00:13 < bridge> nice job 00:50 < bridge> ty 07:18 < bridge> Nice @kollpotato 07:22 < bridge> going to a farmer job finally? 07:48 < bridge> Yea I sent him into the melon garden 07:49 < bridge> gumo btw 08:07 < ws-client> https://zillyhuhn.com/cs/.1751868437.png 08:07 < ws-client> gm 08:17 < bridge> gm ^.^ 08:22 < bridge> the heck. 08:33 < bridge> how can a moderator know IDs of players? 08:35 < bridge> cl_nameplates_id 1 08:35 < bridge> you can type status in rcon as well 08:36 < bridge> There's also a scoreboard option 08:36 < ws-client> @Jupstar ✪ why is the time_get cached per tick? performance? or to be able to compare the values? Seems a bit complex to me 08:37 < ws-client> i am sure someone tryint to benchmark his tick functions got proper trolled by that 08:56 < bridge> Idk, wasn't my idea 09:02 < ws-client> oh i thought 09:02 < ws-client> hm 09:02 < ws-client> https://github.com/ddnet/ddnet/commit/aede6df2ca3b86b5b8c197731f0ca9d3c63031f5 09:03 < ws-client> @0xdeen why u do dis? 09:09 < bridge> Brother that was 11 years ago :KEKW: 09:10 < ws-client> older than the average Multeasymap enjoyer 09:11 < bridge> maybe to have the exact same time within a tick 👀 09:11 < ws-client> @jxsl13 for what xd 09:11 < bridge> my uneducated 2 cente, you are welcome. 09:11 < ws-client> it just breaks benchmarking and if you want to know if its the same tick look at the tick counter 09:11 < bridge> for calculations and stuff based on time? 09:12 < ws-client> ah so you use the timer for stuff that should be tick dependent 09:12 < ws-client> instead of using the tick 09:12 < ws-client> i see yea that might be it 09:13 < bridge> wtf i discovered a bug while implementing `broadcastid` 09:15 < bridge> why do some commands specify `v[id]` in usage and some use `i[id]`? 09:16 < bridge> there is the `pResult->GetVictim()` method 09:16 < bridge> but why does it exist anyways 09:16 < bridge> what's the point 09:17 < bridge> if it can be just `pResult->GetInteger(0)` 09:17 < ws-client> v[id] is more correct 09:18 < ws-client> not all commands have been ported yet 09:18 < bridge> what does v stand for again, value? 09:18 < ws-client> victim 09:18 < ws-client> i is integer 09:18 < bridge> actually? 09:18 < bridge> We tf 09:18 < ws-client> and v is supposed to be a client id 09:18 < ws-client> also accpets magic values such as ``all`` 09:18 < bridge> so does `pResult->GetVictim()` do anything special? 09:18 < ws-client> yes 09:18 < bridge> `all`? 09:18 < ws-client> ``all`` ``me`` and more idk 09:19 < bridge> it returns 0 for negative numbers 09:19 < ws-client> ``kill_pl all`` kills all 09:19 < bridge> which i would say is incorrect 09:20 < bridge> okay so i was right to use `pResult->GetVictim()` then 09:20 < bridge> but negative numbers should return an error for sure 09:31 < ws-client> i didnt look at the code but the idea might be that a victim can be used as array index for players without error checking because it can not be out of bounds 09:32 < ws-client> does it even call the callback when you provide a -1 victim? 09:38 < bridge> Callback? 09:38 < bridge> The command runs with client id 0 09:40 < bridge> ^ @kebscs 09:42 < bridge> So basically you can't know if `pResult->GetVictim()` returned 0 because the user typed in 0 or if it returned 0 because the user typed in a negative number 09:45 < bridge> So for example `muteid -42 10` mutes player with id 0 for 10 seconds 09:46 < bridge> And i think it shouldn't run the command 09:50 < bridge> Chiller you can also request the time directly uncached 10:07 < bridge> Old fastcap shouldn't send the flags 10:07 < bridge> If you're making your own fastcap then you can remove the flipping 10:28 < bridge> https://cdn.discordapp.com/attachments/1390347378372120576/1390681117241249984/image.png?ex=68692437&is=6867d2b7&hm=cba41823b5bd7838db72b221b8b77d2a77f3d41cd0fac725b7b0ff74ce3740cb& 10:28 < bridge> https://cdn.discordapp.com/attachments/1390347378372120576/1390681352436846673/image.png?ex=68692470&is=6867d2f0&hm=4fc1b62c4e2811856fcf1f594b658be9006fc5986bc928b71170f6d7e7bd5741& 10:28 < bridge> https://cdn.discordapp.com/attachments/1390347378372120576/1390681600852885595/image.png?ex=686924ab&is=6867d32b&hm=6d04295e0419f930cd8891fcd0b45eb7df54102f85f12b5047582a6b0ab62500& 10:28 < bridge> https://cdn.discordapp.com/attachments/1391481421717180557/1391484812228825090/image.png?ex=686c10b7&is=686abf37&hm=af8ac7d4180cf1bbffa66bae387e3c4e9c794f0b05b99adc296a642919c31d77& 10:29 < bridge> @Discord Mod 10:34 < bridge> how? 10:36 < bridge> oh that's nice 👍 10:38 < bridge> I assumned the server has no control over weapon flipping 10:39 < bridge> but ofc it has, since it loads the entities on map load 11:25 < bridge> chillerdragon: how the hell do snapshot uuids work 11:26 < bridge> ive read 2 different implementations 11:26 < bridge> ur protocol wiki 11:26 < bridge> even asked ai 11:26 < bridge> i have no idea what i am doing 11:48 < bridge> wait i think im cooking now 11:52 < bridge> can deleted item keys have uuids 11:53 < bridge> probably not 11:54 < bridge> ~~you are baked potato now?~~ 11:54 < bridge> yeah, using `all` as a victim works exactly as it should lol 11:54 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391718926714540102/image.png?ex=686ceac1&is=686b9941&hm=5857f786333c10ea6b2b006794088fc211b705c7da4d368bb6aa7bd187de7c05& 11:54 < bridge> no im just cooked 11:54 < bridge> cooked potat 11:58 < bridge> hot potato 12:12 < bridge> i think this is all wrong 12:12 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391723628101632080/image.png?ex=686cef21&is=686b9da1&hm=8cbdaf9ab048d497f841cd809f1165b1fd1fd65149d3ed332aa3cc0b832be944& 12:12 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391723628629983334/image.png?ex=686cef22&is=686b9da2&hm=d5f9c50cea7710a3401ced71f5bce8ecfacd629cbaefb22f1ab725ffb911d51d& 12:13 < bridge> how is y negative lmao 12:15 < bridge> you can jump out of the map, y negative, means up above the map 12:15 < bridge> I think it was about 200 blocks in each direction 12:15 < bridge> maybe 300, one was the render limit 12:19 < bridge> yep y can be negative 12:19 < bridge> but not in my case 12:19 < bridge> well im not deleting items from the snapshot maybe thats the issue 12:22 < bridge> I'm not sure how far you've read the docs, are you aware of the snap deltas? 12:23 < bridge> yes i am 12:23 < bridge> im applying the delta value to the original snapshot 12:24 < bridge> but do you do it on a new copy of the original snapshot? 12:24 < bridge> in demos the deltas build a chain, over the network they are independant of each other iirc 12:26 < bridge> wait i dont think i do 12:26 < bridge> i store the first snapshot as X, get the delta snapshot and apply it to X 12:26 < bridge> so that X is the current snapshot data 12:27 < bridge> i thought thats how snapshots work 12:27 < bridge> over the network, deltas might get lost. that is why later deltas can't rely on earlier deltas 12:28 < bridge> yeah i was wondering about that 12:28 < bridge> bcus i know ddnet can recover from snapshot packet loss 12:28 < bridge> wasnt sure how xd 12:28 < bridge> so each delta is a delta to the original snapshot, not a continuation of the last delta 12:29 < bridge> thanks i am going to try that now 12:29 < bridge> gl :) 12:29 < bridge> ty 13:14 < bridge> well one issue is gone 13:14 < bridge> another one came 13:15 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391739280690385018/image.png?ex=686cfdb5&is=686bac35&hm=56d1358ec497d1d98195c6fb155961e291722a1b07aa7f343f1341ddd4b6dc3f& 13:15 < bridge> snapshot item size is in billions 13:15 < bridge> https://tenor.com/view/yes-yes-sir-yayy-kataman-gif-12260883688244422951 13:30 < bridge> No way 1 year old pr merged 13:34 < bridge> 💪 13:57 < bridge> 8+- 14:11 < bridge> now we don't need a mouse anymore, we can just type a map 14:12 < bridge> Hey cursor, map me a blockmap 14:12 < bridge> checked out the feature and its pretty jank 14:13 < bridge> I wonder, why he didn't use ASCII offsets 14:13 < bridge> like we have 255 fields, and some control character could just be used for the arrows then 14:14 < bridge> its just a-z 0-9 and only works with tiles that are set up like the teeworlds default font 14:14 < bridge> so aip gores font doesnt work 14:14 < bridge> yeah exactly 14:14 < bridge> also ctrl+t is weird keybind 14:15 < bridge> making a font to letter generator would be very easy, if we'd just use ascii offsets 14:15 < bridge> and it starts placing text at same place your mouse was so you kinda have to aim beforehand 14:15 < bridge> making a font to letter/tileset generator would be very easy, if we'd just use ascii offsets 14:15 < bridge> and after you press enter you cant go back 14:16 < bridge> that's actual jank 😦 and I hoped you could press cntrl enter for next line 14:16 < bridge> i mean enter goes to new line 14:16 < bridge> but no way to go back to the previous one 14:19 < bridge> nvm arrows work ig 14:32 < bridge> @patiga yippe the data is correct now 14:32 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391758753086312458/image.png?ex=686d0fd8&is=686bbe58&hm=f77d0a7157cbbe9f9ecc73d8296b56348a4237561fa4753589b50d6e77e27ebb& 14:32 < bridge> but only once 14:32 < bridge> then it just stops updating xd 14:33 < bridge> Pog 🎉 15:52 < bridge> Sadly Heinrich insisted on fixed layout not supporting custom font offsets. I had two proposals for that implemented over the years 15:53 < bridge> @essigautomat: wdym by ascii offsets 15:54 < bridge> I proposed a pr once where error messages could be silenced for all and kept for single. Was rejected. Its output is not nice but yes it works as it should 15:56 < bridge> You mean the ddnet extensions? They get registered. In the beginning of the snap you have a few items with the uuid as payload the define the type id that will be used for that uuid during this session. All new snap items with that type id are of type matching that uuid 15:59 < bridge> Wdym by original snapshot? The network uses deltas to previous snapshots. The data loss problem is solved by specifying the base snapshot by gametick. The client either has that exact snapshot or a full snapshot is resend as baseline on fatal error. 16:00 < bridge> You can navigate before typing. Where else’s should it start? 16:01 < bridge> i thought you would kinda place it like a tile 16:01 < bridge> or idk 16:01 < bridge> You can do so 16:01 < bridge> how 16:01 < bridge> Move your mouse like you would with a tile and then go into text mode 16:01 < bridge> it insta places it after i press ctrl t 16:02 < bridge> Move first 16:02 < bridge> Where is the difference 16:02 < bridge> yea its not intuitive i mean 16:02 < bridge> i might not notice the text indicator 16:02 < bridge> I see 16:02 < bridge> To me it was 16:02 < bridge> But yea have the mouse become a floating cursor could also work 16:03 < bridge> No hard feeling on it. Did not try the ux could be as good 16:03 < bridge> But your feedback is a bit late you had like 4 years to test it xd 16:03 < bridge> the pr got lost 16:03 < bridge> too old 16:03 < bridge> Now saying because of that it’s pretty jank is rude .\_. 16:04 < bridge> dw its a good addition 16:04 < bridge> not sure who writes a lot of text in maps 16:04 < bridge> but ig itll be easier now 16:04 < bridge> Only rare adventure maps 16:04 < bridge> But even I typing out single words will be using it 16:04 < bridge> I usually type at least one word per map xd 16:05 < bridge> But I won’t be able to use it because of fixed offsets bruv 16:05 < bridge> Makes me so mad that the feature is basically useless for so many font mapres 16:05 < bridge> While it would be so easy to be flexible 16:17 < bridge> https://www.commfront.com/pages/ascii-chart 16:17 < bridge> 16:17 < bridge> what developer doesn't know ascii tables? 16:26 < bridge> omg thats why some type ids are so huge 16:27 < bridge> thank you the protocol genie 16:27 < bridge> i love u 16:27 < bridge> lmao now i have to fix it again 16:27 < bridge> epic 16:27 < bridge> thats why it was stuck 16:33 < bridge> @essigautomat: I know ascii but I don’t know what you propose exactly 16:36 < bridge> I am not sure about the full snapshot on error tho. Would have to look that up again. You don’t need to worry much about it in the beginning. @kollpotato you build a client right? Just store a few snapshots and their game tick. And when you get a new one delta it against the one from storage based on the delta tick 16:37 < bridge> https://chillerdragon.github.io/teeworlds-protocol/07/system_messages.html#NETMSG_SNAP_delta_tick 16:37 < bridge> sounds epic, gonna do that now 16:37 < bridge> because i only store one snapshot now 16:37 < bridge> wasnt sure was delta tick even did xd 16:38 < bridge> is there any library that i can get player info etc from a server, like teeworlds-info npm lib but it only gets server name-maxplayers-ping info 16:38 < bridge> Now you know! Now you pro @kollpotato 16:38 < bridge> @vahemaaa: curl 16:38 < bridge> :D just use the http json master 16:38 < bridge> u proer 16:38 < bridge> me still nub 16:39 < bridge> no u 16:39 < bridge> thank you so much kral! 16:45 < bridge> I propose to use the tileset offset as ascii offsets 16:48 < bridge> @essigautomat: why? 16:48 < bridge> You mean like place A at 0x41 in the mapres? 16:48 < bridge> so we can support special character if we want to AND generate tilesets from fonts directly in an easy manner 16:48 < bridge> I don’t think there is a single mapres that has these offset 16:49 < bridge> like what custom layout does this have now? Do we only support A-Z in upper case and 0-9? 16:49 < bridge> So it wouldn’t even be compatible with any font that’s out there right now 16:49 < bridge> Yes 16:58 < bridge> Just update the mapres that aren't in the standard format. Overtime it'll fix itself 17:01 < bridge> There are more mapres in or close to the alignment including official ddnet ones that I used . There are 0 in ascii. And I’m not sure designers even know about the feature or its offsets 17:02 < bridge> Mine already works. Ascii does not 17:02 < bridge> And for what? That it’s easier to create fonts? It’s not hard to align them in the way they are 17:02 < bridge> If you want to automate it it’s not hard either 17:03 < bridge> Punctuation would be cool but then even less coverage can be guaranteed by existing mapres and not everyone is used to QWERTZ anyways 17:07 < bridge> no, most are used to QWERTY :owo: 17:09 < bridge> Definitely don't use ascii, the standard format should be something that is actually already used by some things at least 17:16 < bridge> chillerdragon: rust borrow checker slows me down a lot xd 17:17 < bridge> i should have used scratch 17:19 < bridge> i either have to redesign the entire client or have about 4 nested match statements resulting in 8 tabs 17:26 < ws-client> thanks @learath2 tell that to assa xd 17:27 < ws-client> @kollpotato rust moment 17:27 < bridge> indeed rust moment 17:27 < bridge> blazingly fast borrowing error 17:42 < bridge> @robyt3 could you test if the notification pr builds for you 17:43 < bridge> i replaced winrt with abi,wrl so maybe it doesnt even need the turned off exceptions 18:04 < ws-client> https://zillyhuhn.com/cs/.1751904186.png i have no custom skin and all skin downloads off and am on a 0.6 server xd 18:06 < bridge> lmao 18:34 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391819710504632370/GvRLqdaXAAcQo5G.png?ex=686d489d&is=686bf71d&hm=dfc689c87f1c2cbe9766a65f20efdb027346323e868702cc46ac6920e559a3a2& 18:35 < bridge> I most certainly did not know this was possible, wtf 18:36 < bridge> my rust cant do that 18:37 < bridge> I guess it is clever 😄 18:37 < bridge> Why would you want to do that tho 18:38 < bridge> To demonstrate that you can 18:39 < bridge> I think you can edit the AST tree in rust 18:39 < bridge> So idk 18:40 < bridge> Seems like that could make this possible 18:40 < bridge> Probably, rust macros are stupid powerful 18:41 < bridge> https://github.com/rust-lang/rfcs/issues/1579#issuecomment-1398724803 yep 18:42 < bridge> Doesn't even need macros 😄 18:45 < bridge> Btw why aren't type systems more powerful? 18:45 < bridge> its hard 18:45 < bridge> What if i want an integer in the range -1 to 2137? 18:45 < bridge> rust has got one of the best ones 18:45 < bridge> That would be an awesome type 18:45 < bridge> make a wrapper type for i16 18:45 < bridge> and validate with the constructor function 18:45 < bridge> Or i feel like sometimes i want struct members to be optional only in certain situations 18:46 < bridge> enums! 18:46 < bridge> Okay but like the type system could also help ig 18:47 < bridge> A struct { 18:47 < bridge> A: T 18:47 < bridge> B: T 18:47 < bridge> } 18:47 < bridge> Can be casted safely into 18:47 < bridge> struct { 18:47 < bridge> A: ?T 18:47 < bridge> B: ?T 18:47 < bridge> } 18:49 < bridge> in rust you would make a method for this 18:50 < bridge> but in c/c++ i suppose you can just cast it 18:50 < bridge> I mean in C optionals don't exist 18:50 < bridge> Only null pointers 18:51 < bridge> yes i love dereferencing null pointers in c 18:51 < bridge> null pointers save a whole one byte comparing to rust 18:51 < bridge> clearly a win 18:52 < bridge> Does it? 18:52 < bridge> How do optionals work 18:52 < bridge> I assume an additional bool can work 18:53 < bridge> But if you have an optional u8 18:53 < bridge> But you know it's never 0 18:53 < bridge> Then 0 can mean no value 18:53 < bridge> yes rust has niche optimization 18:53 < bridge> im not a pro in all that stuff but i know it exists 18:53 < bridge> So `non-zero u8` is actually a useful type to have 18:54 < bridge> thats why rust is the best 18:54 < bridge> and every other languages sucks 18:54 < bridge> i sound like a cultist 18:55 < bridge> Rust sucks 18:55 < bridge> Rust gives you a lot of time to talk on discord while your code is compiling lol 18:55 < bridge> std::optional 18:55 < bridge> its c++ 18:56 < bridge> In cpp i have no idea how it works 18:56 < bridge> yeah gazillion dependencies have to be compiled from scratch 18:56 < bridge> But there probably is a way to cast optional T into T 18:56 < bridge> c/c++ just have shared objects 18:57 < bridge> sounds unsafe 18:57 < bridge> why compare 50 year old c to rust 18:57 < bridge> because i have free will 18:57 < bridge> C is awesome 18:57 < bridge> Even tho it sux 18:58 < bridge> I use zig a lot recently 18:58 < bridge> @learath2 did you see what he wrote 18:58 < bridge> C is awesome I agree 19:00 < bridge> https://tenor.com/view/cpp-dev-bad-developer-code-gif-12287264072566252661 19:01 < bridge> xD 19:01 < bridge> First of all, not trivial, operations between ints of different ranges would lead to very different possible output ranges. 19:01 < bridge> Second there are some languages where it is a thing, like Ada 19:02 < bridge> 1. It's really hard 19:02 < bridge> 2. Someone did it anyways 19:02 < bridge> wow its really a thing 19:02 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391826839168618657/image.png?ex=686d4f41&is=686bfdc1&hm=8c9bddbd51ff110d3334576d010985dff9a753c2bfe40b6a461af99e4d1fb214& 19:03 < bridge> that does not look productive 19:03 < bridge> and no one else really bothered with it afterwards, so it is hard and it doesn't get enough use to make it worth compiler peoples times 19:03 < bridge> There is wuffs by google that also has types that have ranges, (even more generally contracts tbf) 19:04 < bridge> There are the same problems in traditional languages too 19:05 < bridge> Because u8 * u8 could need u16 19:06 < bridge> In C pretty much everything works and it's not obvious what's going to happen 19:06 < bridge> Well we already know how to work with that. There is hardware support to handle it properly. What happens when a 0...10 overflows? The overflow bit isn't set so just to detect it you need branching everywhere 19:07 < bridge> Well we already know how to work with that. There is hardware support to handle it properly. What happens when a 0...10 overflows? The overflow flag or the carry flag isn't set so just to detect it you need branching everywhere 19:08 < bridge> I am not sure i understand what you mean 19:09 < bridge> There are flags on the CPU level 19:09 < bridge> But how can i detect overflows in code? 19:09 < bridge> doesnt overflow cause an interrupt? 19:09 < bridge> and then somehow it catches that interrupt and handles how it wants 19:09 < bridge> I don't think so 19:09 < bridge> Idk 19:10 < bridge> rust in debug mode can just print it, in release build it lets the value go to 0 19:10 < bridge> The compiler can insert overflow detection 19:11 < bridge> But idk how it works 19:11 < bridge> Does not compile 19:11 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1391828887071101049/output.txt?ex=686d5129&is=686bffa9&hm=b774827654db112d71a4c92fda229e26ed32cf9a6279715bc986c9a873518778& 19:13 < bridge> To check that your u8 * u8 overflowed is trivial. To check that your 0...10 * -3...300 overflowed is not, it requires an extra subtraction. So there is a performance penalty. Furthermore the multiplication of those will result in a -3...3000, the division will result in a -infty...10, the addition -3...310. It gets annoying very quick 19:14 < bridge> Well okay i understand 19:14 < bridge> Overall, it's a feature not many people want with many small quirks 19:14 < bridge> It's just an interesting idea 19:14 < bridge> So it's hard to get a compiler person to care enough to implement it 19:14 < bridge> intrinsics 19:14 < bridge> You should check out wuffs 19:14 < bridge> rust has overflowing ops that returns a bool 19:14 < bridge> https://docs.rs/num-traits/latest/num_traits/ops/overflowing/trait.OverflowingAdd.html 19:15 < bridge> https://github.com/google/wuffs 19:15 < bridge> Zig requires explicit casting a lot of the time 19:16 < bridge> @pilonpl more than ints with ranges its more interesting to have ints with values you know wont be used 19:16 < bridge> this is heavily abused by rust 19:16 < bridge> for example we have NonZerou64 19:16 < bridge> Yeah i talked about it 19:16 < bridge> this allows niche optimizations on the zero value 19:16 < bridge> But maybe other compilers can do that too? 19:16 < bridge> https://doc.rust-lang.org/std/primitive.i8.html#method.overflowing_add 19:16 < bridge> https://doc.rust-lang.org/std/primitive.i8.html#method.overflowing_add 19:16 < bridge> Maybe LLVM itself can do that 19:16 < bridge> https://doc.rust-lang.org/std/num/type.NonZeroU64.html 19:17 < bridge> i dont think llvm does that, i dont know any operation that allows u to give such info on top of my head 19:17 < bridge> i worked a bit with llvm 19:17 < bridge> first of all, this opt is used in enums 19:17 < bridge> So how do optionals work usually? 19:17 < bridge> llvm has no enums 19:17 < bridge> u implement them urself 19:18 < bridge> optionals would be enums i guess which llvm also doesnt implement 19:18 < bridge> Any ideas how to fix? 19:18 < bridge> in llvm u would make them by "interpreting" the bits based on a field 19:18 < bridge> enums are really just structs 19:18 < bridge> (tagged unions) 19:18 < bridge> I thought maybe .Ref but I see ReleaseAndGetAddressOf also fails 19:18 < bridge> https://mapping-high-level-constructs-to-llvm-ir.readthedocs.io/en/latest/basic-constructs/unions.html 19:19 < bridge> isnt that just pipes? 19:19 < bridge> https://discourse.nixos.org/t/pre-rfc-pipe-operator/28387 19:19 < bridge> I have no idea how this stiff works 19:19 < bridge> ```llvmir 19:19 < bridge> ; basic type definition 19:19 < bridge> %Foo = type { i8, [8 x i8] } 19:19 < bridge> ; Variants of Foo 19:19 < bridge> %Foo_ABool = type { i8, i8 } ; tagged with 0 19:19 < bridge> %Foo_AInteger = type { i8, i32 } ; tagged with 1 19:19 < bridge> %Foo_ADouble = type { i8, double } ; tagged with 2 19:19 < bridge> 19:19 < bridge> ; allocate the first Foo 19:19 < bridge> %x = alloca %Foo 19:19 < bridge> ; pointer to the first element of type i8 (the tag) 19:19 < bridge> %0 = getelementptr inbounds %Foo, %Foo* %x, i32 0, i32 0 19:19 < bridge> ; set tag to '1' 19:19 < bridge> store i8 1, i8* %0 19:19 < bridge> ; bitcast Foo to the right Foo variant 19:19 < bridge> %1 = bitcast %Foo* %x to %Foo_AInteger* 19:19 < bridge> ; store the constant '42' 19:19 < bridge> %2 = getelementptr inbounds %Foo_AInteger, %Foo_AInteger* %1, i32 0, i32 1 19:19 < bridge> store i32 42, i32* %2 19:19 < bridge> 19:19 < bridge> ; allocate and initialize the second Foo 19:19 < bridge> %y = alloca %Foo 19:19 < bridge> %3 = getelementptr inbounds %Foo, %Foo* %y, i32 0, i32 0 19:19 < bridge> ; this time the tag is '2' 19:19 < bridge> store i8 2, i8* %3 19:20 < bridge> ; cast to variant and store double constant 19:20 < bridge> %4 = bitcast %Foo* %y to %Foo_ADouble* 19:20 < bridge> %5 = getelementptr inbounds %Foo_ADouble, %Foo_ADouble* %4, i32 0, i32 1 19:20 < bridge> store double 1.337000e+03, double* %5 19:20 < bridge> ``` 19:20 < bridge> ```llvm 19:20 < bridge> ; basic type definition 19:20 < bridge> %Foo = type { i8, [8 x i8] } 19:20 < bridge> ; Variants of Foo 19:20 < bridge> %Foo_ABool = type { i8, i8 } ; tagged with 0 19:20 < bridge> %Foo_AInteger = type { i8, i32 } ; tagged with 1 19:20 < bridge> text bomb 19:20 < bridge> I started implementing lox programming language from that book someone recommended in #off-topic 19:20 < bridge> https://learn.microsoft.com/en-us/archive/blogs/ericgu/minus-100-points this is a good link whenever someone is wondering why a language doesn't have X feature 19:20 < bridge> i have an adversion to clicking M links 19:21 < bridge>