01:21 < bridge> @0xdeen: the debug render option in the release log belongs to assa not me .\_. 01:21 < bridge> And automatic quad clipping is also assa instead of ryo I assume 01:23 < bridge> Ah you already fixed it on GitHub. I still see it on \#announcements tho 01:25 < bridge> I suppose using a pr like process for the changelog is overkill but maybe let one of the currently active mergers double check it 01:25 < bridge> what, you mean he wrote Ryo instead of Assa??? 01:25 < bridge> what 01:26 < bridge> Oh, he fixed it from Github 01:27 < bridge> in his defense, there are a bunch of engine changes in this patch 01:27 < bridge> I guess in his defense the merge commits are confusing and also trolled Heinrich 01:28 < bridge> But anyone looking actively at the prs knows that it’s assa coding quads and ryo doing the merged 01:28 < bridge> But anyone looking actively at the prs knows that it’s assa coding quads and ryo doing the merges 01:28 < bridge> ryo stealing all the credit, smh 01:28 < bridge> :kek: 01:29 < bridge> I wonder how manual the process is. I have the changelog generation basically automatic in ddnet-insta 01:29 < bridge> It filters the merge commits and finds the correct author 01:31 < bridge> But is it reliable? 01:31 < bridge> No 01:31 < bridge> ofc lmao 01:31 < bridge> Xd zhn had to set some weird git name and it pinged the wrong person lmao 01:31 < bridge> ig It's better that Deen does it manually 01:32 < bridge> But at least it’s a name zhn chose instead of a different maintainer 01:32 < bridge> wait, who xd 01:32 < bridge> gerdoe-jr had as commit author something like vlad 01:32 < bridge> So the changelog said @vlad 01:33 < bridge> Which happens to be a GitHub account owned by someone else 01:33 < bridge> oh shit, git oopsie 01:49 < bridge> ChillerDragon: yo 01:49 < bridge> how does your irc bridge work? 01:49 < bridge> I'm curious 02:55 < bridge> @ryozuki can code on your OS segfault? 03:15 < bridge> Chillerdragón::YY: :OO: 03:27 < bridge> chillerdragön 04:15 < bridge> @chillerdragon new bug with ddnet-insta -- you can shoot yourself with laser lol 04:34 < bridge> @chillerdragon new bug with ddnet-insta v2 -- you can shoot yourself with laser lol 04:34 < bridge> Another bug is when you suicide with nade, you respawn instantly 05:02 < bridge> And the `sv_welcome` thing was problem in 2.0, but not in 1.7 it seems 06:20 < bridge> it's heinrich's bridge 06:20 < bridge> but he just has a websocket that sends messages through irc 06:27 < bridge> ohh 06:34 < bridge> do you mean the kernel code or user run programs? for user run programs, yeah and you need to handle that ,with the page fault interrupt, there u can for example terminate the process and cleanup it. 06:34 < bridge> if u mean at kernel level, yeah and for lot of reasons, because interactiong with lot of stuff x86_64 has requires asm or registers that if u set a value wrong can segfault 06:34 < bridge> those interactions still happen inside unsafe code 06:35 < bridge> do you mean the kernel code or user run programs? for user run programs, yeah and you need to handle that ,with the page fault interrupt, there u can for example terminate the process and clean it up. 06:35 < bridge> if u mean at kernel level, yeah and for lot of reasons, because interactiong with lot of stuff x86_64 has requires asm or registers that if u set a value wrong can segfault 06:43 < bridge> webhook 06:52 < bridge> i meant chiller 06:53 < bridge> o 08:17 < bridge> gumo 08:25 < bridge> @pathos3005: did you read the docs about the reset.cfg it explains how to disable that 08:26 < bridge> @pathos3005: the welcome message is not in the code. It has to be your config 08:26 < bridge> @pathos3005: the suicide sounds bad I will investigate thanks for reporting. You can also open issues on GitHub so we don’t have to spam this channel 09:40 < ws-client> **** @pathos3005 I can not reproduce the missing delay. https://github.com/ddnet-insta/ddnet-insta/issues/422#issuecomment-3252332442 09:40 < ws-client> **** What did you set this config to? ``sv_self_kill_respawn_delay_ms`` which gamemode did you use? 10:08 < bridge> ``` 10:08 < bridge> 0.025111] Syscall: sys_write(fd=1, buf=0x402000, count=23) 10:08 < bridge> [0.025378] Hello from user space! 10:08 < bridge> 10:08 < bridge> [0.025580] called exit with 42 10:08 < bridge> ``` 10:08 < bridge> im back with user space programs 10:08 < bridge> this time with a super clean way 10:11 < bridge> (elf loaded) 10:14 < bridge> awesome 10:15 < bridge> no custom executable format :feelsbadman: 11:38 < bridge> i had a weird showthought, the cpu is like the engine and the LAPIC is the gearbox 11:38 < bridge> elf is p good imhi 11:39 < bridge> elf is p good imho 11:39 < bridge> supporting elf means compilers like gcc and rust work 11:40 < bridge> Not like you couldn't add that part to the backend 11:41 < bridge> the hard part of generating efficient machine code is already done for you... though I'm sure there's a lot of work done to maintain alignments and such in the actual file 11:47 < bridge> !remind 5760 test if long map setup times cause client getting stuck in loading screen waiting for server response 11:47 < chillerbot12> Helo Assa I will remind you at Mon Sep 08 2025 11:47:25 GMT+0200 (Central European Summer Time) about your matter again. 12:03 < bridge> There's already an open issue about this, but it hasn't consistently been reproduced yet 12:30 < bridge> i believe this is the cause behind the stuck valgrind Pipeline back in another PR 12:32 < bridge> Doesn't look like the timed-out message is about the client joining though 12:33 < bridge> Or is the connecting itself not checked so this is the next test that fails? 12:34 < bridge> the check that failed was on server side waiting for the client joined message 12:35 < bridge> I also had this once in a build client, but it was not reproducable 12:36 < ws-client> **** @kebscs do we have a plan for #10712 ? 12:36 < chillerbot12> https://github.com/ddnet/ddnet/issues/10712 12:39 < bridge> Idk how to fix 12:39 < bridge> Maybe the test is written poorly 12:40 < ws-client> **** sadge 12:40 < bridge> @chillerdragon "Klammer" in german can mean bracket, parenthesis, clip, clamp, paperclip, clothespin ... 12:40 < ws-client> **** germans are nuts 12:41 < ws-client> **** runde klammern, eckige klammern, gezwörbelte klammern 12:41 < bridge> I think we are animals, not nuts 12:41 < ws-client> **** rawrrr 12:41 < ws-client> **** du bist mir so ein klammer affe 12:42 * bridge steals your PR 12:43 < bridge> Do we care about changelog accuracy btw? 12:44 < bridge> you are missing "geschweifte Klammern" e.g. {} 12:47 < ws-client> **** @Assa i do 12:47 < ws-client> **** proper credit is the only payment volunteer contributors get 12:47 < ws-client> **** especially the user facing stuff should be correct 12:48 < ws-client> **** @Assa i tried to be funny by calling it "gezwörbelte" 12:48 < bridge> Looks like a teehistorian bump happened in the meanwhile. It probably just needs a rebase 12:48 < ws-client> **** replyer discord andy 12:49 < bridge> yeah the debug render frature is attributed to you, so you are now responsible for that 12:50 < bridge> yeah the debug render feature is attributed to you, so you are now responsible for that 12:50 < ws-client> **** i am the responsible merger yes. All bugs are on me. 12:50 < ws-client> **** But the feature credit should fully go to you 12:51 < bridge> The GitHub pr one can be work in progress. Nobody sees stuff on discord. But the changelog is also posted in \#announcements and the website and even visible to all players in the client news. There the credit should be correct. 12:52 < bridge> The GitHub pr one can be work in progress. Nobody sees stuff on GitHub. But the changelog is also posted in \#announcements and the website and even visible to all players in the client news. There the credit should be correct. 12:52 < bridge> I ask because I gave deen a list of corrections which got mostly ignored 12:52 < bridge> yes deen does that sometimes he gives a big disconnected vibe 12:53 < bridge> I don’t think that he doesn’t care it’s just that it got lost. He is 10x engineer with busy life who runs entire ddnet with a time budget of 20 seconds per day 12:54 < bridge> Maybe one day he can let someone else in to help him with the holy task of compiling a changelog 12:54 < bridge> yeah I know, and I am not angry or anything 12:55 < bridge> Yea you are famous anyways 12:55 < bridge> I just wanted to support him in that 12:55 < bridge> But imagine a first time contributor waiting to be plugged in the client to see someone else’s name in the end 12:55 < bridge> famous for what? Afaic maybe in zCatch 12:55 < bridge> Your name is all over the changelog 12:56 < bridge> <01000111g> for King's Leap aswell 12:56 < bridge> You are top ddnet contributor 12:56 < bridge> true ❤️ 12:56 < bridge> am I? Didn't do that much yet 12:57 < bridge> did you see the mapsize thing from patiga for this map? xD 12:58 < bridge> Yes you are 12:58 < bridge> I wonder how quad arts can be smaller than embedded png files 12:59 < bridge> <01000111g> no, was it an extreme reduction? 12:59 < bridge> @kebscs: can you try rebasing? I doubt that the teehistorian bump causes compile errors on outdated branches. But I blindly trust lerato. 12:59 < bridge> no I mean this @01000111g 12:59 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1413116373604765788/FileSizeStats.png?ex=68bac2ad&is=68b9712d&hm=f3d1ed3154683935edf14b2e95fdaee5fc3ae94bddb63d431f500ea6e40c0cef& 13:00 < bridge> Patiga is heaven sent. 13:00 < bridge> Patiga the goat 13:00 < bridge> map is > 50% just sound 13:02 < bridge> like the bell sound is a 1 second opus file, uses as much space as a mapres 13:05 < bridge> !remind fix callback hell by moving findnext to outer scope 13:05 < chillerbot12> Helo chillerdragon I will remind you at Fri Sep 05 2025 13:05:37 GMT+0200 (Central European Summer Time) about your matter again. 13:08 < bridge> Oh actually the people bumping teehistorian, are you guys fixing the tests? 13:08 < bridge> Yes 13:09 < bridge> It’s in the pipeline. Without adapting the test no CI 13:09 < bridge> Hm, I wonder why that pr you mentioned in the morning isn't passing CI then 13:09 < bridge> And we merge only green CI this is a stable project! 13:10 < bridge> Yes that’s what we all wonder 13:10 < bridge> It’s fancy c++ code idk I’m not a developer 13:10 < bridge> But shouldn’t be teehistorian version related 13:10 < bridge> iirc creating the elf file is the linker job, im writing a OS not a linker 13:11 < bridge> maybe some day ill make an entire custom toolchain but first i need a proper os 13:18 < bridge> Just rebase it first, if it still doesn't pass I'll look at it 13:19 < bridge> > Maybe one day he can let someone else in to help him with the holy task of compiling a changelog 13:19 < bridge> 13:19 < bridge> * removed virus by @chillerdragon 13:20 < bridge> as if there is no changelog by git diff creator 13:20 < bridge> it has a manual pass, and whoever is doing that pass can do whaetevr 13:21 < bridge> im sure me and chiller and atter can atest to change log being a right pain 13:21 < bridge> im sure me and chiller and tater can atest to change log being a right pain 13:21 < bridge> oh btw theres this new feature i just forgo 13:21 < bridge> oops thats not meant to be there 13:21 < bridge> whats important enuf!??!?! 13:22 < bridge> @sollybunny: \* removed herobrine 13:22 < bridge> https://www.conventionalcommits.org/en/v1.0.0/ 13:22 < bridge> 13:22 < bridge> @chillerdragon if we enforce this commit style its easy to do a changelog automatically 13:22 < bridge> * hide easteregg 13:22 < bridge> @ryozuki: I do that in ddnet-insta actually 13:22 < bridge> yes yes 13:23 < bridge> With conventional commits and bash script 13:23 < bridge> I proposed this before, I don't remember why but @0xdeen wasn't all that positive about it 13:23 < bridge> actually u can enforce it on github pr titles only, not commits 13:23 < bridge> and it works 13:23 < bridge> But I still manually filter out the boring commits 13:23 < bridge> iuts not gonna be easy with that tree jesus 13:23 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1413122296415129801/git-flow--welcome.png?ex=68bac832&is=68b976b2&hm=01d6d12a95dfcad1b019b62beb13b1a3457f15b4cb4c5687db315fad984c0def& 13:23 < bridge> we can also make PRs start with emojis as I had to do at my old work place 13:24 < bridge> one of the most famous rust tools https://github.com/orhun/git-cliff 13:24 < bridge> 🔥 idea 13:24 < bridge> we should do like kernel, patches by mail to filter out nobo devs 13:24 < bridge> /s 13:24 < bridge> :cammostripes: I was joking 13:25 < bridge> xd 13:25 < bridge> 🚀 optimize quad rendering 📈 13:25 < bridge> He did do that xd 13:25 < bridge> 😂 bcachefs 13:25 < bridge> Check prs 13:25 < bridge> Assa the emoji dev 13:26 < bridge> 🥸 fixed a startline skip bug 13:26 < bridge> ⚡ this PR actually contains an emoji 13:26 < bridge> 💩add backdoor for sp someone 13:27 < bridge> https://cdn.discordapp.com/attachments/252358080522747904/1267876556261560454/EB37683D3A4CDA1932DD5DD1561EC9AF.gif 13:28 < bridge> 🧹 💩 Cleanup: ... 13:28 < bridge> today i asked ai how to enable other cores than bootstrap, it said this 13:28 < bridge> 13:28 < bridge> Wake up APs using INIT-SIMP-SIMP sequence via LAPIC 13:28 < bridge> 13:28 < bridge> i didnt know x86-64 loves simps 13:32 < bridge> I don't think it solves the problem of having to go through the list of PRs, filtering, summarizing and sorting them by importance for the changelog. We could just use GitHub for releases and have it list all merged PRs for a lazy changelog that's less helpful for players, if we wanted to avoid the effort. And deen would still have to summarize the changes again for the in-game news anyhow. 13:34 < bridge> I agree, but he probably hopefully doesn't do this fully manually, does he? 13:38 < bridge> How else would he do it? There's no way to automatically determine the importance of changes. 13:38 < bridge> It might make it easier if he gets a list of all prs first though. I think currently he just stares at all the commits 13:41 < bridge> Yeah, I guess he looks at the commits which causes the author confusion because of merge commits 13:42 < bridge> I'd also like to rewrite the in-game news to be more flexible, but then it would have to automatically transform them into the old format for backwards compatibility 13:43 < bridge> When are we sneaking in chromium? 13:44 < bridge> :monkaS: 13:46 < bridge> I'd add the formatting as json objects, allow the news to scroll, and add an identifier so typos in the news can be updated without making the News button green for players 13:46 < bridge> didnt work 13:47 < bridge> Did you run the tests locally btw? Is it just a CI issue? 13:48 < bridge> I think it tried it locally and couldn't reproduce it 13:48 < bridge> I think I tried it locally and couldn't reproduce it 13:48 < bridge> I wonder if strlen becoming constexpr actually somehow modifies teehistorian files, do we have length prefixed stuff in there? Maybe some UB? 13:51 < bridge> :poggers2: 13:51 < bridge> my friend UB 13:51 < bridge> its married to C++ so we cant miss it 13:56 < bridge> doesnt fail for me 13:57 < bridge> but msvc doesnt fail on ci either 14:02 < bridge> Ok, let me find a cafe where I'm allowed to laptop, then I'll check 14:09 < bridge> roby is on mingw 14:24 < bridge> Hello, I've encountered an error during importing the record_maps.sql into the database: 14:24 < bridge> ERROR 1180 (HY000) at line 40: Got error 1 "Operation not permitted" during COMMIT 14:24 < bridge> the line 40 is INSERT INTO "record_maps" VALUES and the map info below 14:24 < bridge> 10.11.13-MariaDB-0ubuntu0.24.04.1 Ubuntu 24.04 14:24 < bridge> I've tried to execute it with sudo, but the question seems not on it 14:27 < bridge> Does your db user have write access? 14:30 < bridge> I've succeeded importing the record_mapinfo.sql before 14:31 < bridge> mysql -uroot teeworlds < ddnet-sql/record_maps.sql 14:31 < bridge> (under root bash) 14:33 < bridge> Can reproduce with MSYS2 in release mode 14:39 < bridge> Oh, it's not a test failure, the testrunner flat out fails to link 14:39 < bridge> Oh, it's not a test failure, the testrunner flat out fails to compile 14:40 < bridge> https://nnethercote.github.io/2025/09/04/faster-rust-builds-on-mac.html 14:41 < bridge> @0xdeen for u 14:42 < bridge> Commenting out `WriteBuffer(vBuffer, aTimeBuf, str_length(aTimeBuf));` fixes it for some reason 14:43 < bridge> This buffer is not constant, but apparently the compiler thinks it is 14:43 < bridge> Huh, compiler is rarely wrong about things being constant, it's usually on the conservative side with those 14:44 < bridge> As in it'd usually rather assume it's not constant then be wrong 14:44 < bridge> maybe u casted a non const to const:? 14:44 < bridge> (im ootl) 14:46 < bridge> Ah, managed to reproduce it here too 14:48 < bridge> <0xdeen> Thanks, but I mostly build on a Hetzner server 😄 14:55 < bridge> Hm, I wonder if we are hitting an stl defect here 14:59 < bridge> Or a gcc defect maybe 14:59 < bridge> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105746 it was broken some time ago for some cases 15:01 < bridge> Replacing the `str_length` with `__gnu_cxx::char_traits::length` in this line also fixes it apparently 15:03 < bridge> Hm, theirs is static constexpr too, I wonder why that doesn't optimize the same way triggering Warray-bounds 15:04 < bridge> btw I think the issue is the line below the one you are messing with, it has compile time constant str_length 15:05 < bridge> yeah, that's what the compiler output indicates as well, but changing that lines seems to change the optimization behavior 15:06 < bridge> does clang has the issue 15:06 < bridge> No 15:06 < bridge> clang ftw 15:07 < bridge> I think it's just a gcc defect given other people reported issues with vector too, perhaps everything we append before that is constant length, that makes gcc make assumptions about the size of the buffer 15:07 < bridge> Inlining the size 295 instead of `str_length(PREFIX2)` gives the same issue 15:08 < bridge> Seems unrelated to the `str_length` change itself 15:08 < bridge> That's what is already happening, I think it might work if that value is not constant 15:09 < bridge> Yep 15:10 < bridge> Make `PREFIX2` volatile and it doesn't happen 15:10 < bridge> `vBuffer.resize(OldSize + DataSize + 1);` also fixes it :nouis: 15:11 < bridge> Anyway, it's related to the str_length change, it allows the optimizer to take a more optimal path which triggers the bounds issue 15:11 < bridge> (doesn't for me) 15:12 < bridge> > gcc version 15.2.0 (Rev8, Built by MSYS2 project) 15:12 < bridge> > gcc (Gentoo 14.3.0 p8) 14.3.0 15:14 < bridge> ➜ ~ gcc --version 15:14 < bridge> gcc (Debian 14.2.0-19) 14.2.0 15:14 < bridge> my debian is close to ur gentoo lol 15:15 < bridge> sys-devel/gcc rolls especially slowly on stable gentoo, it can break a lot of things to update your system compiler 15:16 < bridge> yesterday i asked ai how can a data transfer differentiate a bit stream full of 0s from no data, i learn about 8b/10b encoding and derivatives 15:16 < bridge> rly interesting 15:17 < bridge> you could have asked this to me instead, that's why I exist 15:17 < bridge> i was at bed about to sleep and had the sudden urge to know this 15:17 < bridge> i didnt know if u would answer before i slept 15:17 < bridge> xd 15:17 < bridge> ``` 15:17 < bridge> The core issue: 15:17 < bridge> If you send a long string of 0s (or 1s), the receiver can lose synchronization because there are no voltage transitions to extract timing from. 15:18 < bridge> How PCIe solves it: 15:18 < bridge> 8b/10b encoding (PCIe 1.x/2.x): 15:18 < bridge> 15:18 < bridge> Every 8 data bits gets encoded as 10 bits 15:18 < bridge> The encoding guarantees at most 5 consecutive identical bits 15:18 < bridge> Maintains DC balance (roughly equal 0s and 1s over time) 15:18 < bridge> Example: 8 bits of all zeros 00000000 becomes something like 1001010001 15:18 < bridge> ``` 15:18 < bridge> this is a part of the answer 15:18 < bridge> then asked about usb 15:18 < bridge> USB uses similar concepts but different specific encoding schemes depending on the version: 15:18 < bridge> USB 1.x/2.0 (Low/Full/High Speed): 15:18 < bridge> 15:18 < bridge> Uses NRZI encoding (Non-Return-to-Zero Inverted) 15:25 < bridge> https://rustcurious.com/elements/ 15:29 < bridge> Hm, I couldn't reproduce it in a small program :/ 15:29 < bridge> tbh not surprised with how finnicky it is, it's probably an extremely specific line of optimization that's bricking it 15:55 < bridge> @robyt3 so I think this is a gcc defect anyway, what would you like to do about it? I worry the defect might be the optimizer rather than the warning, as in the optimized code now truly invokes UB 15:58 < bridge> Would it help the GCC maintainers if we'd find a minimal example? Or how do we motivate them to fix it? 16:00 < bridge> I.e. an example without unions or move constructor overloading like in the bug report 16:01 < bridge> If we find a minimal repro we can report it on the gcc mailing list. I highly doubt they would bother if we report it with our huge codebase, they probably can't be assed 16:02 < bridge> I tried to extract everything relevant, but couldn't get it to issue the same warning, and idk how to inspect the generated IL to investigate further 16:07 < bridge> Could tools like creduce help to automatically reduce the program? 16:07 < bridge> Never heard of this, so idk 16:07 < bridge> https://github.com/csmith-project/creduce 16:10 < bridge> I didn't even think something like this could exist, sounds very useful if it can be used for this 16:18 < bridge> USB will add some 1s or some 0s in a constant data transfer 16:20 < bridge> iirc, inserted every 6 same bits 16:20 < bridge> yep 16:20 < bridge> well depends on version iirc 16:20 < bridge> im currently struggling on making animated tikz-timing table for I2C, then i'll do SPI and then USB 16:21 < bridge> USB 1.x/2.0 (Low/Full/High Speed): 16:21 < bridge> Uses NRZI encoding (Non-Return-to-Zero Inverted) 16:21 < bridge> A '0' bit causes a voltage transition, a '1' bit maintains current voltage 16:21 < bridge> Bit stuffing: After 6 consecutive '1' bits, automatically insert a '0' bit 16:21 < bridge> This guarantees transitions at least every 7 bit periods 16:21 < bridge> Start-of-packet and end-of-packet use special signal states (SE0, etc.) 16:21 < bridge> 16:21 < bridge> USB 3.x (SuperSpeed): 16:21 < bridge> Uses 8b/10b encoding just like early PCIe 16:21 < bridge> Same DC balance and transition guarantees 16:21 < bridge> Differential signaling pairs like PCIe 16:21 < bridge> 16:21 < bridge> (the message) 16:22 < bridge> i just want itt o be next monday 16:22 < bridge> i got vacations 16:22 < bridge> holidays 16:22 < bridge> idk the word 16:22 < bridge> im gonna enjoy my european rights 16:33 < bridge> @blazulite https://sollycluncdev.sollybunny.xyz/twinbop.html 16:36 < bridge> Viruz 16:44 < ChillerDragon> souly there is quoting issue in tclient 16:44 < ChillerDragon> if first player in tab complete has a space it quotes the name 16:44 < bridge> ? 16:45 < ws-client> **** https://zillyhuhn.com/cs/.d8082225-9934-4d80-915a-6af76ff0015c.png 16:45 < bridge> epic fail 16:45 < ws-client> **** fr 16:46 < bridge> @chillerdragon doenst happen for me 16:46 < ws-client> **** wat u do? 16:47 < ws-client> **** come solofng 16:47 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1413173529926701066/image.png?ex=68baf7e9&is=68b9a669&hm=7ccf8d0aecbddf454981d5ecf73b3648811e75987e09c8b726fa42c035cb7a7d& 16:47 < bridge> tab 16:47 < bridge> just tab 16:47 < bridge> also does it for me 16:47 < ws-client> **** https://zillyhuhn.com/cs/.f44fc19f-e064-4cf1-a6c5-26b4eaf7553f.png 16:47 < bridge> wtf 16:48 < ws-client> **** thats chillerbot-zx tho xd 16:48 < ws-client> **** switching screen while typing bugged the renderer xd 16:51 < ws-client> **** also the console selecting is offset 16:51 < ws-client> **** also i need two attempts to move the client to another screen 16:51 < ws-client> **** the first time it bounces off xd 16:52 < ws-client> **** using super+arrow keys 16:52 < ws-client> **** in ``gfx_fullscreen 3`` 16:52 < ws-client> **** i hate all these bugs that might also be gnome or xorg related there are so many of them 16:55 < bridge> @chillerdragon change to sdl2 instead of sdl2-compat 16:55 < bridge> if thats a thing 16:59 < ws-client> **** also sdl2-compat is the future 16:59 < ws-client> **** when finish sdl3 pr souly 17:00 < bridge> @chiller Take a string range as input and write start and end as output. Returns true if it is a valid range and false otherwise. 17:00 < bridge> i dont get it 17:01 < bridge> sdl2-compat is easily the stupidest engineering decision I have ever heard in my life. Especially when sdl2 is still getting patches 17:01 < ws-client> **** @kebscs hm 17:01 < bridge> Arch's decision to make it provide sdl2 by default is even stupider 17:01 < ws-client> **** @kebscs it parses a string range to two integers 17:02 < ws-client> **** if you dont like the design dont do it i guess 17:02 < ws-client> **** i thought it would be cool 17:02 < bridge> so string "abcd" would return a and d? 17:02 < ws-client> **** return false 17:03 < bridge> i dont get it 17:03 < ws-client> **** maybe u get rossbits comment? 17:04 < ws-client> **** the method takes as input a string such as ``"2-6"`` and then returns the two integers `2` and `6` 17:04 < ws-client> **** but you can only return one value so the actual return value is a bool for success 17:04 < ws-client> **** and the parameters are pointers where the values will be written to 17:04 < ws-client> **** thats a bit wonkey C returning 17:04 < bridge> complicated for 0 17:04 < ws-client> **** you can also go full bloated c++ approach using rossbit approach 17:05 < ws-client> **** yea if u think its complicated dont do it 17:05 < bridge> whats rossbit 17:05 < ws-client> **** Or more modern: return an std::optional>. 17:05 < bridge> ok i refreshed and see 17:05 < bridge> What are you guys implementing? 17:05 < bridge> rossbit is probably a human. 17:05 < ws-client> **** A* 17:05 < ws-client> **** https://github.com/ddnet/ddnet/pull/10819 17:06 < ws-client> **** i feel like parsing ranges could be something reusable in a method 17:06 < bridge> idk `aReason[0] >= '0' && aReason[0] <= '5'` is easy to understand for me 17:06 < ws-client> **** makes also reading the vote code easier 17:06 < ws-client> **** yea idc much 17:06 < ws-client> **** if u dont like it dont do it 17:07 < bridge> Ah, we have system.cpp now. You can definitely go for the std optional std pair 17:07 < ws-client> **** ew lerato c++ enjoyer 17:07 < ws-client> **** what happend to the C senpai enjoyer of goto 17:08 < ws-client> **** lerato got infected by std 17:08 < bridge> Actually that's not great api, you want to return the rest of the string after parsing out a range, so it's usable for other stuff too 17:08 < bridge> I still enjoy goto, even in my C++. I'm a huge advocate of responsible goto usage. Modern control flow primitives made us all lazy 17:08 < ws-client> **** i mean the range is also limited to 1-5 so yea idk 17:09 < ws-client> **** reuseable might be trash anyways 17:09 < ws-client> **** but testable and moving the code away from the votes 17:09 < bridge> i wouldnt care if it worked 17:10 < bridge> robo would kill me tho 17:10 < bridge> Replacing a library that is still getting patches 17:10 < bridge> With an emulation of it is insane 17:10 < bridge> eah 17:10 < bridge> but i wouldnt care if it worked 17:10 < bridge> which it doesnt 17:10 < ws-client> **** but u guys also blame all bugs on the compat layer 17:10 < ws-client> **** without any proof 17:10 < bridge> Yeah, I lost that battle in this project sadly. Goto makes such nice patterns, just to emulate a fraction of its power you need to invent exceptions 17:11 < bridge> without any proof 17:11 < bridge> my mouse keeps escaping the window in sdl2 compat but in sdl2 it doesnt 17:11 < bridge> There are at least 2 bugs that I know of caused by it. They were confirmed by people linking to actual sdl2 manually 17:11 < bridge> literally unplayable 17:11 < ws-client> **** .-. 17:11 < bridge> So it is worth checking 17:11 < bridge> yes 17:12 < bridge> I had to disconnect all my other monitors and set ddnet to full screen xd 17:13 < bridge> now I'm single monitor enjoyer with all my workspaces 17:13 < bridge> This is the demo for it since I can't paste demos on GitHub. To clarify, this was v2.0. We were in a CTF game and everybody experienced the same bug. 17:13 < bridge>