DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.org/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2024-05-21 00:00:00Z and 2024-05-22 00:00:00Z

LuKron

is it possible to use a cloudfare tunnels when hosting a ddnet server on my pi?

cloudflare tunnels supports udp, just prob requires you to have WARP on client side

<ChillerDragon> omagawd @MilkeeyCat where stream recording!

02:23

<ChillerDragon> Omg i have hot reloading working and also accessing CCollision from the plugin. But not both at the same time ._. that should be possible right? Everything is possible.

ur hot reload works via dlopen right

<ChillerDragon> If I compile with -rdynamic it fails on the crypto dependencies

02:24

<ChillerDragon> yes

what are you gonna do about windows

<ChillerDragon> there is some compability wrapper for dlopen

justatest

<ChillerDragon> but not sure if i ever bother to support windows tbh

02:25

<ChillerDragon> It can't find _Z10md5_finishP11MD5state_st which demangles to md5_finish(MD5state_st*)

02:25

<ChillerDragon> maybe i have to link ssl into the shared object

02:25

<ChillerDragon> but the linker refuses to link unused libraries. Anyways it seems like the wrong approach.

02:26

<ChillerDragon> I am not even using crypto stuff. Its just that i can't lazy load with dlopen otherwise the hotreload does not work ._.

are you hot reloading every dynamic library

<ChillerDragon> i am only reloading the one thing i built

02:27

<ChillerDragon> void *pHandle = dlopen(pFilename, RTLD_LAZY); this works with hot reloading resolving the symbols lazy

02:27

<ChillerDragon> ah wait no this works without hot reloading xd*

02:28

<ChillerDragon> i have to use RTLD_NOW for hot reloading which then fails on IJob which is fixed by -rdynamic which then fails on md5_finish

02:28

<ChillerDragon> ffs idk what else to tr

i guess i dont understand ur problem atm

<ChillerDragon> do you understand my goal? :D

02:31

<ChillerDragon> i need one of your genius hints or i rq here

you are trying to have one library which you can load on the fly

02:32

for whatever purpose

<ChillerDragon> yep

02:32

<ChillerDragon> its for faster iteration while bot development

yeah

<ChillerDragon> @admins legit server side bots justatest

justatest

02:33

<ChillerDragon> and it needs access to CCollision to see its surrounding tiles

they have that in WinUI because it's much better for fast prototyping/changes

02:33

but is it your version of ddnet or the plugin which is failing to link against md5_finish? also I am not familiar with specific dlopen behaviors but the loading behavior seems like it shouldn't matter too much

<ChillerDragon> my ddnet ui hot reloading project was such a success so i thought i can do anything xd

02:34

<ChillerDragon> i guess its my lib i load that fails

02:34

<ChillerDragon> or the dlopen call actually

it's a runtime error?

<ChillerDragon> because i tell it to load all symbols now

02:34

<ChillerDragon> yes

02:34

<ChillerDragon> well i mean i have it all runtime compile time link time

02:34

<ChillerDragon> which error do you want xd

for the same error? how would you ever get to loading it in runtime if it can never succeed to generate at compile time

<ChillerDragon> xd

02:35

<ChillerDragon> nah i mean i tried a bunch of things and collected a bunch of errors

02:35

<ChillerDragon> idk if i am on the right path but right now i seem to be stuck at a runtime error that fails on dlopen

02:36

<ChillerDragon> but i actually blame a linker issue for it

02:36

<ChillerDragon> idk what i am doing here

if you can't expose md5 summing through ddnet then i'd advise linking it statically in your "plugin"

02:38

but bigger picture i think u need to take a walk or something xd

<ChillerDragon> dude i did a 2h walk over this

ah

<ChillerDragon> this is me progressing 0 since yesterday

i can try and help u more interactively after dinner

<ChillerDragon> lemme record epic clip

02:42

<ChillerDragon> of how it should look like

<ChillerDragon> https://tube.zillyhuhn.com/video.php?t=hot_bots.mp4&u=chiller

02:42

<ChillerDragon> is my internet too slow or is the video bugged? xd

seems fine for me

<ChillerDragon> omg it doesnt load at all for me

02:43

<ChillerDragon> anyways you can see how a recompile instantly changes the walking direction of the tee

02:44

<ChillerDragon> so that works. But it also says that the current tile is -69 which is my fallback if the collision stuff did not work so thats sad

ah

<ChillerDragon> i can get that to work but then it does not hot reload anymore

02:47

<ChillerDragon> this has to be possible rigth?

yes

<ChillerDragon> ok i managed to fix the md5 thing but it does not hot reload tee_thinking

tee_thinking

02:59

<ChillerDragon> so wrong path i guess

[ddnet/ddnet] Pull request opened: #8387 Whitelist class names starting with I

Fixes clang issues being detected in interface classes such as class IJob Fixes the following clangd error: Invalid case style for class 'IJob' (fix available) !image

ws-client

<ChillerDragon> so wrong path i guess

did you figure that part out

<ChillerDragon> yea i can now build link and even load with dlopen and find the symbol with RTLD_NOW and -rdynamic

04:17

<ChillerDragon> it can access collision but it just does not reload

good

04:17

hmm

<ChillerDragon> it stops finding my function as soon as i delete the shared library file

04:17

<ChillerDragon> but when i create it again and it finds it again it calls into the old code somehow :D

04:17

<ChillerDragon> the one i deleted from my harddrive xd

04:18

<ChillerDragon> seems like as soon as i activate -rdynamic and compile my plugin from within the ddnet code base it somehow caches the result of dlsym

04:18

<ChillerDragon> or better it never finds it in the external library but within it self

04:19

<ChillerDragon> i am this close to passing state down as a void pointer and use magic offsets

04:20

<ChillerDragon> to be able to receive a pointer to CCollision it needs collision.h which further depends on vmath.h and protocol.h and if i properly want to link it also on collision.cpp which well depends on everything

04:21

<ChillerDragon> this would be so cool if it worked but idk what else to try

you are trying to use the implementation header but you should be using the interface header

04:23

ICollision

04:23

if it doesn't exist then that might be an issue

<ChillerDragon> does not exist

chillerdragon: I didn't select option to save vods so no recording kek

kek

linking problems over, ChillerDragon?

morning

shitty little terminal map viewer

Image attachment

troll

1

louis

shitty little terminal map viewer

looks cool

nice

06:42

seems like the wrong medium though

ideally i'd have some sort of window app that automatically updates the map whenever the file is updated

06:43

i wanted a quick/easy way to view maps for another project of mine

@Patiga wrote a map viewer tool

06:48

it's easily compilable IIRC

06:49

https://gitlab.com/Patiga/twgpu

Patiga / TwGpu · GitLab

Renderer for Teeworlds and DDNet maps (more to come)

06:50

sample usage: cargo run --release --bin=twgpu-map-inspect -- ~/.teeworlds/maps/Stronghold.map /usr/share/ddnet/data/mapres/

ouch, I should really merge the new-approach branch, as I need to tell everyone to switch to it

07:01

eh I'll do that today

github-merge-queue[bot]

[ddnet:master] 2 new commits

ab64d64 Whitelist class names starting with I - ChillerDragon b4a41ea Merge pull request #8387 from ChillerDragon/pr_clang_format_interfaces - def-

07:16

[ddnet/ddnet] Pull request closed: #8387 Whitelist class names starting with I

<ChillerDragon> @heinrich5991 i am still stuck banging my head at the same problem since over 12 hours now xd

07:40

<ChillerDragon> i red the entire source code of a hot reloading lib and a bunch of man pages and have been try and erroring for quite some time

07:40

<ChillerDragon> @MilkeeyCat you gotta stream again and record this time

okay. first thing: check with readelf whether the symbol is exported

07:41

ChillerDragon

<ChillerDragon> well i can link

then find out why/why not

07:41

but apparently not link at runtime

<ChillerDragon> but it does not run the new code when i update the shared object

07:42

<ChillerDragon> it also claims it finds the symbol at runtime

07:42

<ChillerDragon> dlsym

07:42

<ChillerDragon> i got no error messages anymore :( i fixed them all

the problem is fixed?

<ChillerDragon> no it does not reload

07:42

<ChillerDragon> but without error :(

07:43

<ChillerDragon> maybe my fixed for the error broke the logic

07:43

<ChillerDragon> fixes*

07:43

<ChillerDragon> im a bit lost rn

thats epic where can i buy it?

Replying to @louis tw-map-cli.gif

<ChillerDragon> yo @heinrich5991 why does antibot have these weird function annotations?

ChillerDragon: im on my way to becoming a gitfluencer like u so itll be on gh/l-ouis when i polish it

can you say which annotations you mean?

<ChillerDragon> sec

ChillerDragon: what happens when you try to hot-reload? what do you expect to happen instead?

<ChillerDragon> the api thing in front of the declarations

08:00

<ChillerDragon> i tried so many things sometimes when i delete the shared library it shows errors and when i compile a new one it calls the old code

08:00

<ChillerDragon> or it doesnt error at all an keeps calling the old code

sounds like it doesn't unload

08:01

but what would help are the concrete symptoms right now, so that we could start debugging

<ChillerDragon> yes

for the uninitiated (like me):

c++
ANTIBOTAPI int AntibotAbiVersion();

defined as

c++
#define ANTIBOTAPI DYNAMIC_IMPORT

defined as

c++
#ifdef CONF_FAMILY_WINDOWS
#define DYNAMIC_EXPORT __declspec(dllexport)
#define DYNAMIC_IMPORT __declspec(dllimport)
#else
#define DYNAMIC_EXPORT
#define DYNAMIC_IMPORT
#endif

I’m back

so only relevant on windows

08:03

have you verified that it unloads (with gdb), ChillerDragon

08:03

how do you even unload?

<ChillerDragon> sry im a bit busy sometimes but im still here thanks for helping

<ChillerDragon> @louis followed you i check my github feed 4 times a day i am waiting for your publish

08:13

<ChillerDragon> @heinrich5991 i unload with dlclose and that works wonderfully if i compile without -rdynamic and strip out the collision code

08:14

<ChillerDragon> but then the shared object cant read data from the map but hot reloading works

08:14

<ChillerDragon> ah __declspec is windows only

08:15

<ChillerDragon> i dont worry about windows for now

08:15

<ChillerDragon> but on stackoverflow i have seen a bunch of visbility annotations maybe i need some of those

08:17

<ChillerDragon> how can i verify it is unloaded? @heinrich5991

can you bring the thing into some state where we can start debugging?

08:17

info proc mappings

08:17

in gdb

<ChillerDragon> i mean i kinda know its not unloaded

08:18

<ChillerDragon> i called dlclose and deleted the shared object

08:18

<ChillerDragon> then dlopen and dlsym and it still runs the code from that shared object

08:18

<ChillerDragon> ok lemme se

if you know that it's not unloaded, it's probably best to figure out why it's not unloaded ^^

<ChillerDragon> idk what else to do than call dlclose

08:19

<ChillerDragon> info proc mappings is huge

08:21

<ChillerDragon> maybe its never loaded in the first place

08:21

<ChillerDragon> thats also a theory of mine

verify your theories!

08:21

that makes debugging a lot easier

08:21

because you then know which to pursue further and which not

<ChillerDragon> ye idk how

then figure that out. in this specific case, I gave you the answer

<ChillerDragon> i dont see my file name in info proc mappings

but you see it before the dlclose?

<ChillerDragon> i am one step after dlsym right now

08:26

<ChillerDragon> and there does not seem to be anything

08:26

<ChillerDragon> can i grep xd

08:27

<ChillerDragon> maybe i should test that with the working hot reload first

08:27

<ChillerDragon> idk how info proc mappings work and if dlopen shows up in there and if yes how

there's /proc/<pid>/maps which contains the same things, I think

08:27

you can grep that

<ChillerDragon> ah nice

08:29

<ChillerDragon> and it also freezes when i break with gdb i guess

yes

<ChillerDragon> ah i didnt step enough before

08:30

<ChillerDragon> i see it in gdb now

08:32

<ChillerDragon> ou this is smooth grep twbl /proc/$(pgrep DDNet-Server)/maps

08:33

<ChillerDragon> interestingly it shows my lib after dlclose when hot reloading is working

how do you know hot reloading is working? (edited)

08:33

how do you recompile the library?

<ChillerDragon> i have two ways of compiling it

08:34

<ChillerDragon> one where hot reloading works and one where my lib can access the collision pointer but cant hot reload

08:35

<ChillerDragon> i recompile both of them by running make in the servers build directory

08:35

<ChillerDragon> i know it is working because i can change the code and see the updates live in game without the server restarting

do you remove the library before recompiling it?

<ChillerDragon> sometimes

08:36

<ChillerDragon> i tried it all by now xd

can you check if hot reloading still works when you do that?

08:36

because maybe you're not hot-reloading, but simply overwriting stuff in-memory

08:36

^^

<ChillerDragon> works

08:37

<ChillerDragon> eee

08:37

<ChillerDragon> proc panic

08:37

<ChillerDragon> xd

proc panic?

<ChillerDragon> https://paste.zillyhuhn.com/s4

08:38

<ChillerDragon> it works and all xd

08:38

<ChillerDragon> i think you are right

08:38

<ChillerDragon> it hot reloads seemingly correct

08:38

<ChillerDragon> but now when i check the maps again it says deleted

08:38

<ChillerDragon> omg 0iq

08:39

<ChillerDragon> the grep matched the directory not the lib file xd

08:39

<ChillerDragon> but still something got deleted

08:39

<ChillerDragon> ah now i grepped the right thing

seems like your linker moved some stuff away before overwriting it? do you use mold?

<ChillerDragon> no that looks healthy

08:40

<ChillerDragon> ye i think i use mold

so maybe it's deleting the file before overwriting it?

<ChillerDragon> maybe the mold stuff is not really related my grep is flawed because of the folder name

but what is /home/chiller/Desktop/git/twbl-ddnet/build/.mold-hqv1pr?

08:40

I doubt you linked that

08:40

so it's probably mold moving away the old library to a temporary name before deleting it?

<ChillerDragon> https://paste.zillyhuhn.com/0c

08:41

<ChillerDragon> here is a better one where it actually shows the lib when loaded

08:41

<ChillerDragon> then i delete it and let it reload

08:41

<ChillerDragon> then its gone

08:41

<ChillerDragon> probably some weird artifcat im not sure if that mold thing is related to anything

check what it says about the address where the lib was previously loaded

08:41

yea, I agree, seems ot be an artifact

<ChillerDragon> you mean x/s in gdb or something?

08:42

<ChillerDragon> (gdb) x/s 0x7ffff7fb7000

08:42

<ChillerDragon> 0x7ffff7fb7000: <error: Cannot access memory at address 0x7ffff7fb7000>

08:43

<ChillerDragon> wow my web chat sanitized away the entire < error thing >

no, in the mappings

<ChillerDragon> a

to see if there's still a mapping there

<ChillerDragon> grep 7ffff7fb6000 /proc/$(pgrep DDNet-Server)/maps comes back empty

08:45

<ChillerDragon> grep 7ffff7fb /proc/$(pgrep DDNet-Server)/maps

08:45

<ChillerDragon>

7ffff7fbb000-7ffff7fc3000 rw-s 00000000 103:02 19352181                  /home/chiller/.teeworlds/ddnet-server.sqlite-shm

08:45

<ChillerDragon> so nothing right?

08:45

<ChillerDragon> idk if my grep made sense xd

so nothing, yes

<ChillerDragon> it seems to use the exact same mapping when i load it again

how do you determine that?

<ChillerDragon> https://paste.zillyhuhn.com/JA

08:47

<ChillerDragon> i did another lib change and recompile

08:47

<ChillerDragon> let it step after the load

08:47

<ChillerDragon> dumped the maps and compared it with the one from earlier (https://paste.zillyhuhn.com/0c)

but how do we figure out if it's the old or the new lib in the process?

<ChillerDragon> okay now that i feel comfortable with maps i guess its time to inspect how the non working one behaves

08:48

<ChillerDragon> @heinrich5991 it printed the new text

08:48

<ChillerDragon> so that works pretty sure

ah, so it works?

<ChillerDragon> yes its the working one

perfect

<ChillerDragon> lemme recompile to broken one

then let's understand the non-working one

<ChillerDragon> proc mappings is pretty hackerman

08:50

<ChillerDragon> okay before the first load no hits

08:50

<ChillerDragon> after the first load it finds it in the mapping

08:50

<ChillerDragon> and now i gotta afk xd

08:51

nice cliffhanger

<ChillerDragon> tell me about it xd

09:03

<ChillerDragon> ye and if i step to the dlclose its gone cleanly tee_thinking

tee_thinking

09:03

<ChillerDragon> lemme try to delete it and then step into the load xd

09:04

<ChillerDragon> okay it fails to load. thats weird i swear that didnt happen earlier

what's the error?

<ChillerDragon> that the file i try to dlopen is not on my harddrive

09:05

<ChillerDragon> which makes sense because i deleted it

09:06

<ChillerDragon> but i am so sure that didnt happen earlier it would just keep happily calling its code claming it loaded it succesfully

09:06

<ChillerDragon> anyways now i recompiled it so the file is back and then gdb hinted something interesting

09:06

<ChillerDragon> warning: Temporarily disabling breakpoints for unloaded shared library "./libtwbl_bottick.so"

09:07

<ChillerDragon> and now it happily called the old code somehow

09:07

<ChillerDragon> the one i deleted from my hard drive :D

09:07

<ChillerDragon> and the maps are empty

09:07

<ChillerDragon> so unloading works

09:07

<ChillerDragon> but loading doesnt?

09:08

<ChillerDragon> but dlopen() throws no error neither does dlsym() they just find old stuff even tho there is nothing in the memory maps

09:08

<ChillerDragon> or how is this /proc/maps thing called? memory maps? ram?

09:09

<ChillerDragon> gg i spent the entire day on it with 0 progress

[ddnet/ddnet] Pull request opened: #8388 Update Polish translations

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-ad...

when you load it again, you say it's not in /proc/<pid>/mappings?

<ChillerDragon> yep

but it's executing some code? so there must be something in /proc/<pid>/mappings

<ChillerDragon> but dlopen() an dlsym() pass without any error

what do they return? where do they point?

<ChillerDragon> and its executing code i deleted

into which mapping

<ChillerDragon> fp=0x7fffe456fc14

09:12

<ChillerDragon> thats my function pointer

so into which mapping does it point?

<ChillerDragon> does it work like that?

yes

<ChillerDragon> is %p in c++ directly mapped to the mappings?

09:13

<ChillerDragon> smooth

%p is mapped to the pointer value

09:13

and the mappings show you pointer ranges

09:13

so you won't find your pointer, but you'll find a range that includes it

<ChillerDragon> ah wait

09:14

<ChillerDragon> i understand now

09:14

<ChillerDragon> its the same pointer

09:14

<ChillerDragon> it points into the same region

09:14

<ChillerDragon> which is empty in maps

and what is mapped in that region?

<ChillerDragon> thus never overwritten by anyone

09:14

<ChillerDragon> nothing

but then calling that pointer would cause a segmentation fault

<ChillerDragon> so probably freed memory that still has the old data right?

no

<ChillerDragon> rly?

yes

<ChillerDragon> hm

09:15

<ChillerDragon> but i am happily calling that xd

09:15

<ChillerDragon> lemme gdb inspect that

can you print the memory around it using gdb?

<ChillerDragon> 0x7fffe456fc14: <error: Cannot access memory at address 0x7fffe456fc14>

09:15

<ChillerDragon> xd

09:15

<ChillerDragon> im telling you

09:15

<ChillerDragon> i am using non mapped memory

then you're maybe not calling it?

09:15

no, you're not, something else is off

<ChillerDragon> i see the printf result

go to the point where the library functon is called

09:16

and step into it, see where you end up

<ChillerDragon> nice

09:16

<ChillerDragon> ok

09:17

<ChillerDragon> ah my gdb checking was flawed that was before load

09:18

<ChillerDragon> hmm my gdb stepping skills are not the best

09:18

<ChillerDragon> i feel like i am single stepping the entire libc

n(ext) goes to the next line

09:18

s(tep) goes into functions as well

<ChillerDragon> ye

09:19

<ChillerDragon> next oversteps everything

09:19

<ChillerDragon> and step is going deep

09:20

<ChillerDragon> stepi worked

stepi is stepping one instruction

<ChillerDragon> hm memory is mapped again

It's easier if you use some editor integrations

<ChillerDragon> ah i guess i compiled xd

Gdb 14 supports dap, I use it with nvim-dap

whats the menu in spec called

<ChillerDragon> bruv ok lets go again

when you hold right shift (by default (i think))

<ChillerDragon> selecting tees to spectate?

yes

<ChillerDragon> Console()->Register("+spectate", "", CFGFLAG_CLIENT, ConKeySpectator, this, "Open spectator mode selector");

09:23

<ChillerDragon> spectator mode selector xd

weird name

09:24

suggestion: add specvoted to that menu

<ChillerDragon> omg i key slipped

09:25

<ChillerDragon> i wish prefix+space had a undo -.- https://zillyhuhn.com/cs/.1716283493.png

Embedded image

09:28

<ChillerDragon> window moving in tmux is honestly fucked

09:30

<ChillerDragon> okay windows back

09:30

<ChillerDragon> uhm so what am i looking for now heinrich i forgot xd

09:31

<ChillerDragon> so it does map something now

09:31

<ChillerDragon> but the old stuff

you claimed that it's calling a function that is not mapped

<ChillerDragon> then it un maps it again

09:31

<ChillerDragon> yes i did claim that

but that's not the case?

09:31

now

<ChillerDragon> when i checked that it was before the call to dlopen

ok. so what's the current problem?

09:32

with the second call to dlopen, it somehow opens the old file again?

<ChillerDragon> it now seems kind of sane

09:32

<ChillerDragon> it fails nicely when i delete the file

yea, sounds good

<ChillerDragon> but when i compile it again it still prints the old message

so let's find out why it does

09:33

check what it maps when you dlopen again

<ChillerDragon> you mean the function pointer?

no, the /proc/<pid>/maps

<ChillerDragon> wdym check?

09:34

<ChillerDragon> how do i examine it?

check what filename it tells you after mapping it again

<ChillerDragon> /home/chiller/Desktop/git/twbl-ddnet/build/libtwbl_bottick.so

09:35

<ChillerDragon> thats from the maps

09:35

<ChillerDragon> and that file exists

can ls -l the relevant symlink in /proc/<pid>/map_files/<addr_range>?

<JStar> chiller u still have the problem?

can you somehow check in gdb which version of the library you have mapped in the program?

<ChillerDragon>

lr-------- 1 chiller chiller 64 May 21 17:36 7fffe4400000-7fffe44af000 -> /home/chiller/Desktop/git/twbl-ddnet/build/libtwbl_bottick.so

09:37

<ChillerDragon> @JStar ? xd

09:37

<JStar> @ChillerDragon u get a crash when hot reloading?

09:37

<ChillerDragon> oh you mean from a few weeks/months back?

09:37

<JStar> i dunno, i read this chat

09:37

<ChillerDragon> no that was ddnet_hotui thats working smoothly now i use it all the time

09:38

<JStar> and u said u have trouble

09:38

<ChillerDragon> i am now since yesterdayish in deep trouble while developing hot reloading for bots

09:38

<ChillerDragon> luckily heinrich is being very helpful now

09:38

<JStar> do you automatically hot reload, or manually press a button?

09:38

<ChillerDragon> @heinrich5991 i can let it tick and call the callback and then print not sure how reliable that is but then its printing the old message

09:39

<JStar> anyway, linux touches the file (creates an empty file) before writing the content. so better be careful with automatic hot reload

09:39

<ChillerDragon> @JStar i am not that far yet for now im trying to get it to work. When comes later. As of right now i reload every tick. (its blazingly fast suprinsingly) later i might add a command or look at the timestamp of the file like i do in ddnet_hotui

09:39

<JStar> u might end up loading a partial file

09:39

<ChillerDragon> ye ik

(that's why you should mv the file over)

09:40

(instead of incrementally writing it)

<ChillerDragon> but i compile the file once manually then i reload every tick @JStar :D

09:40

<ChillerDragon> those are all details for later

09:40

<ChillerDragon> first it has to work

09:41

<JStar> i dunno heinrich, i think i even had trouble with mv

09:41

<JStar> ok maybe only with cp

09:41

<JStar> then nvm

mv is atomic

<ChillerDragon> @heinrich5991 so what do you think? is it printing the old message a reliable check of what is mapped?

cp isn't atomic by default

ws-client

<ChillerDragon> @heinrich5991 so what do you think? is it printing the old message a reliable check of what is mapped?

ChillerDragon: can I see the source code somewhere?

<JStar> how do u even deal with global variables @chiller

not sure. depends on how you print the message. where does the message come from? static string? should be safe™. but something is off, so I'm not sure if I'd treat it as a safe signal if it says it's still the old library

09:44

what does the message say, ChillerDragon?

<ChillerDragon> my git work tree is a bit dirty i can tmp push that

I'll also have to leave for 2 hours or so, soon

09:44

but I'm interested in continuing

<ChillerDragon> https://zillyhuhn.com/cs/.1716284626.png

Embedded image

09:45

<ChillerDragon> ye i wanted to be on the beach since 40 mins

09:45

<ChillerDragon> lemme push to tmp branch

09:47

<ChillerDragon> https://github.com/TeeworldsBotLib/ddnet/tree/tmp_heino

GitHub - TeeworldsBotLib/ddnet at tmp_heino

DDraceNetwork, a free cooperative platformer game. Contribute to TeeworldsBotLib/ddnet development by creating an account on GitHub.

09:47

<ChillerDragon> here is all the code

09:48

<JStar> chiller first finish work

09:48

<ChillerDragon> @JStar for now i only wanna see the map so i pass a pointer to CCollision into my callback

09:49

<ChillerDragon> wot work jstar

09:49

<JStar> finish ur hot reload lib

09:49

<JStar> then beach

09:49

<ChillerDragon> dude im 9 to 5 tw dev

09:49

<ChillerDragon> its beach time since 50 mins

09:49

<ChillerDragon> tw life balance

09:50

<JStar> nice but it's not about time

09:50

<JStar> u finish the work then beach

09:50

<JStar> hot reloading is epyc, u need ur reward

09:50

<ChillerDragon> @heinrich5991 yes the string is static and it says UwU as you can see in the screenshot

09:51

<ChillerDragon> @JStar i wont finish this today :(

09:51

<JStar> wtf chiller only skimmed the code, but how does it even work at all xD

09:52

<ChillerDragon> same as the old ddnet hotui

09:52

<ChillerDragon> veri simple

09:52

<ChillerDragon> just dlopen then dlsym

just to make sure: ChillerDragon: when you start the server with the new version, does it actually print the new stuff?

09:52

in the first dlopen

<ChillerDragon> and if smart also dlclose

09:52

<ChillerDragon> thats all

just to make sure that we're actually modifying the correct library

<ChillerDragon> @heinrich5991 ye server restart uopdates

09:53

<ChillerDragon> but i compile and link it into like 3 places

09:53

<ChillerDragon> its all a bit complicated in my grug brain

09:53

<ChillerDragon> i also somehow smashed the entire ddnet code base into the shared object

09:53

<ChillerDragon> and have two cmake projects

09:54

<JStar> :lol

that sounds very complicated

<JStar>

lol

no way your changes work

09:54

<ChillerDragon> ikr

maybe start with a single C file that just prints something ^^

09:54

when linked as a shared library

09:55

oh, that worked, right?

<ChillerDragon> yea kinda

"kinda"?

<ChillerDragon> but then it does not have access to ccollision

09:55

<ChillerDragon> hot reloading works

09:56

<ChillerDragon> thats why i have two libs

09:56

<ChillerDragon> one that falls back to fake collision

09:56

<ChillerDragon> and that one can hot reload

09:57

<ChillerDragon> but i think even that "lite" version breaks if the server is compiled with -rdynamic

09:57

<ChillerDragon> yep seems so

09:57

<ChillerDragon> adding -rdynamic "caches" hot reloading and makes it very non hot

09:58

<JStar> bro how does the code u posted even work at all

09:58

<JStar> i no understand

09:59

<JStar> im no understand

09:59

<ChillerDragon> whats unclear?

09:59

<ChillerDragon> bra im lagging af

10:00

<ChillerDragon> thank you so much for your time @heinrich5991 i rq for now

I'll also be gone for now

<JStar> yeah sry chiller

10:02

<JStar> i guess i am simply no genius like u

10:03

<JStar> ok i understand it now lmao

Pro

<JStar> chiller: how do u pass collision

If you understand send fix hehe

<JStar> i'm blind

Types.h

10:05

And then in character.cpp

<JStar> wtf bro

Ikr

<JStar> u create a new ccollision struct?

Just for mocking

10:06

When used outside of the server

10:06

In unit test mode

<JStar> ah ok

Oh and right now it’s also nice because that mock collision works with hot reload

10:07

But ideally real collision would hit reload too

<JStar> so your problem is only that it crashes if u pass ccollision?

Never crashes

<JStar> but ? i didnt read full convo sry

10:07

<JStar> just read hotreload and trouble and i was hooked

It just does not load new stuff with dlopen when rdynamic is set

10:08

It only loads once

<JStar> i see

10:08

<JStar> i wish ur epyc web client would have search function

And I need rdynamic on the server side otherwise it can’t find Ccollison

10:08

It has

10:08

Activate plugin

10:09

In settings and then reload page

<JStar> lmao

10:09

<JStar> so epyc

Xd

<JStar> u can also force visibility of symbols differently, cant ya?

That’s the mysterious visibility keyword potentially

10:10

I have seen it around

10:10

Idk what it does

10:10

It can be used in c++ code but idk if it’s related

10:10

Also everyone uses rdynamic with dlopen

10:10

That shouldn’t cause issues

10:11

I mean i use it client side too

<JStar> i enabled ur epyc plugin. now i have two smileys to open the smiley popup

10:11

<JStar> nice

Xd

<JStar> cant find search

Use plugins at own risk

10:12

One is search

10:12

I didn’t bother to add texture

<JStar> ah lol xDDD

ux

<JStar> it cant find it, does it have an age limit?

I may have reused parts of the emoji code

10:13

It only searches in the history known to the backend

10:13

Which is capped at like idk 800 msgs

10:13

Because bloat

10:14

Old stuff I grep in irc logs xd

10:14

It’s a works on my machine kind of chat app

<JStar> mh ok, i see, yeah. makes no sense, u are sure it never reloaded?

10:20

<JStar> anyway dragon, don't give up, just rewrite ddnet to be modular and stuff

:w

huh I was not in vim there hey

hey

justatest

3

ChillerDragon: could you give reproducing instructions so I can try locally?

how to quickly calculate the number of hours for ALL TIME? and not in a year?

Image attachment

12345

how to quickly calculate the number of hours for ALL TIME? and not in a year?

Loop through every object & total_hours += hours_player

Avolicious

Loop through every object & total_hours += hours_player

through every object = quickly?

12345

through every object = quickly?

Its json, what do you expect? Speed out of a sudden?

kek

1

Avolicious

Its json, what do you expect? Speed out of a sudden?

Maybe there is some special code that will calculate?

cache every time you calculate it for ceratain player, you can't speed it up any other way

JSON is slow per design, its human readable. Either ask if deen can add a "total_time" to the root json or you have to parse through every object

imagine protobuf ddnet api

12:53

justatest

Avolicious

JSON is slow per design, its human readable. Either ask if deen can add a "total_time" to the root json or you have to parse through every object

it's very stupid that no one thought of this, a pretty obvious thing

12345

it's very stupid that no one thought of this, a pretty obvious thing

Because nobody asked for it I guess

there's no need in total hours as if it's still not here

Why bloat JSON if there is no use-case

hey avo btw

zhn

hey avo btw

Hey

1

zhn

there's no need in total hours as if it's still not here

That’s it, keep quiet, otherwise it will start again “this person is clearly stupid”

12345

That’s it, keep quiet, otherwise it will start again “this person is clearly stupid”

and still you did ask about how to parse it faster

But now for real, how slow is you CPU?

12:56

I guess looping through each item isnt that time consuming. Time complexity seems reasonable

i bet this guy uses standard python's json module and its kinda slow

python orjson is a good replacement tho

12:59

but even the normal python standard json module is fast

12345

it's very stupid that no one thought of this, a pretty obvious thing

its not like you want to fetch that every day do you? :D - i just tried it, took less than 3 seconds

12345

Maybe there is some special code that will calculate?

yes i suggest using a scripting language such as python or javascript instead of doing it manually with a scientific calculator

troll

1

12345

how to quickly calculate the number of hours for ALL TIME? and not in a year?

@12345 An alternative would be to mirror our race database and then use SELECT SUM(time) AS Time FROM record_race WHERE Name = %s; to get your desired result

louis

yes i suggest using a scripting language such as python or javascript instead of doing it manually with a scientific calculator

Even better, do it on paper

louis

ideally i'd have some sort of window app that automatically updates the map whenever the file is updated

twgpu-map-inspect now does exactly this, in case you still need it :)

1

15:01

you can use --vsync to limit the framerate. without that, it currently renders at max fps (edited)

[ddnet/ddnet] Issue opened: #8389 Black screen on Steam Deck

Reported by cheese0613, using SteamOS (works with Arch), not sure if the logs are related: !image

I think Steam Deck used to work, anyone around who had it working before? https://github.com/ddnet/ddnet/issues/8389

Black screen on Steam Deck · Issue #8389 · ddnet/ddnet

Reported by cheese0613, using SteamOS (works with Arch), not sure if the logs are related:

@cheeser0613 Can you try launching with dbg_gfx 4 and share the console output?

Robyt3

@cheeser0613 Can you try launching with dbg_gfx 4 and share the console output?

in launch options?

yeah

wait for few min i want to double check that all the log that i snap is related to ddnet

yeah i think the only error that shows in the screenshot means nothing because it does show before i launch ddnet

[ddnet/ddnet] Pull request opened: #8390 Minor refactoring of fs_chdir and fs_getcwd functions

Remove unnecessary fs_is_dir check in the fs_chdir function. The SetCurrentDirectoryW and chdir functions would already include such necessary checks anyway. Add remark about the fs_chdir and fs_getcwd functions not being thread-safe on Windows to documentation. Assert that GetCurrentDirectoryW was successful instead of logging an error message. This function should only potentially fail if the current working directory was changed by another thread between the two calls, ...

16:18

[ddnet/ddnet] Pull request closed: #8388 Update Polish translations

16:18

github-merge-queue[bot]

[ddnet:master] 2 new commits

b63380f Update Polish translations - def- 4473217 Merge pull request #8388 from def-/pr-polish2 - Robyt3

hi dev, can u help me, how to change size of this white lines?

Image attachment

12345

how to quickly calculate the number of hours for ALL TIME? and not in a year?

use rust simd json crate and have a cpu with avx512

16:50

https://edgarluque.com/blog/zstd-streaming-in-rust/

Parsing compressed files efficiently with Rust

Sometimes you need a bit of a stream.

16:50

read this its ddnet related

16:50

gigachad

gigachad

1

troll

1

бирафа

hi dev, can u help me, how to change size of this white lines?

why do you need to change borderline width (edited)

Robyt3

@cheeser0613 Can you try launching with dbg_gfx 4 and share the console output?

I find out that getting the full log from console is harder than I expected, steam console inside the steamdeck will replace everything with "..." if the text is over display and the way to break this situation is to connect the developer kit on pc. To connect it, steam deck itself must run a dev software which call "SteamOS Devkit Service". This software itself is pre-installed for the steam deck developer kit (basically you just ask valve to ship you a free steamdeck for dev purpose) and auto run everytime when the device launch. If the developer didnt receive the kit, valve also provide a solution to build a low budget device and provide a installation for the same steamOS. Heres the hard part, "SteamOS Devkit Service" do run on the dev set steamdeck and selfbuild steamOS that valve provide, but if developer decide to buy himself a retail set steamdeck, it seems like it just not going to run and the CEF remote function will not going to work in order to access the steam console from pc even if valve clarify that the dev set steamdeck is same as what they deliver to consumer (edited)

117.68 KB

that one's new for me:

Image attachment

18:38

it kept on downloading and eventually connecting tho

it fell back to game server download

бирафа

hi dev, can u help me, how to change size of this white lines?

right click on the layer then change the width and height, but it will extend to right and bottom, next time consider ask the mapping related question on #mapping channel

Image attachment

meloƞ

that one's new for me:

chicken pie 2 justatest

justatest

justatest

justatest

kek

1

good map!

[ddnet/ddnet] Pull request opened: #8391 Allow econ to succeed, even if ipv6 econ fails to bind to socket

It seems there is a problem with ECon, where the server somehow tries to bind to wrong IPv6 address, making the bind fail. This PR allows ECon to succeed if at least one of the two sockets (IPv4 or IPv6) could be opened. I did some preliminary debugging and found that default bind address (localhost) at some point got properly resolved to ::1, but when trying to bind to it, the address seems wrong. I guess it somehow got mangled up before the actual bind call. This needs some further deb...

Does anyone know how to convert 0.6 skins to 0.7 properly? I tried out this https://www.teeworlds.com/forum/viewtopic.php?id=11527 It kinda worked but the proportions are not matching and the skins look then awful Ingame. Also colors are way darker when you join red/blue team

SkinMe script: convert old skins to the new format (Page 1) — Graph...

SkinMe script: convert old skins to the new format (Page 1) — Graphics — Teeworlds Forum — Everything Teeworlds!

that is simply bcs 0.7 skin system works so different

19:12

not bcs of the script

Ok so converting won‘t work, I‘ll have to recreate the graphics ?

19:15

I mean how they‘ve done it for default skins

default also looks different

19:16

is your original skin in a high resolution? (edited)

Jupstar ✪

default also looks different

Yes but did they make whole new graphics for cammostripes, etc. or converted them

maybe the script only works with 256x128 skins

David

Yes but did they make whole new graphics for cammostripes, etc. or converted them

ah i guess they used the original inkscape files and modified them a bit

19:17

but yeah i assume they are not converted

19:17

but actually created

Not sure about the skins systems in general but I only have a png version of mine

yep the original inkscape files are private sadly

19:19

nobody has them

19:19

we recreated some skins in inkscape: https://github.com/ddnet/ddnet-data-svg/tree/main/skins

ddnet-data-svg/skins at main · ddnet/ddnet-data-svg

Contribute to ddnet/ddnet-data-svg development by creating an account on GitHub.

maybe I'll have to trace it with vector program

Image attachment

19:19

seems to be low resolution

seems like i also once wrote a converter: https://github.com/Jupeyy/teeworlds_skin_converter_06_to_07/tree/ff0b63f8a33e18723e5ab5c46fcaa2d5c89465a4/bin

teeworlds_skin_converter_06_to_07/bin at ff0b63f8a33e18723e5ab5c46f...

Convert skins from 0.6 to 0.7. Contribute to Jupeyy/teeworlds_skin_converter_06_to_07 development by creating an account on GitHub.

is there any docs on the 0.7 skins like what size the hands have to be and how to place them?

i think i mostly fixed eyes, bcs they looked so weird

19:21

havent used that in 4 years tho

yes with the script i used the eyes where to much placed on the top they had to be lowered

19:21

and the body was to low so it was going "under" the feets

u can try mine, no idea if it still works

I will try tomorrow

i guess it's drag & drop or smth like that

Can it run on mac?

uh, i dunno, if u can compile on your mac

19:22

why not

19:22

else probably no

David

maybe I'll have to trace it with vector program

looks easy to redraw though

19:22

justatest

zhn

looks easy to redraw though

yep thats what I wanted to do maybe as well, but still I need like a template of the 0.7 "original" skins or whatever that I know where to place the parts of the skin and what sizes the images needs to be

David

yep thats what I wanted to do maybe as well, but still I need like a template of the 0.7 "original" skins or whatever that I know where to place the parts of the skin and what sizes the images needs to be

https://github.com/teeworlds/teeworlds/issues/2271

Secondary Skin file format · Issue #2271 · teeworlds/teeworlds

0.7 introduced a new skin system, separating each body part to allow deeper customisation by combining them. This was a good decision I think, and most players love it. The other day, @iTzMeRafa pu...

libpng error: [00][00][00][00]: invalid chunk type btw u skin also seems buggy

19:26

this is what my converter created

21.71 KB

Jupstar ✪

this is what my converter created

looking better than that what the other converter did. I'll maybe just redraw it couldn't be that hard doing it in a proper vector program

thought there is an quick fix. Got that skin back then from a colleague

github-merge-queue[bot]

[ddnet:master] 2 new commits

3af8c00 allow ipv4 econ to succeed, even if ipv6 econ fails to open port - hardliner66 7c4d3a5 Merge pull request #8391 from hardliner66/fix_econ - heinrich5991

19:37

[ddnet/ddnet] Pull request closed: #8391 Allow econ to succeed, even if ipv6 econ fails to bind to socket

murpi

@12345 An alternative would be to mirror our race database and then use SELECT SUM(time) AS Time FROM record_race WHERE Name = %s; to get your desired result

that's different, I think. yours is the time of finihsed races, the other is the online time as determined by the serverbrowser

Did you know US customs can manually search any electronic device at a border crossing? No warrant, no probable cause, no suspicion. If after the manual search they suspect anything they can copy your entire device and retain it for an undisclosed amount of time with no duty to delete it.

22:27

If you refuse to give them your password and that makes them suspect something they can seize the device and hack into it or copy it encrypted. If you are foreign national they can just flat out deny you entry just for that

22:28

They can also store and utilize any keys you give them, wild stuff. A truly tyrannical level of violation of privacy that I haven’t heard of before

meloƞ

good map!

no

Learath2

They can also store and utilize any keys you give them, wild stuff. A truly tyrannical level of violation of privacy that I haven’t heard of before

welcome to united states

Learath2

They can also store and utilize any keys you give them, wild stuff. A truly tyrannical level of violation of privacy that I haven’t heard of before

Surely this doesn't apply if you're a citizen, with the fourth amendment (edited)

23:45

Not that it makes it acceptable, it's still bad to be doing that to foreigners (edited)

Peter0x44

Surely this doesn't apply if you're a citizen, with the fourth amendment (edited)

You are in for a surprise. It does

23:47

First Amendment arguments against this insanity were outright rejected by all courts. Fourth Amendment does indeed help a bit, some courts have ruled that while they can seize and clone your devices (by exception granted to borders and upheld by the supreme court iirc), they can not compel you to reveal the password without warrant

I meant to say fourth amendment Edited

Peter0x44

I meant to say fourth amendment Edited

I elaborated on both as far as I understand the relevant law/rulings

those things often also apply 100 miles from the border, i.e. to 67% of the US population

I lived under 30km from the Mexican border in Texas, there wasn't much checking from what I remember

23:50

But I left when I was 9 Anyway, yes, the USA is really not far from fascism Plenty would argue it is

Peter0x44

I lived under 30km from the Mexican border in Texas, there wasn't much checking from what I remember

Remember during the BLM protests there were talks of federal agents straight up abducting people in vans?

23:51

That’s the rule they were invoking when sending in CBP agents, 100 miles from any border including the ocean is very “technically” fair game

I guess if you want more horrible US actions: https://en.wikipedia.org/wiki/Murat_Kurnaz This guy got tortured for 5 years in Guantanamo

Murat Kurnaz (born 19 March 1982) is a Turkish citizen and legal resident of Germany who was held in extrajudicial detention by the United States at its military base in Kandahar, Afghanistan and in the Guantanamo Bay detention camp at Guantanamo Bay Naval Base, Cuba beginning in December 2001. He was tortured in both places. By early 2002, inte...

23:52

They will treat you even worse if you aren't a citizen

Peter0x44

I guess if you want more horrible US actions: https://en.wikipedia.org/wiki/Murat_Kurnaz This guy got tortured for 5 years in Guantanamo

Obama said he'd shut down the "facilities" there but never did

There are many horror stories from how the CIA operated in Afghanistan, but this search and seizure thing could be such a normal part of your normal travel

23:55

E.g. all my whatsapp messages and discord messages can become a part of the US intelligence db next time I come along

23:56

It just takes me looking a little too disheveled or anxious, bam manual search, then the dude sees some sus anime pic that one weirdo keeps posting in #off-topic, bam all my phone dumped into the system, forever

~~they probably already are~~

23:57

Discord at least definitely

23:57

There is of course the CIA and the NSA doing their "nefarious" things

Peter0x44

I guess if you want more horrible US actions: https://en.wikipedia.org/wiki/Murat_Kurnaz This guy got tortured for 5 years in Guantanamo

I can’t believe I did not know of this one, I’m a bit of an enthusiast of US crimes against humanity

23:58

I bet I’ll get flagged next time because they monitor discord and think that me talking about this is sus

Exported 643 message(s)