Time Nick Message 03:28 MTDiscord Cool 05:06 MTDiscord lower resolution 05:10 MTDiscord Never 05:10 MTDiscord But you will see something soon 05:10 MTDiscord And you will not have expected it 05:11 MTDiscord https://tenor.com/view/piano-playing-piano-spongebob-spongebob-playing-organ-organ-gif-21786918 06:07 MTDiscord Another blind speed run of a game 06:31 MTDiscord Pretty close actually 09:44 sfan5 @jordan4ibanez did you disable post-processing 09:44 sfan5 that can eat a lot of fps even if isn't doing anything apparently 09:53 erle luatic if #15477 is indeed security relevant, server owners (who do not necessarily use the latest git HEAD) and packagers would *probably* want to know the impact and which versions are affected, to decide if it is necessary to cherry-pick it. 09:53 ShadowBot https://github.com/minetest/minetest/issues/15477 -- Network: Fix serialization version checks by SmallJoker 09:54 MTDiscord erle I think so 09:54 MTDiscord But I have a solution 09:54 erle ? 09:54 MTDiscord https://tenor.com/view/what-stare-cat-kitty-gif-3565065 09:54 MTDiscord it's a secret 09:56 erle in particular i would like to know, because of the whole “intel integrated GPU are the enemy and deserve huge frame rate loss if they can't do shaders or can't do them fast enough” thing (because it means i am collecting patches that i need to apply to 5.10 and earlier versions) 09:57 erle like if this is e.g. a remotely triggered crash introduced post-5.10 it's different than if it has been there forever 09:58 erle and i am pretty sure debian might want to know too, if it affects anything 09:59 MTDiscord Intel IGPU is horrible 09:59 MTDiscord It's horrible on my mac mini 2014, a piece of poop 09:59 [MatrxMT] <🇬regon> My Intel integrated GPU worked fast with Minetest before 5.7/5.8 09:59 erle you keep saying that but as long as i get 30fps it's “playable” 10:00 erle 🇬regon what change made it worse? 10:01 [MatrxMT] <🇬regon> I don't know 10:01 [MatrxMT] <🇬regon> Just it worked fast before I upgraded from 5.6 to 5.8 10:02 MTDiscord How long ago did you update to 5.8 10:02 erle 🇬regon what fps loss are we talking about? 10:02 MTDiscord What os you run 10:03 [MatrxMT] <🇬regon> Debian 9 10:04 MTDiscord Holy shit 10:04 erle 🇬regon if you test the slow and fast versions and activate wireframe mode, do you see if the slow version merges identical faces into planes horizontally? 10:04 erle or not? 10:05 erle i ask because it *used* to do this: https://forum.luanti.org/viewtopic.php?p=300317&sid=1cff65da6a08dbe31e8210d191d3d0b4#p300317 10:08 [MatrxMT] <🇬regon> https://matrix.org/_matrix/media/v1/download/matrix.org/aYGnobjOuGgUwFGYxuxRYhnH 10:08 MTDiscord That might have caused strange issues under certain texture scenarios. Or maybe someone just bulldozed it on accident I dunno 10:08 [MatrxMT] <🇬regon> https://matrix.org/_matrix/media/v1/download/matrix.org/CGpCBPIEadiUWcSKmTXOtuOX 10:09 [MatrxMT] <🇬regon> Yes they are not same 10:09 [MatrxMT] <🇬regon> @blockhead256:tchncs.de can you please resend this images? 10:10 erle jordan4ibanez i know in which scenario it causes strange issues … but i think it was removed because no one remembered if it was even relevant or so and on *their* hardware it did not make any difference. it made for a slight fps drop on mine. 10:11 MTDiscord Hmmmm 10:11 MTDiscord Then perhaps it can be re-added as an option! A great success 10:11 erle bold of you to think that stuff that only benefits crappy hardware will be added 10:11 [MatrxMT] <🇬regon> https://matrix.envs.net/_matrix/media/v1/download/matrix.org/CGpCBPIEadiUWcSKmTXOtuOX 10:11 erle 🇬regon i can not open the images, please upload them to https://mister-muffin.de/paste 10:12 [MatrxMT] <🇬regon> https://envs.net/_matrix/client/v1/media/download/matrix.org/aYGnobjOuGgUwFGYxuxRYhnH?allow_redirect=true 10:12 erle and post the link 10:13 [MatrxMT] <🇬regon> https://matrix.envs.net/_matrix/media/v1/download/matrix.org/aYGnobjOuGgUwFGYxuxRYhnH https://matrix.envs.net/_matrix/media/v1/download/matrix.org/CGpCBPIEadiUWcSKmTXOtuOX These are working? 10:14 [MatrxMT] <🇬regon> Because they work for me 10:14 erle yes they are 10:15 erle 🇬regon the drawtime is basically the same, but the dtime jitter is bizarre 10:16 MTDiscord Hmm 10:16 MTDiscord Is that running on one or two cores? 10:16 erle 🇬regon okay, next thing: can you install qapitrace? it is a program to capture opengl app output. you can see how many drawcalls a frame has. 10:17 erle 🇬regon for my intel integrated GPU there was a limit of draw calls where it became dog slow and software rendering was faster. maybe you also encountered a limit? 10:17 erle 🇬regon the trace can be shared too and replayed on other GPUs, it is only the rendering output of the app. 10:17 erle well, the opengl instructions 10:18 MTDiscord Because if it is. It might even be the os task scheduler messing with the process and putting it behind in the queue at the wrong time 10:19 MTDiscord Before we get too crazy. Maybe you can try to run with vblank_mode=0 luanti and change the nice value of the process to boost it wayyyy up 10:20 sfan5 did you try disabling post-processing in settings? 10:21 MTDiscord I dunno if Debian 9 supports feral gamemode, but if it does, you can add that in between the vblank parameter and the luanti command to run with as little limitations as possible. Gamemode also allows you to set the nice value on launch 10:21 sfan5 also transparency_sorting_distance = 0 10:22 erle what is “feral”? 10:23 MTDiscord https://github.com/FeralInteractive/gamemode 10:25 MTDiscord I especially use this in Transport Fever 2 so when things get crazy the game is still somewhat playable. And I use it in forza horizon 5 to run it maxed in 4k out with Raytracing for much better frame times 10:25 MTDiscord Maxed out in 4k lol 10:25 MTDiscord It is the sauce on the steak to make it extra great 10:26 erle 🇬regon do you know how to use git bisect? 10:27 [MatrxMT] <🇬regon> No 10:28 erle 🇬regon okay what about 5.7 does it also have low fps? 10:29 [MatrxMT] <🇬regon> I don't know, I didn't install it 10:29 [MatrxMT] <🇬regon> But the images that I send were not same 10:34 erle 🇬regon yes in the newer version triangles are not merged 10:37 erle 🇬regon so about qapitrace, can you install it and do two traces of the same scene? 10:37 erle i'd love to know if the drawcall count is different much 10:39 [MatrxMT] <🇬regon> All package managers that I installed haven't this (guix and apt) 10:39 MTDiscord My only question is, why are you running Debian 9 lol 10:41 erle 🇬regon can you install apitrace? 10:43 [MatrxMT] <🇬regon> Yes, I will install it now 10:48 [MatrxMT] <🇬regon> How to run it? 11:08 MTDiscord Yes it's as bold as a cup of fat free milk. Minetest doesn't even get my gpu to like 20% unless I start doing ridiculous stuff 11:10 MTDiscord But I'm not complaining. That's good. But also, that's bad. But it's also not the game's fault because hardware like this really didn't exist when the bottom of the engine was made 11:12 MTDiscord Roller is lurking in these here woods 11:13 erle jordan4ibanez regardless of what GPU you have, “the game is only using 20% of my GPU” is like complaining about how “we only use 33% of our traffic lights” 11:13 MTDiscord That means that the game has a horrendous cpu bottleneck because I'm running with vsync off 11:14 erle how much of your CPU is it using? 11:16 MTDiscord I dunno, this is a new cpu, I'll have to benchmark it again 11:17 erle 🇬regon “apitrace trace ” followed by the program you want to trace i think 13:23 MinetestBot 02[git] 04Desour -> 03minetest/minetest: Add a setting to group transparency sorted triangles by buffer (#15115) 13df4e70b https://github.com/minetest/minetest/commit/df4e70b2c7631317430bd78061507270d23db67b (152024-11-28T13:22:53Z) 13:48 [MatrxMT] <🇬regon> Apitrace types only that on both versions and nothing more: 13:48 [MatrxMT] <🇬regon> ``` 13:48 [MatrxMT] <🇬regon> user@user:~$ apitrace trace /opt/env.sh /opt/minetest-5.8 do /opt/minetest-5.8/bin/minetest 13:48 [MatrxMT] <🇬regon> apitrace: loaded into /usr/bin/apitrace 13:48 [MatrxMT] <🇬regon> apitrace: unloaded from /usr/bin/apitrace 13:48 [MatrxMT] <🇬regon> apitrace: loaded into /bin/bash 13:48 [MatrxMT] <🇬regon> [text not from apitrace] 13:48 [MatrxMT] <🇬regon> apitrace: loaded into /opt/minetest-5.6/bin/minetest 13:48 [MatrxMT] <🇬regon> [text not from apitrace] 13:48 [MatrxMT] <🇬regon> apitrace: tracing to /home/[username]/minetest.trace 13:48 [MatrxMT] <🇬regon> apitrace: attempting to read configuration file: /home/[username]/.config/apitrace/gltrace.conf 13:48 [MatrxMT] <🇬regon> apitrace: warning: unknown function "glSpecializeShader" 13:51 erle lmao 13:51 erle i did not konw that the matrix bridge could get thrown out xD 13:52 erle matrix is BANNED 17:41 shaft Can I detect player collision in a simple manner somehow? 17:41 shaft I tried hijacking on_step but it don't slap. 18:01 MTDiscord you can't 18:01 Krock shaft: player collisions are handled client-side. 18:01 shaft ok. I'll do something else 18:02 Krock although you could spawn an object where the player stands, hide it from everyone (observers) and use that to do physics 18:02 Krock it's a hacky workaround but could work to some degree 18:12 shaft For what I'm doing that's overkill. Checking for collisions is probably too 18:13 erle shaft what ARE you trying to do? 18:14 shaft I'll show you when I'm done 18:15 erle if you don't say what, it's difficult to help 18:59 shaft It says in the api documentation that register_on_leaveplayer "Does not get executed for connected players on shutdown." but it does. At least in singleplayer. 19:02 shaft Yep. It's wrong. Someone should update it. 19:03 MTDiscord someone could be you (feel free to make a pr) 19:29 shaft Does the doors mod support larger doors than 1x2 19:29 shaft ? 19:39 cheapie shaft: IIRC https://content.luanti.org/packages/FaceDeer/castle_gates/ can do larger ones if you're willing to put up with it 19:40 shaft No, I was just wondering whether I can be sure that doors:hidden is always the node above the base door node 21:21 cheapie When the server has mesecons but not digilines: https://cheapiesystems.com/media/images/screenshot_20241128_151532.png 21:28 cheapie FWIW it's based around the "MESEbus" protocol that I came up with a while ago - it's just a simple-ish synchronous (DDR) serial bus, pin A is clock and pin C is data. The LuaC in the bottom right runs the thing. 21:31 shaft I experienced some very weird behaviour and was debugging it for hours only to find out it was because i was using an an undeclared variable as index in table.remove() 21:32 shaft Despite this I like Lua. 21:35 MTDiscord shaft: (1) do you not read warnings? (2) configure luacheck properly and use it, don't have this problem ever again 21:36 shaft No, I should have 21:38 sfan5 warnings are invisible unless you read debug.txt :) 21:39 Krock chat_log_level = warning 21:39 MTDiscord sfan5: i mean there's an okay chance they have found the setting for making them appear in chat or are starting luanti in the terminal.. 21:39 sfan5 i am not sure how discoverable our settings are honestly 21:39 MTDiscord random related thought: we should either have better advice on who should set which settings or "setting presets" 21:40 MTDiscord Development tab 21:40 sfan5 or just a "developer mode" 21:40 MTDiscord Gasp, I love it 21:41 MTDiscord yeah "developer mode" could be one set of presets 21:41 Krock connected_glass = true <- this should become a default 21:42 MTDiscord You're right 22:07 shaft Why does Luanti not provide table.clear? LuaJIT even has an optimization for that. I can't require table.clear because of mod security 22:07 shaft Would be an easy enhancement 22:09 MTDiscord shaft: Because no one implemented it. Partially because it's not very important. 22:10 shaft You do it a lot and it would save time 22:11 MTDiscord shaft: These kinds of things can be, without major issue, done by pure Lua libraries like modlib. Hence it isn't very important. 22:12 MTDiscord There is a small case to be made because of the LJ thing but most of the time that doesn't matter either. 22:12 MTDiscord Not to mention that usually, when you clear tables, you can (and maybe should) instead replace the entire table with an empty table. 22:20 shaft I think it's faster. 22:22 MTDiscord shaft: faster than what? 22:22 shaft creating a new table 22:23 MTDiscord than the empty table thing? it can avoid reallocs in theory, but often it doesn't matter (i remember trying it once) 22:26 MTDiscord as said, we have much bigger fish to fry than stuff like this typically 22:27 MTDiscord if somebody cares about it enough to open a PR, then i'll look at it, because at that point it becomes less work for me and it demonstrates a certain level of importance 23:43 erle i am not sure if my memory is deceiving me, but i feel like that “move your head to look somewhere else, terrain starts being rendered (or loaded?)” wasn't always like that. is there some optimization that tries to send only mapblocks a player is looking at or moving towards? if so, when was it introduced and can i configure it server-side? 23:45 erle i found a funny bug in mineclonia: if you click on a boat that is on top of water with an empty bucket, you get to sit on the boat AND the boat starts sinking. amazing! 23:46 shaft Now I need a name. What's a good name for a mod that automatically opens and closes doors? 23:51 MTDiscord door-o-matic 23:57 erle shaft auto_door 23:57 erle shaft automatic_door 23:57 erle shaft smart_doors 23:59 shaft auto_door was my first thought too. 23:59 erle shaft if your mod does work look like this, i am disappointed: https://inv.nadeko.net/watch?v=o2ObCoCm61s