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 2020-10-12 00:00:00Z and 2020-10-13 00:00:00Z

[ddnet-maps:master] 1 new commit

9a8bd3b M Sunny Land, A UpNDown Coop - ddnet-maps

Hey @deen

04:43

What is "sixup connections"?

04:44

I googled it, but found a site for study in europe

04:44

nothing more about programming or network main

[ddnet/ddnet] Issue opened: #3072 e84d3a2 introduces performance problems and graphics tearing on my toaster

e84d3a2 makes the game jittery all the time, even though my fps (cl_showfps 1) doesn't drop at all. It hurts my eyes and affects my ability to play the game. There's also occasional tearing. I tried capturing it on video but it's kinda hard to notice in the video for some reason. I used git bisect to determine that this is the commit that introduces the jittery-ness and graphics tearing. @Jupeyy can you please post the specs of the pc you test this with? keep in mind tw is a "[A retro mult...

Obiwan: It's our way of providing Teeworlds 0.7 net protocol compatibility in DDNet Server

Unsigned let's you think about the math u do. They nice

06:55

Google doesn't even give a reason. So thumbs down on that

"Bjarne Stroustrup says that unsigned ints are error prone and lead to bugs"

06:57

@Deleted User the creator of c++ says things too

06:57

xd

In my mind the main danger comes from the very common value 0 being one-off to an underflow. much higher chance of going wrong than with int64_t

yeah

[ddnet-web:master] 1 new commit

321c761 Remove Umbra from ddnet mods (stays pvp mod) - def-

well teeworlds is the best argument against this. unscalable, full of undefined behaviour, negative array accesses that arent checked

07:16

int to floats, floats to int without thinking

07:16

ever wondered why we dont have a full 360° cursor in teeworlds

wdym 360° cursor?


    m_HookDir.x = pObjCore->m_HookDx / 256.0f;
    m_HookDir.y = pObjCore->m_HookDy / 256.0f;

07:17

bcs we just kill the precision

07:17

if that would be 1024, tw would be much nicer already

o_o

move ur cursor very slowly

07:17

with hook line

07:17

and u'll notice

is that why some hooks thru corners from far away dont work well

07:17

because of precision

if the hook is long enough probably,

wow

07:18

u just saved me

07:18

alot of time thanks

u cannot change this

07:18

the client uses it for network

yes thats why it saved time xd

ok xD

i thought i messed something up

07:18

didnt consider that it could be limitation of tw protocol

one thing is true tho, try not not mix data types

07:20

but i dont see an argument against unsigned

07:20

if unsigned underflows u are fucked

07:20

so u need to care much more than with ints

07:20

signeds*

array[-1] and array[2147483648] will both crash

yeah bcs -1 is probs converted

07:21

but (ptr - 1) might be a legit memory address

oh true

and thats undefined by your source probably

07:22

not invalid read write, but undesired

and undefined ye

Image attachment

07:31

we prefer todo this instead of not program like this xd

~~wizard~~

The bounds checking on gcc is absolutely shit, we are in the right to silence it

i dunno std::array for example typedefs its size_type to size_t, so the [] operator would require unsigned

08:15

(what i am just saying is, the compiler probs prefers to warn than to create user undesired behaviour)

@Learath2 To be fair, C++ standards commitee is in an impossible situation. They want 100% backwards compatibility, new features without any cost, and modernize the language at the same time.

Exactly, it's clear to any sane person that they have to give one up

But they wouldn't be C++ if they did that. Other languages fill the other niches

(I'd give up the backwards compatibility, with one major revision of time to fix deprecated stuff)

Hello Python 3, my good friend

but isnt backwards compatibility what makes c++ stay for so long

Unmaintained code can be compiled with old compilers

The idea in C++ is instead to use something like clang-tidy to enforce modern stile in new code bases and to try to clean up old ones

Pythons upgrade was such a mess because they didn't want to commit to it. They let the userbase draaaaaag it on

09:23

Oh and that guy the benevolent dictator for life, he was just too benevolent

Oh, they wanted to commit. But the user base was strongly against it, and the same would happen in C++ but much stronger

yeah

09:23

c++ fanatics

09:23

like learath

09:23

monkalaugh

What are you on about? I don't even like C++

you like C right (edited)

@deen users don't know what's good for them, that's why we have committees making decisions

09:25

Especially the average python user

Not that it says much, but I was a bit shocked to see C move to number 1 on TIOBE Index: https://www.tiobe.com/tiobe-index/

yeah

09:25

i wonder why it is first

mostly because Java went down, not C up, but anyway

Huh, I wonder where the java devs went

09:26

Js maybe?

PHP going up justatest

justatest

09:26

damn wordpress

09:26

i think java went down due to oracle

09:26

doing that paid thing

What did they do?

you have to paid to have a LTS version

09:28

or something like that

09:28

pay*

[4:25 AM] deen: Not that it says much, but I was a bit shocked to see C move to number 1 on TIOBE Index: feelsamazingman

feelsamazingman

feelsamazingman

feelsamazingman

https://dev.karakun.com/java/2018/06/25/java-releases.html

Do I need to pay for Java now?

This post gives an overview of the new Java release train as it was announced by Oracle. Plus, the article provides some important information and hints how you should handle new Java releases in the future and helps you to decide if you need to buy commercial Java support or ...

c best

@Learath2 they are speeding up java development

09:28

Starting with Java 9 most Java version will only have a lifetime of 6 months. Updates won’t be provided for the version after this period. Even if you have a commercial contract with Oracle you won’t get any additional updates for Java 9 or 10 after these 6 months. Plus, all 3 years a LTS (long term support) version of Java will be released. The first version with LTS support will be Java 11 that is scheduled for fall 2018. This LTS version will provide a commercial support with updates for 8 years.

09:29

as u know companies want stability

09:29

so they seek LTS versions

09:29

i think

09:29

then u have to pay

09:29

xd

09:29

idk maybe i misunderstood

PHP going up justatest

justatest

i know some php

09:29

it isnt really good

The most important fact is that updates for LTS versions can only be accessed and used by people that buy commercial support at Oracle. My experience is that today most companies that develop Java based software do not have commercial support since the release train of Java was quite „friendly“ and an adoption of the newest version was easily possible.

09:29

this ^

I doubt the commercial support being paid is an issue

before new java releases were rare

Java 11 is soooooooo stable

but now they are speeding

Even the open source version, pretty much everyone uses that

btw i think java 15 is modular or smth

09:30

yeah

09:30

but those wont receive updates

09:30

if u donth ave commercial

09:30

support

multiline strings were added with ` in java 12 (edited)

09:31

lol

Oracle JDKs may only be used in production if you buy the commercial support. I

09:31

also this

what do you expect from oracle than wanting money

u can no longer use oracle jdk freely

well that commercial thingie really sucks

most minecraft stuff is still java 8

09:32

i wish they updated

09:32

coding spigot is quite fun

java 8 is enough for a java dev i suppose

Well do you really need updates to java 11?

my son will rise up and conquer the world

Image attachment

09:33

monkalaugh

i have both 8 and 11 but 8 still does my stuff

OpenJ9 maintains LTS releases too and you can use them

there are good featurs in new java versions

09:33

getting stuck in 8 is n ogood

its just enough

09:33

i dont say its bad

09:34

nor good

Java's streams are so nice to work with

java 9 intoduces modules

maven exists

They really made functional programming quite palatable and usable

i agree with that

i dont like ppl who are content with "enough"

09:35

weak mindset

it just does what i need

09:35

i wouldnt need a higher version than 11

yeah why software gets updates

09:36

stupid developers

to get better ofc

09:37

but i see no reason to update while there is no need

@deen can you put in the readme the dependencies needed for videorecorder on debian atleast? they arent there

what are the dependencies? ffmpeg?

libx264-dev libavfilter-dev libavdevice-dev libavformat-dev libavcodec-extra

10:13

libavutil-dev

10:13

im finding them

then you make a PR

yeah, i thought u knew all them tho

10:14

also

10:14

cmake complains this By not providing "FindSQLite3.cmake" in CMAKE_MODULE_PATH this project has

10:14

but then it says

10:14

-- Checking for module 'sqlite3' -- Found sqlite3, version 3.27.2 -- Found SQLite3: /usr/include

10:14

which makes no sense to me

10:14

(this on debian)

maybe @heinrich5991 knows better about cmake problems ^

[ddnet/ddnet] Pull request opened: #3073 Add video recorder depedencies

https://ryozuki.xyz/ddnet-scan/

10:21

i set this up

10:21

ill make daily scans

I'm not a big fan of daily scans. I think we should fix them and enable clang-tidy with those checks in github so we notice on the PRs directly

10:22

and not only after they are merged

yeah but we should fix all first

10:23

otherwise the prs will get a x tick right

10:23

and its not their fault

Yes, but you could also mark all existing ones with an exclude if you don't want to fix them immediately

10:24

// NOLINT(clang-analyzer-core.DivideZero)

[ddnet:master] 2 new commits

06ede17 Add videorecorded depedencies - edg-l 8577990 Merge #3073 - bors[bot]

10:27

[ddnet/ddnet] Pull request closed: #3073 Add video recorder depedencies

well i dont know much about configuring the CI and stuff

make it work locally and we can get it done

10:28

basically just copy the normal build and run it with clang tidy on

10:29

and in the .clang-tidy file set WarningsAsErrors: '*'

10:30

And Checks: -*, clang-analyzer*. You could test if that results in the same failures you're seeing with clang-analyzer

https://cmake.org/cmake/help/latest/prop_tgt/LANG_CLANG_TIDY.html#prop_tgt:%3CLANG%3E_CLANG_TIDY

10:33

looks like cmake has clang-tidy integration

10:33

https://cmake.org/cmake/help/latest/variable/CMAKE_LANG_CLANG_TIDY.html

[ddnet/ddnet] Pull request opened: #3074 Fix skin load

Guess this makes more sense

yeah, so -DCMAKE_CXX_CLANG_TIDY=ON -DCMAKE_C_CLANG_TIDY=ON should work

[ddnet:master] 2 new commits

ec16abd Fix skin load - Jupeyy 9b153b1 Merge #3074 - bors[bot]

10:43

[ddnet/ddnet] Pull request closed: #3074 Fix skin load

that variable doesnt work like that

10:44

u set it in cmake: set(CMAKE_CXX_CLANG_TIDY clang-tidy -checks=-,readability-)

10:44

and then when u run ninja

10:44

it shows the warnings

10:44

and there are quite a lot

10:44

xd

set it just to clang-tidy and use a .clang-tidy file as I described

10:45

not all checks make sense for us, start with the clang-analyzer ones

[ddnet/ddnet] Pull request opened: #3075 Initial clang-tidy checks

treating it as error wont show all the warnings tho

10:55

src/game/collision.cpp:1030:2: error: Variable 'f' with floating point type 'float' should not be used as a loop counter

10:55

well this makes sense

10:56

oh

10:56

the CI doesnt have clang-tidy installed

Hm, we could grep for warnings in the output or use make --keep-going

10:57

just add it in the installation at top

where do put keep-going hm

in the call to make, but I think we don't call make explicitly but make cmake call it

11:01

cmake --build <dir> -- --keep-going

11:01

so just add the -- --keep-going at the end

@deen why put .clang-tidy in src?

11:05

i think cmake detects it fine in the base dir

Hm, I guess it's fine

ur github comment was a bit weird formatted

Just thought it's more specific in src, but we don't really have other source code around anyway

thanks to markdown xd

oops, fixed

@deen its full of unknown warning option

I can give it a look, is your pr up-to-date?

i mean the CI btw

11:11

yea

11:12

localy it shows warnings correctly

ok

11:13

Why did you set clang-tidy always on for every build?

well im testing

I think the idea is to have a separate clang-based build

11:14

where are the errors then?

when u compile

11:14

or what errors u mean

you said locally it works, where are the failures in CI

it doesnt recognize the warnings added by clang-tidy i guess

what doesn't?

11:15

where can i find the failures?

https://github.com/ddnet/ddnet/pull/3075/checks?check_run_id=1241803867

Initial clang-tidy checks by edg-l · Pull Request #3075 · ddnet/ddn...

I think we should decide which warnings ignore and which to keep, and maybe fix them in this pr

11:15

look at

11:15

build debug

11:15

anmd scroll down

11:16

its full of unknown warning option

That is a clang-tidy warning

rly?

11:17

localy it looks like this to me

11:17

Image attachment

also, is it building with gcc? we should be using clang for that I guess

i think heinrich can do this better

11:18

im no cmake magician

I can also try to finish it up if you want

go for it

[ddnet/ddnet] Pull request opened: #3076 Add clang-tidy to CI and enable clang-analyzer checks (superseded #3075)

470 warnings treated as errors

The actual work will be fixing those or at least NOLINT(clang-analyzer-optin.cplusplus.VirtualCall) them so we can enable this.

oh nice

[ddnet/ddnet] Pull request closed: #3075 Initial clang-tidy checks

Float as a loop counter is a first for me

@Learath2 wdym by first for me

11:42

as in top priority? or as in u never saw it

Never saw it before

xd

I hope we don't actually use it to index anything

I'll try to fix the clang-tidies we have for now

looks like lot of work

See my update in https://github.com/ddnet/ddnet/pull/3076, mostly all the same

yea

Oh, we should probably exclude external stuff too

11:44

I see 1 warning in wavpack atleast

yeah, will try

"Failed to finalize filesystems. Ignoring" Thanks systemd

11:53

I always prefer corrupt filesystems over waiting a couple seconds or actually troubleshooting

https://discordapp.com/channels/407308363031117832/490150878934990850/765175509142470666

frozen

if anyone can hep :(

After spending almost a year with systemd and archlinux, I have to say, I'm quite unimpressed

12:05

It's all packaged for even a goof to be able to use linux, but it's a rolling release. A rolling release always has issues and archlinux gives you no way to fix them

i have yet to find a issue tho

12:05

and im using it since a year too

Well do you get the failed to unmount oldroot on shutdown?

also ur quite wrong usually when there is a problem with a package update they post a news

12:07

https://www.archlinux.org/news/ghostpcl9532-2-and-ghostxps9532-2-updates-require-manual-intervention/

They decided the best course of action, is to try to unmount and if it doesn't work, ignore it

12:08

It tries to synchronize my filesystem fails, just ignores it

12:08

During boot I get like 10 driver issues, none of them have been fixed even though they've been reported for weeks

ah and that is systemd fault?

If it wasn't a PITA, I would just build the kernel without the overzealous drivers trying to attach to hardware that they are not for

in what way are driver issues arch fault

If it wasn't a PITA to build an initrd with systemd, I would just fix the stupid script that's trying to unmount filesystems that aren't even mounted

and "weeks" is a small time imho

@Ryozuki it's a rolling release that gives me no way to fix it myself. It would never happen in debian, because even though they give you no way to fix shit, they follow the kernel sooo slowly it doesn't matter

12:11

On gentoo it's absolutely trivial for me to just not build these drivers I don't need

omg man

12:11

just install the lts kernel

how can i increase the fps

u have solutions

12:11

u just dont know them

cl_increasefps 1

linux-lts

My solution is to use an old kernel on my rolling release distro?

12:11

Thanks

hell

12:12

use linux-lts44

12:12

from aur

12:12

if u want

12:12

how is a new kernel driver bug arch linux fault?

12:12

ur insane

noby biggest stalker

u wish

12:12

i was reading the chat

stalker

of who

12:12

lol

thats what a stalker would do

no u

It's not their fault that the driver is broken, it is their fault that they provide no way for me to not have the broken driver

they do proivide one

12:13

use a older kernel

12:13

and wait until its fixed

12:13

u can also make ur own aur package

Anyway, some people just don't care about any of this and they are free to use archlinux and shit over other peoples opinions online

and configure it how u like

@Learath2 you can try void linux, install older kernel series https://docs.voidlinux.org/config/kernel.html#kernel-series

Kernel - Void Linux Handbook

did u know u can configure and use pkgbuild files to ur own liking?

also runit really fast

@nori xdd

12:14

thats not his problem

I'll just go back to gentoo where I can just build my system however I want

cuz i actually told him about using older kernel

12:14

you can do that

12:14

the fact that u dont know or arent willing

12:15

is urs only

I am not willing to spend time building a whole package to customize my kernel, yes.

https://wiki.archlinux.org/index.php/Kernel_module#Blacklisting

12:15

lol

12:16

why do i even talk ur just set in stone already

12:17

use ur allmighty gentoo and done

Exactly, I don't get why you are throwing stuff at me. I tried, I don't like the unflexible nature of this distro. Opinions are opinions

i go back to watching anime

12:17

cuz it triggers me u say u cant when u can

12:17

u just dont like t he way u can

Exactly, It's very inconvenient IMHO. It might be convenient enough for you

also why are u even using arch linux if u have systemd

12:18

are u a masochist

12:18

xd

12:18

hate*

Because people tell me I'm not allowed to have an opinion about systemd before trying it

ah sure its that

IMO that's like saying I'm not allowed to hate on nazis before being a nazi. You can disagree with something without trying it, shocker

12:20

Anyway, now that I'm done with my experiment and I can freely rant. I'll be heading back to gentoo

im not sure ithats

Or maybe I'll try one more on the way

i wouldnt use a distro just to prove someone wrong

I didn't do it to prove someone wrong. I just geniunely wanted to know what people saw in systemd and a binary rolling release distro at the same time

http://www.linuxfromscratch.org/lfs/

12:21

just use this 4Head

12:21

monkalaugh

That's just too much work. Gentoo gives me almost the same level of customization without spending half a decade putting together my own distro

well cya until u finished compiling (edited)

12:22

monkalaugh

Did you know what with a well configured kernel, you can even use the computer while it's compiling?

yeah im joking

idk if its already reported, but: http://prntscr.com/uxqr1n

Embedded image

12:36

when i spec and my time is over 1hr, hindus skin is over his time

yes, fixed

12:36

will release fix now

[ddnet:15.1-devel] 1 new commit

0aa28fe Version 15.1.2 - def-

13:13

[ddnet:master] 1 new commit

fedb031 Version 15.1.2 - def-

[ddnet-web:master] 1 new commit

41dd1a8 Version 15.1.2 - def-

[ddnet-web:master] 1 new commit

55ae83a css: Make it harder to download thumbnail in forum (by Soreu) - def-

[ddnet/ddnet] Pull request opened: #3077 [WIP] UI Elements

Build will fail bcs of clang-format, i'll update this comment to explain.

Am I the only one who has to double-tick "borderless window" every start of a non-fullscreen 15.0.5 DDNet client?

14:25

As it starts as border windows regardless of the setting is loaded correctly to an ui.

14:25

Image attachment

Weird, sounds like an SDL bug

14:28

oh wait

14:28

we only create borderless window on mac, that seems wrong

[ddnet/ddnet] Pull request opened: #3078 Make initial borderless work again

Mistakenly disabled in https://github.com/ddnet/ddnet/pull/1418

No, in linux borderless windows works fine. It's just it ignores the setting on load. I mean it loads setting in UI but doesn't apply it to the game window @deen

14:35

If uncheck/check it after load it applies.

14:36

yeah bcs its only ignored when initialized

14:36

Exactly. Please don't remove borderless windows from linux build

14:36

he did the opposite

14:36

I cannot use full-sized as they give me more problems.

14:36

he added it again

14:37

Nice, thanks for adding it

14:37

So... please leave borderless windows as it is. I will just silently leave

no need to remove it.

14:38

I will tolerate this small bug. k thanks bye

14:38

?

14:38

As I understand you were going to remove borderless window functionality from linux build.

14:38

Please don't.

14:38

dude

14:39

read my sentences carefully

14:39

i don't know who did the oppisite and what. Still don't get it

14:40

see the github popup xD

14:40

ah... my brains filters github popups

14:40

thanks!

if unsigned underflows u are fucked

@Deleted User if you use something to index into an array, it doesn't really matter whether it's an unsigned or not, right?

15:04

I think -1 and MAX_INT behave the same for accidental out of bounds reads/writes

15:04

MAX_UINT*

15:05

they both access the array element before the first one

15:05

@Deleted User what do you think about blacklisting opengl drivers like chromium does? https://discordapp.com/channels/252358080522747904/293493549758939136/764996392300707851

Are you sure? MAX_UINT would probably go over

15:07

Ah, array indexes are signed aren't they

15:08

Actually not really sure, the standard seems to just say "integral type"

c style arrays should be ptrdiff

15:10

@heinrich5991 i only mean, that -1 might not crash while 2^64 will crash for sure

15:10

@heinrich5991 generally ok, but i dunno if we even require this anymore, the fallback catches everything rn i think

@Deleted User why will 2^64 crash for sure?

15:12

2^64-1 will access the previous array elemetn the same way as -1 as far as I can tlel

mhh yeah converted to ptrdiff probs

15:12

tho, e.g. for iterating

15:12

it will crash for sure xd

but for iterating, -1 actually does the sane thing

while with signed it might not even enter the loop

int i = 0; i < -1; i++

yes

15:13

thats bad, might create undesired things

ok fair

15:14

but it's unfair to say that the problems in teeworlds come from signed integers

not directly

15:14

but i still would use int for alot of stuff

the problems in teeworlds come from missing bounds checking or from C/C++'s problems of not automatically bounds checking

thats fair too

15:15

ofc rust would be nicer in general, but the question is, how would other devs react to that

15:16

or better: with infinite time, c is never better than rust

15:18

and yeah, scalability is something else too

15:18

i know it might sound crazy to have support for 16k textures, but we laugh about this now, not in 20years xd

Most any compiler I know of will catch the infinite iteration because unsigned btw

15:23

gcc, clang, icc and msvc all have warnings about it

it cannot predict user input

15:23

load a tw map, make the camera go somewhere insane

15:23

smth like this

I'm sure there are compilers/settings that can make this even more paranoid to cover for the user input case

or negative zoom lvl

15:24

smth that isnt really expected

gcc already does a lot with trying to predict the range of values a variable can ever be

well we had a bug

15:24

with the camera going outside the kill borders

Avoiding a data type because people used it wrong back when gcc4 was current is a bad idea imho

https://github.com/ddnet/ddnet/pull/2768

Clamp kill border count by Jupeyy · Pull Request #2768 · ddnet/ddnet

fixes #2766 Appearently the center of the camera can even go outside of the kill border(like even left from it) I also notice that sometimes when zooming out, and then in again the zoom level gets ...

15:25

we caught it bcs it uses unsigned, and crashed bannzays game

15:25

ofc this wouldnt be insanly critical with signed

15:25

it would just not render something at all

looks like learath still havent seen jupstar ui pr justatest

justatest

15:25

war incoming

@Ryozuki i made it as simple as the tw gui btw

What is unsigned there anyway? m_Zoom?

i havent added something insane, and the old ui is still in tact

yeah i know

15:26

im ok with it

@Learath2 it renders a unsigned amount of tiles

Ah

15:27

How did it crash the game anyway?

mhh, wait gonna look into the code

https://github.com/ddnet/ddnet/pull/3076/commits/f0807845e8ee762182a3085bc752992a2dbd5916

15:28

can this for some magical reason

15:28

change game physics?

15:28

justatest

15:29

i wish github links didnt include a fcking massive image embedded everytime

15:29

so annoying

@Ryozuki I welcome a UI cleanup, it's been in a horrible state for so long

15:29

Though I am a little concerned about a capturing lambda in the hot path

yeah i also thought about that

15:30

is it me or github also shows the tabs however it wants?

Compilers are actually great with them nowadays, gone are the early days of wasteful captures but still, needs investigation and thorough review

ah

15:31

its dark theme fault

15:31

damn

@Ryozuki github does that, you can set it to a value with the dark theme though

in RenderBorderTileEmulation i implemented the bordertiles in a loop, just like a opengl driver would do it, so it creates a specific amount of tiles to render

@Ryozuki github does that, you can set it to a value with the dark theme though

@Learath2 where

How did you install the theme? Stylish?

stylus

Image attachment

1

u in vs?

15:33

debug mode?

@Ryozuki

Image attachment

this happens a lot in china already, not a good sign.

oh

15:34

thought its just curl does support https xD

ohhh

15:34

it has customization

15:34

damn this is good

happened to me in windows debug mode

I assumed cloudflare would mitigate the effects of the great firewall

maybe add a mirror in a china server

i think cloudflare are heavily restricted already

I'm not really very familiar with how one routes based on geolocation

when that happens, can you ask them to go to the URL manually?

15:35

in a browser

which one

15:35

i don't know the full url

https://info2.ddnet.tw/info

also does the client retry?

I think it does when you refresh, not sure though

might just be outbound QoS by ISPs at high stress.

15:37

but it is 11pm so might not be. anything outbound is slow between 6pm - 9pm i think

what QoS outright blocks the request to a HTTPS page with low traffic?

by outbound i mean out of china

does cf not have servers in china?

thonk

they might dropped it, not blocked completely

tcp retries if packets are dropped

then i have no idea

15:38

doggo_lol

15:39

they didn't respond, it might be fixed itself

I'm not really very familiar with how one routes based on geolocation

@Learath2 some months ago i got millions of traffic from china to my website and i had to geoban them

https://www.cloudflare.com/network/

Data Center Map | CDN Pop Locations Map | Cloudflare

The larger the network, the better performance we can deliver to our customers. Our DNS Server Map highlights our 155+ data centers around the world.

justatest

seems like it does have servers in china

i think china servers are handled by Baidu, don't know how it works. let me check.

@Ryozuki in this case we'd need to route before it hits our servers, that sounds impossible

15:40

I don't really know how CDNs do it, the clients don't have a way of choosing the closest server

yeah, you need to call cf as a enterpirse customer to use china server

15:41

https://www.cloudflare.com/network/china/

Cloudflare China CDN | Fast, Secure Browsing | Cloudflare

For the Cloudflare China CDN we've partnered with Baidu to create a global content delivery network, with fast website speeds, secure browsing, and China DDoS protection.

An ICP license is issued by China's Ministry of Industry and Information Technology (MIIT)

oof, plus you need a ICP

Lol

i have one tho

15:42

it is outdated, but i can renew it

You are a licensed internet content provider in china?

Guys, how do you fight spam on your forum? Do you have some kind of extensions or automatic solution or it's hand labor of post-moderation.

ye

i combat it by not having a forum

15:43

greenthing

it is content provider, not internet provider, i have it for teeworlds.cn

You can choose to place your private SSL keys in our China data centers

Guess what will happen to your private keys

baidu_wink

wink

monkalaugh

thonk

@Deleted User Premoderation on registration

but thats since the forum is not active

15:44

as much*

cloudflare business is 200$/m

oof

cloudflare enterprise is contact us/m

15:44

probably more than business

tbh i can try mirroring some of ddnet on teeworlds.cn if i get my shit together

Contact Us/m is usually businessspeak for way out of your budget unless you have investors

@TsFreddie mirroring isn't really possible, it's dynamic content

15:45

I guess you could mirror the static content and use the dynamic one if it's contactable

If ddnet was popular enough I bet they would let us have it for free as a PR stunt. "Helping the open source community"

1

bluekitty

1

thanks, @Learath2

not sure

15:46

they quoted us 5k$/m for game ddos protection IIRC

thonk

info.json is probably fairly static, maybe a short cache

no

15:46

it's different per player

15:47

it contains your finished maps

15:47

so basically no way to cache

rip

@heinrich5991 we just don't give them enough exposure at this size. It'd be a waste to help us for cheap

ah

thonk

then..

15:47

i'll let tell them to restart the client then, that worked so far (edited)

15:47

lol

You see how they instantly took in brian krebs for free after akamai dropped him, who were carrying him for free too?

[ddnet/ddnet] Pull request opened: #3079 Reset renderflags directly after use

created by 5ebabe2ece3d6b7628c01491445fbcea7b003f3c

It's all about exposure for these businessppl

thonk

or maybe made a warning a banner with a retry button instead of a full-on popup.

15:49

it scares ppl i think

15:49

and it is red too

is it bad to be red

We could attach a severity to the warnings and color the popup accordingly

arent chinese very red

15:49

the flag

15:49

etc

ddnet-info.json failing to download is definitely not as fatal as red warning would imply

thonk

but you don't stare at the flag every day on your pc

why not

why would you

@TsFreddie they can just refresh on the ddnet tab btw, that should retry getting ddnet-info.json

i see

troll

15:50

cuz u love ur country

but still, probably change it to some other color

15:50

doggo_lol

Yeah I'll add it to my TODO

and possibly no popup

No popup would make it a tad hard to indicate the warning in a generic way

true.

Can you create an issue about it so I don't forget and someone else can take a whack at it if they want to?

saddog

or you can make the refresh button red or something

I thought the exact same thing, but keeping that generic is quite hard

15:53

I'd need to check the warnings list before rendering any component

15:54

Since components don't really exist until rendered I don't see any other way

[ddnet/ddnet] Issue opened: #3080 Make the ddnet-info.json warning less fatal looking

is github really slow right now or it is just china

works for me

works for me

feelsbadman

feelsgoodman

i have a proxy, but both proxy and direct connection is pretty slow

16:01

also this

Image attachment

monkaS

it does look like a avatar ngl

16:05

nvm, github isn't going through my proxy because it is technically not blocked

16:05

it works now. thx

@Moderator Robert blocking on ger1:8312

1

[ddnet:master] 2 new commits

2d88d44 Reset renderflags directly after use - Jupeyy 8f65839 Merge #3079 - bors[bot]

16:27

[ddnet/ddnet] Pull request closed: #3079 Reset renderflags directly after use

https://releases.llvm.org/11.0.0/tools/clang/docs/ReleaseNotes.html

16:52

clang 11.0 out

16:52

poggers

16:53

clang’s AST now improves support for representing broken C++ code. This improves the quality of subsequent diagnostics after an error is encountered. It also exposes more information to tools like clang-tidy and clangd that consume clang’s AST, allowing them to be more accurate on broken code.

i have clang 12

what

16:54

@Deleted User how x

16:54

d

https://apt.llvm.org/

16:54

u can add devel source

ah 12 is in development

yes xd

i use arch btw

16:55

lol

does arch not have dpkg and apt ?

lmao no

16:55

it has pacman

16:55

and AUR helpers

i mean optionally

u can install that version im sure

16:56

with clang-git pkg or smth like that

16:56

Image attachment

16:56

see

ez

but why u want the dev version

16:57

do u test new llvm stuff?

16:57

xD

well c++20 features, and also i had a bug some day and i was annoyed xd

16:58

since then i didnt deactivated it xd

https://clang.llvm.org/cxx_status.html

16:59

ah

@heinrich5991 u obviously know that, but, if you not use size_t(on 64bit) there is ofc still a case where unsigned int is different but to be fair, i'd probs use size_t most of the time when i work with ptrs directly, and then this ofc is equal to -1

Image attachment

yea, with 32bit integers on 64bit platforms, -1 as an index is different between signed and unsigned

makes me wonder if a 128bit compiler(emulated) would find some very rare bugs

[ddnet/ddnet] Pull request closed: #3044 [Draft] Add Head Jump

[ddnet:master] 2 new commits

200c4fc Fix alignment issues in CCommandBuffer - Learath2 d006b40 Merge #3068 - bors[bot]

17:34

[ddnet/ddnet] Pull request closed: #3068 Fix alignment issues in CCommandBuffer

[ddnet:master] 2 new commits

f85db90 Make initial borderless work again - def- 3244eba Merge #3078 - bors[bot]

17:49

[ddnet/ddnet] Pull request closed: #3078 Make initial borderless work again

Alguém BR

[ddnet/ddnet] Pull request closed: #2811 Retry SQL write statements

[ddnet/ddnet] Pull request opened: #3081 NaNs

Ofc not all NaNs are terrible, the most important thing is, that these NaNs arent used as integers, or other critical behaviour i added feenableexcept(FE_INVALID); to test this. tested a bit, the game felt the same, the shotgun bug is fixed by this(ofc undesired) Maybe we can collect them and see what can be fixed

@deen u dont know the shotgun bug? XD

no

thats funny

20:26

bcs its used in brainduck i think

there are no physics bugs

xDD

everything is as it is supposed to be

tho we could think about fixing it like heinrich did with grenade

if the shotgun bug is the one where u get speed to one side with spec

20:36

or when u go one direction and shot a side

20:37

they are used in maps

20:37

also shotgun bugs are fun

20:37

pls dont remove them

20:39

https://youtu.be/aYCpB0v4Mcg?t=97

20:39

fun

hes trolling xd

20:42

deen always trolls xD, he just makes it so perfect that u think hes not

deen is a unkown troll

21:08

omg a typo

21:12

and btw stop "fixing" features

21:12

whats next fixing the wallhammer?

NaN = not a number, so, basically this is non existing

21:14

it never existed

21:14

what shotgun bug?

@noby im legit and no

confirmed by duck.com https://zillyhuhn.com/cs/.1602537337.png

Embedded image

kclient has zoom and tracers and radar

21:17

so ban u

oof

chillerbot-zz.exe

21:18

kek

who owns http://rafabrony.com/ ?

Welcome, on Rafael's Brony Network [DashNet]

Welcome, on Rafael's Brony Network [DashNet]

21:26

Embedded image

21:27

similar sites ranking to teeworlds.com lol https://zillyhuhn.com/cs/.1602537961.png

Embedded image

whats tracers and radar

https://www.alexa.com/siteinfo/teeworlds.com

teeworlds.com Competitive Analysis, Marketing Mix and Traffic - Alexa

What marketing strategies does Teeworlds use? Get traffic statistics, SEO keyword opportunities, audience insights, and competitive analytics for Teeworlds.

Exported 668 message(s)