00:22 < bridge> but hes back 00:33 < bridge> yo how do i statically link the ddnet libs? 00:33 < bridge> yo how do i statically link all the libs ddnet needs? 00:48 < bridge> Why does the executable even when compiled with HEADLESS_CLIENT still depend on things like libGLEW and libSDL? 00:48 < bridge> any quick way to fix it? 00:57 < bridge> how are the linux releases for ddnet compiled? 00:57 < bridge> they don't rely on system deps do they? 01:03 < bridge> Definitely not 01:04 < bridge> <0xdeen> They rely on some system dependencies 01:04 < bridge> they do AFAIK 01:04 < bridge> not all of them though 01:04 < bridge> <0xdeen> but we use a very old system to build it, so that any supported Linux should be fine 01:08 < bridge> yea i was compiling a custom headless client for a server where i can't install packages. well i keep getting missing shared libs like sdl2. any way to solve that? 01:09 < bridge> yea i was compiling a headless client for a server where i can't install packages. well i keep getting missing shared libs like sdl2. any way to solve that? 01:09 < bridge> `cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=On -DHEADLESS_CLIENT=On -DPREFER_BUNDLED_L 01:09 < bridge> IBS=On -DVULKAN=Off -DVIDEORECORDER=Off -DWAVPACK_CLOSE_FILE=Off` 01:09 < bridge> this is my current cmake command 01:10 < bridge> ```cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=On -DHEADLESS_CLIENT=On -DPREFER_BUNDLED_L 01:10 < bridge> IBS=On -DVULKAN=Off -DVIDEORECORDER=Off -DWAVPACK_CLOSE_FILE=Off``` 01:10 < bridge> this is my current cmake command 01:12 < bridge> ``` 01:12 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:12 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:12 < bridge> libSDL2-2.0.so.0 => not found 01:12 < bridge> libopusfile.so.0 => not found 01:12 < bridge> libopus.so.0 => not found 01:12 < bridge> libogg.so.0 => not found 01:12 < bridge> ``` 01:12 < bridge> output of ldd DDNet | grep "not found" 01:13 < bridge> ``` 01:13 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:13 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:13 < bridge> libSDL2-2.0.so.0 => not found 01:13 < bridge> libopusfile.so.0 => not found 01:13 < bridge> libopus.so.0 => not found 01:13 < bridge> libogg.so.0 => not found 01:13 < bridge> ``` 01:13 < bridge> output of `ldd DDNet | grep "not found"` 01:16 < bridge> <0xdeen> You are building on too new of a machine, you need an older one for the glibc error 01:17 < bridge> <0xdeen> There is probably a bug where our build system still links all these libs even though it's a headless client 01:17 < bridge> yea its not implemented correctly. 01:18 < bridge> <0xdeen> the headless client was just for testing in CI I think 01:18 < bridge> but i wish it would link them statically atleast tho. is that possible? 01:19 < bridge> <0xdeen> copy the .so files to the same directory and run with `LD_LIBRARY_PATH=.` I guess 01:19 < bridge> well i guess that works xd 01:19 < bridge> <0xdeen> and fix it in our build system, these dependencies should not be linked in for headless client 01:20 < bridge> <0xdeen> search for `if(CLIENT` in CMakeLists.txt, most of the `find_package` etc should not run for `HEADLESS_CLIENT` 01:21 < bridge> you would need to completely exclude all sdl code in the src if HEADLESS_CLIENT is defined tho no? 01:21 < bridge> <0xdeen> hm, might be more complex, not sure 01:22 < bridge> what are you trying to use the headless client for? ^^ 01:22 < bridge> heyo, I want some input for twmap. I recently changed the compression backend to zlib-ng, as this sped up the parsing of maps with a factor between 2 and 3. 01:22 < bridge> One kinda downside is that now the maps twmap produces arent the exact same anymore, but that is fine imo(?) 01:22 < bridge> With the same compression level 6 (number between 1 and 9), the files are slightly bigger, which is ofc negative. 01:22 < bridge> I was thinking about which level I could choose alternatively (higher -> smaller files, but takes longer) 01:22 < bridge> So I made this table, which was made with all brutal maps, and the different compression backends 01:22 < bridge> the times are `time/usr time`, in this case the `usr` time is the significant one 01:22 < bridge> ``` 01:22 < bridge> size load load+save 01:22 < bridge> zlib 231M 34s 174s 01:22 < bridge> zlib-ng6 238M 12s 67s 01:22 < bridge> zlib-ng7 225M 89s 01:22 < bridge> zlib-ng8 216M 119s 01:22 < bridge> zlib-ng9 209M 186s 01:22 < bridge> ``` 01:22 < bridge> imo, the compression level 7 looks reasonable. what do you think? 01:23 < bridge> ah, I removed the non-usr time in the table 01:23 < bridge> how does the zlib-ng* compression affect zlib decompression timese? 01:24 < bridge> i.e. for the maps compressed with zlib-ng*, how long does it take to load them with zlib? (as an approximation for the ddnet client) 01:24 < bridge> 👍 will test 01:24 < bridge> <0xdeen> I think zlib-ng6 looks best 01:25 < bridge> <0xdeen> compared to ng7: 25% faster and only 6% larger file size 01:25 < bridge> <0xdeen> Can we use zlib-ng in DDNet? Would it be faster there? 01:26 < bridge> hm, my motivation for this investigation was that Welf noticed that `$optimize` in the testing channels sometimes increased the file size, which was the result of the backend change 01:27 < bridge> does anyone know what is this client? its on a chillerdragon website 01:27 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1270523573735391286/image.png?ex=66b402c9&is=66b2b149&hm=2b1e95deb5ce12e0bc184aeb0b4308170d4d7131c97b11c867a832adebe35c6f& 01:27 < bridge> <0xdeen> You can also respond to Welf with "small file size increase doesn't matter much, map loads faster now" 01:28 < bridge> <0xdeen> but I'm ok with ng7 too 01:28 < bridge> should be possible, not sure if it would come with other issues 01:29 < bridge> well the faster load is only in the twmap library right now, where it doesn't matter all that much I think 01:29 < bridge> maps are rarely saved, but downloaded very often, so smaller file sizes could be neat 01:30 < bridge> doesn't seem to be affected, also 12 seconds for ng to load the compression level 9 maps 01:34 < bridge> this works ig. 01:34 < bridge> only issue is glibc now: 01:34 < bridge> ``` 01:34 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:34 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.38' not found (required by ./DDNet) 01:34 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.38' not found (required by ./libSDL2-2.0.so.0) 01:34 < bridge> ./DDNet: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.38' not found (required by ./libSDL2-2.0.so.0) 01:34 < bridge> ``` 02:16 < bridge> you need to downgrade ur host or upgrade ur server ig 02:24 < bridge> that's annoying and I don't know a good solution for it 02:24 < bridge> it just shouldn't be so hard™ 😦 02:27 < bridge> is docker an option? 02:42 < bridge> the solution was find a third pc with a lower version and move all the files there to compile it lmao 02:48 < bridge> despite its drawbacks, i think statically linking glibc would still be a good solution for this 03:18 < bridge> how... 03:18 < bridge> i would 03:18 < bridge> if i knew 03:19 < bridge> ```./DDNet: error while loading shared libraries: libpulse.so.0: cannot open shared object file: No such file or directory```??? 03:19 < bridge> from SDL 03:19 < bridge> FK 03:21 < bridge> right 03:22 < bridge> u can use ldd to find out all runtime dependent lbs 03:23 < bridge> well 03:23 < bridge> those needed by ld.so 03:23 < bridge> and static linking is stupid 03:24 < bridge> and static linking libc is stupid 03:25 < bridge> there's no abstraction for it in cmake so it's essentially a matter of providing the right compiler/linker args for the right system/config 03:26 < bridge> you may want `-static` 03:26 < bridge> there are more tho 03:26 < bridge> u may wanna just try it out 03:27 < bridge> am i a developer? 03:27 < bridge> do you develop? 03:27 < bridge> this is a public chat 03:27 < bridge> i tend to dabble in the arts 03:27 < bridge> then yea 03:28 < bridge> im honored 03:28 < bridge> but to be purple i think u have to commit to ddnet 03:28 < bridge> i believe this is my game 03:28 < bridge> i have a whole 5.5hrs you know 03:28 < bridge> nice 03:28 < bridge> nice 03:29 < bridge> change something crucial in readme :troll: 03:29 < bridge> actually 03:29 < bridge> if u are a capable dev there should still be some pretty simple chores 03:29 < bridge> lots of unsolved issues rn 03:29 < bridge> woah woah woah slow down there bud, i said dev not capable dev 03:29 < bridge> fair 03:30 < bridge> xD 03:30 < bridge> i'd say try it anyway 03:30 < bridge> so you are a /dev/ 03:30 < bridge> i develope deseases 03:30 < bridge> i should grab some signals from ya 03:30 < bridge> i develop deseases 03:31 < bridge> i develope deseases 03:31 < bridge> i've known taylor for 5 mins and i already trust him with my block devices 03:31 < bridge> i appreciate that friend 03:31 < bridge> already dmed him my private keys 03:31 < bridge> o geez 03:31 < bridge> why is there a lil heart next to my name 03:31 < bridge> gg teerosrc leaked 03:32 < bridge> xD 03:32 < bridge> it's an icon Discord automatically adds to members who are new to a server 03:32 < bridge> everyone loves you 03:32 < bridge> oh no, ddnet will fall to ruin 03:32 < bridge> how lovely 03:32 < bridge> i concur 03:32 < bridge> i purr 03:33 < bridge> i 03:33 < bridge> cant think of anything witty 03:33 < bridge> damn 03:33 < bridge> maybe the lack of sleep 03:33 < bridge> ah ye its 4 am 03:33 < bridge> eepy time 03:33 < bridge> my clock has said 8:45am for the past 2 weeks 03:33 < bridge> gn everyone 03:34 < bridge> only 2:30 for me 03:34 < bridge> mine says 1:33 rn because i think whenever i resume from suspend systemd does some dumb shit 03:34 < bridge> mine says 1:33 AM rn because i think whenever i resume from suspend systemd does some dumb shit 03:34 < bridge> but it is 7:34 pm 03:34 < bridge> where abouts do you live 03:34 < bridge> smells like france or spain 03:35 < bridge> im mountain time 03:35 < bridge> nope, guess again 03:35 < bridge> so USA 03:35 < bridge> so I'm from the USA 03:35 < bridge> funny enough i3 freezes after suspend id you try to use input device 03:35 < bridge> i have a mate on holiday in the USA as we speak 03:35 < bridge> english? 03:36 < bridge> oh no my vocabulary may have given away where i'm from 03:36 < bridge> lol 03:36 < bridge> i did watch movie and suspended it midway xd 03:36 < bridge> launched pc again 03:36 < bridge> and watched until end 03:36 < bridge> I'm rather bored 03:36 < bridge> ye same 03:36 < bridge> so it stuck right after i touched mouse 03:37 < bridge> the sensible thing to do in this situation would normally be to sleep as i have important things happening tommorow and it's 2:37am 03:37 < bridge> hm so you are zombietroll timezone neighbor 03:37 < bridge> i'm not sure 03:38 < bridge> i could be however i do not have he foggiest clue as to what that is 03:38 < bridge> i could be, however i do not have he foggiest clue as to what that is 03:38 < bridge> its gb 03:38 < bridge> come again? 03:38 < bridge> oh great britain 03:39 < bridge> nah gorgeous bahrain 03:39 < bridge> have you ever made contact with a live rat? 03:39 < bridge> ye 03:39 < bridge> theyre pretty big 03:39 < bridge> was it a pet or a wild rat 03:39 < bridge> especially if you compare their sizes with mice 03:39 < bridge> ofc wild 03:40 < bridge> my father had 2 pet rats when i was younger 03:40 < bridge> their little tails were really odd 03:40 < bridge> imagine 2 pet rats racing 03:40 < bridge> i was a bafoon for not making that happen 03:40 < bridge> silly 8 year old me 03:40 < bridge> unfortunately 03:41 < bridge> are you also a developer of DDnet? 03:42 < bridge> oh the heart is because i'm new, not because everyone loves me. That's slightly disappointing. 03:50 < bridge> You, my friend, have lied to me. I'll have you demoted for that. 05:19 < bridge> why do i always miss the tomfoolery moments in every single server im in 05:19 < bridge> 😩 06:05 < bridge> Oh boi thats old. I think I never even tried that client haha. Its based on teeworlds 0.6 so its pretty outdated these days. If I remember correctly it added fun cosmetics such as more particles when tees die. 06:06 < bridge> or i confuse it with Goreworlds 06:40 < bridge> Maybe that's a better idea xd 06:43 < bridge> It's not that wrong if it works :justatest: 07:45 < bridge> but isn't it wrong? for example if the struct has 3 fields of type `u8` and I want to write to the first one(for example struct offset is 0), if i use struct offset + field offset, it would write to `rbp - 1` but it actually has to write to `rbp - 3` 07:45 < bridge> it has to write in reverse order, if that makes sense 08:07 < bridge> are u even taking in consideration alignment? xd 08:07 < bridge> no, aligment for noobs 08:10 < bridge> hmm idk what u struggling with, but yes make it all 0-based, structs fields are usually all laid sequentially, even nested structs, you kind of "flatten" them in memory, i dont think it should be that hard 08:10 < bridge> the only thing a bit hard is to think about alignments 08:11 < bridge> https://tenor.com/view/smh-gif-21608426 08:12 < bridge> why the gif 08:12 < bridge> xd 08:14 < bridge> - stack offset: where your stack starts 08:14 < bridge> - struct offset: the offset from your stack offset to where the first field of the struct is 08:14 < bridge> - field offset: the offset from the start of your struct offset to where the field is 08:14 < bridge> 08:14 < bridge> addr = stack offset - struct offset - field offset 08:14 < bridge> read (size of field) on addr 08:14 < bridge> its - cuz stack on x86 grows downwards (iirc its morning need coffee) 08:15 < bridge> if u wanna take alignment in consideration, the only thing u would change is make the struct offset and field offsets differ from their sizes if needed 08:15 < bridge> cuz u need padding 08:28 < bridge> definitly 6 or 7, the jump from 7 to 8 has a way bigger increase in time. 08:52 < bridge> DDNet-musl when? 08:52 < bridge> What do you mean not the same anymore? They get different sha sums when being resaved with twmap? Or the same map created with twmap is different than the one with the ddnet client? 08:53 < bridge> I took that personally... 08:54 < bridge> hi please help. Why cant i update my ddnet since 18.01 08:54 < bridge> Learath gg help 08:54 < bridge> In general I appreciate if the files do not change. I track everything in git at all times. And use a dirty working tree to monitor unexpected changes to files. 08:54 < bridge> sorry Learath2 08:54 < bridge> This Client Versions Auto updated is broken, Download Off the website 08:54 < bridge> This Client Versions Auto updater is broken, Download Off the website 08:54 < bridge> i dont want to have to redownload off of site each time 08:54 < bridge> Not each time, Just once ^^ 08:55 < bridge> Its fixed again, it was that one Release where it broke 08:55 < bridge> oh 08:55 < bridge> xDD i wasnt reading 08:55 < bridge> lets see 08:55 < bridge> No worries jimmy, gimme Credit on your next Song and we're good :greenthing: 08:55 < bridge> https://tenor.com/view/kurt-angle-stare-100-yard-stare-meme-gif-251436459491629941 09:00 < bridge> Why are you reversing the field order? Even if you are doing that (which I think you shouldn’t), you’d be calculating the offsets at compile time for your fields 09:03 < bridge> is `rbp - 1` correct for first element? 09:04 < bridge> Yes 09:04 < bridge> Well it’s your language, you can lay out your structs however you damn well please 09:05 < bridge> what the hell is this then :feelsbadman: 09:05 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1270638825554841612/image.png?ex=66b46e1f&is=66b31c9f&hm=c9acce757a5685bc5b0281253c80549f238ad9df7d531ae7525078581d3c6645& 09:11 < bridge> what the hell is this then :feelsbadman: why value 1 isn't moved in `rbp - 1` 09:11 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1270638825554841612/image.png?ex=66b46e1f&is=66b31c9f&hm=c9acce757a5685bc5b0281253c80549f238ad9df7d531ae7525078581d3c6645& 09:11 < bridge> Why do you care what C does? 09:12 < bridge> i want to be able to call C functions xd 09:12 < bridge> when they make a c++ std lib :feelsbadman: 09:12 < bridge> (never) 09:15 < bridge> Well if you want to lay it out like that, then have the offset of the struct point at the end 09:22 < bridge> your first problem is calling the struct and variable the o face 09:22 < bridge> chillerdragon: yeah different sha sums 09:22 < bridge> :justatest: 09:22 < bridge> the speedup is just so darn significant 09:23 < bridge> ```c 09:23 < bridge> struct foo { // stack offset 0xC 09:23 < bridge> uint32_t first; // field offset 0x0, stack offset 0xC - 0x0 = 0xC 09:23 < bridge> uint32_t second; // field offset 0x4, stack offset 0xC - 0x4 = 0x8 09:23 < bridge> uint32_t third; // field offset 0x8, stack offset 0xC - 0x8 = 0x4 09:23 < bridge> } 09:23 < bridge> ``` 09:25 < bridge> do you actively resave the file and check for changes? 09:32 < bridge> To be fair. Maybe I misread your image. The lowest address is where you want to be pointing with your offset for all types on a little endian stack 10:01 < bridge> Does anyone know how to solve being banned for using a VPN? 10:03 < bridge> Use #✉-create-a-ticket and click on `Admin-mail`, they will Help you Out ^^ 10:15 < bridge> Thanks for taking time and explaining xd I.. came up with something which seems to work xdddd 10:16 < bridge> Please don't have 1-based offsets, I beg of you 10:16 < bridge> i made it 0-based 10:16 < bridge> <0xdeen> Do we have anyone around who has this graphics bug where with Vulkan backend some text is invisible? https://wiki.ddnet.org/wiki/GFX_Troubleshooting#Some_text_is_invisible_with_the_Vulkan_backend If so, can you try out https://ddnet.org/downloads/DDNet-vulkan-win64.zip and see if it fixes it? 10:19 < bridge> I'll see if i can downgrade the GPU of my father 10:22 < bridge> okay different take @heinrich5991 @0xdeen @jxsl13 chillerdragon 10:22 < bridge> map files are typically distributed and read many many times more than they are written. 10:22 < bridge> this is ofc highly true for released maps, but I think even also true for maps shared in the map testing channels. 10:22 < bridge> the only case I can currently think of, where maps are not really shared, is maps being auto-saved in the editor 10:22 < bridge> 10:22 < bridge> we have the trade-off between speed and size, where we need to find some kind of sweet spot 10:22 < bridge> so, as the load+save time with zlib and zlib-ng9 is comparable, why not go there? 10:22 < bridge> we distribute map files potentially thousands of times, so 2.25x* the effort on saving between zlib-ng7 and zlib-ng9 isn't that huge, for a ~9% reduction in file size 10:22 < bridge> 10:22 < bridge> *load+save time subtracted by the load time, then divided for percentage 10:23 < bridge> here's the abomination btw `stack_offset - struct_size + field_offset + field_type_size` looks bad but at least works 😬 10:24 < bridge> This doesn't look too wrong, but why are your offsets not precalculated to point at the end? 10:25 < bridge> Offsets to the wrong side of a value are pretty useless, the cpu can't do anything with those, you'll pretty much always have a + size wherever you use them 10:26 < bridge> `field_offset`? 10:27 < bridge> Yes, why doesn't `struct_offset - field_offset` point to the lowest byte of the field directly? 10:30 < bridge> probably it's because of how I calculate the offset :\ 10:30 < bridge> ```rust 10:30 < bridge> pub fn offset( 10:30 < bridge> &self, 10:30 < bridge> ... 10:30 < bridge> ) -> Result { 10:30 < bridge> let mut offset = 0; 10:30 < bridge> 10:30 < bridge> for (field_name, type_) in &self.fields { 10:31 < bridge> if name == field_name { 10:31 < bridge> break; 10:31 < bridge> } 10:31 < bridge> 10:31 < bridge> offset += type_.size(arch, scope)?; 10:31 < bridge> } 10:31 < bridge> 10:31 < bridge> Ok(offset) 10:31 < bridge> } 10:31 < bridge> ``` 10:42 < ws-client> @patiga so what makes the shasum change? Is there no way whatsoever anymore to recreate the same map file? 10:42 < bridge> a different zlib compression backend 10:42 < bridge> I will probably try to get the same change into ddnet 10:42 < ws-client> but the zlib backends will match between each other? 10:43 < ws-client> and if its also in ddnet everything matches again? 10:43 < bridge> different zlib backends don't produce the same binary data 10:43 < ws-client> but same do? 10:43 < bridge> yes 10:44 < ws-client> Ah then its fine with me 10:44 < bridge> nice :) 10:45 < ws-client> but yeah ideally the switch in ddnet would also actually happen :p 10:45 < bridge> yea 😇 10:46 < bridge> Mh, I actually don't get it. This looks fine. Maybe I'm stupid 10:47 < bridge> nah, it's my skill issues xdd 10:50 < bridge> Several people asked over the past months if we can enable it for KoG too, is there any way to get it into the client or is there proper support planned or no support planned for 3rd party mods? 10:50 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1270665428024295425/indicate_finish.png?ex=66b486e5&is=66b33565&hm=5bc555f7f98575decd321816cb6a7824aa6432a623b8a376bc9e5794ed1bca56& 10:51 < bridge> btw its not as low level but this is how llvm gets the struct/array/etc addresses at compile time https://llvm.org/docs/GetElementPtr.html 10:51 < bridge> btw its not as low level but this is how llvm gets the struct/array/fields addresses at compile time https://llvm.org/docs/GetElementPtr.html 10:51 < bridge> https://llvm.org/docs/LangRef.html#getelementptr-instruction 10:52 < bridge> ```c 10:52 < bridge> For example, let’s consider a C code fragment and how it gets compiled to LLVM: 10:52 < bridge> 10:52 < bridge> struct RT { 10:52 < bridge> char A; 10:52 < bridge> int B[10][20]; 10:52 < bridge> char C; 10:52 < bridge> }; 10:52 < bridge> struct ST { 10:52 < bridge> int X; 10:52 < bridge> double Y; 10:52 < bridge> struct RT Z; 10:52 < bridge> }; 10:52 < bridge> 10:52 < bridge> int *foo(struct ST *s) { 10:52 < bridge> return &s[1].Z.B[5][13]; 10:52 < bridge> } 10:52 < bridge> 10:52 < bridge> The LLVM code generated by Clang is approximately: 10:52 < bridge> 10:52 < bridge> %struct.RT = type { i8, [10 x [20 x i32]], i8 } 10:52 < bridge> %struct.ST = type { i32, double, %struct.RT } 10:52 < bridge> 10:52 < bridge> define ptr @foo(ptr %s) { 10:52 < bridge> entry: 10:52 < bridge> %arrayidx = getelementptr inbounds %struct.ST, ptr %s, i64 1, i32 2, i32 1, i64 5, i64 13 10:52 < bridge> ret ptr %arrayidx 10:52 < bridge> } 10:52 < bridge> ``` 10:53 < bridge> do you have benchmarks for the missing load times :? 10:54 < bridge> they are same for all zlib-ng compression levels 10:54 < bridge> as far as I can tell 10:54 < bridge> 12s? 10:55 < bridge> yes 10:55 < bridge> I think there are no real maps as big as 200+MB 10:56 < bridge> is it possible to like find the biggest ddnet map and have those benchmarked 10:56 < bridge> or an average of a few big maps 10:57 < bridge> I cannot really imagine how those times will be with real maps 10:57 < bridge> why was I banned for vpn even though I don't have it? 10:57 < bridge> those stats are from all brutal maps 10:57 < bridge> hm 10:58 < bridge> how would it impact someone using an editor ._. 10:58 < bridge> but twmap is not used in the editor, I guess 10:58 < bridge> How to set `has_finishes: false` to true for the KoG community on info.ddnet.org? 10:59 < bridge> would definitly make sense to use the highest compression. 10:59 < bridge> then 11:00 < bridge> I don't know if the editor saves asynchronously, if yes, then it wouldn't even really have an effect there 11:04 < bridge> does one want to save asynchronously at all. Save -> quit -> map saved in a broken state? 11:05 < bridge> interesting point, I don't know what happens there 11:05 < bridge> at least the autosaves in the background should be asynchronous though 11:05 < bridge> hmhm 11:11 < bridge> https://github.com/khonsulabs/pot 11:11 < bridge> @heinrich5991 did u know it? 11:11 < bridge> ``` 11:11 < bridge> $ cargo test --example logs -- average_sizes --nocapture 11:11 < bridge> Generating 1000 LogArchives with 100 entries. 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | Format | Bytes | Self-Describing | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | pot | 2,627,586 | yes | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | cbor | 3,072,369 | yes | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | msgpack(named) | 3,059,915 | yes | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | msgpack | 2,559,907 | no | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | bincode(varint) | 2,506,844 | no | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> | bincode | 2,755,137 | no | 11:11 < bridge> +-----------------+-----------+-----------------+ 11:11 < bridge> ``` 11:17 < ws-client> @avolicious imo that would belong in a per community json api endpoint. So every community has their info endpoint. But that would mean every launching client would do a bunch of http requests to all communities so thats a bit meh 11:17 < ws-client> also heinrich said communities should not have an advantage 11:19 < bridge> We intended to, but I don't think we figured out how we can expose this feature to 3rd parties 11:19 < bridge> if you can provide us with a list of (playername, finished map) pairs, we can include them in the info.json 11:19 < bridge> Mh, this would work for kog since we don't have an overlapping map pool. But it's just a stopgap 11:20 < bridge> nah, we'd save (community, player name, finished map) tuples 11:20 < bridge> We can add map hash if needed, there are some maps on both networks 11:22 < bridge> Still not my very favourite tbf, so do all communities give us an endpoint that we fetch that from whenever we get an info request? 11:22 < bridge> So the only param you need is the username? is the username encoded somehow? urlencode? 11:22 < bridge> opposite way around. we provide an endpoint to add a pair (player name, finished map) to the community 11:23 < bridge> we store tuples (community, player name, finished map) and do a DB query when we get a request 11:23 < bridge> this way, we don't need to make a request and leak no player data 11:24 < bridge> But in exchange we have to keep a db of all communities finishes 11:24 < bridge>