Time Nick Message 01:05 erle Intel HD (3000) is in fact capable of rendering dynamic shadows, as well as any other shaders. It just happens to be slow as hell. 01:06 erle similar to my general experience with hardware that *nominally* can do opengl 2.0 or 2.1 – it's really slow on the shader codepath 01:06 erle like, the i915/i945 is easily outperformed by mesa software rendering 01:07 erle (if you shade enough shading) 01:08 erle i think anyone who claims that the shader code path makes the game playable on hardware just because that hardware nominally supports opengl 2.0 is either lying to others or themselves. 01:09 erle i have just tested something and basically on the same hardware (that nominally supports opengl 2.1) the shaderless code path has a drawtime of 6 to 8ms and the shader code path – for the same scene, no effects – has a drawtime of 60 to 70ms 01:10 erle which is the difference between “playable” and “less than 15fps, into the trash it goes” 01:11 erle mesa even reduced advertised opengl 2.x for some hardware (e.g. integrated intel GPUs) for such a reason https://www.phoronix.com/news/Mesa-i915-OpenGL-2-Drop 01:11 erle > The reasoning for dropping from OpenGL 2.1 to OpenGL 1.4 for i915 is due to Chrome and other applications using really slow code-paths with these newer extensions available and thus provide a worse experience. 01:15 erle btw, i am pretty sure 15fps is basically software rendering if your CPU is not occupied with anything else 01:16 erle even for “garbage hardware” 03:40 Blockhead256 my "bouncer" is https://irc.minetest.net :P 03:40 Blockhead256 forums are down again.. but before they went.. "108 guests" in the stats at the bottom of the page 03:40 Blockhead256 that's definitely bad bots, right? No way 108 people were actually browsing the forums at this time of day, 03:41 Blockhead256 (when the Americans are nearly all in bed and the Europeans are fast asleep, and a lot of the rest of us are at work) 03:41 MTDiscord the americans aint in bed yet 03:42 Blockhead256 west coasters less likely so, and of course the people who stay up to midnight or so.. plenty of people in NYC are in bed already 03:59 Mantar only 8pm here in California 04:13 MTDiscord its 11 pm here 04:14 MTDiscord and im late time for whole country (eastern) 04:14 Blockhead256 I'm glad I brought all 3 of you out to comment, but have any of you been using the forums as a guest? or is it, as I was trying to point out, bots? 04:15 MTDiscord no but i am going to bed like immediately 04:25 MTDiscord goodnight 04:25 Blockhead256 good night :) 05:20 Blockhead256 Luanti should probably move to the XDG dirs instead of continuing to pollute home directories https://github.com/minetest/minetest/issues/15382 05:25 MTDiscord I dunno man, it's easier to type cd .minetest when you open your terminal 05:26 MTDiscord Plus, don't you want the user to feel at home when playing the game? 05:26 MTDiscord https://tenor.com/view/gary-spongebob-spongebob-meme-drum-gif-13282653510173043160 11:54 erle https://github.com/minetest/minetest/pull/15210#issuecomment-2384314803 11:54 erle > If performance with the same set of visual effects is worse with shaders enabled than with shaders disabled, please file a bug report. 11:54 erle apart from the predictable response being “oh your hardware must be garbage”, i am banned from filing bug reports 11:54 erle so much for that 11:56 erle luatic but if you are keeping a list, the Radeon X1400 is one GPU that can do shaders and the engine unplayably slow when they are activated 12:59 specing That GPU is old enough to drink 13:31 ROllerozxa meanwhile the chromebook which I got to keep for basically free from my old school (now corebooted and runs regular linux) runs minetest perfectly fine with shaders enabled. no performance drop compared to having shaders disabled. and since it supports all the way to OpenGL 4.6 it could probably take advantage of whatever cool shader based performance improvements that may come when the shaderless codepaths can be completely axed 13:32 erle ROllerozxa can you explain what about the shaderless code path inhibits performance in the shader code path to help me understand that? 13:32 erle also, “making it work better on hardware i have while dropping support for hardware others have” is a bit of a myopic stance 13:33 erle ROllerozxa i don't think i own a computer that can do opengl 4.6 even, are you very young? 13:33 erle like, did you go to school in like the last few years? that's the better question 13:35 specing Only apples are the real winners in a potato contest. 13:35 specing erle: has the driver situation improved in those 19 years or is it still a disaster (blobs, proprietary init)? 13:42 MTDiscord https://cdn.discordapp.com/attachments/749727888659447960/1302991325875929160/Screenshot_2024-11-04_at_8.41.35_AM.png?ex=672a20bc&is=6728cf3c&hm=c4b006818700aad777e7715e67972777a67a0cd8f554f622f588df98c0c38b7f& 13:42 MTDiscord Reading the message in Discord, the "New message" indicator bar caused me to read "apples are the real winners in a potato contest" completely out of context. 13:43 MTDiscord causing me to imagine some farmer sneaking disguised fruit into some state fair vegetable-growing contest. 13:55 ROllerozxa erle: well my point is that there is a big flood of old chromebooks right now, especially ones that have dropped support from google, and IT departments just want to get rid of them through any means necessary. but any x86-based ones can be corebooted and run regular linux for anyone who needs a computer on the cheap. the specs on most chromebooks aren't very great but what they are is accessible so as long as minetest runs on these 13:55 ROllerozxa , which they will do even with the shaderless pipeline dropped, then being able to play minetest remains accessible. 13:55 ROllerozxa also yes I guess I am very young, 19 years old 13:59 erle ROllerozxa are you saying that old chromebooks are the new “old thinkpads”? 14:00 rubenwardy We absolutely won't be supporting 19 year old hardware anymore. It's not about the power of the devices, it's about the support of more modern standards that can improve performance and increase capabilities. Where we draw the line depends on what is available in different graphics specifications and how many of our users remain supported 14:01 erle again, what about the shaderless code does actually inhibit making the shader code better? i mean it's not like you can have feature parity (that went out when i proposed to make shaderless shadows and it was rejected) 14:03 erle like, is that code somehow intertwined too much? 14:04 specing chromebooks do not have a track record of being quality machines, rubenwardy 14:04 specing ROllerozxa:* 14:04 * specing prefers paying a little bit more for a used ThinkPad 14:05 erle how much are used chromebooks vs used thinkpads? 14:05 specing good question 14:07 erle specing i have no idea about the internals of the radeon driver, but it works with debian? it needs a BIOS update though to not lock up at boot, as the driver seems to lock up the machines if the BIOS is too old 14:08 specing rm -r /lib/firmware -> no driver 14:10 erle specing it seems that firmware-amd-graphics is indeed non-free in debian (but redistributable obv.) 14:11 erle Description-en: Binary firmware for AMD/ATI graphics chips 14:11 erle it contains all the radeon blobs 14:11 specing Yes 14:11 specing 2024 and radeons still have no drivers 14:12 specing and, if you have a free bios, they don't even enumerate 14:12 specing because they require running a proprietary blob on PC init 14:12 erle well the solution to that is intel integrated GPU? or what? 14:12 specing Yes 14:13 specing up to (and perhaps including) skylake. After that, it's firmware hell just the same 14:13 MTDiscord alternative solution: don't care 14:13 erle luatic “i don't care” is the default stance 14:14 specing My GMA4500XHD runs blobless, with free init 14:15 MTDiscord erle: and for a reason, it's a decent stance 14:16 erle luatic can *you* maybe explain to me what in the fixed pipeline inhibits making the other thing bettter? like, where they are intertwined? 14:16 specing minetest devs have a hidden pact with gpu manufacturers to do planned obsolescence of old hardware 14:17 erle specing i think “devs tend to only care about hardware they have and sometimes they like to deprecate stuff of people they dislike” is a less conspiracy take on the situation 14:18 erle specing also if that is true where is the blockchai^H^H^H AI integration? :--DDD 14:18 erle can saturate much more GPUs that way!!! 14:18 specing That's really odd that noone did blockchain integration yet 14:18 specing mine coins while mining nodes.. perfect combination 14:19 erle i bet some assclown server owner did that already some time 14:19 specing and for that you'd need to optimise the rendering so maximum resources can be devoted to crypto mining 14:19 erle “install this CSM mining shitcoins while you mine nodes” 14:19 specing that'll happen when SSCSMs happen 14:19 erle i recently learned that when on skyblock server you share resources the cops come and threaten to grief your island. 14:20 erle “don't give resources to new players”, “do not share your island” 14:20 MTDiscord erle: I can't give a comprehensive answer yet but from what I've seen so far it's largely a thing of "it lets us afford not to care about the shaderless path which makes things much simpler". It may be debatable whether we could just roughly leave the shaderless code be though it isn't always trivial. 14:20 erle luatic i said before that i volunteer to keep it working (which would necessitate that it is not too much in the way of other things) 14:21 MTDiscord The whole modern shader pipeline thing isn't even just about performance, it's about correctness. The fix for alpha blending is well known, but only usable via shaders, and our workarounds for the legacy pipeline are incomplete, hacky, a pain in the ass to maintain, don't perform well, and are completely unrelated to the correct implementation. 14:21 specing I think that banning erle is a good thing 14:21 specing we need a continuation fork away from proprietaryhub 14:21 erle lmao 14:21 MTDiscord specing: weakest point in the history of points lmao 14:21 erle specing has adam12345 released the minetest-but-with-irrlicht thing already? 14:22 MTDiscord of all the points you could've made for what erle could do better, you chose to pick "they could migrate to a FREE git hoster!!1!" lmao 14:22 MTDiscord The idea of erle becoming a maintainer for a legacy-pipeline-focused fork, and allowing the core devs to focus on a shader-based pipeline ... actually sounds pretty good to me. 14:22 specing erle: no idea 14:22 erle Warr1024 yeah but for that you need the code to not be in the way of the core devs or else they hate it 14:22 erle oh, fork 14:22 erle i see 14:23 MTDiscord minetest-but-with-irrlicht sounds like a horrible idea. good luck with that. 14:23 specing but we really need a strong copyleft competitor to Microsoft Luanti 14:23 MTDiscord ? 14:23 MTDiscord erle: in theory, at least for a while, it might be feasible to just merge upstream improvements that DON'T touch the renderer into a fork and remain API-compatible. 14:24 ROllerozxa specing: I don't care if it is quality or not if it's a laptop with a somewhat functional battery I can carry with me and it's all I could get my hands on 14:24 ROllerozxa if you care about the "quality" then you likely have the money to get yourself a "proper" computer, whatever you'd think that is 14:24 MTDiscord I also never throw away old hardware until it either physically breaks, or I don't have the space to use it, and I also really like having future versions of software continue to work on them indefinitely. The virtual elimination of built-in obsolescence is a feature I love from FOSS ... but it obviously has certain limitations. 14:25 erle Warr1024 btw i would believe the correctness angle more if patches addressing correctness issues were appreciated more. remember the gamma discussion? or the mipmap improvements? i don't think correctness is *fun* for developers 14:25 specing ROllerozxa: used thinkpads aren't that expensive, but you're unlikely to get one free 14:25 erle just in case anyone now cares about gamma: https://blog.johnnovak.net/2016/09/21/what-every-coder-should-know-about-gamma/ 14:25 MTDiscord I've got plenty of 32-bit stuff, or stuff with tiny amounts of RAM, that are finding narrower and narrower uses as time goes on. Gaming is one of the first things I tend to lose and have to move to newer hardware. 14:26 erle warr1024 does nodecore use any advanced features that require new graphics? i am not too well acquainted with it 14:26 erle recently i played nodecore skyblock again and … remembered why i stopped nodecore skyblock 14:26 erle i praise and curse kimapr for making it hahahaha 14:26 MTDiscord anyways i gotta go now, be back in a couple hours 14:26 MTDiscord Even if my old machines don't stop being useful at any specific point, though, it doesn't mean I'm never forced to get new hardware for new use-cases. FOSS still makes this as painless as possible, and I can get away with very cheap refurbs of machines that are a few years old, so the amount of economic privilege it requires is way less than in the proprietary world. 14:27 MTDiscord NodeCore doesn't require fancy features, but it does benefit from a few if you can enable them, like the auto-brightness shader making it a bit easier to navigate in caves without having to fine-tune your gamma settings. 14:27 erle well gamma is broken anyway isn't it? 14:27 erle or has that changed? 14:28 erle Warr1024 i think within the next 5 to 10 years or so the usefullness of x86 machines will sharply decine, because of 32 bit issues 14:28 MTDiscord However, NodeCore also has a lot of design compromises to account for the fact that certain basic rendering features, like alpha blending and depth buffering, only KINDA work, and if MT were to modernize its pipeline to the point where those could be fixed, I'd probably eventually rely on them actually working. 14:28 erle oh, where do you want to rely on blending? 14:29 erle IIRC blending semitransparent textures will *never* look right unless the gamma issues are addressed and that was decided as “won't do” 14:29 MTDiscord I wouldn't say gamma is "broken" so much that gamma isn't "gamma" 😆 It's a functioning way to adjust the brightness of middle-intensity lighting areas without changing the zero or maximum points. It's probably not calibratable, but at least it lets you work around the "my midtones are too dark" issue. 14:30 MTDiscord Places I want to be able to use alpha blending are things like liquids, entities, and particles. For instance, having particles be visible when viewed through a liquid surface. 14:30 erle Warr1024 left is gamma-correct blending https://blog.johnnovak.net/2016/09/21/what-every-coder-should-know-about-gamma/img/opacity.png 14:31 erle oh, you are talking transparency, not “colors/brightness is wrong” 14:32 erle Warr1024 like order independent transparency? 14:32 MTDiscord When I say "blending is wrong" I don't mean "they used the wrong amount of A and B to create C", I mean "when A goes behind B it's fucking gone" 😆 14:32 MTDiscord yeah, OIT would be great 14:33 erle Warr1024 let me test this 14:33 erle i thought it was fixed, but maybe i remember wrong 14:41 erle Warr1024 you are correct! i can't see particles outside of liquids. so no underwater pummeling i guess? 14:43 MTDiscord The way things are, it's not fixable, because we rely on face sorting, and you can only sort faces that are in the same lists, and we have different lists for different purposes. Sorting all faces would require us to break apart the mapgen meshes, throw them into a single list with particles, entities, etc. and sort them all. Every frame. Then deliver all that to the rendering pipeline after having lost any reusable batching across 14:43 MTDiscord frames. 14:44 erle batching lol 14:44 erle are particles batched by now? 14:45 MTDiscord I ... think so? As of 5.10 or something, there are a few things that are batched that weren't before... 14:46 erle well at least then lagging out people by means of particles is no longer a concern 14:46 erle hacked clients have for years now had an option to disable rendering particles at runtime 14:46 MTDiscord much less of a concern :-) 14:46 erle to prevent getting lagged out 14:47 MTDiscord I actually almost went and wrote a server-side mod that would redefine add_particlespawner to track how many total expected particles there were and start reducing counts to try to keep it within a reasonable total limit across all effects. 14:48 erle Warr1024 question: why is sorting by distance necessary? like, can't you just first draw all the opaque objects and then draw all the transparent ones from back to front? is the issue with that that the display lists are by object type and not if it is transparent or not? 14:49 erle keep in mind i have no idea about anything 14:49 MTDiscord sorting is necessary because of the "from back to front" part. 14:50 erle so if you had order independent transparency, then not 14:51 erle i wonder why particles are visible through glass, but not through liquid 14:52 MTDiscord because glass uses clip transparency instead of blending. They affect the depth buffer differently. 14:53 erle oh, so glass is either transparent or not? 14:53 erle pixel-wise i mean 14:53 MTDiscord tbh clip transparency is a completely different performance train wreck (you have to do a texel lookup before you can tell if the pixel will be rendered, so looking through the transparent parts of a clip-transparency texture still incurs a lot of the rendering costs as if you actually drew it). 14:54 MTDiscord yeah, that's what "clip" transparency is. No blending or anything. If alpha >127, it becomes completely opaque, if <=127, it's fully transparent and the RGB and depth buffers aren't updated. 14:55 erle Warr1024 what do you think of this approach? https://vtk.org/Wiki/VTK/Depth_Peeling 14:55 erle is it too slow? 14:56 MTDiscord Reading the "Cons" section, that might be a problem, yes, but the "limited to recent OpenGL" thing might ironically be worse in some ways. 14:57 MTDiscord Like, all this started as an argument about "why are we dropping support for legacy rendering pipelines" and that doesn't sound like it'd avoid that. 14:57 erle no, i just wanted to know what you would do *instead* 14:58 erle or propose instead 14:58 erle also, am i correct about no underwater pummeling? 14:58 erle or hmmm 14:58 erle let me check something 15:01 MTDiscord you can pummel underwater but if you're not underwater yourself, you won't see the particles unless they breach the surface. 15:03 erle yeah i just checked, particles underwater are visible, but not unless you are ALSO underwater 15:04 erle Warr1024 also keep in mind, the page talking about “recent GPUs” was last updated in 2010 (not that it matters, but i found it funny :P) 15:08 MTDiscord The underwater thing is basically "if an alpha-blended surface exists between you and the particle" If you're not underwater, and the particle is not underwater, but there's a waterfall between you, it's a no go there too. 15:09 erle yeah i checked that out with devtest just now 15:09 erle btw, in case i have not mentioned it: for *windows* the “these machines are too weak” thing is absolutely correct. for some reason you can get much more 3D performance out of old hardware using linux. 15:10 erle (i am not going to install windows to substantiate it, but fleckenstein once tried and failed) 15:10 erle probably because mesa is pretty good at backwards compat 15:11 erle (last time i asked about a mesa bug about my old intel GPU, the answer was “are you using the old driver? there is now a new one!”) 15:12 erle Warr1024 do you have any kind of note collection that describes the engineering decisions of nodecore? given you seem to dot all the i's and cross all the t's (as they say), i'd like to read it. 15:12 erle or is it the source comments like “keep this compat hack until we require engine version $X” that are the only public hints? 15:13 erle and chat comments ofc 15:14 MTDiscord https://gitlab.com/sztest/nodecore/-/blob/master/docs/releng.txt 15:14 MTDiscord There are actually a few things in there I need to update, the MT/luanti rename being the least of them 😄 15:15 erle https://gitlab.com/sztest/nodecore/-/tree/master/docs this is interesting, thanks! 15:18 erle Warr1024 if you ever do a talk on nodecore, please tell 15:18 erle i am always interested in projects where a single person tries to have a coherent design 15:19 erle the discussions in e.g. mineclonia are also much more deep than i see from other projects, but ultimately it's more a team effort to have more features (i.e. not as strict and not as narrow of a vision) 15:19 erle Warr1024 it reminds me of this https://en.wikipedia.org/wiki/Film_d%27auteur 15:22 erle Warr1024 also thank you far taking my question seriously and answering it! 15:24 sfan5 >you have to do a texel lookup before you can tell if the pixel will be rendered, so looking through the transparent parts of a clip-transparency texture still incurs a lot of the rendering costs as if you actually drew it 15:24 sfan5 hm? 15:25 sfan5 a texel lookup is needed anyway for texturing and apart from 'discard' being bad for performance on some mobile gpu's clip alpha is good because the depth buffer will do the sorting for you 15:26 sfan5 ...actually I'm talking past your point 15:26 sfan5 but "drawing something and making it transparent is worse than not drawing something" isn't the widom it seems 15:40 MTDiscord There's no such thing as a large, complex project that's run by a single person; a challenging enough project will just surface internal conflict within that person. You'll have arguments between past-you, present-you, future-you, and between you-as-a-designer, you-as-a-programmer and you-as-a-systems-engineer. 15:40 MTDiscord The way I see it, you can continue to pretend that you can keep it all sorted out in your head, or you can accept reality and start taking notes 😄 15:41 MTDiscord Also, Conway's Law is very interesting when you realize that it applies to individuals as well. 15:53 erle warr1024 look, things like this is why i want to know about your engineering decisions/process 16:05 MTDiscord if you make good git decisions you don't have to take that many notes 16:07 erle sick burn hehe 16:12 erle are you by any chance opposed to 109k line commits cora? :D 16:17 MTDiscord to do good git you first have to git good 16:33 MTDiscord > are you by any chance opposed to 109k line commits cora? 😄 uh yes obviously 16:34 MTDiscord i am even kind of opposed to 109k line PRs lol 16:35 MTDiscord however i guess this is not the place to talk about good git decisions 😛 16:35 erle leave your git history hygiene at the door pls 16:36 MTDiscord (or it will be removed by force) 16:38 erle by force squash hahaha 16:38 MTDiscord i really do not understand the reasoning behind that 16:39 MTDiscord i mean if the git history of a pr is total suck you can always say "let's squash this" on a case by case basis 16:39 MTDiscord but ultimately you mostly just remove information for, if anything a little aesthetic gain 16:40 MTDiscord and the thing such a policy encourages people to make total suck git history 16:41 erle grorp btw, i see you work on texture filtering. i have noticed that ALL fonts are slightly blurry on my system now. but only with unifont (my preferred font) i get absolutly awful rendering. 16:41 erle well, they are slightly blurry in minetest/luanti 16:41 erle while they weren't before 16:42 erle or maybe they were and my eyes/hardware/software changed, who knows ;) 16:42 erle (i am pretty sure it worked at some point in the past) 16:42 MTDiscord my friend keeps implying luanti is an exotic fruit not a video game i might make a mod with luanti trees 16:42 MTDiscord or does it grow on bushes? 16:43 erle make a mob, the luantick 16:43 erle it does nothing, except if a player is standing on a rug 16:43 erle then it pulls it 16:43 erle ^_^ 16:43 MTDiscord lol 16:44 MTDiscord don't be so mean you will infuriate the luantciks :p 16:44 erle luna chicks 16:47 erle does anyone have any guesses why repixture is a great game but has very few servers? i found one that said it encountered an error when trying to connect 16:48 MTDiscord bc agpl :p 16:49 MTDiscord > encountered an error when trying to connect i think that means it crashed 16:50 erle corarona seems i need to make an AGPL mod download mod that sends the modified source code DIRECTLY into the users brain 16:55 [ specing: skylake (for GPUs) is fine, it only requires blobs for h265 decoding which you don't need and didn't exist before so you're not losing anything that previous generations have 16:56 [ I think it's since alder lake-p that the blob is actually required 16:56 [ according to https://wiki.archlinux.org/title/Intel_graphics#Enable_GuC_/_HuC_firmware_loading 16:56 MTDiscord erle idk if that will help i think the issue is more like server owners feel "nah don't wanna get involved in licensing bs and agpl makes me" 16:57 specing server owners feel "I wanna locally patch luanti and agpl prevents that, boooo AGPL!" 17:02 [ cheapie: I get the reference, I think. (dunnet has Robert Toukmond and Thomas Stock, but where do the middle name initials come from?) 17:39 MTDiscord A 100kLOC commit is like the size of what you'd get if you smashed like 4 or 5 large, complex projects into a single repo, and then squashed the entire repo into a single commit. I think I can say I'd be opposed to 100kLOC commits too. 18:05 MTDiscord I'm not the biggest fan of it either. Now when I'm working on Irrlicht stuff the history I get is "Copy irrlichtmt to /irr/" 🙃 18:16 erle luatic for future reference, the correct way to do this is called a subtree merge and several people (me included ofc) pointed that out before. it was ignored by the-powers-that-be. 18:16 MTDiscord erle: i'm aware 18:17 erle well, just like “git add -p”, i often encounter people who do not know it 18:17 erle (names have been withheld to protect the guilty) 18:18 MTDiscord "git add -p" has become obsoleted by powerful editors 18:18 erle wdym? 18:18 erle emacs magit or so? 18:19 MTDiscord VS Code or its FOSS variants 18:20 MTDiscord as well as your average IntelliJ IDE, though I don't like those 18:21 erle i bet this is software that uses like 1GB for a single text file or so 18:21 erle one of my favourite tests for an editor is “open a large XML file (like multiple megabyte) and see if the syntax highlighting makes the computer lag” :D 18:23 MTDiscord obviously "does it make a computer lag" isn't a very useful metric when there are a wide variety of computers, and when you have to nail it down to a specific computer or range of computers, that kind of reveals the underlying assumptions. 18:24 erle well if i have 50MB of XML (and i once had multiple of those files, years ago, at work) it should be possible to edit this on any thinkpad from the last 20 years 18:24 [ vs code is electron junk 18:24 MTDiscord honestly I've never used add -p for the same reason I never use the "powerful editor" equivalent: there's not much value in me making a commit that doesn't correspond to code that I've actually written and probably tested. 18:25 erle i think “git add -p” is more a tool that makes people not commit their editor configs or accidental changes ^_^ 18:28 MTDiscord For editor configs, the correct tool is probably .gitignore. For accidental changes, the correct tool is probably aborting the commit, carefully fixing the accidental changes (after possibly making a stash copy to ensure you don't accidentally revert the non-accidental changes), and then testing it again to make sure the accidental changes weren't more needed than you assumed, and THEN committing 😆 18:30 erle you are arguing in favor of being more careful to a chatroom in which the most polite reply to “a recent change broke something” is considered to be “well, you have garbage hardware” 18:31 erle snark aside, i do actually teach people “git add -p” so they stop doing “git add *” or whatever leads to their config files going into commits 18:33 [ you don't even need -p for that 18:33 MTDiscord The thing about -p and editor files is that the editor files tend to stick around, so -p only solves it for that one commit. It's too error prone to be a long term solution. 18:34 [ add to gitignore and use git add ., or add files individually 18:34 erle yes. gitignore is 18:34 erle technically you just need discipline, but i can not replace the people who i show something 18:34 erle i can just teach them 18:34 erle some will appreciate it, some will not 18:34 erle and sometimes they teach me something too! 18:34 erle i love those cases 18:34 MTDiscord It's fine for somebody who wants to do their first and possibly only contribution to a project and doesn't want to take the time to set things up properly, but by like the 2nd or 3rd time they do it, they really should do it the right way, and there's a question then about whether you'll be around to give them the right advice at that time. 18:34 erle true 18:35 erle i shall mention .gitignore the next time i do it 18:35 erle thanks! 18:35 MTDiscord Sometimes having discipline is about putting in the work when it's needed, and sometimes it's about putting in the work now because you know you won't always do it when it's needed later. 18:36 MTDiscord lol, yeah, "if you find yourself using git add -p again later, you should probably look up gitignore" is probably a good addendum 😄 18:37 erle Warr1024 besides air_equivalent (which i think should be improved instead of removed, so that it actually counts as air for some engine functions), which things did nodecore do that ended up on the chopping block (or close to it)? 18:46 MTDiscord You mean like things I added to the game but was forced to remove due to engine changes? Or do you mean experiments I tried and later removed because they didn't work out well? 18:47 erle the first 18:49 MTDiscord I don't remember ever having clearly had a case of that. 18:50 MTDiscord I've had features I've backed out because engine support I expected to be there was buggy or incomplete. 18:50 MTDiscord For example, skybox navigation was significantly curtailed due to the texturemod memory leak problem, and pickpocketing I canceled when I discovered that raytrace can't hit attached entities. 18:51 MTDiscord er, server-side raytrace can't, even though client-side you can still point at and punch them. 18:52 MTDiscord There have been changes I haven't agreed with, like when shadows were first added and they were 100% in end-user control with no way for a game to disable them, and that caused people to create nonsensical shadows in nodecore that they claimed to be a game bug, not an engine/config bug. 18:52 MTDiscord Situations like that though have generally been resolved. 18:55 MTDiscord sub tree is good howver you have to get used to doing bisect skip when you end up in a subtree commit 18:55 erle corarona that's good advice! 18:56 erle wsor4035 you tagged issue #15289 – in case you care about it: the submitter of the issue is mistaken about this being a bug and “fixing” it would break a bunch of things at the map border, e.g. schematics that are placed at boundaries and have a postprocessing step that expects the entire schematic to be presented etc. pp. 18:56 ShadowBot https://github.com/minetest/minetest/issues/15289 -- minetest.get_node() returns real nodes for coordinates outside the map (but adjacent to it) 18:56 MTDiscord also rebasing a subtree branch becomes more or less impossible 18:57 MTDiscord i am pretty sure erle is the only one who really knows whats happening at the map border 18:58 erle nah, many people know implicitly and i am pretty sure paramat knows (or knew?) 18:58 MTDiscord although they have bestowed some of that wisdom to me 18:59 erle “know implicitly” as in “they write code that works because the behaviour at the map border is carefully engineered by whoever made it, so that it looks not too different from anything else unless you go WAY too far out” 19:00 MTDiscord if the map wasn't so tiny it wouldn't even matter that much 19:00 MTDiscord but as is everyone with a tiny bit of patience can easily reach it 19:01 erle i remembered when i visited a map border on a server with a boat to go out of bounds 19:01 erle but there already was a boat out of bounds 19:01 erle (this was not a time travel story) 19:02 MTDiscord lol 19:02 erle best map border bug was obviously the crashbutton 19:02 MTDiscord yes ❤️ 19:02 erle now i wonder if that ever worked with mesecons hmm 19:03 erle would be a shame etc. 19:03 erle brb testing 19:03 ireallyhateirc Hmmm I'm reading the bug report and it seems that it's simply a manifestation of the +/- 1 node overgeneration which causes the other bug: https://github.com/minetest/minetest/issues/9357 19:04 MTDiscord erle and cora being cute 19:04 MTDiscord https://cdn.discordapp.com/attachments/749727888659447960/1303072315441156156/crashbutton.png?ex=672a6c29&is=67291aa9&hm=345938ea04e6a7b6a152a772b696fd2050a91441376f27d0659139060916cd97& 19:04 erle oh lol i forgot that 19:04 MTDiscord i blew you up accidentally shortly after iirc 19:04 erle well we did want to blow up the map border 19:05 erle but we accidentally blew up the entire server 19:05 erle and myself 19:05 MTDiscord hehe 19:05 erle brb checking out if mesecons is bugged 19:05 MTDiscord this sounds like a typical historic mcl bug though 19:06 * SwissalpS imagines mesecons being bugged as in spy movies, sending home info about players 19:07 MTDiscord well certain games send info about players to all other players on certain occasions unrelated to mesecons though 19:43 erle i was not able to crash anything using mesecons, but i only tried a few minutes 21:40 erle I'm not the biggest fan of it either. Now when I'm working on Irrlicht stuff the history I get is "Copy irrlichtmt to /irr/" 🙃 21:40 erle ah, yes, that just happened to me too 21:42 erle hahaha 21:43 erle and then i end up at “Move source/Irrlicht/ to src/” so i have to figure out the magic incantation that makes “git log” rename-aware 21:43 erle for reference: git log --follow "${FILENAME}" 22:31 erle i found a fix for “ Textures of TGA Type 1 with color format A1R5G5B5 seem to have their transparency bit ignored.” 22:35 erle if someone can guess what it is, they will be as amused as i will be 23:02 cheapie [: The middle initials were just selected at semi-random, it didn't sound right without them. 23:03 cheapie (my justification at the time was D for dunnet and E for emacs, but that's about as good as random) 23:11 wsor4035 erle: dont care about your spam 23:13 MTDiscord I don't think erle's remark was spam at all, though it was perhaps slightly rude to ping you just because you added a tag to that issue 23:15 MTDiscord yeah, i would say to post it on github as it might be relevant, but oh wait, there fortunately banned there 23:15 MTDiscord more so pinging for just tagging something is rather ridiculous 23:16 MTDiscord (unrelated to the tagging itself) 23:21 [ when will luanti move away from github to something freedom-respecting, like codeberg? 23:22 erle wsor4035 i thought you were the last person to interact with the issue, therefore someone interested in it 23:29 erle also you did not notice that it was not a bug, even though the topic has been discussed to death, so i thought i was helping. i was obviously mistaken. 23:29 MTDiscord no, im a triager, i label issues. 23:29 MTDiscord occasionally i comment if i have something of value 23:30 MTDiscord ones i care about are usually the ones i will follow up on/create myself 23:30 MTDiscord *follow up on outside of tagging/triaging role