DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/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 2022-10-01 00:00:00Z and 2022-10-02 00:00:00Z

NovaShock tested them and it's super laggy for him as well, seems like a bad option

ZA servers have so much jitter (edited)

louis

Click to see attachment 🖼️

I love doing the crossword but it's usually too hard for me because it's just so american

Robyt3

GitHub About still lists ddnet.tw, should also be changed to ddnet.org?

@heinrich5991 @deen ^

fixed

ty, just noticed i apparently can do that too xD

I found another one on github, also fixed

louis

Click to see attachment 🖼️

issou

good luck

I always wondered; what are the 15.1-devel to 15.5-devel branches in ddnet?

which branches xd

hmm, did someone delete them?

08:44

I also remember seeing thsoe

yeah just deleted them ^^

I think these had our releases

we use tags usually

ah, they're still in the tags

08:46

the development of these 15.1.x, …, 15.5.x versions happened in these branches

08:48

deen branched off these versions so we could still PR to master, but new features wouldn't go into these releases

Ah, that makes sense.

08:53

Also, CodeCov? Is that a code safety checker?

I think it's a coverage checker

08:54

if it is, it tries to record which paths of the code run during tests

So you can learn how much of the code your tests COVer

Right. I always wondered what it's like. Especially the whole headless client "open client 1, record demo" thing.

We don't do proper unit testing so our coverage will be very low compared to other projects

and for the backends its hard anyway

Who deleted the bitcoin offer? I was going to become rich

wat

08:59

was there a spam msg? xd

09:00

irclog doesnt forget

09:00

i always see stuff popping up here, wtf going on xD

seems to be automated

09:00

the deleting

ok ^^

09:02

could be the new way of discord chat annoyance if its not discord deleting them, but the bots themself

[ddnet-web:master] 1 new commit

0b7ace1 Add 500 € (!) donation by Mıss Fʋггצ - def-

6

heartw

9

poggiesmindblown

poggers

GitHub

Click to see attachment 🖼️

when #announcements xd

poggersrow

[ddnet-web:master] 1 new commit

057a98e qshar.com -> kog.tw - def-

Kog went from a common tld to a country one, we went the opposite direction

i mean .tw is pretty cool

I liked the .tw too tbf

09:47

The two character ones look pretty cool

.de

DD-Nett

The .tw was fine for everyone who was aware that we started as a Teeworlds mod. For new players it's probably just confusing

and chinese ppl get mad xDD

They only get mad about Taiwanese flag ingame, I haven't heard anything about domain name

i mean we have negative steam reviews bcs of that flag, completly redicilous, steam should delete them and ban the accounts

why should they ban the accounts?

I'm not sure. If they feel seeing the Taiwanese flag in scoreboard during gameplay annoys them so much, maybe it's worth a bad review from their PoV

i doubt thats not against the TOS

it's not like they are saying anything wrong in their review

writing reviews based on political views

why would that not be allowed? "there's nudity in the game, and I don't like that"

bcs its ridicolous that simple

I don't share that opinion

well u'd remove the taiwan flag if enough ppl give negative reviews?

10:29

that's really stupid imo

you seem to say that everything that gives bad reviews needs to be changed

10:30

but I think we actually do hide it in the chinese client

we do? thats already sad enough

thats racist

https://help.steampowered.com/en/faqs/view/6862-8119-C23E-EA7B

Steam Support :: Rules and Guidelines For Steam: Discussions, Revie...

10:33

"Religious, political, and other "prone to huge arguments" threads"

10:33

so its against their tos

10:33

and steam should delete them

that seems unenforceable

10:34

and unenforced

10:35

I saw plenty of mentions of ukraine or gay rights on steam

well we don't call it country flag in the UI so we don't judge about taiwan in any way and taiwan has an official flag, there is nothing political from our side so these reviews should be deleted yes. I know what you mean, but if the game does not mention gay rights in any kind, and somebody posts a review (be it negative or positive) "gay rights suck/are awesome" or smth like that then it should be deleted (edited)

I can see your point, the reviews are not really related to the game and about a topic that invites flamewars

Nah actually hiding a flag for certain groups of people is just weird

[ddnet/ddnet] Issue opened: #5899 Question when searching server lists

I can't login my discord,so I have to report here.Sorry for that.

Some Chinese players said that they can't see servers with other mods(like MMOTee,InfClass).I had these questions,too. I saw:

Abroads' servers without DDNet could show successfully.
I had Teeworlds Cilent(v0.6.5),and I could see and join Chinese Infc servers in it.

I heard DDNet are using own master servers now.Is this the reason of the question?Thanks for help.

@Beelee @NovaShock Try new South Africa server, it's stable for me

deen

@Beelee @NovaShock Try new South Africa server, it's stable for me

thanks the servers seem to be stable for everyone so far

[ddnet/ddnet] Pull request opened: #5900 Set coverage flags to separate coverage (fixes #5898)

https://docs.codecov.com/docs/flags

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](h...

[ddnet/ddnet] Pull request opened: #5901 FIX: Announce gametype "Gores" & parse them correctly in the client

Adding new gamemode "Gores" to the client with its own color https://i.gyazo.com/65b587ffe87f68cb98f86ee3a6e21166.png

Checklist

[x] Tested the change ingame
[x] 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
[x] Changed no physics that affect exis...

Hello, enumeration declaration inside protocol_ex are they connected in a strict order? If there is an offset on any side of the client or server? What would be the behavior in this case?

its generated by python

13:12

iirc

I'm writing about https://github.com/ddnet/ddnet/blob/dbe6d8cb293ea7452c0d4f6613b60f591bbd7c48/src/engine/shared/protocol_ex.h

ddnet/protocol_ex.h at dbe6d8cb293ea7452c0d4f6613b60f591bbd7c48 · d...

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

13:15

I can kind of see that it is, but I'd better check

[ddnet:master] 2 new commits

89739ef Set coverage flags to separate coverage (fixes #5898) - def- 7f8408e Merge #5900 - bors[bot]

13:20

[ddnet/ddnet] Issue closed: #5898 Consider using Codecov flags to separate unit test and integration test coverage

13:20

[ddnet/ddnet] Pull request closed: #5900 Set coverage flags to separate coverage (fixes #5898)

Kurosio

Hello, enumeration declaration inside protocol_ex are they connected in a strict order? If there is an offset on any side of the client or server? What would be the behavior in this case?

Neither the server nor the client cares about the actual order. These are translated to uuids at the protocol level

Learath2

Neither the server nor the client cares about the actual order. These are translated to uuids at the protocol level

Okay, thanks, I kind of figured out how it works with debugging server and client

[ddnet/ddnet] Issue opened: #5902 Opening demos (and presumably maps) with spaces in the filename by dragging on the ddnet executable doesn't work anymore

Reported by teini94. I'll bisect. Was probably introduced recently by me.

yahoo.de is missing A records: https://yahoo.de/

[ddnet/ddnet] Pull request opened: #5903 Fix unknown command handler for filenames containing spaces

Pass the original string to the unknown command callback instead of the parsed command, as the latter ends at the first whitespace, which breaks for unknown commands (filenames) containing spaces. Closes #5902.

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
[ ] Consider...

[ddnet:master] 2 new commits

9d1f393 Fix unknown command handler for filenames containing spaces - Robyt3 243acf5 Merge #5903 - bors[bot]

18:33

[ddnet/ddnet] Issue closed: #5902 Opening demos (and presumably maps) with spaces in the filename by dragging on the ddnet executable doesn't work anymore

18:33

[ddnet/ddnet] Pull request closed: #5903 Fix unknown command handler for filenames containing spaces

[ddnet/ddnet] Pull request closed: #5092 Rework HTTP to use curl-multi interface

Is discarding the return of a write somehow insane? Am I missing something?

19:34

write(2) that is

if you don't care about whether it worked or not, it's okay

I don't have much in the way of resolving the situation if it didn't

you can always crash the process. but I agree that gnu is insane here by not recognizing the (void) idiom

19:36

I think we can turn that warning off

I can just return it upwards and ignore it there, maybe that way it'll stop warning

heinrich5991

I think we can turn that warning off

In most cases where it's used it's actually useful. Like a socket() or fopen() being ignored

19:37

But a write, I honestly don't know what they were thinking

I agree with them there that crashing is preferable to ignoring, if we think that it should always succeed

19:37

however, I disagree that (void) shouldn't silence the warning

19:37

I think clang handles this correctly

19:37

so we could silence the warning on GNU, I guess

19:37

but that's hard, I guess

So, should I assert or not that's the question

do we think it can ever fail?

19:42

if not, I'd write an assert

19:42

but I don't have a strong opinion on it

19:42

it's just how I'd do it

how bout at least add an dbg msg

I'm fairly sure it can't ever fail without a catastrophic failure in the kernel or gross user error

sounds like a should u check the return of a malloc type of error^^

Oh there is one insane fail condition that ignoring would make sense in. If the pipe is full. Since all the info is to be discarded anyway there is no point writing, might as well ignore it, there is already stuff there to trigger select() to return

19:47

How the pipe would get full is beyond me though

ah

19:47

then it's fine to ignore I guess

19:47

idk

19:48

do whatever you want

can we pls remove this

Image attachment

20:04

its no error and nobody cares

20:04

and it spams the console hard

Image attachment

20:06

Image attachment

xD

yea, these shouldnt be in console_output_level 0

Jupstar ✪

can we pls remove this

change dbg_msg to log_debug or log_trace

heinrich5991

change dbg_msg to log_debug or log_trace

can u bcs u added it xd

xd

no i can do it too, i'm just mad that SDL stuff never works under windows so i rq coding again

20:21

apparently u cannot kill a window under windows after only 1-2 frames were rendered... i cannot imagine it otherwise

Jupstar ✪

apparently u cannot kill a window under windows after only 1-2 frames were rendered... i cannot imagine it otherwise

It's the windows way, stuff that's completely arbitrary from the perspective of the user

i'm already mad that u cannot just open a msg box lmao xd

20:23

windows fullscreen is so weird, windows has no more control over exclusive fullscreen windows

20:23

and the drivers apparently dont have enough control over the windows compositor to render stuff

20:23

so basically they cancel each other out

Honestly, that's how I understand exclusive fullscreen on any OS. Isn't that like the point? The app that has exclusive fullscreen basically controls everything rendered on the screen

mhhhhhhhhhhhhh

20:26

i dunno

20:26

linux solved it and it still has better input times

It's already bizarre to me that you seem to be able to render stuff on top of exclusive fullscreen apps under X

but imagine it like this

So bizarre that I wasn't even sure if what linux has nowadays is even exclusive fullscreen

u dont render stuff from WM u have 100% exclusive there is a popup u have fullscreen(looks like before) but not exclusive

20:27

if u dont get 10 seconds black screen like on windows when going from fullscreen to minimize it should easily work

20:28

i mean it might not be 10000000=% exclusive fullscreen but some if(hasevent) check cannot be a reason to not do it like linux does it

Problem as far as I understand it is that that used to entail a mode change for the longest time, I don't even know when the requirement really went away

this can never cause so much perf decrease

Jupstar ✪

u dont render stuff from WM u have 100% exclusive there is a popup u have fullscreen(looks like before) but not exclusive

Like when you get a popup the app that had exclusive fullscreen now becomes borderless fullscreen on linux?

Learath2

Problem as far as I understand it is that that used to entail a mode change for the longest time, I don't even know when the requirement really went away

i dunno, there is the same amount of information going to the monitor

Learath2

Like when you get a popup the app that had exclusive fullscreen now becomes borderless fullscreen on linux?

i mean just for the few moments another event is there, but yes

20:30

i think linux also calls it direct scanout

20:30

ok wayland calls it like that

20:31

but they still force vsync

20:31

evil

What I thought was that in this special mode that is exclusive fullscreen, there is no intermediate handling anything, it's like GDI handing the control of the entire framebuffer to your application. Or an example in linux would be going from tty to X, even when this doesn't involve a change of mode it still gets that black screen in between.

20:32

I should investigate it more, I honestly don't know much about the lowest layer of the graphics stack in either linux or windows

Learath2

What I thought was that in this special mode that is exclusive fullscreen, there is no intermediate handling anything, it's like GDI handing the control of the entire framebuffer to your application. Or an example in linux would be going from tty to X, even when this doesn't involve a change of mode it still gets that black screen in between.

i just wonder why? i mean if X does nothing than just do a simple if(event) check for every swap in buffers it should be fair enough shouldnt it?

20:33

in the end if x writes the framebuffer or other software does that really matter?

20:33

they can simply switch whenever they like

I thought the idea was avoiding an extra copy of the framebuffer. E.g. your software copies to a buffer in x, x copies the buffer to the driver, vs your software copies it directly to the driver

20:35

Again, I'm not sure, that's just the mental model I had for it for the longest time when thinking about the difference between exclusive and windowed fullscreen

Learath2

I thought the idea was avoiding an extra copy of the framebuffer. E.g. your software copies to a buffer in x, x copies the buffer to the driver, vs your software copies it directly to the driver

ok but then, why would x even get the framebuffer in first place, why not let the driver handle the x code

20:36

i see them more as the same in the end its some call to the gpu kernel drm stuff

Well the WM needs to decorate the window, cursors need to be drawn over it, there might be popups to render above it

20:37

If X isn't inbetween, e.g. you have direct access to the driver, nothing can be rendered on top

yes i understand that, but then again just switch ur app between app only and window manager + framebuffer of the windowed game

20:38

but tbh... why would a video mode change ever cause such blackscreens anyway? is the monitor processor so slow?

20:38

i mean we are in 2022 nobody can tell me init 4mb buffers takes 3 seconds

That part I have 0 clue about, it might have something to do with legacy signalling standards

20:39

Like need to wait 100 vblanks before sending any new video so CRT tvs can catch up or sth

lmao

20:39

well who knows

20:39

i ofc also dunno what linux does internally, or what monitors do

20:39

but from my naive point of view i dont see a reason for it

20:40

at least not such times

Or maybe there are some signalling constraints, the monitor has to sync up to the signal properly, for which some sort of clock recovery needs to take place, that can take a while to settle

if u not changing resolution i'd also just render 16bit colors in the 32bit colorspace bcs why should this cause a video change

No change in resolution usually does result in no blackscreen nowadays, so something did get better somewhere along the way

yeah but windows for example does it, everytime for exclusive fullscreen, for some ppl even 3 seconds or more

20:41

now we at the start of our discussion again i know

20:41

but i simply cannot imagine a good reason

Jupstar ✪

yeah but windows for example does it, everytime for exclusive fullscreen, for some ppl even 3 seconds or more

I thought that didn't happen for you, does it?

but windows updates also take 10-20 minutes that's also not logical to me ^^

Learath2

I thought that didn't happen for you, does it?

on windows for me it does it for like half a second

20:42

but yes i also get this

20:42

on linux i dont get any of that

20:43

and funnily enough, we meassured input lag with a arduino light sensor and linux was faster or equally fast as windows in ddnet so it's not like linux trades anything for it

20:43

(the light sensor has 8khz refresh rate)

Jupstar ✪

but windows updates also take 10-20 minutes that's also not logical to me ^^

This one I can actually explain, it's the shitty unefficient format of the delivered updates, they come in these small chunks that independently install and each create a rollback point so at any error during the installation it can be undone

Learath2

This one I can actually explain, it's the shitty unefficient format of the delivered updates, they come in these small chunks that independently install and each create a rollback point so at any error during the installation it can be undone

ok, i just wonder why they dont improve it xD

20:44

this cant be the best microsoft came up with

20:44

after 25 years of making OS

No need to, it works, it worked for decades, they just do the updates when people don't use the computer and they heavily background it incase they have to coincide with computer use

20:45

It's also a profit seeking entity, if faster updates isn't something people would pay extra for, why bother?

everytime i start windows i can a big update.. so probs all 6 months xDDD

20:45

and then the grant waiting starts

Learath2

It's also a profit seeking entity, if faster updates isn't something people would pay extra for, why bother?

you are absolutly right

20:46

thats probs their logic

Well do you boot into windows every day? Most of their users do, so they probably get smaller wait times at boot instead of one huge chunk like dual booters which make up a very small amount of microsoft clients

[ddnet/ddnet] Pull request opened: #5904 Less logging when loading maps on expected path

!screenshot-20221001@224631

Checklist

[x] Tested the change ingame
[x] 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 tha...

Learath2

Well do you boot into windows every day? Most of their users do, so they probably get smaller wait times at boot instead of one huge chunk like dual booters which make up a very small amount of microsoft clients

yeah ^^

Anyway, I'm actually interested in how windows and linux handle exclusive fullscreen differently, I'll dig a bit into it

Learath2

Well do you boot into windows every day? Most of their users do, so they probably get smaller wait times at boot instead of one huge chunk like dual booters which make up a very small amount of microsoft clients

At work I still saw this problem: Person comes back after long vacation, important meeting on first day, and Windows update decides to run at that time and person can't present for 20 minutes

20:48

or are you talking about the time before update runs?

Learath2

Anyway, I'm actually interested in how windows and linux handle exclusive fullscreen differently, I'll dig a bit into it

nice, i can just tell ya, with KDE compositor with some effect that requires bit of GPU work, it's defs slower if the window renders some overlay, e.g. alt+tab overlay so it defs changes in between them "silently"

deen

At work I still saw this problem: Person comes back after long vacation, important meeting on first day, and Windows update decides to run at that time and person can't present for 20 minutes

Hm, don't big corporations just do their updates completely on patch tuesdays? I thought that was how they kept the corpospace happy

20:50

IT does updates on tuesday night through group policy, policy restarts computers completing the full update, then shuts them all down.

20:51

Though I must admit I've never been in any corporation of significant size

They came a bit later, but the updates still nagged me and if you don't react quickly enough they happen automatically

20:55

and can't be turned off because of company policy I guess

deen

They came a bit later, but the updates still nagged me and if you don't react quickly enough they happen automatically

Hm, maybe just suppress them until the very end of the work-day?

21:06

How on earth did I break the integration test? :/

They only fail on CI too, that's excellent

[ddnet:master] 2 new commits

55ff665 Less logging when loading maps on expected path - def- 3da9aeb Merge #5904 - bors[bot]

21:43

[ddnet/ddnet] Pull request closed: #5904 Less logging when loading maps on expected path

deen

@Beelee @NovaShock Try new South Africa server, it's stable for me

server is much better, thank you

21:51

@deen can we add a gores maps to the server?

21:53

like KoG, easy, main, hard etc

No, you have to talk to Avolicious about that

22:22

it's run separately from DDNet

Exported 227 message(s)