00:02 < bridge> some maps use it 00:09 < bridge> chillerdragon: do u still play ? 00:15 < bridge> i sometimes see chiller on kog 01:07 < bridge> some new AI results from coding competition, openai destroyed deepmind lol, and solved the problem that none of the human teams got. 01:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418010429552070707/G1EF-ega0AAm9Ie-1.jpg?ex=68cc909f&is=68cb3f1f&hm=84fb5936664d638e680acd46d18a5b38c7066e59f1a64609b900c47eccf86fa2& 01:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418010429887877161/unnamed.webp?ex=68cc909f&is=68cb3f1f&hm=c1963a8b5f0d5166c7ac20988363b9b5de667abafabbde770284487aab186c6b& 01:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418010430231806013/image.png?ex=68cc909f&is=68cb3f1f&hm=1dfeda98ed0d1cff607435ec2b3b408ebdfb641750c54ba9893dcaa79f7168fe& 01:10 < bridge> <8_9z> how can i dawnload kog maps?? 01:11 < bridge> <8_9z> https://cdn.discordapp.com/attachments/293493549758939136/1418011382233956473/IMG_20250918_020955.jpg?ex=68cc9182&is=68cb4002&hm=a75626de89e60a89ee80daaae07e270edd8597adca73eec895ac22b445041d4c& 01:11 < bridge> $kog 01:12 < bridge> to be fair to the human teams, they have 3 people with 1 laptop, with a time limit to solve 12 problems. So it's kinda embarrassing that google couldn't beat them. 01:15 < bridge> apparently just regular GPT5 solved 11/12 problems, but they had to use "special model" for the hard one. 01:18 < bridge> zamn 01:23 < bridge> How many instances of GPT were they competing against? 01:23 < bridge> the humans you mean? 01:24 < bridge> yeah I'm wondering if there were multiple bots running in parallel 01:24 < bridge> the humans weren't competing with the AIs directly 01:24 < bridge> oh ok 01:24 < bridge> it's not like the 1st place human team lost because openai did better than them 01:24 < bridge> I've seen a couple of these competitions with AI specific categories 01:25 < bridge> where they do compete in parallel 01:25 < bridge> but from what they've said I'm presuming they did all 12 problems in parallel because why wouldn't they, not that it really matters because you can just add up all the times to figure out how long it would take if they did it sequentially. 01:25 < bridge> they still win sequentially 01:26 < bridge> yo 01:29 < bridge> they're very light on the details but from what they've said it *sounds* like for each problem they probably had many many instances of GPT5 with super long resoning budget to solve it, and then they had another model look at all the solutions that those made to pick one to submit 01:29 < bridge> they're very light on the details but from what they've said it *sounds* like for each problem they probably had many many instances of GPT5 with super long reasoning budget to solve it, and then they had another model look at all the solutions that those made to pick one to submit 01:30 < bridge> I played minesweeper all day and didn't do shit. What do I do 01:32 < bridge> https://www.theblackbox.games/ 01:33 < bridge> should compare with the total number of tokens somehow 01:38 < bridge> I think counting tokens is a bad way to think about it. It's a black box that takes a certain amount of compute and time to give an answer. You shouldn't care about how it gets there other than the time and compute it takes. 01:40 < bridge> Just compare energy consumption and lose against ppl 01:41 < bridge> idk 01:41 < bridge> eeh, but then comparing against humans is kinda not useful imo 01:41 < bridge> burgers take a lot of energy 01:41 < bridge> its impressive that they were able to solve all 12 problems though 01:42 < bridge> people said comparing chess AI to humans wasn't fair because the chess AI didn't really understand how to play chess, it just brute forces a lot of positions. 01:43 < bridge> true 01:44 < bridge> hm i guess i'm trying to say comparisons like that sometimes feel misleading like they're trying to say gpt5 can think better than a human expert 01:46 < bridge> I mean again, this was a hard format for the humans, it's also *just* the smartest university students in the world. not the smartest competitive programmers, who are significantly better. 01:47 < bridge> oh wait is that true 01:47 < bridge> im not really tapped into the competitive programming world 01:48 < bridge> make no mistake tho, there's probably less than 20 humans who can actually solve competitive programming problems that the best models can't do with ultimate resources 01:48 < bridge> also i wonder what the special model looks like xd 01:49 < bridge> if ai can do all of this whern can it make me a goddamn sandwhich 01:50 < bridge> I'd guess that it's fundamentally similar to GPT5, but just their next training run that they haven't published yet, or they mixed in more competitive programming/math problems than the regular GPT5. They're all just iteratively improved versions of the idea that they had with o1 01:50 < bridge> GPT is really just o5 01:50 < bridge> if ai can do all of this whern can it make me a goddamn sandwhich 01:50 < bridge> () 01:50 < bridge> GPT5 is really just o5 01:51 < bridge> insert a circle in the traingle factory but its all the gpt versions and theyre not evne on the conveyer belt theyre just there 01:52 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418021810863214592/image.png?ex=68cc9b39&is=68cb49b9&hm=274ab5589b96077dd0722a27416178226e8310d2b8e284690b0e0a4094b8ca41& 01:58 < bridge> also I saw someone say that the reason none of the teams solved problem 3 was not because they couldn't but because it's a really time consuming problem that requires a lot of code, so they all skipped it. but the LLMs can write 1k loc in a few seconds so they had an advantage. 01:59 < bridge> the actual hardest problem had a few human solves 02:03 < bridge> Simulation problems 02:05 < bridge> can any dev help me? 02:05 < bridge> https://discord.com/channels/252358080522747904/342454474117218334/1418024244826210367 02:05 < bridge> i tried to do it myself 02:05 < bridge> failed miserably 02:07 < bridge> I love gambling 02:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418025666338623601/image0.png?ex=68cc9ed0&is=68cb4d50&hm=05efa6898e16b144ad4632348a7a40ea8753b0901dd218f547a4847bae9cd63c& 02:28 < bridge> I'm so stupid I could've just checked how many bombs were left 03:44 < bridge> I have never done competitive programming but it seems like a somewhat useless pursuit to me 03:44 < bridge> I saw the libstdc++ maintainer complaining about some data structure library he wants to deprecate and the only people complaining about it are competitive programmers who need it to fit in some rules about what tools preinstalled that you are allowed to use 03:55 < bridge> I mean it's kind of objectively useless if you isolate just the task of competitive programming, it's not like they're creating something useful. It's a sport, or game, however you want to call it. 04:27 < bridge> I will competitively sleep now gn 04:31 < bridge> @jupeyy_keks quick question. If I upload a texture to the gpu is there a way to update single pixels without reuploading the whole texture? I feel like I have asked this before but I forgot the answer. 04:33 < bridge> same 04:33 < bridge> not rly that's why it's hard 04:34 < bridge> glTexSubImage2D is how you do it in opengl 04:34 < bridge> I still think it's quite slow and not recommended 04:35 < bridge> I'm almost positive that's because it converts it back to "software form" for alteration 04:36 < bridge> best analogy I have is SDL_Surface vs SDL_Texture 04:37 < bridge> the former's image data lives in CPU managed memory and it worked in by the CPU, the latter will only hold a handle 04:38 < bridge> the former's image data lives in PU managed memory and is worked on by the CPU, the latter will only hold a handle 04:38 < bridge> the former's image data lives in CPU managed memory and is worked on by the CPU, the latter will only hold a handle 04:38 < bridge> yes and I think most APIs also have the notion of an "immutable texture" that you cannot change 04:39 < bridge> that's the difference between glTexStorage and glTexImage 04:39 < bridge> apparently it helps some things work faster if the GPU knows you cannot write the texture 04:39 < bridge> that's the difference between glTexStorage and glTexImage I think 04:39 < bridge> apparently it helps some things work faster if the GPU knows you cannot write the texture 04:39 < bridge> definitely 04:40 < bridge> probably varies across different APIs too opengl exposes some weird stuff more modern APIs don't try to do because it's not performant or transparent enough 04:44 < bridge> that just exists on a spectrum relating to openGL version tbh 04:46 < bridge> i think most simple programs would be performant in modern OpenGL as vulkan if they were managing their resources and shaders the same way as GL does in vulkan 04:47 < bridge> the difference just exists in the level of control they give you, so vulkan has the potential to be more performant but the latest open GLversion is probably gonna have all of the same features which don't rely on that explicitness to be worthwhile 04:48 < bridge> i think most simple programs would be equally performant in modern OpenGL as vulkan if they were managing their resources and shaders the same way as GL does in vulkan 04:48 < bridge> i think most simple programs would be equally performant in modern OpenGL as vulkan if they were managing their resources and shaders the same way as GL does within their vulkan code 04:49 < bridge> i bet there's a marginally less slow way to do it vulkan specific lol 05:16 < bridge> You could upload the pixels you want to change to the gpu in a small buffer and then use a compute shader to modify the texture. 05:16 < bridge> If you really need it to live in gpu memory 07:42 < bridge> Sure, but if the texture is in use you have to expect race race conditions 07:48 < bridge> @louis.place: @soulyvevo I play Teeworlds an average of 4 hours a day I would say. I played less when I started playing. And I took a 1 year break in 2024 but not by choice. I would call my self quite active! 07:49 < bridge> It’s all about the game 08:09 < bridge> -^ Tested out on Nightly. 08:16 < bridge> gm nix users have you ever experienced nixos-rebuild taking over 90GB at multiple points throughout evaluation (frees between) only for it to segfault? 08:17 < bridge> :justatest: 90gbs? 08:17 < bridge> the segfault is like the icing on the cake 08:17 < bridge> 64 of real memory, 24GB of swap 08:18 < bridge> previously the build ended thanks to the OOM killer so I closed everything else 08:18 < bridge> I even have zram enabled 08:18 < bridge> zamn 08:19 < bridge> seems like it has to be a bug but the fact it survives and goes back down in memory usage makes it non obvious 08:20 < bridge> unless the nix process garbage collects itself but that'd be crazy and kinda gross considering it's just c++ 08:20 < bridge> I tried nh but obviously I can't do everything needed as my user, so I need to elevate 08:21 < bridge> I usually elevate nixos-rebuild with sudo but nh insists you call it as your user and it will handle calls to sudo 08:21 < bridge> but it just misses one and the build fails anyway 08:21 < bridge> permission denied 08:22 < bridge> ngl that's some black magic shit for me lmao 08:22 < bridge> I have a friend who nags about me trying out Nixos to get more familiar with c++ but what you mentioned now actually discourages me lmao 08:22 < bridge> :kek: 08:23 < bridge> well nixos has been the most predictable OS I've had to date 08:24 < bridge> it's obvious unable to fix a lot of issues with linux itself but compared to other distros it seems to have the best set of packages maybe ever 08:24 < bridge> kek 08:24 < bridge> gentoo ftw 08:25 < bridge> and if it's not yet up to date in the unstable channel, you can often very easily specify an exact version 08:25 < bridge> morning 08:25 < bridge> gm 08:25 < bridge> mornin ☕ 08:25 < bridge> a nixos-rebuild takes a lot more time compared to the package upgrade & reboot most distros have 08:26 < bridge> but I definitely do not have the patience to emerge as often as I uld need to 08:26 < bridge> but I definitely do not have the patience to emerge as often as I would need to 08:26 < bridge> u know what i do? 08:26 < bridge> i run a script that updates and then shuts down 08:26 < bridge> i run it when i go to sleep 08:27 < bridge> my PC is loud and ~2m from my head when in bed 08:27 < bridge> gotta have it off 08:27 < bridge> well atleast i dont get segfaults 08:27 < bridge> or the lazier way, autorun it on startup 08:27 < bridge> yea 08:27 < bridge> :justatest: 08:27 < bridge> no, that doesnt make any sense? 08:27 < bridge> xd 08:28 < bridge> u should recognize tho it is equally likely for a segfault to slip into emerge, as it's c afaik 08:28 < bridge> ik, cuz you must reboot in order to have changes apply 08:28 < bridge> but like, do I manually run it? xdd 08:28 < bridge> nix is a pretty complicated process in comparison 08:28 < bridge> no brother.. its because i want to use my pc not compile 08:28 < bridge> oh, wait it compiles? uuugh 08:28 < bridge> ok im out 08:28 < bridge> it takes a very long time 08:28 < bridge> suddenly it reminds me of what I do in Rust 08:28 < bridge> cya later 08:28 < bridge> cya 08:28 < bridge> icant 08:29 < bridge> until recently you had to compile every single thing from scratch, excepting the kernel if you were lazy 08:29 < bridge> anything that's changed gets recompiled on update 08:29 < bridge> yep 08:30 < bridge> there is no partial build cache for basing it only off changes either, afaik it has to be a fresh rebuild every single time 08:31 < bridge> I mean, on release mode it makes sense 08:31 < bridge> there's also no uniform way to do that plus it might not have great returns 08:33 < bridge> well without trying to restructure... cargo, you have no other ways to speed up compilation, unless you either just abandon libraries and make your own, or some black magic that I don't know of yet 08:33 < bridge> tbh the nixos-rebuild && systemctl poweroff is fairly common for me 08:34 < bridge> isn't there a queue on what to run before it powers off? 08:34 < bridge> I'm still talking about nix but I will acknowledge that rust build times are pretty bad 08:35 < bridge> oh kek 08:35 < bridge> not as slow as many super modern cpp features will make it tho 08:35 < bridge> after experiencing ddnet-rs, I can confirm 08:35 < bridge> and rustis inherently complicated to compile 08:35 < bridge> it's slow 08:36 < bridge> think about all the extra evaluation logic and hyper specific parsing errors u get in rust 08:36 < bridge> I bet without all the safety features that overhead the compilations, I would have errors on 90% of my workspace 08:37 < bridge> agreed 08:37 < bridge> But like, what if you don't want that? xd 08:37 < bridge> if your only experience beforehand is in interpreted languages then I can understand the pain but compared to langs like C where the compiler gives so much less of a shit it's logical 08:37 < bridge> u can use unsafe blocks or..... another language 08:38 < bridge> I could 08:38 < bridge> But like 08:38 < bridge> Rust :kek: 08:38 < bridge> rust is architectured around it's safety features 08:38 < bridge> rust is architectured around its safety features 08:39 < bridge> Gboard smart enough to log every keystroke for the NSA but not smart enough not to autocorrect "its" to "it's" when it's grammatically incorrect 08:39 < bridge> Google services in a nutshell 08:40 < bridge> Btw, about the map editor issue I reported, who made changes with Destructive Mode? 08:44 < bridge> its still mostly on llvm tho 08:44 < bridge> its just that rust has a compile the world model 08:44 < bridge> I don't think clangs are so bad 08:44 < bridge> cpp uses shared libs 08:44 < bridge> yea 08:45 < bridge> most of that can be cached or vendored but the former is fragile and the latter is more setup 08:45 < bridge> also by using the compile the world model rust has local lto always 08:45 < bridge> yea they call it PGO when in cpp compilers they are technically not the same term 08:46 < bridge> im not talking about pgo 08:46 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418125947818737675/image.png?ex=68ccfc35&is=68cbaab5&hm=8e48817db006131bf65c33106c54b0e77c79ea67b3df15965811dabf0485337f& 08:46 < bridge> even with debug / off u have a local lto 08:46 < bridge> false 08:46 < bridge> even with false u have a local lto 08:46 < bridge> omg i cant type 08:47 < bridge> ok the local lto is only within the crate 08:48 < bridge> i would like to see a firefox in rust to compare compile time 08:48 < bridge> firefox has to use jumbo builds to improve compile time 08:48 < bridge> hopefully that means they're constexpring everything 08:49 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418126718010396682/image.png?ex=68ccfced&is=68cbab6d&hm=9643c6beaaa74574a579001e98c7b00f9fc1a0f3df30b9140d77b4b374927032& 08:50 < bridge> I didn't know about that but it makes sense 08:50 < bridge> would not be necessary with modules 08:59 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418129229224349738/IMG_9234.jpg?ex=68ccff43&is=68cbadc3&hm=1a39884dd54870b6140e1b19c45c460183e035f3b33b3abd5b38a5cd1e4ca348& 08:59 < bridge> delightful 09:42 < bridge> morming 09:43 < bridge> nornimg 09:47 < ws-client> **** @robyt3 can you explain this code? then i can send merge https://github.com/ddnet/ddnet/pull/10937/files#diff-523c7cb184ab750e026099047cf54bd74271130410325fabf89e78ad38ed8112R57-R65 09:47 < ws-client> **** i understand `if(Length > 0)` 09:48 < ws-client> **** a it checks if it ends in newline to avoid doing double newline because `log_log` also adds a newline 09:50 < ws-client> **** i wonder if this can be moved to str.cpp but i dont think so 09:50 < ws-client> **** ruby has a method that does exactly that its called `chomp` 09:51 < ws-client> **** ``"foo".chomp # => "foo"`` 09:51 < ws-client> **** ``"foo\n".chomp # => "foo" `` 09:51 < ws-client> **** ``"foo\n\n".chomp # => "foo\n"`` 09:57 < ws-client> **** also whats the state on these bois holding me back from receiving ungodly amounts of donation money https://github.com/ddnet/ddnet/pull/10892 https://github.com/ddnet/ddnet/pull/10836 https://github.com/ddnet/ddnet/pull/10827 https://github.com/ddnet/ddnet/pull/10793 10:06 < bridge> llm says 10:06 < bridge> > Key Steps for Buffer-to-Image Copy (most common for CPU-driven updates): 10:06 < bridge> Allocate a staging buffer and map it. 10:06 < bridge> Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:06 < bridge> Record a command buffer: 10:06 < bridge> Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:06 < bridge> Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:06 < bridge> Transition the texture back to its original layout. 10:06 < bridge> Submit the command buffer to the queue. 10:06 < bridge> 10:06 < bridge> I was thinking of how I would do minetee. 10:06 < bridge> > Thanks for not calling it a feature 10:06 < bridge> Lmao, you should thank me for not calling it shit ugly spaghetti code 10:07 < bridge> llm says 10:07 < bridge> > Key Steps for Buffer-to-Image > > Copy (most common for CPU-driven updates): 10:07 < bridge> > Allocate a staging buffer and map it. 10:07 < bridge> > Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:07 < bridge> > Record a command buffer: 10:07 < bridge> Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:07 < bridge> > Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:07 < bridge> Transition the texture back to its original layout. 10:07 < bridge> Submit the command buffer to the queue. 10:07 < bridge> 10:07 < bridge> I was thinking of how I would do minetee. 10:07 < bridge> llm says 10:07 < bridge> > Key Steps for Buffer-to-Image > > Copy (most common for CPU-driven updates): 10:07 < bridge> > Allocate a staging buffer and map it. 10:07 < bridge> > Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:07 < bridge> > Record a command buffer: 10:07 < bridge> > Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:07 < bridge> > Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:07 < bridge> Transition the texture back to its original layout. 10:07 < bridge> Submit the command buffer to the queue. 10:08 < bridge> 10:08 < bridge> I was thinking of how I would do minetee. 10:08 < bridge> llm says 10:08 < bridge> > Key Steps for Buffer-to-Image Copy (most common for CPU-driven updates): 10:08 < bridge> > Allocate a staging buffer and map it. 10:08 < bridge> > Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:08 < bridge> > Record a command buffer: 10:08 < bridge> > Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:08 < bridge> > Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:08 < bridge> Transition the texture back to its original layout. 10:08 < bridge> Submit the command buffer to the queue. 10:08 < bridge> 10:08 < bridge> I was thinking of how I would do minetee. 10:08 < bridge> llm says 10:08 < bridge> > Key Steps for Buffer-to-Image Copy (most common for CPU-driven updates): 10:08 < bridge> > Allocate a staging buffer and map it. 10:08 < bridge> > Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:08 < bridge> > Record a command buffer: 10:08 < bridge> > Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:08 < bridge> > Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:08 < bridge> Transition the texture back to its original layout. 10:08 < bridge> Submit the command buffer to the queue. 10:08 < bridge> 10:08 < bridge> I was thinking of how I would do minetee. 10:08 < bridge> What motivates me to do PRs? 10:08 < bridge> 1. Read some code 10:08 < bridge> 2. Almost get a heart attack 10:08 < bridge> 3. fix/refactor/optimize 10:09 < bridge> llm says 10:09 < bridge> > Key Steps for Buffer-to-Image Copy (most common for CPU-driven updates): 10:09 < bridge> > Allocate a staging buffer and map it. 10:09 < bridge> > Write your pixel data to the buffer (e.g., for a single pixel, write the RGBA values at the correct offset). 10:09 < bridge> > Record a command buffer: 10:09 < bridge> > Transition the texture to VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. 10:09 < bridge> > Use vkCmdCopyBufferToImage with a VkBufferImageCopy specifying the target region (e.g., imageOffset for the pixel’s x, y, z coordinates and imageExtent for a 1x1x1 region). 10:09 < bridge> > Transition the texture back to its original layout. 10:09 < bridge> > Submit the command buffer to the queue. 10:09 < bridge> 10:09 < bridge> I was thinking of how I would do minetee. 10:10 < bridge> why post llm output slop 10:13 < bridge> https://webassembly.org/news/2025-09-17-wasm-3.0/ 10:15 < bridge> With these new features, Wasm has much better support for compiling high-level programming languages. Enabled by this, we have seen various new languages popping up to target Wasm, such as Java, OCaml, Scala, Kotlin, Scheme, or Dart, all of which use the new GC feature. 10:15 < bridge> 64-bit address space. Memories and tables can now be declared to use i64 as their address type instead of just i32. That expands the available address space of Wasm applications from 4 gigabytes to (theoretically) 16 exabytes, to the extent that physical hardware allows. While the web will necessarily keep enforcing certain limits — on the web, a 64-bit memory is limited to 16 gigabytes — the new flexibility is especially interesting for non-we 10:15 < bridge> finally 10:15 < bridge> when we will see the end of js 10:21 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418149869037686814/image.png?ex=68cd127c&is=68cbc0fc&hm=0d3c194d917960fadb010f9bdbf8bdcbae47fc6ee8ef9d32d747c720865925a2& 10:21 < bridge> xd 10:22 < bridge> wtf how 10:22 < bridge> ? 10:22 < bridge> device 9999 guid 060606 10:23 < bridge> ah thats my fake partition 10:23 < bridge> its a memory filesystem 10:23 < bridge> oh xd 10:23 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418150376087228437/image.png?ex=68cd12f5&is=68cbc175&hm=c3127a5be22606255440fe2f360ca0bb1b4fdd2181ab06b361290d4a8db6920d& 10:23 < bridge> 1p0 means device 1 partition 0 10:23 < bridge> rly I've never used linux 10:23 < bridge> dev 0 p 0 is mounted at root / so its not here but i should add it here maybe idk 10:24 < bridge> well this isnt linux 10:24 < bridge> its edos 10:24 < bridge> iim totally making this up xd, well the device id matches the order of found active ahci ports 10:24 < bridge> and partition is from parsing gpt or mbr 10:25 < bridge> lovely 10:30 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418152100113940511/image.png?ex=68cd1490&is=68cbc310&hm=dc04cfbf05461af143e56c5e6e06a7f9a87895cd4d917c11f5d75bf22539d8b5& 10:30 < bridge> mountception 10:30 < bridge> does linux allow mounting same partition in multiple places? 10:30 < bridge> xd 10:34 < bridge> I believe so 10:34 < bridge> just a reparse point or w/e 10:50 < bridge> found out in my counrty it's cheaper to buy through amazon than on local virtual shops 10:50 < bridge> I mean 10:50 < bridge> a raspberry pi 5 full kit is at around ~300 USD here 10:50 < bridge> but the same product on amazon is like ~120 USD?? 10:51 < bridge> And even without taxes?? 10:51 < bridge> that's so crazy 10:51 < bridge> I mean imported stuff 10:51 < bridge> I mean imported products 10:55 < bridge> Looks correct. My sentence is still true 11:08 < ws-client> **** @Jupstar ✪ I had to update submodules when switching into your new prs branch <:justatest:572499997178986510> 11:24 < bridge> it still lacks good threading, the biggest issue for targetability. 11:25 < bridge> the last proposal that was getting close to finished got denied and now they're back to nothing in terms of threading proposals. literally like 3 years setback 11:29 < bridge> honestly idk why this isn't a higher priority for them, they seem to not care about it that much despite the huge demand and unofficial implementations that people made just to have something 12:03 < bridge> why isnt it just pthreads? 12:03 < bridge> Of course 12:04 < bridge> Isn't there a flag to mount to multiple places even ? 12:06 < bridge> i dont think mounting to multiple places is a good idea, running 2 fs for a simple fs may work but for more advanced ones updates will cuuse lots of divergence in state and dada you just lost data 12:06 < bridge> syslink is much better 12:07 < bridge> not really, the only issue you may run into is if you mount with different perms 12:07 < bridge> i dont feel like different perms will have that much of an effect 12:08 < bridge> i cant really test this because theres not a way for you to do it 12:08 < bridge> mount --bind is probably not similar to mounting twice 12:09 < bridge> mounting is just saying "your root is relative to this path" and thats it 12:09 < bridge> idk how linux does it but i detect if a path is in a mounted path and then strip the prefix and pass to the fs handler as if it were / 12:10 < bridge> for example if its mounted at /mnt and i do ls /mnt, the list files is received by the fs mounted at /mnt with a path of / 12:10 < bridge> So a fancier chroot for the fs? 12:10 < bridge> idk how linux does it tbh 12:10 < bridge> 2 fs handlers wont be able to communicate any locks 12:10 < bridge> i explained how i do it 12:11 < bridge> its the same fs handler 12:11 < bridge> it has to be 1 which is shared 12:11 < bridge> i have a router and fs handlers, the router handles mount points 12:11 < bridge> and u can have multiple mount points to same fs 12:11 < bridge> These aren't a native concept, emscripten maps them to web workers which require a browser 12:12 < bridge> wel not yet, but why cant they be 12:12 < bridge> pthreads is posix threads 12:12 < bridge> winshit doesnt have those 12:12 < bridge> windows can run pthread code 12:13 < bridge> Also theres huge limitations to the workers as threads workaround. 12:13 < bridge> web workers are disgusting 12:13 < bridge> You have to bend over backwards to target it 12:13 < bridge> web sucks 12:14 < bridge> If they didnt ditch the wasi threading proposal it could not suck 12:16 < bridge> Microsoft is maintaining a cosmic eldritch horror to make C# compile to web worker right now. 12:16 < bridge> And its a buggy mess 12:16 < bridge> Microsoft is maintaining a cosmic eldritch horror to make C# threads compile to web worker right now. 12:17 < bridge> ``` 12:17 < bridge> As the last step of walk_component(), step_into() will be called either directly from walk_component() or from handle_dots(). It calls handle_mounts(), to check and handle mount points, in which a new struct path is created containing a counted reference to the new dentry and a reference to the new vfsmount which is only counted if it is different from the previous vfsmount. Then if there is a symbolic link, step_into() calls pick_link() to deal with 12:17 < bridge> ``` 12:17 < bridge> 12:17 < bridge> From kernel.org 12:18 < bridge> https://www.kernel.org/doc/html/v6.14-rc2/filesystems/path-lookup.html 12:22 < bridge> Dunno what u mean chiller 12:26 < bridge> @chillerdragon did you just do that for me? xD 12:39 < ws-client> **** @Assa wat 12:40 < ws-client> **** @Jupstar ✪ your prs are based on old master branch :C 12:43 < bridge> the clineitem constructor PR looks very much needed for the hook collision line PR ^^ 12:43 < bridge> 12:43 < bridge> Also I just commented on it 😛 12:43 < ws-client> **** yes that idea came to me while i reviewed your pr 12:51 < bridge> need to check i didnt break rendering 12:51 < bridge> ! 12:59 < bridge> I am pretty sure you did break rendering the laser heads 12:59 < bridge> I am pretty sure you did break rendering the laser 13:00 < bridge> oh and the progress spinner 13:00 < bridge> Tja 13:02 < bridge> ~I am pretty sure you did break rendering the laser~ 13:02 < bridge> ~~I am pretty sure you did break rendering the laser~~ 13:08 < bridge> i cant test because it takes 500000 minutes to compiiile 13:08 < bridge> im only compiling game-client 13:08 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1418191947712954378/image.png?ex=68cd39ad&is=68cbe82d&hm=9e7b457cc6ceef345e60af59005660f42d88ca9f4833b0c876c2a5d4fb04c876& 13:08 < ws-client> **** download more cpu 13:08 < bridge> i started after the pr 13:08 < bridge> )::: 13:08 < ws-client> **** also ninja is slow 13:08 < ws-client> **** use bam 13:08 < bridge> bam? 13:08 < ws-client> **** xd 13:09 < bridge> I am pretty sure you don't need over a month to compile :justatest: 13:09 < bridge> HYPERBOLE 13:09 < bridge> !!!! 13:09 < bridge> can cmake even make a bam 13:09 < ws-client> **** no bam can make a cmake 13:09 < bridge> err 13:09 < bridge> !remind 50000 @sollybunny test your compiled client 13:10 < bridge> bad bot 🙁 13:10 < bridge> Sowwy :c 13:10 < ws-client> **** @Assa almost did same joke 13:10 < ws-client> **** but didnt want an annoying reminder for it xd 13:10 < ws-client> **** !ping 13:10 < ws-client> **** rip you overflowed ryos reminder code 13:10 < bridge> well whats a joke without the punchline 13:10 < bridge> rust crashed instead of ovreflowing 13:10 < bridge>