DDraceNetwork - developer

ChillerDragon BOT 2023-02-18 07:13:13Z

freenodes?

I keep getting crashes in my ddnet server fork no idea how to debug it. It says "Broken pipe" full crash here https://paste.zillyhuhn.com/oD

07:25

The way i understand it is that it timeouts while doing masterserver stuff for whatever reason and then the socket dies and it somehow keeps trying to send there. Does that make sense? But then this is not a official ddnet bug is it? So it means it somewhere got introduced during merges hmm

ChillerDragon

I keep getting crashes in my ddnet server fork no idea how to debug it. It says "Broken pipe" full crash here https://paste.zillyhuhn.com/oD

Do you run your server with gdb or in debug mode? What version of curl is it used?

i build in debug mode and then run the server with gdb does it print the curl version? Anyways when running curl --version it says libcurl/7.74.0 so i assume thats what the server uses too

07:49

whatever my debian 11 repos currently offer

Is the stderr missing? Maybe look in a debugger where it wants to write to

08:02

It's either file io or network or logging i guess

Idk how to look that up only thing i see is fd=19 but that file descriptor doesnt help me much does it?

Guess no

08:08

Stack trace

ah i see when i get the backtrace i should see that yea missed the chance ill check again in

08:08

idk a few weeks when it happens again :D

08:09

from a old backtrace i know its here but not sure which line because its outdated trace https://github.com/DDNetPP/DDNetPP/blob/cd8991efa291dd8e74354ce3635426403721d77b/src/engine/shared/http.cpp#L188-L308

DDNetPP/http.cpp at cd8991efa291dd8e74354ce3635426403721d77b · DDNe...

DDraceNetwork + city + block! Based on ddnet (www.ddnet.tw) which is based on teeworlds (www.teeworlds.com). - DDNetPP/http.cpp at cd8991efa291dd8e74354ce3635426403721d77b · DDNetPP/DDNetPP

08:09

wait pretty sure its here https://github.com/DDNetPP/DDNetPP/blob/cd8991efa291dd8e74354ce3635426403721d77b/src/engine/shared/http.cpp#L295

DDNetPP/http.cpp at cd8991efa291dd8e74354ce3635426403721d77b · DDNe...

DDraceNetwork + city + block! Based on ddnet (www.ddnet.tw) which is based on teeworlds (www.teeworlds.com). - DDNetPP/http.cpp at cd8991efa291dd8e74354ce3635426403721d77b · DDNetPP/DDNetPP

08:11

but no idea whats the problem w curl i mean thats basically the same code ddnet uses so the issue has to be somewhere else

is there a way to trigger a whole graphical refresh of the game without reinitializing the graphics backend (edited)

09:44

09:44

im doing this cuz i don't like the way ddnet does the rounded corners & it works for the GUI but persistent things like some components of the hud are rendered responsibly & don't update without being notified of a change

You are not using ddnet. ui_rounded_corners has never existed for ddnet. And if it did, I would advocate for its immediate removal.

i made it

09:53

i'm not trying to pull request or anything

Ewan

i made it

i don't know how that was not obvious by my statement

edgy

Then you basically answered your own questions: check if the variable was changed and then update the rects

get it? cuz edges

09:55

rect

I can't think of an easy way, you'll just have to pepper some if statements everywhere

bummer

lerato sprinkling italian pepper

is there an existing mechanism to fire a callback on a settings change

yes, but for this you probably need to remember the old value wherever you render a rect and then check if the old value != the new value

I think you can add a conchain. We do it for sv_map e.g.

Robyt3

yes, but for this you probably need to remember the old value wherever you render a rect and then check if the old value != the new value

why would i need to do that?

Robyt3

yes, but for this you probably need to remember the old value wherever you render a rect and then check if the old value != the new value

Doesn't the conchain happen before the actual callback?

otherwise how to you "push" the updated value to all components that draw a rect? you'd first need a list of all components drawing rects (edited)

i wanted to do a graphical refresh for this reason

09:59

because it shouldn't really require a restart. it would just be that

It shouldn't but it does :/

maybe you could hack it together by calling OnWindowResize, which should invalidate all cached quad and text containers

Robyt3

maybe you could hack it together by calling OnWindowResize, which should invalidate all cached quad and text containers

I was just about to say that

10:00

Maybe that would work

not worth it. i'll just store its first value forever & require a restart (so it's at least consistent)

10:01

im doing this because the way ddnet does rounded rectangles is weird. they're absurdly large and they don't pad the contents to accommodate. you just have to kinda hope it doesn't clip lol

10:01

and they're not antialiased for whatever reason. could be a number of factors but any way i don't like it

nothing is antialiased except round corners AFAIK, but you need to enable MSAA

yeah this is different from in-game AA & should be handled as such

Absurdly large? Hm iirc they fit exactly inside the rect you pass to the function

i just mean the radii are too large for my taste

10:03

nothing broken about the size

just multi each corner radius by 0.5f before drawing nouis

Ah, yep it's an old teeworlds thing, back then it was the cool trend

10:04

I wanted a sharper ui but iirc people said it lost character so we kept it as is

sad lol

Learath2

Doesn't the conchain happen before the actual callback?

are there docs for conchain?

Nope, but it's not very hard to use, just grep for Chain or ConChain or maybe sv_map

https://www.youtube.com/watch?v=2rk5bXUPhlE

Prudrugtiq

Teeworlds programming #006 - chaining config variables

10:09

video docs poggers

i agree but it's not for me

10:10

i don't have that kind of attention span lol

10:10

ill grep

Check how it's done in the code, e.g. search for ConchainFriendlistUpdate

what's the highest level CComponent?

10:16

that i could call OnWindowResize() on & have it trickle down

Call it on CGameClient, which should propagate it to all components

awesome

10:17

thanks

10:17

is there a global instance or is it static

All components have a pointer to it with GameClient()

nice

10:20

void CMenus::ConchainRoundedCorners(IConsole::IResult *, void *,
                                    IConsole::FCommandCallback, void *) {
  GameClient()->OnWindowResize();
}

this is weird, i'm getting a can't call member w/o instance error, as if CMenus didn't inherit from CComponent? (edited)

10:21

and sorry if this is bad form & i'm not populating something i need to or whatever. i see some pfnCallback action above but im trying to get it laid out first

Did you also add it in menus.h?

yeah

You need to call pfnCallback so the config value is changed (the original command handler)

oh this is my bad

10:23

i copy pasted & the def in the header is static

10:29

got it working. thanks

Robyt3

[ddnet/ddnet] Pull request opened: #6343 Simplify `io_write_newline` error handling

Instead of returning the number of bytes written, which are platform specific, return true on success and false on failure, so no platform specific code is required when checking the result.

Checklist

[X] 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 a...

bors[bot]

[ddnet:master] 2 new commits

3173aee Simplify io_write_newline error handling - Robyt3 686b265 Merge #6343 - bors[bot]

14:53

bors[bot]

[ddnet/ddnet] Pull request closed: #6343 Simplify `io_write_newline` error handling

@ChillerDragon I didn't get if you found the issue. The thing is that I had the same ('crash' with exit code 141 after SIGPIPE received). It turned out that DDNet does not handle SIGPIPEs. Maybe what you need is this commit: https://github.com/ddnet/ddnet/commit/199da2410df05ccdce4f98d41628725699c41080#diff-83a6b32fd0c42ef297d20d7ffbca30bd6945a6a78dc48a8df0d1130375477011R76 Also note that some tools catch the signals before the application, thus e.g. if you run gdb ./DDNet-Server you can get a stop on that (or another otherwise suppressed) signal.

Prevent OpenSSL SIGPIPE (fixes #3513) · ddnet/ddnet@199da24

by ignoring the signal ourselves See https://curl.se/libcurl/c/CURLOPT_NOSIGNAL.html

15:16

(Thanks to @heinrich5991#9827 for giving me the clue about the signal ignoring, I could spend a lot of time trying to find the "fix" (in quotes because IMO it is not a proper fix). I was running the server via catchsegv and it seems (but I'm not sure) that the tool catches SIGPIPE before the app could ignore it.)

15:40

31 hours of compiling with MSVC

sup, how do i convert this into a std::any_of? its a bool array All examples online are vectors and i couldnt get it work here (edited)

@deen could you update the twmap-tools for the ddnet discordbot?

Vy0x2

sup, how do i convert this into a std::any_of? its a bool array All examples online are vectors and i couldnt get it work here (edited)

True could have asked this guy thx

Robyt3

[ddnet/ddnet] Pull request opened: #6344 Add `reset` command to server

Move reset command from client to console, so it's also available on the server.

Checklist

[X] 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 wit...

@Kaffeine interesting thanks. But the fix you sent if from jan 2021 pretty sure my servers were running with that patch already since i merge with upstream more than once a year

16:39

is from*

16:41

@fokkonaut pretty overkill readme for a project where like what 3 people have access and one of them is you :D sorry idk where to dm you these days

Maybe it becomes opem source in the future (?)

:o

a readme is supposed to help a new person interested in the project understand what the project is about (edited)

16:43

you could add one more to the project anytime in the future

wowo leak

or as fokkonaut said, open sourcing it

thats basically what i wanted to hear hehe :)

16:44

poggies fddr open src

16:44

confirmed

A small readme/documentation has been planned for years already

ikr

idk never was motivated but started today

im surprised u actually did it hehe

A little

16:44

xd

16:45

Usually I am good at writing, Im just lazy

okok mr nobel winning author

closed source bad

btw i saw you started typing out cmd docs i highly recommend automating that

16:46

i think ddnet or vanilla even have tools for that

ChillerDragon

btw i saw you started typing out cmd docs i highly recommend automating that

Yea I know, thats why I didnt start it yet

just for the look of the table

16:47

But Im not sure whether I want to document it in detail or simply the desc from ingame

https://github.com/ddnet/ddnet-scripts/blob/master/client_wiki/commands.py

ddnet-scripts/commands.py at master · ddnet/ddnet-scripts

The magic scripts that make DDNet run. Contribute to ddnet/ddnet-scripts development by creating an account on GitHub.

16:47

i think dis

16:48

and dis https://github.com/ddnet/ddnet-scripts/blob/master/client_wiki/settings.py

ddnet-scripts/settings.py at master · ddnet/ddnet-scripts

The magic scripts that make DDNet run. Contribute to ddnet/ddnet-scripts development by creating an account on GitHub.

lol cool

but what do i know i typed out 10k lines html for network docs eventho network code is generated with py scripts that could also spit out html if changed xd

16:50

big pro

Ryozuki

closed source bad

Anti-bot? Sometimes there are reasons for "closed source".

chiller would still argue

16:52

Kaffeine

Anti-bot? Sometimes there are reasons for "closed source".

nein

16:53

thats just security via obfuscation

not even obfuscation

16:54

obfuscation is hiding the information behind cryptography

16:55

antibot can't be deobfuscated (because it isn't obfuscated) (edited)

insert villager sound

is there a difference between open source and closed source onbgy anti bot?

16:55

axaxaxax

chillerdragon: cant imagine its been 5 years coding now

half decade pro haxor

ChillerDragon

is there a difference between open source and closed source onbgy anti bot?

lul, open sourcing it would lead to obfuscation

16:56

still secure

no I think what he meant was

16:56

that if it was good it wouldn't need to be closed source

Nono xf

xd

16:57

ye fddr bad so it closed src

16:57

axaxax

ongby src code = hf reading

ChillerDragon

ye fddr bad so it closed src

hardcoding detection vectors of public clients is bad

He is trolling

nobys c looks like assembly

xD

16:57

f3

optimization

its new language

16:57

its NobyC

don't steal my memes

default

hardcoding detection vectors of public clients is bad

but I understand why you do it

16:59

it's hard to come up with a 100% detection method that doesn't rely on that information

@ReiTW ok lemme think of susmin new: "obi wan c noby" meh only works in german :D

17:00

well doesnt rly work in eng or ger xd

lul

kenoby works in german (edited)

17:00

or broken english

A B KE

He aint using the ABC, he has his own NoByC

watrfak

Thats why no one can read it

17:01

ax

obfuscation doesn't imply cryptography

Ob-Ewan C noby

Ewan

obfuscation doesn't imply cryptography

you know what I mean though

17:02

you can't deobfuscate something that isn't obfuscated

Chuck Norris can

damn it

17:02

LOL

17:03

fokko are you ever going to have a server for the US?

donate for one

Hmm, probably not in near future

17:03

I would need to re do some things on the serverside to work with MySQL

17:03

otherwise no db con

i see

I wished I could just set one up rn

netfs + txt db > mysql

17:04

change my mind

Lol

17:05

Just use mails to send the data and later scan the data and paste it into txt documents and use fifo to tell the server new data arrived

17:05

scan the letter

lol (edited)

stonks

17:05

80ct porto

xd

computers were a mistake my db is a unpaid intern that has to remember all data

in what units is the m_Vel?

17:16

is it in pixels?

in what units is pixels?

int

xd no i mean what are pixels how long are they?

17:19

i assume m_vel is in the smallest unit of m_pos

17:19

which was like what 32 a tile or something?

17:19

64?

32 yeah

pixel seems screen resolution dependend idk how that would work

Not necessary screen pixel

i do not for a fact but i highly assume vel is measured in the smallest unit a tile has

ChillerDragon

pixel seems screen resolution dependend idk how that would work

yea but pixels still could get used

17:21

what you see on screen is affected by the zoom factor

thats just confusing to me :D

default

yea but pixels still could get used

I at least could list a few games that do that

17:22

i wouldn't use it for velocity though

17:22

velocity in most cases got no unit

according to ddnet wiki a tile is 64x64

but is just representing your position change in a specific time

so i assume a vel of 32 makes you travel half a tile per tick?

17:23

idk

default

but is just representing your position change in a specific time

it's a vector that gets affected by gravity too

isnt velocity measured in distance unit per time unit?

17:23

generally speaking

ChillerDragon

according to ddnet wiki a tile is 64x64

pixels, yup, but what if m_vel doesn't represent pixels kekw

oh wait no

17:24

60x60 units xd

17:24

wat

what

Solid tiles (unhookable, hookable, hookthrough) have a square hitbox of 60x60 units.

17:24

https://wiki.ddnet.org/wiki/Collision

Collision

is that even true

kinda of misinformation?

internet never lies

17:25

wiki always facts

ChillerDragon

wiki always facts

X years old wooker

gerdoe

kinda of misinformation?

I mean that was written by nondevelopers that try to explain collision