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 2020-10-27 00:00:00Z and 2020-10-28 00:00:00Z

[ddnet/ddnet] Issue opened: #3199 Feature Request: /timesnear

Cousins with the /teamrank command, except allowing you to figure out who you [probably] finished a map with on team 0. Ex usage: /timesnear d3fault *** 5 years and 9 months ago, 4:20.69 <-- same info as /times d3fault Skeith +0.01 (4:20.70) Broken -5 (4:15.69) [etc, giving 5 nearest finishes] The plus and minus indicate whether they finished before you or after you, but the list is sorted via distance from your time... because the tees that finished nearest your time are most l...

has anyone else had this bug where the game randomly freezes and then starts flashing angrily and you have to kill it?

00:29

has been happening to me for a few last updates at random

well nobody wants to debug xd

I could debug but it happens seemingly whenever it wants to and I'm not doing anything special

it only happens on windows and on intel devices

that would be correct

so we arent sure if its the ddnet code or the intel driver since windows update

00:31

if you can reproduce, you can enable asan in visual studio, to find a memory leak

well you know I'm coding-illiterate

louis even has it ready, he just needs to reproduce the crash once xd

well to me it happens just every few days for no reason

yeah, but if you want join trashmap with multeasymap(non halloween) louis said it happens there always (edited)

it freezes, then I alt tab out the game and after tabbing back in it flashes black until I kill it

00:33

to me it happens on any map

00:33

often on my home server when testing own maps or on any ddnet map

@louis wanna debug xd

00:33

@Ravie you could ofc try a newer intel driver, but seems like the new one has other issues

it's fine, this isn't much of a problem to me but just reporting

00:34

I don't play the game enough to have this kill a run lol

yeah, i'd love to know whats causing it, but yet only intel users reported it

00:34

and only windows

00:34

and only since a few weeks xD

hm I've had similar things happening for maybe a few months

00:35

I think at one point it was flashing white instead but I might be remembering it wrong

then it could be the ddnet code ofc

00:35

but i cannot reproduce on amd hardware under linux

huh

and under linux u have pretty good tools to find such errors

what I can say is, this time it happened when someone joined the server

00:36

but this isn't always the case since this also happens when I'm completely alone on lan

our only current finding is that something is wrong in prediction code, sadly deen caught it randomly, so we don't know when it happened xD

btw ive always had it near the start of a map

00:37

or when i first join a srv

@louis i'd really like to know if its still reproducable with trashmap with that multeasymap

00:38

if you want i can setup a trashmap quickly

sure

trashmap.ddnet.tw:8575 pw is Test

@Ravie https://devblogs.microsoft.com/cppblog/wp-content/uploads/sites/9/2019/10/ASan-Blog-Post-Image-0.png

Embedded image

@Deleted User

01:37

ok

01:37

i have it frozen

01:37

and i have asan on

01:37

(i think)

01:37

now what

01:37

do i pause it?

Yes

01:38

Asan should have output smth

01:38

In the output

its a bunch of exeption thrown for memset

01:39

do i rly have to dig thru it to find asan

Just copy the whole output

01:39

Ctrl a

its literally just tons of exception thrwon at index 0x0000 smth

01:41

im looking thruw hole thing

Yeah but the asan output looks very clean

01:42

It should say smth like address

01:42

Or sanitize

01:43

but use ctrl f

01:43

just search the output

If the addresses don't resolve, you are missing llvm-sanitize in your path

i was in windows and tested it btw

llvm-symbolize*

just using the SLN project and added /fsanitize=address was enough

Or the symbols

01:46

@Deleted User that wouldn't link properly for me no matter what I tried

dont have anything

strange, i also used a clean vs 2019 tho

01:47

then its not inside of ddnet

Or it's not a memory issue

yes

01:47

but must be actually

01:47

else the 10000 triangles make no sense

@Deleted User try ubsan

i tried everything under linux

Maybe we are overflowing sth

01:48

I mean on windows

@louis just to be sure, can you search the file graphics_threaded.cpp

for what

on the right of vs u can search it

01:48

we'll quickly do an overflow to check if asan was working

i wih u guys could just be on my comp and do it for me

01:50

but im nto gonna remote access

Embedded image

01:50

top right

01:50

search solution explorerer

01:50

just search the file there

ye

inside the file search for


mem_copy(Cmd.m_pVertices, m_aVertices, VertSize * NumVerts);

and replace it with


mem_copy(Cmd.m_pVertices, m_aVertices, VertSize * NumVerts * 200000);

01:51

then compile and run

ok

it should crash and output stuff from asan

01:53

@Ravie got a SLN file?

Image attachment

where?

@louis looks good

01:53

@Ravie or how did u compile it

01:53

with cmake gui?

nah built it from within vs

mhh i think that didnt really work @Ravie

01:53

but louis tested anyway

01:54

and asan showed nothing

01:54

and now we proved that asan was on, bcs the overflow by ourself was triggering

i feel like my comp has like an obscure bug and provides negative results

01:54

justatest

i wonder if it also crashes with opengl 1.x

01:55

gfx_opengl_major 1

01:55

one intel user said, he even had a crash with gl 1.x

wait how is it not an issue in gamecode if its so reproducible on spexific maps

well this tool finds all overflows, and it has to be an overflow, since your debug showed a invalid variable

02:00

so it must have been overwritten by something

02:01

but if its not detectable by asan it might not be reachable by asan e.g. by the intel driver

02:02

tbh that would make most sense, deen and me already used asan and ubsan quite often lately, the only bug we found lately that could have caused something was a bug reported by asan

02:03

if its inside opengl and intel knows about it dbg_gfx 1 would probably show it

02:03

my amd driver and the mesa software renderer doesnt show anything tho

so i need to change my drivers till i find one that works?

@Ravie on the right see the DDNet project, right click properties add this: https://devblogs.microsoft.com/cppblog/wp-content/uploads/sites/9/2019/10/ASAN-Blog-Post-Image-2.png then set the ddnet project as start project

Embedded image

02:05

build and run

02:05

@louis thats my fear yes

02:05

@louis dunno if u ever used linux, e.g. from a bootable usb stick but you could try it

done

intel is open source udner linux

ye i have

02:06

how would that help

@Ravie when running it will probably crash inside a "memset", just ignore that one

02:06

@louis i assume intel drivers under linux are better, bcs heinrich and deen use intel both

xd well i cant use linux just to play ddnet

yeah i can understand that

02:07

i dunno, they release a driver once per month or smth

02:07

maybe the next driver will be better

maybe ill try linux again to see if fl studio support isnt total trash anymore

yeah thats the downside of linux, its not well supported desktop wise from the industry

so should it throw an error somewhere when it crashes?

asan should output alot of stuff in output bottom right @Ravie

oh yeah

but not the memset thing

wait what

it really has to contain a stacktrace and address sanitizer in the name (edited)

there's a lot of access violations

yes but they false positives or smth

The thread 0x289c has exited with code 0 (0x0).

its not the report from asan

02:22

thats also not a bad thing

02:22

just a info

alright so how do I filter out asan

just ctrl + f and search address

it's a million entries long list of violations

or sanitizer

02:23

or "sanitize" to be safe

nothing

yeah

02:23

same for louis

nothing for sanitize and for address it only found irrelevant ones

and louis even tried if asan is enabled

02:24

so i could be wrong, but the chance that its not inside ddnet code is not small

02:24

i'll wait for deen tomorrow, maybe he has in idea

did my messages in game go through?

yes

02:26

the client works

02:26

thats also what others reported

02:26

its really just the visuells broken

These sanitizers are all useless if we can't trust them...

02:33

On linux I know that asan won't miss anything, I know that ubsan will never generate a false positive

02:33

On windows god knows

ikr

02:33

but on linux it doesnt show anything either xD

yeah this bug I don't know what causes it

02:34

the fact that it's reproducible for some people is odd to me

02:34

@Deleted User do you know which generation of intel cpus we are getting trouble with?

yes

02:34

only since 2017 gpus

02:34

late 2017

I have a couple different generations with iGPU I can test on the weekend

i even tested some 10 years old intel and it had no problems xD

meh I don't have a kaby or coffee lake either

02:36

I'll try a haswell and a broadwell

actually mine seems to be from late 2016

02:50

i3-8145u with hd graphics 620

The one I tested was an i3 ivy bridge or something like that

02:51

Ash yeah 620

02:51

I think Louis has 6xx too

Image attachment

alienouis

1

nou

2

03:00

justatest

Hi! sorry to disturb with that again, but looks like nobody really understand my problem, tho i'd love to be able to fix it myself I have no idea how :

I'm on macOS, using the official DDNet from the website.

So, the problem is about the skin fetcher, the downloadedskins folder really gather the fetched skins from the database, but the problem is that i don't see those in the client at all. Neither from the skins menu and ingame. I could totally just download the database and put them in the data/skins folder, but i'd have to do it at each little update ddnet releases In Windows I know that it behaves correctly, so the problem is really on mac Thanks for reading

07:05

Image attachment

07:06

(the downloadedskins folder from the "Application Support/Teeworlds" folder, that have the downloaded skins)

You are not supposed to see them (edit: in menu) iirc. Download the skin your want and put it in skins folder instead of downloadedskins (edited)

08:20

The downloadedskins folder is for downloading other ppls skins when you encounter any. It does not download the entire database.

but you should see the skins if someone is using them ingame. Is that working @Neben ?

08:27

maybe someone with Mac can reproduce it, but I remember it worked for @Learath2

oh he said in game

08:28

my bad

@Neben can you run client from command line and check console output?

when your merge with ddnet upstream fails because the "save_replay" (demo replays) is already defined in your codebase :justatest:

Can I just replace the latest .exe in my old folder to get the update?

That can break things

09:35

You should always update everything

Bruh

i redownload the .app when i want to update and it works fine for me

Nice, but broken shaders or missing textures won't fix it

[ddnet:master] 2 new commits

090b094 Stop embedding rgb images - Patiga a050370 Merge #3197 - bors[bot]

09:50

[ddnet/ddnet] Pull request closed: #3197 0.7 map converter - Stop embedding rgb images

@deen Ya, here it is : https://pastebin.com/jTiNngXJ

[2020-10-27 11:34:28][engine]: running on unix-macosx-amd64[2020-10...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

Idk if this was already suggested, but what about adding a button to launch a local server and some gui for its config?

There is a button for the server

11:06

just no gui for config

[ddnet/ddnet] Pull request opened: #3200 Use IntsToStr function from gamecore.h

was copy pasted

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 if it works standalone, system.c especially
[x] Considered possible null pointers and out of bounds array indexing
[x] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addr...

Embedded image

11:27

lmao

[ddnet/ddnet] Pull request opened: #3201 Map optimizer

Wrote a map optimizer(very aggressive, explain later why). It reduces map sizes by ~40% (optimized the whole ddnet-maps repo and it was around 40% smaller afterwards) Test and idea by mind, credits to him What it does, and why I called it aggressive:

It removes all unused tiles from embedded images
It clears transparent pixels and dilates them as 3D textures, instead of simple 2D textures.
It uses zlibs best compression

So yeah, it changes the output map(keeps the source ...

We have a small style check problem in main branch @dev

@Deleted User r u still down for that language ui thing?

what language ui thing xD

omg

11:45

we discussed it yesterday

u want to write it?

xd

yes

11:45

the .teestyle xD

we need to discuss this a lot before doing anything like this

[ddnet/ddnet] Issue opened: #3202 RFC: A DSL for generating User Interfaces

To know what DSL is: https://en.wikipedia.org/wiki/Domain-specific_language This came up one day on discord. One of the ideas is to create a specific language that compiles to C++ teeworlds code, so it's not interpreted and thus it has no overhead. Here are some resources: http://dinosaur.compilertools.net/ Since we are already using python, we could implement it with the PLY library. We should also use relative units instead of absolute ones, we ...

is there no minimal css3 + html lib?

its lot of work and if its not wanted and well thought it will be bad

11:46

here

11:46

a issue to discuss

11:46

justatest

i'd prefer a simple css + html lib, bcs thats easier to use xD

then say it there

well

11:47

lets say it like this

i prefer a focused language

if u really interested in it

11:47

i'd defs be interested in ur impl

xd

but dont want to plan style language just so that nobody ever implements it bcs its too much work xD

11:48

or do u want to compile it to the current ui code?

i am interested in making a language

11:48

so this is a interesting project

11:48

xd

11:49

@Deleted User to current ui code was the idea

11:49

but maybe modify the ui code to make compilation easier

11:56

@Deleted User fix style in ur pr nobo monkalaugh

monkalaugh

11:56

src/game/client/render.cpp:180:43: error: code should be clang-formatted [-Wclang-format-violations]

guess who created that error @Ryozuki xd

11:56

https://github.com/ddnet/ddnet/commit/05d6201a72579d73547e79af0ed4773a19cdbf86

xdd

but yeah, theses whitespaces.. always the same xd

[ddnet/ddnet] Pull request opened: #3203 Fix style in render.cpp

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 if it works standalone, system.c especially
[ ] 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-addresssanitizer--u...

12:19

[ddnet/ddnet] Pull request closed: #3200 Use IntsToStr function from gamecore.h

12:19

[ddnet:master] 3 new commits

cce9174 Use IntsToStr function from gamecore.h - Patiga 1253215 Fix style - Patiga 333218e Merge pull request #3200 from Patiga/pr-remove-duplicate-function - def-

[ddnet/ddnet] Pull request closed: #3203 Fix style in render.cpp

12:27

[ddnet:master] 2 new commits

1347537 Fix style in render.cpp - def- dd7da19 Merge #3203 - bors[bot]

@deen i'm confused, the ddnet download page still has graphic tools with borderrem/add etc. but your commit removing them is tagged as 15.1.3, is the zip on the website outdated?

hi guys

15:41

(save the best demo of each race) this make record automata all time right? but if you die or not finished map. would lost the record , some time i get good shot but because me not finished the map this useless

Image attachment

yes

2

15:41

so my suggestion , put any method for save the save automatake even if you die (edited)

Yes that would be dope.

16:10

Or at least some button with save anyway or something. And not just drop the demo with death

we have 2 options (edited)

16:13

auto demos and replay demos

16:13

thats enough

@Deleted User yeah, I haven't updated that in a while

oh ok didnt know they updated manually

@hussainx3 @Luca save_replay seconds in F1

17:02

And you have to enable replays

ah I didn't know it was possible yet

the exec myServerconfig.cfg doesnt work for me

[ddnet/ddnet] Pull request opened: #3204 Much more readable Call Vote menu and code

I find "Call Vote" menu useless af right now, because it's simply impossible to read it. So when you want to call vote for a map you open website and find the one you want, after that making /map skipping the "Call Vote" menu completely. It's useful for calling for random map btw, but anyway too ugly to look at. OLD LOOK !callvoteold NEW LOOK ![callvotenew](https://user-imag...

merge #3190 pls

https://github.com/ddnet/ddnet/pull/3190

Fix chat message margin and background rounding by Banana090 · Pull...

Just the same as #3165 but from new branch and with one clear commit. I will not make any new messy PRs and close it anymore, sorry for that. Thanks for ChillerDragon for helping a noob like me wit...

why

18:06

it still collides with my pr

it's becoming too old, I want to delete this branch and update everything to match master xD

u dont need to touch that branch

18:07

u can also just close the pr, bcs my pr changes the margin again, we just havent decided yet what we want to use appearently

Image attachment

just do git checkout -b new_branch and ignore the other one

this is like 1000000 commits behind now

it doesnt matter xD

18:08

havent deleted a branch since like 2 years bcs im too lazy

justatest

1

terminal master race

Image attachment

oh its christmas already

wot

18:17

its halloween

looks like a half christmas tree

true

18:17

xd

18:24

@Deleted User this one is even better

Image attachment

18:24

git config --global alias.lg "log --color --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit"

18:24

a alias for git log

18:24

"git lg"

too much % for my brain

thats why u use alias

Where can I check if my static variable is true and if so - DoButton?

19:17

Needs to happen every frame in CMenus

you know you can google a bit

Google for ddnet code

19:19

To know where u fkin call frame logic updates

you should learn more about static variables

Ofc

they not only exist here :)

@Дядя Женя if(variable) DoButton() ?

Am I really that dumb in english?

19:41

I wanted to know where you do all the input logic

19:41

Already figured it out

you really make people want to help you

But to make it clear: I've created Color Picker. The code works that way: 1. You call for something like RenderPicker(pColor), it renders simple Square with the color you provided 2. This method checks, if you pressed this square. If you do, it makes static Color Picker be "linked" to this color and turning Color Picker on, if it wasn't enabled 3. Here I was when asked the question Now I should check every frame if the Color Picker enabled and if it is - draw it for "linked" color

19:46

Didn't really make it clear, but whatever...

19:46

wait for the best color picker in ddnet client

19:47

and ban Ryozuki

color picker for what

everything. For colors?

yeah but which colors

19:48

for tee it makes no sense

Something like you have in editor

19:48

for tee it makes no sense

@Deleted User why?

bcs it clamps the light value

?

with HLS values u can modify colors differently than wiith RGBA

19:49

and it clamps the light value there

Color Picker is hsl, isn't it?

depends on what u call a color picker

19:50

but normally its rgb (edited)

Image attachment

then make sure for tees that the values that cant be selected arent even shown

19:51

else it just confuses ppl

Does HSL have different color range from RGB?

thats not the point here, the problem is tee colors are clamped

the color picker crosshair in the editor is hard to see when its on greyish values btw

Crosshair is shit, I'm making my own color picker anyway (edited)

19:53

If I'll do it good enough, I'll apply it for editor I think

19:53

thats not the point here, the problem is tee colors are clamped

@Deleted User oh, from 0 to 100...

19:53

To not let it be black and white

no

I see

@Дядя Женя e.g. u cant have black skins

i think half of the lightness values are not allowed

but learath is the color guy, he knows all the HLS tricks

It just have slightly clamped range

19:54

Should be easy to fix

ye

I'll try to just scale that square of picker

19:55

I mean... Doesn't matter, thank you for pointing it out

Image attachment

20:38

ew hsl

Now I should guess what colors refer to what values in vec4

wait thats not hsl

And now I just realised it uses RGB, so should make new method for drawing HSL nice

Embedded image

21:25

Embedded image

nice

21:27

I knew which part of the code you were talking about before looking at the code snippet you posted

the server list doesnt load for me

22:29

ah

22:29

restart fixed it

Exported 356 message(s)