in block bug" do you mean the ghost entity problem, and that players get removed? 02:49 hmmmm it's not as large as the first one, but there's an /even further drop in performance/ sometime after 43fcfbf 02:49 hmmmm est31, yeah 02:49 est31 thats a very hard one 02:50 est31 you cant reproduce it without making your CPU get 100% for a couple of seconds 02:50 est31 this makes debugging with valgrind impossible 02:50 hmmmm HEAD no shaders: 23/8 02:50 hmmmm alright 02:50 hmmmm the second performance regression is shaders only 02:50 hmmmm the first one is no doubt related to the S3DVertexTangets 02:51 hmmmm i think it might be a smart idea to first try running the shaders through a GLSL optimizer 03:09 est31 neoastetic has raised the request to disable minimap on their server 03:10 est31 and we should perhaps allow i 03:10 est31 t 03:10 est31 one way to do it would make the client ignore minimap related textures for client side texture packs 03:10 hmmmm or you could include the minimap as part of the hud 03:11 hmmmm disable-able through the api 03:11 est31 that sounds good too 03:15 hmmmm https://github.com/kwolekr/minetest/commit/7b171ea2be0e476d7cdc9300b53ba86a9f694161 03:15 hmmmm pushing in 5 minutes 03:20 est31 reading 03:20 est31 please dont push while i do 03:20 est31 so you want to enforce the style for single lined methods too? 03:21 est31 like { FATAL_ERROR("FIXME"); } 03:21 est31 because I've seen it in the test framework too 03:21 hmmmm whoops 03:21 hmmmm too late, it's been past 5 minutes 03:22 hmmmm don't like it? well I thought that's the way it works around here 03:22 hmmmm that's the way things seem to be going 03:32 est31 hmmmm, youve pushed after I've asked you to wait 03:32 est31 thats not nice 03:32 est31 even after youve said "too late" 03:33 hmmmm maybe you and the rest of the people who are guilty of it should drop this "pushing in 5 minutes" bullshit 03:33 est31 hmmmm, do what instead? 03:33 hmmmm wait for approval 03:33 est31 make a PR where I have to get 3 core devs to agree? 03:33 hmmmm minetest isn't going to spontaneously combust if you don't push some trivial thing within 5 minutes 03:33 est31 for a single line commit 03:34 hmmmm how about 30 minutes 03:34 est31 its hard to wait for approval if nobody is around 03:34 est31 your PR then rots for weeks 03:34 hmmmm well then maybe you should wait for people to be around 03:35 hmmmm nonsense, it only rots if people are actively pushing code 03:35 hmmmm if people are actively pushing code, then in theory, they should be on IRC or somewhere you could get in touch with them 03:38 est31 you can get in touch with me 03:38 est31 I always check irc before I push after 5 minutes 03:39 hmmmm can we agree on extending it to something like 30 minutes? 03:40 est31 what bout 15 03:40 hmmmm if you're going to negotiate with me i should've started at 60 03:40 est31 sn has requested a 15 minute move too 03:40 est31 hehe 03:40 hmmmm if i knew you were going to* 03:40 hmmmm nope, 60 minute waiting period, firm 03:40 hmmmm take it or leave it 03:41 hmmmm i can go down to 55 because you seem like a nice guy 03:41 hmmmm 55 is the best i can do 03:42 est31 17 not a second more 03:43 hmmmm lemme call up my buddy who knows all about SCM merge waiting periods. 03:43 est31 ? 03:43 hmmmm pawn stars 03:51 hmmmm https://www.youtube.com/watch?v=wc32_VHXMU0 03:52 est31 heh 03:57 hmmmm I think we could generalize Mapper for things other than the minimap 03:57 est31 like 03:58 hmmmm map preview before creating a world 03:59 hmmmm dumping an image of a map from the command line 04:36 VanessaE [07-02 22:26] looking at a wall and having the FPS sag is an indication that the occlusion culling algorithm blows 04:36 VanessaE I've been saying this for like 2 years now 04:37 hmmmm oh believe me, i know about it too 04:37 hmmmm but everybody is more concerned about flashy new features than fixing what we already have 04:37 VanessaE so it would seem. 04:39 hmmmm and then we need to work on fixing the new features' problems 04:39 kaeza since we are talking about graphics, anybody else having this problem? http://i.imgur.com/VM4axoy.png 04:40 hmmmm no 04:40 kaeza (notice the red nodes at the bottom) 04:40 hmmmm are you an ATi user? 04:40 kaeza no, Intel 04:40 kaeza it only happens at certain angles 04:41 VanessaE that reminds me: 04:41 kaeza it didn't happen a month or so ago, if that helps anything 04:41 kaeza guess I'll have to bisect :/ 04:41 VanessaE water (and other) alpha switching between "normal" and fully opaque depending on view angle needs fixed. badly. 04:42 VanessaE it'll even sit there vacillating while the pause menu is open 04:42 hmmmm Kaeza, I would recommend you start at 43fcfbf 05:05 kaeza hmmmm, the previous commit to that one works; I'm now building 43fcfbf 05:05 kaeza I wish `make -j3` worked :/ 05:06 kaeza it just locks up the whole shell window while doing nothing (I blame mingw32-make) 05:10 kaeza hmmmm, I can confirm that 43fcfbf broke this 05:10 hmmmm :) figures.. redirect complaints to RealBadAngel I guess 05:11 kaeza well, you just saved me the work of pinging him ;) 06:32 kaeza http://i.imgur.com/RN5JmlH.png :( 06:33 est31 kaeza, might be a regression cheapie pointed out 06:34 est31 #2815 06:34 ShadowBot https://github.com/minetest/minetest/issues/2815 -- Texture corruption 06:34 cheapie No, the one I pointed out is the broken textures when that parallax thingy is on. RBA insists it isn't broken. 06:34 est31 that means nothing 06:35 est31 usually RBA either insists it isnt a bug or its your drivers 06:35 cheapie At any rate, kaeza's problem seems to be related to shaders failing for some reason (causing the red). 06:36 est31 kaeza, which commit do you run? 06:36 kaeza est31, see #-dev logs 06:36 kaeza specifically: h_mmmm, I can confirm that 43fcfbf broke this 06:37 est31 does it work on latest master 06:37 kaeza no 06:37 est31 I'm sorry if i sound like rba but his last commit did some fixes to shaders 06:38 kaeza I'm at d75009a75853bd188f132e93165622c5e1315f70 ("Add android tools environment to gitignore") 06:38 kaeza eh that is maybe old 06:38 est31 try more recent 06:38 est31 yea 06:45 kaeza est31, nope, still happens on HEAD 06:46 kaeza (now at 7b171ea2be0e476d7cdc9300b53ba86a9f694161 "Fix code style from recent commits and add misc. optimizations") 06:49 est31 ok 08:20 rubenwardy RealBadAngel reminds me of this: http://thatwasfunny.com/top-20-programmers-excuses/239 08:21 rubenwardy :P 08:21 rubenwardy I'm just testing to see if I can see a regression 08:21 rubenwardy I also share hmmmm's viewpoint. http://irc.minetest.ru/minetest-dev/2015-07-02#i_4306698 08:27 rubenwardy Irrlicht log: 0:111(24): preprocessor error: syntax error, unexpected EQUAL 08:28 rubenwardy 2015-07-03 09:26:43: ERROR[main]: generate_shader(): failed to generate "nodes_shader", addHighLevelShaderMaterial failed. 08:28 rubenwardy https://gist.github.com/rubenwardy/d4a4b1edf1d239b450e7 08:28 RealBadAngel kaeza, i assume youre using shaders, yes? 08:29 RealBadAngel i need to make simplified version of the shaders for opengl 2.1 like kaeza have 08:29 kaeza RealBadAngel, https://github.com/minetest/minetest/issues/2866 (relevant system info there) 08:29 rubenwardy I have opengl 2.1 too 08:30 RealBadAngel 2.1 have 512 instuctions limit per one fragment shader 08:31 RealBadAngel i will have to disable relief mapping for example in order to run shaders on 2.1 08:32 rubenwardy I have 55-60 FPS staring into a wall with chunks loaded behind it, shaders enabled, but paralax, normal mapping, etc off. 08:32 kaeza RealBadAngel, yes, "shaders" checkbox only; nothing else 08:34 RealBadAngel kaeza, do you got any errors in console? 08:36 kaeza nothing interesting: https://gist.github.com/kaeza/266ec9246f53f292daff 08:39 Hunterz hi RealBadAngel, pls add new feature to minimap: https://github.com/minetest/minetest/issues/2868 08:40 VanessaE "The site ahead contains harmful programs" 08:40 Hunterz lol 08:41 kaeza RealBadAngel, I got this on debug: https://gist.github.com/kaeza/a8386f26b0242670d675 08:43 RealBadAngel kaeza, thats normal, we dont use any geometry shaders 08:43 kaeza oh ok then 08:45 RealBadAngel Hunterz, add them in overlay texture for minimap. Its texture pack stuff. 08:45 VanessaE RealBadAngel: actually I think those are HUD elements. 08:45 RealBadAngel VanessaE, you already made a texture with one direction, "N" 08:46 VanessaE RealBadAngel: yeah but that was hard-coded into an image. I think what Hunterz is talking about is HUD text overlays that are always upright 08:47 RealBadAngel Hunterz, does VanessaE know what are you thinkin about? 08:48 Hunterz i think about characters with world directions like North, South... 08:49 Hunterz when rotate map, rotate characters too 08:49 est31 thats already doable now 08:49 est31 just change the texture 08:49 est31 make a texture pack 08:50 RealBadAngel VanessaE, can you hand Hunterz that texture you made? 08:51 RealBadAngel i need to reboot, brb 08:53 kilbith https://github.com/minetest/minetest_game/pull/550#issuecomment-117806901 08:53 kilbith err... 08:53 kilbith it's too late to implement enet now, release is in one month 08:54 est31 we can do that after the release or in a separate branch 08:54 est31 nrz should fix some of his bugs though 08:54 VanessaE [07-02 22:36] I don't think that minetest is ready for a release in a month 08:55 kilbith sapier already did a branch with enet 08:55 kilbith but none has relayed this 08:56 est31 VanessaE, with your creative map, is there a way to provoke the ghost entity bug? 08:56 VanessaE est31: usually just signing on in the spawn area is enough 08:57 VanessaE make sure your items-per-block limit is sufficiently high 08:57 kilbith or dropping large areas of sand/gravel is usually enough for reproducing it 08:57 est31 is it right after you join, or do you have to stall there? 08:57 VanessaE est31: right after you join. 08:58 VanessaE kilbith: interestingly I was not able to reproduce the same results as the live server, though the errors DID happen. they just weren't "active" the way the server is. 08:59 est31 VanessaE, well I think hmms main concern is stability 09:00 est31 once the bugs are fixed, we can release 09:00 VanessaE perhaps/ 09:00 est31 if it were hmms descision, 0.4.12 wouldnt be released 09:00 VanessaE but depending on who you ask, there's a shitton of bugs to fix.. 09:00 est31 we would still have 0.4.11 09:00 kilbith ^ 09:01 VanessaE well to be fair, I wouldn't have let 0.4.12 out the door either, were the decision solely up to me. 09:01 kilbith i think it's matter of paradigm somehow 09:01 est31 the two texture regressions are RealBadAngel's responsibility I think 09:04 est31 and 0.4.12 was a "bugfix" release 09:04 RealBadAngel est "two texture regression" ? whats that? 09:05 est31 #2815 and #2866 09:05 ShadowBot https://github.com/minetest/minetest/issues/2815 -- Texture corruption 09:05 ShadowBot https://github.com/minetest/minetest/issues/2866 -- Graphical regression. 09:05 RealBadAngel first of all i need to detect what opengl version client is using, and if its <3.0 enable kinda "lite" version of shaders 09:06 est31 you think one of the issues is caused by that? 09:06 RealBadAngel 2815 is caused by parallax mapping, thats flaw of all displacement mapping algorithms 09:06 RealBadAngel havent found any workaround for that since really its not a bug 09:07 RealBadAngel i do have an idea how to make it insvisible tho, i need to add dynamic offset to the texture that will depend on view angle 09:08 RealBadAngel est31, https://github.com/minetest/minetest/blob/master/client/shaders/nodes_shader/opengl_vertex.glsl#L38 09:09 est31 ok 09:09 RealBadAngel 2866 is what i said just above, its opengl 2.1 and too few instructions per fragment shader 09:09 RealBadAngel so lite shaders are needed 09:09 RealBadAngel remember what i said about unrolling loops? 09:10 RealBadAngel thats the cause 09:10 RealBadAngel shader may not look like 512 instructions at all 09:11 RealBadAngel i learned that when fixing amd issues for krock 09:12 RealBadAngel loop with 4 steps compiled ok, with 5 thrown error Fragment Shader not supported by HW 09:14 RealBadAngel so, for <3.0 i will make separate folder with lite shaders with only basic funcionality. 09:26 kilbith so is there a real performance benefit to switch the stairs in meshnode ? 09:32 RealBadAngel yes 09:33 RealBadAngel less vertices to handle 09:36 RealBadAngel atm stair made out of nodeboxes have 16 vertices, meshnode would have 12 vertices 09:36 kilbith because of that i guess : https://lut.im/KlSTVki3/sTKX4GDA 09:37 kilbith but it's not a huge difference 09:37 RealBadAngel exactly 09:37 RealBadAngel yes it is 09:38 RealBadAngel stair is a common used node, when you have lotsa of them number will grow to pretty huge 09:38 kilbith right 09:39 RealBadAngel ive seen already somewhere rdy object for that 09:39 RealBadAngel im just not sure if it had uv mapping done 09:39 kilbith i can do that in 10 min 09:39 RealBadAngel so please do it, but take care of proper uv mapping of the textures 09:40 RealBadAngel as for stair with facedir = 0 09:40 kilbith uv mapping on that is pretty trivial really 09:41 RealBadAngel i know, just pointing the fact :) 10:02 kilbith 0.9 KB with a simple uv-map 10:02 kilbith not bad 10:18 kilbith also i switched the bed in mesh : https://github.com/minetest/minetest_game/pull/486 10:18 kilbith but PilzAdam & BlockMen were against 10:19 kilbith there was some valid points but a bit fatfetched 10:19 kilbith *farfetched 10:26 RealBadAngel kilbith, as i can see in case of beds placement problems wasnt correctly resolved 10:28 kilbith the reason was that the top was not registered internally and thus falling nodes could be merged with it 10:28 kilbith the top is "ghost" node 10:29 kilbith although i could convert the top & bottom nodes separately 10:34 RealBadAngel kilbith, that could be a solution 10:35 kilbith but you don't have smooth lighting on the whole bed surface in this way 10:35 kilbith unlike one node 10:50 kilbith mesh stairs done : https://lut.im/fN6ltjLy/CmZJ5BQQ 11:03 RealBadAngel can you make a screenshot with highlighted stair? 11:03 Calinou is the texturing the same as before? 11:04 Calinou make sure the textures are not rotated 11:07 kilbith textures are strictly the same yes 11:07 kilbith the UV-map is projection from view respecting bounding 11:08 kilbith https://github.com/minetest/minetest_game/pull/551 11:08 kilbith ^ sfan5 11:08 sfan5 looks good 11:09 RealBadAngel kilbith, can you? 11:09 kilbith yes 11:11 kilbith RealBadAngel, https://lut.im/V8emWc1Y/RgH6jnHM 11:11 RealBadAngel you have put that stairs on snow, highlighting is almost invisible :) 11:11 RealBadAngel please put it on something darker, and make screenie from a bit closer 11:11 kilbith will try on dirt 11:12 kilbith else you can pull my PR 11:13 kilbith RealBadAngel, https://lut.im/v5ZdmAuu/OUG52JfN 11:14 RealBadAngel i will try the PR, highlightin is still flawed, propably due to selection box 11:15 kilbith highlighting is a mesh ? 11:15 kilbith half-transparency is flawed pretty much everywhere 11:27 RealBadAngel it doesnt use regular transparency 11:29 RealBadAngel selection is defined the same way as nodebox was 11:29 RealBadAngel so it has same problem as the model 11:30 RealBadAngel with meshnode you should remove selection box def 11:31 RealBadAngel hmzz, im not sure what will happen then :) propably whole cube 11:31 RealBadAngel need to test it 11:31 RealBadAngel or code getting actual mesh into selection one 11:32 kilbith no if delete the selection_box, you get a whole cube selection 11:34 RealBadAngel ok, then let it stay as is. i need to code copyin the mesh 11:36 est31 Warr1024, seems I want to use an R-tree of one of its variants 11:37 kilbith proller says there's a small mem leak with the minimap 11:37 RealBadAngel he said where? 11:37 kilbith PM 11:38 est31 ? 11:38 kilbith "maybe only on closing game" 11:38 kilbith https://github.com/freeminer/freeminer/commit/cff451c6f9e9ee046aec906df16b47a9adf42d85 11:40 RealBadAngel hes wrong 11:40 RealBadAngel minimap mapblocks get deleted elsewhere 11:41 RealBadAngel he propably screwed things deleting the data in client.cpp 11:41 est31 if (m_minimap_mapblock) delete m_minimap_mapblock; <---- the if isnt needed here 11:41 est31 delete NULL is harmless 11:41 est31 == does nothing 11:42 RealBadAngel runtime blocks are deleted here: https://github.com/minetest/minetest/blob/master/src/minimap.cpp#L74 11:42 RealBadAngel and here on shutdown: https://github.com/minetest/minetest/blob/master/src/minimap.cpp#L135 11:43 RealBadAngel prollers code is obsolete 11:43 est31 lol he even added an if (data) where I said to RealBadAngel before it wasnt needed 11:44 RealBadAngel that comes from lua, there you have to check if it exists in the first place 11:45 est31 https://github.com/minetest/minetest/pull/2814#discussion_r33005742 11:45 RealBadAngel and is the most common reason for mods crashes 11:45 est31 https://github.com/minetest/minetest/pull/2814#discussion_r33005692 11:45 est31 ok 11:46 est31 everyday some new thing to learn 11:46 est31 I myself learnt that from kwolekr 11:48 est31 we can call porting::threadSetPriority(); 11:48 est31 thats perhaps a good addition 11:48 est31 but I dunno about these things 11:48 est31 lets wait what hmmmm sais 11:49 kilbith *says 11:51 est31 dammit youre right 11:55 RealBadAngel for disabling minimap on server simple mod that will always set enable_minimap to false should do the trick 11:56 RealBadAngel thats btw 11:56 est31 ewww 11:56 est31 we should perhaps do some #ifdef here 11:56 Calinou suddenly: modified client 11:57 RealBadAngel indeed 11:57 RealBadAngel but anyway i plan to add lua interface to control minimap anyways 11:57 RealBadAngel so it could be possible to make minimap craftable item etc 11:58 est31 yea 11:58 RealBadAngel or change the minimap look, position and the like 11:58 RealBadAngel this way server will be able to control it for sure 11:58 est31 one should perhaps be abled to display it in formspecs, and in hud 11:58 RealBadAngel but atm its not highest priority task 11:59 RealBadAngel now i have shitload of things to tweak 12:08 Hunterz is there any mod, that add map like in the minecraft using paper and compass ? 12:09 kilbith Hunterz, better to ask in #minetest please 12:09 Hunterz ok 12:10 Hunterz RealBadAngel: do you plan add some kind of waypoints and show them on the minimap? 12:10 Calinou that would be amazing, would bury Rei's minimap forever :P 12:11 RealBadAngel i do plan add there waypoints (theyre already implemented in game) and position of players 12:11 Hunterz reis minimap was wery best map 12:11 Calinou player positions? great for hunger games :) 12:11 Hunterz position of player please controlled by server 12:11 Hunterz fair play on the pvp for example 12:12 RealBadAngel Hunterz, any server controls will come with lua api for minimap 12:12 Hunterz nice 12:12 Calinou you can already see player names anyway 12:12 Calinou I guess any visible player on world will be visible on minimap, that makes sense 12:12 est31 RealBadAngel, perhaps we should add player display the same time we add the lua api 12:12 Calinou but please do add a toggle, in case it becomes too crowded (eg. spawn area) 12:13 Calinou est31, yes there needs a way to make some players fully invisible 12:13 Calinou eg. "Vanish" 12:13 Hunterz visibility of nicknames will be fine controlled by server too 12:13 Hunterz something like tagapi in the bukkit 12:15 Hunterz btw tested latest code and minimap load all chunks around, perfect... 12:16 est31 hrmm 12:17 est31 there exists a pure c++ R* tree implementation on github 12:17 * est31 wonders whether we should implement an R* tree ourselves 12:17 est31 most likely that will trigger ugly bugs though 12:19 Calinou don't reinvent the wheel :) 12:19 est31 implementing an R* tree is non-trivial thatsfor sure 12:20 est31 and from this image you see, R-tree is shit https://en.wikipedia.org/wiki/R*_tree#Performance 12:36 est31 the lib even has an ubuntu package 12:36 est31 this is a good candidate for an optional dependency 12:36 est31 if its there, use it, if not, use a vector instead 12:38 Calinou how about linking the dependency statically? 12:39 est31 ? 12:41 est31 linking statically has nothing to do with optional dependencies 12:41 est31 its more about " 12:41 est31 how hard do I rely on it" 15:08 RealBadAngel hmmmm, here? 18:22 RealBadAngel i think i minimalized displacement mapping texture flaw 18:22 RealBadAngel it is still visible from very close distance, but hardly 18:23 paramat a PR for re-enabling ravines in mgv7 but now applied to large caves #2870 18:23 ShadowBot https://github.com/minetest/minetest/issues/2870 -- Mgv7: Enable large cave ravines by paramat 18:24 RealBadAngel https://imgrush.com/4YAAYbN1TIqR.png 18:24 RealBadAngel and https://imgrush.com/A9mArp4nC8DP 18:25 RealBadAngel i cant do anything more for non tiling textures without deforming the displacement effect 18:38 RealBadAngel even with this, theres slight deformation is certain cases 18:38 RealBadAngel thx to fastfaces and same faces mergeing 19:02 paramat in the close-up screenshot, the rear line of grass pixels is at an angle to the node behind it, that looks bad to me (might be an acceptable cost to some players, personally i don't use shaders) 19:14 hmmmm I feel as if the minimap update was pushed before it was ready 19:14 hmmmm est31 seems too overly eager to approve things 19:16 hmmmm I think it'd be a good idea to either increase the number of core developer approvals for a major commit, or perhaps removing the "workaround" to +1 your own commit 19:17 hmmmm there are just sooo many things to do here 19:17 hmmmm ugh I wanted to spend my time/effort elsewhere, but if this doesn't get fixed now it'll become forgotten 19:29 hmmmm why on earth does the minimap use a mesh buffer anyway 19:35 sfan5 you can approve your own commits? 19:36 RealBadAngel hmmmm, it uses mesh to display minimap 19:36 RealBadAngel mesh can be easily rotated 19:37 hmmmm sfan5: you can't, but that's the popular interpretation that RealBadAngel and est31 seem to have picked up 19:37 hmmmm it's like any gray areas in rules are exploited so that they can push their commit through faster 19:37 hmmmm and with less oversight and more bugs 19:37 kilbith « You can push something to upstream [1] only if two members of the core team [2] or the subsystem maintainer agrees on it » 19:38 kilbith extract of the rules 19:39 RealBadAngel hmmmm, and what again do you have against minimap? 19:40 hmmmm just trying to completely understand everything 19:40 hmmmm all the design decisions involved 19:40 hmmmm i think the basis of the minimap is fantastic 19:40 RealBadAngel then just ask 19:40 hmmmm i can generalize this and make into more than just "that minimap thing at the top right of the screen" 19:41 RealBadAngel i do have further plans with minimap 19:41 RealBadAngel its mapper, it can be used for other things 19:41 hmmmm so what's the point of the MinimapData structure? 19:41 hmmmm it seems to me like all the members in there really should be memberes of Mapper 19:42 RealBadAngel theyre shared between two classes 19:43 RealBadAngel i didnt wanted them to be either private, public or whatever. theyre just in common structure 19:43 hmmmm what is "radar" 19:43 Calinou the cave radar 19:44 RealBadAngel best way to understand it is to try it 19:44 hmmmm I know what it is, but I want to know in your own words what the member MinimapData::radar is supposed to represent 19:44 RealBadAngel dot on minimap gets more green the more air is found in the scan column 19:45 RealBadAngel its bool 19:45 RealBadAngel what it can be if not on/off ? 19:45 hmmmm that doesn't answer the question of what it's supposed to represent 19:45 celeron55 >come and look what's going on in #minetest-dev 19:45 celeron55 >that's bool 19:45 celeron55 :D 19:46 celeron55 keep up the good work i guess 19:46 hmmmm save your eyes, don't bother looking in here. 19:47 * VanessaE peeks in 19:47 RealBadAngel talks why i named a variable like that are rather not productive 19:47 VanessaE (my eyes are already shot, so looking in here should be safe :P ) 19:47 RealBadAngel and thats has nothing to do with either coding style and quality of the code 19:47 hmmmm what if I started writing really shitty code 19:48 hmmmm code that works but sucks, has lots of race conditions and memory leaks, and i named all my variables 'a' 'b' and 'c' and so on 19:48 RealBadAngel name "radar" is shitty? 19:48 hmmmm when somebody asks me what "float c" is, I'll say 19:48 hmmmm it's a float 19:48 RealBadAngel its not called "r" :P 19:48 hmmmm it could be a decimal value 19:48 hmmmm it's pretty vague 19:49 RealBadAngel now , please point what sucks in minimap, where are race conditions and the leaks 19:49 hmmmm I'm fixing them 19:49 hmmmm don't bother 19:49 RealBadAngel and a, b , c etc variables 19:49 hmmmm yea 19:50 hmmmm and then when somebody tells me, "hey, you shouldn't name your variable c, that's confusing!" i'll respond with: "just try it out, it works!" 19:50 hmmmm and then when there's a problem with it I'll say: 19:50 hmmmm "what version is your driver?" 19:51 VanessaE troll.... :P 19:51 hmmmm "that could be the problem, try updating your drivers" 19:51 hmmmm "oh the problem is that you have an AMD cpu" 19:51 hmmmm "AMD cpus aren't good with threads" 19:51 hmmmm christ 19:51 RealBadAngel laugh, but thats sadly true, i have to ask such questions 19:51 RealBadAngel because same code can work on nvidia, and refuse to work on amd 19:52 RealBadAngel and i dont have amd to test it 19:52 hmmmm you asked me what graphics card I'm using after I already proved the problem was not in the shaders. 19:52 VanessaE if you have to ask such questions, RealBadAngel, then you need to also ask, "Is it failing for you (VanE, hmmmm) because I (RBA) am using nVidia-specific extensions in my code?" 19:52 RealBadAngel atm im forced to prepare special shaders only for AMD 19:52 RealBadAngel because i cant make it work for both in the same file 19:53 RealBadAngel read what i just wrote 19:53 RealBadAngel i have to make separate shaders for amd and nvidia 19:54 VanessaE that's not the same thing. 19:54 hmmmm and you never fixed the wieldhand shader 19:54 hmmmm btw 19:54 hmmmm just pointing that out 19:54 RealBadAngel what? 19:54 RealBadAngel do we HAVE wieldhand shader? 19:54 hmmmm nope 19:54 hmmmm that's the issue 19:55 hmmmm i had to disable the shader material for wieldhands because you never implemented that part 19:55 RealBadAngel i wanted to make it but it was rejected 19:55 VanessaE RealBadAngel: what I mean is not #ifdef's that only work on one driver but not another (like the parallax/bumpmapping problem you and I had to work out). I mean code that is explicitly using some feature that only works on nVidia because it's an nVidia-specific vendor extension 19:55 RealBadAngel same as proper extruded items etc 19:55 hmmmm well, then, why was it rejected 19:56 RealBadAngel VanessaE, main problem is that i dont have amd to test the code before 19:56 RealBadAngel hmmmm, it worked 19:57 VanessaE RealBadAngel: too bad I can't just lend you my spare HD6870 :P 19:57 RealBadAngel but what was rejected it was fix to extrusion code 19:57 hmmmm just because code works doesn't mean it's ready to be part of the codebase for everybody else 19:57 RealBadAngel and it was necesary for wielded shader to work 19:57 hmmmm you need to begin understanding this 19:57 hmmmm "works" 19:57 hmmmm works in your very limited set of test cases 19:57 hmmmm i'm sorry, that's a scrappy way of coding 19:58 hmmmm "it works therefore I merge" 19:58 hmmmm any criticism is responded to with "just try it out, see that it works and it makes neat effects" 19:58 RealBadAngel im not design studio with hardware base to test all cases 19:58 RealBadAngel sorry bout that :P 19:59 hmmmm and 95% of bugs in your code are not due to hardware-specific issues 19:59 hmmmm you keep using that excuse when it's not warranted at all 20:01 RealBadAngel whatever, i wont be wasting time on just talkin, since two days im coding fixes for my issues 20:01 kilbith some bugs are present in OpenGL 4.4 that OpenGL 2.1 don't have (eg. wieldhand glitching) 20:02 RealBadAngel https://github.com/minetest/minetest/issues/2866 20:03 RealBadAngel and believe it or not, its not my fault but the drivers :P 20:03 RealBadAngel opengl 2.1 is too old 20:03 VanessaE that's not a driver fault. 20:03 VanessaE that's a RealBadAngel fault :P 20:04 RealBadAngel sure :P 20:04 RealBadAngel i made 512 instruction limit for fragment shader 20:04 VanessaE well. it worked. you committed something. it broke. your fault :) 20:05 RealBadAngel for opengl 2.1 there will be set of lite shaders 20:05 RealBadAngel without advanced effects at all 20:05 RealBadAngel only final color blend, some geometry stuff, bumpmapping and nothing more 20:06 VanessaE inb4 RealBadAngel codes something that breaks on my GPU because my OpenGL version (4.4.13374) is too new or something :) 20:06 kilbith 5 years old GPU uses OpenGL 2.1 with mesa (which limits the latest features of OpenGL API) 20:06 kilbith it's not too old 20:07 RealBadAngel on 2.1 im not able to make parallax/relief mapping to work 20:07 VanessaE too old would, in my opinion, be trying to use a fancy new effect on some 15+ year old GPU 20:07 RealBadAngel with everything in, simply theres not enough instruction space 20:08 RealBadAngel and btw, im coding somethin and all i get now is just blaming me instead of help 20:08 RealBadAngel dont want my code so just say thx bye 20:09 RealBadAngel meanwhile im off for coding, have enough for today 21:42 RealBadAngel VanessaE, not big mistake, opengl 2.1 is not 15 yrs old, its just 9 yrs old 21:43 VanessaE RealBadAngel: well you got the point :P 21:43 Taoki RealBadAngel: Any news on that pull request? 21:44 RealBadAngel im busy fixing issues right now 21:44 Taoki ok 21:58 VanessaE got a new bug that I don't know the source of. users are, very rarely, getting invisible, undefined, no-textured items into their inventories. don't know if unified inventory is at fault or if it's the engine. The one user who reported it managed to give the item to me after a few tries. the item shows up as "Item 65510" (with two spaces) in my player file. 21:58 kilbith VanessaE, can you try this fix ? https://github.com/minetest/minetest/pull/2869 21:58 VanessaE if you mouse over it, there's no tooltip. if you drop a stack of something on it, it swaps with the stack and there's nothing attached to the pointer. 22:00 OldCoder Hi. Starting today I have a mysterious crash every few minutes. No error messages or even segfault messages. Program simply exits. 22:00 paramat hi hmmmm here's a small commit for review #2875 22:01 ShadowBot https://github.com/minetest/minetest/issues/2875 -- Mgv7: Lower base of mountain generation to -112 and define constant by paramat 22:01 OldCoder I updated from 0.4 stable to git current. Problem still occurs. Also did clearobjects and vacuum. 22:01 OldCoder Hate the idea of reverting and problem may simply come back. What known issues cause silent exits without error messages or segfaults? 22:02 OldCoder Note: This is Lord of the Test. But the question is about the core software. 22:02 OldCoder FWIW git-current seems to work. Except for the silent crash persisting. 22:04 kilbith try out with gdb maybe ? 22:04 RealBadAngel VanessaE, cant do anything better than that: https://imgrush.com/Gfc14n2Wn7t8.png 22:05 VanessaE RealBadAngel: looks fine to me I guess, as long as the textures don't slide around 22:05 VanessaE however I don't use parallax mode anyway so it won't bother me :P 22:07 kilbith RealBadAngel, the displacement is amplified on HD textures, how does this look with them ? 22:07 RealBadAngel hard to say 22:08 RealBadAngel it all depends on the heightmaps 22:08 kilbith look this huge displacement with a 64px texture : https://cloud.githubusercontent.com/assets/415827/8275588/3bc239ce-186e-11e5-9ed5-419c2f09dcf1.png 22:08 RealBadAngel whats this texture> 22:08 kilbith from 'streets' mod 22:08 RealBadAngel does it have heightmap? 22:08 kilbith no 22:08 RealBadAngel or its autogen? 22:09 kilbith the screenshot is from cheapie 22:09 kilbith (and the mod is bundled with dreambuilder) 22:09 RealBadAngel i will try it 22:12 RealBadAngel no way 22:12 RealBadAngel im going to disable parallax mapping for non tileable textures 22:12 VanessaE ew. 22:13 VanessaE that texture is tilable but only in one direction.. 22:13 RealBadAngel if you dont believe me ask uncle bout it, but stop calling the code corrupted or similiar 22:13 RealBadAngel parallax and relief mapping works on seamless textures 22:13 RealBadAngel period 22:15 RealBadAngel im going to add to nodedef fields that will describe the tiling. horizontal, vertical, seamless, none 22:15 VanessaE where did I call anything corrupted? 22:15 RealBadAngel its needed anyway for connected textures 22:15 OldCoder I think it's the network issue again 22:15 RealBadAngel you all have called it bug, textures corruption etc 22:16 VanessaE um 22:16 OldCoder con(5/1)RE-SENDING timed-out RELIABLE a million of these messages 22:16 VanessaE no, I didn't call it corrupted. 22:16 RealBadAngel i will also soon be blamed theres gravity on earth 22:16 VanessaE it IS a bug though if the textures slide around. 22:16 VanessaE but not if they're statically displaced. 22:16 RealBadAngel its not a fucking bug 22:16 RealBadAngel this is how parallax works 22:17 RealBadAngel check on wiki 22:17 OldCoder VanessaE, RealBadAngel, anybody; the RESEND RELIABLE error goes into a loop and overflows a stack, I think 22:17 OldCoder But it looks like somebody worked on this code 22:17 VanessaE RealBadAngel: parallax occlusion is not supposed to cause textures to slide around -- but I think you already fixed THAT problem anyway. 22:17 VanessaE the static displacement of textures is a different matter 22:17 OldCoder Same issue that shut down Silvercrab for a while 22:17 VanessaE and I don't think that's a bug at all. 22:17 RealBadAngel they do not slide 22:18 VanessaE all right, I give up 22:18 VanessaE no one's listening to me anyway 23:26 VanessaE here's something new for you guys to ignore: #2877 23:26 ShadowBot https://github.com/minetest/minetest/issues/2877 -- Player files corrupting with "empty" items