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 2020-09-14 00:00:00Z and 2020-09-15 00:00:00Z

What do you think of implementing permamute feature? Some kid just rejoins his dummy and gets unmuted each time he reconnecta it

its add_foe

In console?

yes f1 add_foe name

And how UI muting way does work?

idk it resets

@Deleted User @TsFreddie ah, I think I know the reason. DDNet displays the text with a shadow

ah. but shouldn't it stored the uv of glyphs in the first pass already? so the second pass shouldnt call renderglyph again. sounds like a bug.

05:40

if you are talking about getchar then, ye that make sense.

But isn't the shadow larger than the original?

05:40

yes, about getchar

the are processed in one pass

05:40

in renderglyph

oh, ok

with that grow function

https://gyazo.com/a136f3cd58ee5c16ae79c6c99a8accf7 Doesnt show teamranks after finish

Image attachment

[ddnet/ddnet] Issue opened: #2822 Entities text numbers bug

Image attachment

10:55

freetype stroker for outline is definitely faster

10:55

Image attachment

But if something you changed makes outlines cost more, then switching to a cheaper algorithm to create the outlines is only masking the performance degradation, no?

i think the performance drop recently is a separate issue

10:59

i'm just saying the Grow function should go.

Probably, but be careful not to forget that there is another issue

i mean i'm working on vanilla rn EeveeShy

EeveeShy

11:01

i can port everything i find back to ddnet later

11:02

nvm, i'll take a look at ddnet rn. since it is pretty bad on eki's machine.

Traitor

catGun

thonk

11:03

aren't you the one supporting me rewriting everything in the vanilla discord

is a double agent

it is coming along pretty nicely. it barely drop any frames when rendering dbg_tuning now, which has quite a lot of text

Nice, how much are you caching?

haven't rewrite chat and console yet so we'll see.

11:04

depends. I can't cache anything on UI

11:05

every static text is cached on HUD (edited)

11:05

but immediate mode is already much faster than before

Why can't you cache anything on UI? I'd guess that's where you'll have the most improvement

11:06

Chat, console and menus have the most cacheable text

UI uses a DoLabel wrapper. The way i'm caching quads are just use static CTextCursor. so it can be reused.

11:06

chat console should be fine with caching.

11:07

Menus needs more work to do the cache stuff. but i think it is not the top priority since you usually don't look at menus when playing. (edited)

11:08

And since you have access to every quad position after a text call. Text selection should be easier by just searching quad position in relation to cursor

Can also optimize grow function to be only called once

just use stroker

Did you see my text selection patch for vanilla? I don't really like it at all, feels soo bolted on

you don't need grow anymore

And have u any resource what freetype doing there

what have me resource

What the performance impact is

all i know it is faster than grow in profiler

Yes but what when grow is only called once

but you'll need to make a better convolution algorithm tho?

But the text outline is barely visible

larger text need larger outline

11:13

that's why grow is called multiple times

Yeah

if you only care about chat then ye. you can just call grow once

I know but u know how much it grows

11:13

So I don't see a problem

@Deleted User I'd be very surprised if a text library used by thousands of products is doing worse than us

i don't think a convolution works like that

Optimize it...

11:14

Not leave as is

Why bother optimizing it when it's already optimized for us and in a library that we already link to

^ ths

11:14

besides freetype stroker produce much cleaner outline

@Learath2 if it's interpolating..

thonk

we can do both. and see which is better. but i'm not optimizing that shit personally tho.

I don't get why that's an issue

anyway other than outline, there should be another issue. i'll try taking a look rn

@Deleted User I'd be very surprised if a text library used by thousands of products is doing worse than us

@Learath2 btw, just yesterday we found out that FT_NEW_FACE keeps the file open on windows. so it trades execution time for 16MB of RAM

11:30

so its not really an argument, only math is

keeping file open doesn't mean it is taking memory space

yes

11:31

thats what i said

11:31

it rather keeps the file open instead of loading it

thonk

but fact is, accessing HDDs is insanly slow

Are you sure? That sounds quite stupid

https://github.com/ddnet/ddnet/pull/2818

Load fonts from memory (fixes #2810) by def- · Pull Request #2818 ·...

Missing: Freeing the memory again. But not so important since we keep the same fonts until end of process anyway in our case. @QingGo Could you give this a try from the Github build artifacts? If n...

should i give it a profiler treatment?

yes, you can do it

11:34

but i doubt freetype, used by thousand of ppl, wouldnt do somekind of binary search to find its keys XD (edited)

11:34

so the chance isnt small that it actually loads from HDD causing over 1 second of lag

oh right i don't have a hdd

11:35

nvm

11:36

thonk

but wait

11:36

what do you mean binary search?

11:37

font have charmap. the encoding map should be loaded. and you get the glyph index out of the charmap, and it should be able to just seek once and pull the glyph

11:38

but again, pulling multiple glyphs at once might be slower. but i doubt it would seek the disk more than once to find a glyph

11:39

Truetype and Opentype containers aren't designed to be stupid doggo_lol

doggo_lol

yes

11:39

i know, but why would such a thing cause 1second lag?

it is confirmed?

11:39

by eki?

yes

11:40

he even posted a video

11:40

Only when I receive message full with CJK characters which had not encountered before, I feel lag. Repeated CJK characters message didn't make me lag. I upload a video to youtube https://youtu.be/et_SsWyCd5U?t=27 . The lag is begin at 0:26.

>

And I found maybe most of the older player use old version of DDNet Client, so they did not encount this issue. When most of new player use steam client, they may think it is normal behavior.

lag when people send long CJK chat DDNet client

i know about the lag

11:40

is the fix confirmed?

11:40

to solve the problem?

no, but else, freetype must be really slow finding its glyphs

11:41

and i mean realllyy slooww

you can't just say freetype causes 1 sec of lag when you haven't fully checked your own impl.

i know that the impl is doing a bit more since fallback

11:41

but that would never explain such a huge lag

11:42

and then again, why would a freetype api call be so costly

are you sure there aren't other things holding back the performance in the fallback code?

as i said

11:43

at worst it loads a char twice

Without profiling you are all just making wild guesses

but that wouldnt cost 1 second of lag

but eki doesn't see any lag pre fallback. aren't the freetype calls the same?

yes but the font was like 1mb

11:44

not 16mb

11:44

iterating thorugh it is obviously slow

thonk

and otf is often even smaller than ttf

11:45

so 1mb ttf vs 16mb otf

There is no way they iterate through that many glyphs

no but accessing the file on a HDD is still a problem

11:46

even if they binary search over it

Yeah, let's see if loading it into the memory helps

if its not, we cant use that font probs, bcs then its even slow on memory

11:47

but even with 64k glyphs it should have max. 16 iterations to find the glyph

thonk

but other software using freetype don't have that problem?

well either they use memory fonts

11:48

or u just dont notice it

you know what. i'll nuke the grow function rn and sent it to eki as well

We can also try memory mapping the font file

11:49

Grow doesn't look like something that would cause spikes though

ofc not

not if you grow like 100+ times in one frame

nothing inside a cpu/ram should cause such lags

11:49

it really needs to be a heavy algorithm

but grow is literally O(wh)

grow is slow

bitmap isn't that small

but it iterates over like 50x50

11:50

thats nothing

how about we try it anyway

even if u do it 10000 times

or are you telling me not even bother

i think a rewrite is nice @TsFreddie

Well this kind of spikes happen with sub optimally growing vectors e.g. so you can get it to happen even on the cpu

bcs every wasted performance is still wasted

11:51

@Learath2 yeah but by freetype? xd

thonk

I'd look at our code before I look at theirs

our vectors built just as before

11:52

the only thing changes is, that it loads from 2 freetype faces and that it uses a different font

Revision: I'd look at the profiler output before anything

the profiler says RenderGlyph spend half of its life time on Grow yesterday

11:52

at least on mine impl.

11:52

i'll profile ddnet rn

@TsFreddie yes its 100% slow function, and it should be optimized, i totally agree

11:53

you can do what it does in one call

I literally cut mine from 28ms to 7ms by replacing it with freetype.. but sure if you want to rewrite the thing then do it.

11:54

i'm not spending that time rn

the only thing i fear is, that freetype is doing too well

11:54

we dont need cubic outline

I just want to see if replacing it with freetype would help eki

11:54

if it does then yeah, you can optimize it better

11:54

if it doesn't we look elsewhere

11:55

is it that unreasonable

11:55

or am i being dumb

[ddnet/ddnet] Issue opened: #2823 Border button broke my map frequently

If you not changed layer and pressed on a place where tele/switch number picker stays then you are fucked (because Border resides in this place). I don't think this button is usable nowadays. Please add confirmation message at least or move this button to less frequently used place.

@heinrich5991 wow, you're using wireshark for teehistorian. How cool

Does anyone know FuN~EleFaNt and have some contact details of him?

[ddnet/ddnet] Issue opened: #2824 [Feature request] Editor fill with the tile

Similar to Windows Paint (or other image editor) fill feature (https://youtu.be/tAcRRD9OOqg?t=33) It should fill bounded by unhokable and hookable tile area with a tile from the brush

b3z, do you want to get battlefield mod sources from him? (edited)

@Brokecdx- it takes an hour to refresh

kk

@TsFreddie @Learath2 @Deleted User He confirmed that it's faster when we load it from memory

@TsFreddie so if u want to load from memory for vanilla too

thonk

already did

13:29

@BannZay do u want to fix the entity text?

yes it seem much more better than before (edited)

Image attachment

13:43

is this clear what it is doing?

13:44

i don't want it to be too confusing, since textrender does work differently from before.

@Eki how much better? Is there still lag?

"Silky smooth"

Image attachment

13:48

~~DM leak~~

Nice, we should probably memory map the large file

I prefer reading it in once over mmap

13:50

The most important thing is that we don't start doing disk IO during gameplay

13:51

16 MB in memory is not too bad

Hm, I don't think mmap would lazy load it either

Isn't that the whole idea of mmap?

Ah, it does by default, I was thinking of MAP_POPULATE

13:53

anyway, 16M isn't too bad, malloc should put it somewhere sane too

can we document what is required for using clang format

13:56

i mean, what to execute

13:56

"python scripts/fix_style.py --base HEAD"

13:58

@TsFreddie the funny thing about grow is, you dont even need to change alot xD

13:58

                for(int sy = -OutlineCount; sy <= OutlineCount; sy++)
                    for(int sx = -OutlineCount; sx <= OutlineCount; sx++)

13:59

all i did is passing the outline and changing the check

13:59

still looks normal to me

thonk

@Deleted User yeah, I think when something is wrong, there should be a big fat message telling the developer how to fix it automatically. @heinrich5991 could you take care of that?

14:04

@Learath2 can we just use your warning popup to display all unexpected io permission problems?

I didn't put any limitation on it, idk how jupstar ended up changing it, but it shouldn't be too hard to make the text change

Yeah, I used it already for something

it only displays ingane i think

14:05

oh

I used it in menu I think

ok then it just checks if no other popup is blocking

I used it for "Server executable not found, can't run server"

14:05

in the start menu

Though, popups should probably be a layer on top of all other layers and not part of the menu

@Learath2 also, can we get https://github.com/ddnet/ddnet/pull/2819 in? Otherwise there are no nightly builds

Fix compilation on MinGW Windows by def- · Pull Request #2819 · ddn...

graphics.h:197:36: error: ‘uint8_t’ has not been declared

[ddnet:master] 3 new commits

be90f3f Replaced tile names. - gerdoe-jr a67a8e3 clang-format: fix style - def- 3df1407 Merge #2815 - bors[bot]

14:08

[ddnet/ddnet] Pull request closed: #2815 Renamed tile names.

@BannZay do u want to fix the entity text?

@Deleted User yea, I will try if none started to work with it

i can also do it

14:15

just fear to run the style check xDDDDD

okay then. I guess you'll done it in a better way

14:20

than me (edited)

14:22

why wouldnt we run script on ddnet solution to fix all style issues ? Beacause it leads to big amount of merges? (edited)

i think it changed around 15k lines and few things didnt look well

^^ It creates a massive diff

#2791

https://github.com/ddnet/ddnet/pull/2791

Add .clang-format; require it for pull requests by heinrich5991 · P...

This will help me not getting annoyed about code style.

Not sure if it's better to do it in each change though

14:24

Now it becomes impossible to find the actual change inbetween the style changes

yes

1

@deen you can ignore whitespace changes on github

so it should at least be done in separate commits

just ignore style check xdd

not optimal but you can

14:26

It's either this or we just don't enforce any styling

[ddnet:master] 2 new commits

2c5f01a Fix compilation on MinGW Windows - def- 3392933 Merge #2819 - bors[bot]

14:26

[ddnet/ddnet] Pull request closed: #2819 Fix compilation on MinGW Windows

or do it one large change

fix_style should be run only on new version tag imo (edited)

fixing 1 by 1 annoys a lot of peolpe because they have to fix things they didnt even touch. Just to satisfy build

https://github.com/ddnet/ddnet/pull/2820/files?diff=unified&w=1 if you omit whitespace changes, it looks fine

14:27

We can do the one massive commit if you prefer, that'll kill git blame though

In this case yes,but others no

14:27

git blame supports -w too

Oh, TIL

and maybe you can even somehow tell it to exclude specific changes

14:28

yes, --ignore-rev <rev>

I guess we can just do the entire repo

and even --ignore-revs-file <file> for a list of revisions to ignore

There were a couple issues with the formatting that should be fixed though

I'd give it a try later today to reformat everything and try to keep the changes nicer

Oh btw, I was wondering the performance of conntrack, but it was all too speculative so now I'm coding a conntrack module for tw

@heinrich5991 opinion? ^

14:30

@Learath2 very cool!

I fear it will do more harm than good though

without the large packet?

Yeah, just the current protocol we support

#2792

https://github.com/ddnet/ddnet/issues/2792

Use Github actions' checkout's submodules · Issue #2792 · ddnet/ddn...

teeworlds/teeworlds#2724

The trojan horse they make you install to have remote exams messes macOS up

-                                               size_t GlyphOffset = (OffY) * SlotW + OffX;
+                                               size_t GlyphOffset = (OffY)*SlotW + OffX;

14:41

why does it do this xd

14:41

makes it wrose readable

[ddnet/ddnet] Pull request opened: #2825 Fix entity text & improve Grow

fixes #2822 @BannZay wanna check?

@Deleted User always same number of spaces between operators

and why did it remove the space then?

maybe it thinks ) is enough or counts it as a space

xd, its just strange xD

(OffY)*SlotW OffY*SlotW i guess it finds the former readable xd

Huh, is that a cast?

no it probs was a calc inside the () some day

Rather, is it confusing it with one

14:58

It's impossible to control the spacing between operators

() mean cast? I wonder why result of '(Flags) & CFGFLAG_COLLIGHT' differ from 'Flags & CFGFLAG_COLLIGHT' (edited)

it only means cast, if the text inside the paranthesis is a datatype and right from it is some variable

1

For it to be a cast, Flag needs to be a datatype

sure it isnt (Flags & CFGFLAG_COLLIGHT)?

15:07

bcs & operator has a weak binding

15:07

so its safer to pack them into paranthesis

15:08

https://en.cppreference.com/w/cpp/language/operator_precedence see 10 and 11

15:08

!= would bind before &

static CColVariableData Data = { this, &gConfig.m##Name, static_cast<bool>((Flags) & CFGFLAG_COLLIGHT)

pls use

15:12

backsticks

15:12

discord messes ur msg

15:13

do we rly use { } to initialize structs tho?

15:13

i dont think so

You can, it's fine

ah well its static

15:13

yeah but doesnt fit the code

15:14

triggers me

15:14

xd

15:14

Oh I just noticed now, it's my code

why is there ##

15:14

is it some macro

15:14

?

Yep it's inside a macro

yes

15:15

so we need to cast flags to perform & operator for enum?

15:15

I feels stupid but its still unclear for me

We don't need to cast anything

15:15

Flags is an integer, the enum members are integers, you just & them together

so nothing should change if we remove () from Flags ?

it would

15:17

well it depends xd

15:17

but there could be more than 1 flag

15:18

CF_SAVE | CF_CLIENT what do i know and & binds stornger than |

yes

15:18

but in particular case it will not change? (edited)

if e.g. Flags is a variable (edited)

15:19

it couild be removed

[ddnet/ddnet] Issue closed: #2822 Entities text numbers bug

15:20

[ddnet:master] 3 new commits

8923116 Fix entity text - Jupeyy fbdd66c Improve grow - Jupeyy c0f7a2b Merge #2825 - bors[bot]

15:20

[ddnet/ddnet] Pull request closed: #2825 Fix entity text & improve Grow

thx for the explanations guys, there is no magic as I thought xD

@deen the mapconverter 0_7 should also detect invalid textures, with width of 0 or height of 0

15:32

appearently the client allows them to load

15:33

the client shouldn't even allow this

but how does someone even embedds such mapres lol

i dunno xD

15:34

but that is way too undefined behaviour even for non tile sets

15:35

malloc can return a pointer or null

15:35

pretty undefined xd

@Deleted User done

[ddnet/ddnet] Pull request opened: #2826 map_convert_07: Check for width/height 0

as suggested by Jupeyy

We don't chunk teeworlds packets, right?

16:07

Or rather, teeworlds chunks

16:07

Are chunks always transmitted in one piece is my question

hi, i have the rank from the test server on the new novice map (Inure). when I write /times it says “3 weeks ago” but the map was released yesterday

Hm, it's really hard to write a heuristic for teeworlds packets, almost any random string of bytes can be a teeworlds packet

@Learath2 it is hard to filter cuz udp is connectless and the filter doesnt know if the server considers the client as already connected right

16:27

are ddos made of packets before connection or after?

16:27

i guess it doesnt matter

@deen can i trigger bors btw for ur things?

That's not hard to handle, conntrack adds state

16:28

I'm working on a wireshark dissector too, but I'm having trouble deciding how to figure out something is a teeworlds packet

16:28

I could decode almost anything as a teeworlds packet

they dont have any header or stuff

16:28

?

16:28

xd

16:28

well if a packet consists of cnetobjs u can find the type and maybe match the size of any cnetobj

16:29

i may be talking shit tho idk how it works

We do but it's 3 bytes, first byte has flags and ack, second byte has ack, third byte has the chunk code

16:29

All of those bytes can be any value and the packet would be valid

16:30

@heinrich5991 how did you do it for your old wireshark dissector?

@deleted thanks for report, removed

Ah, he cheated, he only dissects for port 8303 :S

(and it's been long fixed, but was still a problem 3 weeks ago)

ah okey

️ ty (edited)

@Learath2 https://twitter.com/nvidia/status/1305290653761503232

NVIDIA (@nvidia)

NVIDIA to Acquire Arm for $40 Billion, Creating World’s Premier Computing Company for the Age of AI. https://t.co/Pj0lR8W03u

Retweets

4357

Likes

13130

16:55

is this good or bad

16:55

justatest

16:55

justatest

justatest

16:55

Image attachment

16:55

i would have prefered amd

eh, I would have preferred no one bought out ARM

yeah that

16:56

too

Multiple vendors is nice to have

16:56

Monopolies are very meh

but nvidia is the worst

They should give it to me

16:57

I promise to not sell it to anyone

@Learath2 is arm better than x86_64?

16:58

i guess it fits better for the internet of things

16:58

they cal

16:58

call

Eeeeeeh, not really easy to compare

16:59

ARM has a much saner ISA imho

Processors that have a RISC architecture typically require fewer transistors than those with a complex instruction set computing (CISC) architecture (such as the x86 processors found in most personal computers), which improves cost, power consumption, and heat dissipation. These characteristics are desirable for light, portable, battery-powered devices‍—‌including smartphones, laptops and tablet computers, and other embedded systems[3][4][5]‍—‌but are also useful for servers and desktops to some degree. For supercomputers, which consume large amounts of electricity, Arm is also a power-efficient solution.[6]

17:00

$40 Billion tho

17:00

companies have so much money

17:00

insane

17:00

well its american billions

Well, most of it is just given as stock of their own company, not in cash

american billions are thousand millions here iirc

ARM isn't strictly RISC or CISC

arm literally means Advanced RISC Machine

17:01

where do u get that from

Well in thumb mode arm processors can do variable length instructions, which isn't really RISCy

if u want modded minecraft with 60fps u need intel 20k clocking at 8ghz, if u want blender at 5 watt, u need arm xd

that depends on the mods

no thread optimized ones

well i mostly play those kitchen sink packs that are in a void world

17:04

so xd

Apple makes some very strong A64 processors

yes, will be interesting

17:04

but i only believe what comes from many independent sources

Strong enough that they were atleast considering making the next macbook arm, not sure if that'll happen tho

@Learath2 why u use a mac and not linux tho

17:05

do u se some propietary software?

17:05

or u just had a mac around

Kryo 585 is an octa core with a whopping 3.1ghz, it can even do fancy speculative execution

3.1 isnt much tho

@Ryozuki I had a laptop that was dying, someone offered to get me a macbook if I coded them an iOS app, I took him up on the offer

17:07

When I'm back home I'm on my desktop anyway, so this is my portable solution for now

@Learath2 u cannot install linux there? xd

arm founder says that its bad @Ryozuki

Kryo 585 is a phone sized arm chip, fits in a fraction of the actual cpu of the phone even

ah thats nice then

@Ryozuki I can but space is quiiite tight (ssd not upgradable

)

eh

17:09

linux should take less space

17:09

if u completly wipe mac tho

17:09

xd

17:09

u probs can get more perfomance too

There is some software I still need on the mac side

17:11

darktable is very meh compared to lightroom and adobe doesn't feel like supporting linux at all

frozen

17:11

yeah adobe sucks

17:11

gimp best

though it is quickly getting better, maybe next year I'll be able to hop over

lol

17:12

have u tried krita

gimp is no replacement to darktable

https://krita.org/

Krita is a professional FREE and open source painting program. It is made by artists that want to see affordable art tools for everyone.

I'm okay with gimp as a replacement to PS tho, it works great

gimp is not a direct replacement of PS

Lightroom is for photography

well maybe yes

17:12

xd

17:12

but krita has more stuff

17:12

than gimp iirc

I don't illustrate, so I doubt krita's features of much use to me

17:13

I actually wanted to learn how to draw for quite some time, it's just that it seems to take a whole lot of time

17:13

I get demotivated and unfocused sooooo quick, it's really hard for me to pick up new things

if u want to animate, blender is powerful

17:14

they improved that special pencil they have

Greasepen?

17:14

I use blender a bit, but mostly for modeling

https://www.youtube.com/watch?v=lriYv4P4TcE

BLENDER 2.9x - NEW FEATURE - 2D FROM 3D!

17:14

this looks awesome

Blender 2.8 brought so many new things, I'm expecting greatness from 2.9 too

2.9 is out already

17:16

https://www.youtube.com/watch?v=7M1UaDe0qsk

Blender 2.90 New Features in LESS than 5 minutes

17:16

greenthing

17:16

breathtaking

Heh, if I ever get back to italy and do more modelling, I can move to 2.9

17:17

Extrude manifold looks awesome

17:22

Another great looking release

17:23

I love how blender started from "unusable", became "competent", became "usable" and now it's becoming "great"

17:23

It's an open source success story

yeah

17:27

feelsbadman

Image attachment

17:29

"this will give nvidia and stronger mobile presence"

17:29

xD

Why is it ARM that's taking off though

@Learath2 tho ppl are saying arm sold cuz they are not taking off anytime soon

17:30

even worse they falling

We need more work on RISC-V

17:30

A simple ISA is soooo much more fun to work with

[ddnet/ddnet] Pull request opened: #2827 Update russian translation.

Now that I think about it, I think I agree with freetype being flexible. They should have documented the fact that the font isnt loaded though

thonk

Since they gave you the FT_New_Memory_Face, so they might think it is obvious the other one isn't "in memory"

18:09

but ye, they should probably say that some where in the doc

@deen don't like the "one change", especially since we're probably missing something, and it would all need to be reformatted again. I'd prefer to make it easier to fix the style

18:19

@BannZay the fixes for stuff you didn't even touch, is that mostly the binary operators in the config variables?

18:20

@Learath2 not all flag combinations are possible, I think. not all packets can contain a payload

@Learath2 have u heard about https://gpuweb.github.io/gpuweb/ ?

18:22

an example implementation https://github.com/gfx-rs/wgpu

Native WebGPU implementation based on gfx-hal. Contribute to gfx-rs/wgpu development by creating an account on GitHub.

Hadn't yet, looks quite intriguing

@Learath2 from what i understnad, its a api that abstracts metal directx vulkan and opengl, it even has its own shader language spec

18:23

https://gpuweb.github.io/gpuweb/wgsl.html

@BannZay the fixes for stuff you didn't even touch, is that mostly the binary operators in the config variables?

@heinrich5991 mostly yes. (here is another couple of style issueshttps://github.com/ddnet/ddnet/pull/2778/checks?check_run_id=1108826164)

Whats up with the horrendous opengl support?

u mean in the implementation?

Ye

idk

18:24

bear in mind this is new stuff

18:24

xd

What happened to open source brotherhood?

but i guess they dont prioritize it since there is vulkan

18:24

whats open source brotherhood

What is all this directx support triggeredField

triggeredField

justatest

Remove directx

well its unrealistic to not support it

18:25

many gamedevs think it is better

18:25

than opengl

18:25

do u have nitro btw

18:25

who is gifting them

18:25

justatest

I was helping someone with some ddnet thing and he gifted it

justatest

Image attachment

oh

18:26

they say thanks to u?

18:26

i only met rude ppl asking for help

18:26

"hey you, do this for me!

18:26

right now!

18:26

justatest

I help make bots

banhammer

3

@BannZay I guess we can reformat these config files. that would conflict with your PR though, so either you do it, or I do it and you have to fix the conflicts

troll

18:27

what kind of bots

18:27

@plsplsplslol learath closet weeb

ryozoozki out of closet weeb

i only manifesty my weeb against weebs

18:29

@Learath2 https://bevyengine.org/ look how shiny and cool it looks

Bevy - A data-driven game engine built in Rust

Bevy is a refreshingly simple data-driven game engine built in Rust. It is free and open-source forever!

18:29

but sadly all these things are early in dev

18:29

feelsbadman

Data driven game engine sounds like something I would hate

@Learath2 it means using a ECS mostly

but sounds like something you want to have for performance, I think(?)

18:31

I think unity also shifts into that direction

Oh then, I might like it

18:31

I feared it might be some declarative bs

amethyst rs is also good but they are changing stuff drastically rn

18:31

so not a good option rn

why client doesn't render textures when gpu v-ram is small?

Image attachment

what client version r u using

19:16

@m!ki

latest one

that doesnt look like a vram problem

i have no problem with other versions

19:17

like 10.x.x

yeah

19:17

set gfx_opengl_major 1 in f1 and restart

19:17

still has nothing todo with it

19:17

its bcs ur driver is buggy

19:18

normally the new client downgrades your gl version automatically, but it seems ur driver is just lieing

my bad, it was 14.0.3

19:19

but with latest client crash

19:20

while loading it crash

have u used the autoupdater?

i compiled client by myself

then see where it crashes

19:20

would really help

Image attachment

ty, seems like the sample function is buggy

19:23

can u also give me the stacktrace?

19:23

the function caller list

Image attachment

19:24

this one?

ty alot

19:26

@m!ki btw what does gfx_texture_quality say in f1 in version 14.x

0

ah

19:26

set it to 1 and try the new client

19:27

i think its the because the sample function only works with specific texture sizes

19:27

maybe i'll just remove it

that's working now

19:27

ty

[ddnet/ddnet] Pull request opened: #2828 Use resize instead of rescale

rescale doesn't clamp, and looks worse, might be slightly faster tho. So i also renamed the config vars that might cause resizing and/or are not very useful

@BannZay I guess we can reformat these config files. that would conflict with your PR though, so either you do it, or I do it and you have to fix the conflicts

@heinrich5991 I can do it. But I guess its better to run script on repo to prevent people from fixing things in other files and copying wrong style. It was big problem to me earlier. I copied the style that I thought was good from the repo, but later some say that I should fix the copied style

[ddnet-scripts:master] 1 new commit

cdf074f Some ddnet-setup.sh fixes - def-

[ddnet:master] 2 new commits

35e93df Update russian translation. - gerdoe-jr 016e52c Merge #2827 - bors[bot]

22:02

[ddnet/ddnet] Pull request closed: #2827 Update russian translation.

22:09

[ddnet:master] 4 new commits

9a4dd80 Remove rescale function, always use resize - Jupeyy 3eea399 Rename quality texture and compressed texture, because they should not be used - Jupeyy 8b29240 Use POT texture when resizing, when NOTP is not supported - Jupeyy 393f3de Merge #2828 - bors[bot]

22:09

[ddnet/ddnet] Pull request closed: #2828 Use resize instead of rescale

Can anyone assist please

/DDNet: error while loading shared libraries: libSDL2-2.0.so.0: cannot open shared object file: No such file or directory

which OS?

Ubuntu 18

Can you do an ldd ./DDNet?

22:11

Actually, how did you get DDNet?

It lists a whole lot of packages, what am I looking for

22:12

from the website but think I got the linux version since I don't see a ubuntu version

@deen idk how we ship the linux release, do we include libSDL? or should he install it?

install

soooo, I guess sudo apt install libsdl2

I guess I need the Ubuntu versions not the linux?

22:22

Unable to locate package libsdl2

hm try sudo apt install libsdl2-2.0-0

I tried that and seemed to install with unmet peer depencies and fixed using sudo apt --fix-broken install, but still getting same error

can you post the output of sudo apt install libsdl2-2.0-0?

can you post the output of sudo apt install libsdl2-2.0-0?

@heinrich5991 Working now thank you very much, mb had to run the command again after it failed with unmet peer dependencies thanks my brain is dead

nice

Why don't we ship sdl2 library inside instead of manually installing

the idea was that most people will have it installed

22:31

if you played any sort of game before, it should be installed

@Deleted User yes I would like to. Or at least some binary

@heinrich5991 I see that makes sense

did a fresh install and first game I installed is DDNet offcourse since it's the only fun game that exists

I guess flatpak is kinda there to solve this issue

doesn't ubuntu ship with flatpak?

I wonder if there is a nice way to solve this issue with linux, without shipping all the libs

well, we could ship all the libs, and prefer linking to system libs

Something like an installer that can get the package manager to pull in dependencies

the only reason not to do that would be that it increases the download size

How do we even prefer that btw? Is there a linker option?

I'd guess it searches the rpaths in order

Eh I guess we could have /usr/lib as an rpath, but how portable is that I wonder

ah

22:54

no default paths when we have rpath?

Hm, I'm not sure how it works on linux actually, is rpath just added to the search path?

22:59

On macOS, if you want to load a library from the rpath you need to link to the rpath, you can't have it both ways

22:59

Though if you forget to add a rpath entry it will insert /usr/lib there just in case

anyone have data on the oldest still-standing rank 1?

I tried getting it but too tired to SQL

@louis isn't that just the oldest rank 1?

yeah but still-standing so unbeaten

23:59

i mean like i guess technically the oldest rank 1 was the first finish but it's been beaten

Exported 592 message(s)