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 2021-05-06 00:00:00Z and 2021-05-07 00:00:00Z

Deleted User

but maybe ddrace is too fast pace already.. these mario speedruns are pretty basic and easy to understand

have you seen the block clips, wrong warps, pixel perfect jumps, and the 1-1 bonus they have to do xD.

[ddnet-maps:master] 1 new commit

fd80d87 A Gone - ddnet-maps

Image attachment

[ddnet/ddnet] Pull request reopened: #3804 HiDPI-aware Resolution List

Image attachment

07:48

i don't get it

07:48

my cmake is still messed up

did u try to make clean build dir?

07:50

or is that a normal macos error xd

07:52

if you refer to this error: https://discord.com/channels/252358080522747904/342454474117218334/839758984587771924 deen: i guess this is really bcs they just click unsupported resolutions, and select something insane?

07:53

or do you think it can help with this @TsFreddie ? your pr

07:57

i mean it showed them incorrect, but they werent invalid, were they?

@Deleted User i replied in the pr.

@TsFreddie btw just out of interest, do you actually get a lower resolution canvas then on macos? if ur have low dpi set?

08:03

like full hd on 4k screen

08:03

or smthj

at least my mac does that

ok

08:03

kinda interesting approach xd

08:03

but also a bit weird

ye, on windows is weird too

08:05

on windows the desktop width is lowered if you have hidpi off. but it does make a bit more sense.

or is there a difference in windows?

have hidpi off

I mean in the application property, application doesn't know anything about DPI in windows.

thought our hdpi awareness meta data just makes ur window always the same

08:06

or did u only test that on vanilla?

I rewrote the whole thing in ddnet. it is actually a different method.

08:07

only the issue is similar in vanilla

08:07

and our window size doesn't change indeed

08:07

screen size changes

08:08

there are two resolutions in mac os. You pass the window size during init and during resize, and SDL gives you the scaled canvas size back (which we store in m_ScreenWidth)

08:10

I remember seeing a thread somewhere that mentions the hidpi thing is implemented for wayland as well. but i don't know if they work the same as macos.

i find it kinda strange anyway that we need it

08:13

teeworlds is dpi aware in a sense that it always asumes the same height

08:13

resolution independent anyway

08:14

full hd to 4k should only improve image quality not really change anything

oh i didn't change anything to the size that we are communicating with SDL and stuff at all

yeah no i mean that u get full hd canvas on 4k monitor

the resolution list is displaying the wrong numbers that's it.

doesnt that imply that u have lower image quality

yes, if you don't have hidpi on

ah so its the exact oposite?

08:16

so macos reports full hd resolution but actually has 4k canvas?

ye

i see

wait.

guess that makes more sense

i'm a bit confused

i mean the desktop reports full hd resolution, but the opengl framebuffer is 4k?

08:17

thats what i mean

yes

ok

in ddnet we still sees the 4k

ok

but we need to pass full hd to initialize

08:18

which means if someone selects 4k with hidpi on, it goes boom

08:18

so the resolution list is wrong.

08:18

so i was fixing that. there you go

08:19

finally lol

yeah

Image attachment

and this can happen without ppl configuring it manually per app?

09:21

thats already insane

if steam is in lowDPI mode, like an auto update of steam breaks the setting. then steam may launch other games with the dpi setting overriden. Some of my app like QQ restarts itself in low DPI mode if it crashes

09:23

either way the game on the left will just think the desktop is smaller, it's the same thing with switch monitors. You can avoid it just by checking if the resolution is larger than the reported desktop size during launch.

puh ok

09:24

and then windows acts like u told me before

09:24

with the screen size being smaller or smth

ye. mac gives you smaller window with larger canvas, windows gives you smaller desktop with canvas and window at the same size.

09:25

also, some laptop user may change resolution back and forth to save batteries. again, same thing.

its so random

09:25

why cant u just disable it completly xD

09:25

that behavior xD

the windows one?

yes

09:26

i mean SDL wise

because older apps can't scale

09:27

windows will guess which one is high and which one is low. so older apps appears blurry but not insanely small, and newer apps appears at the right resolution.

09:27

SDL doesn't know about windows scaling like you said

TsFreddie

ye. mac gives you smaller window with larger canvas, windows gives you smaller desktop with canvas and window at the same size.

but wouldnt this mean we don't explicitly have to care

09:28

windows always gives a valid screen

09:28

never an invalid

we don't cae

09:28

we care about changing resolution and switch monitors on windows

but the issue in #general is that the canvas is way too big

windows will happily give you window context in any size

ok, so just to be clear u on low dpi mode, set ur resolution to something high, so it looks fullscreen, then u change back DPI settings and then ur window is broken?

09:29

or the opposite

wdym low dpi mode

09:30

in ddnet? or the per app property set to "System"

yeah the steam thing

09:30

which influences ddnet

you were in Application "hidpi" mode, game sees 4k screen. you set a 4k resolution. it looks fullscreen.

ok i think i get it

something breaks the DPI setting, now it is in System "lowDPI" mode, game sees full hd screen, game still launches in 4k resolution, with a 4k window. you can see a quarter of the game.

so does this realisticly happen easily, or is the users input required to make this happen

if clicking update button is user input then it requires user inputs

09:33

but it can happen any time when ddnet isn't running.

09:34

maybe you launched it a week ago. something break during the week. you launch ddnet next week it will be broken.

ok

09:34

atleast this guy said he clicked graphic settings https://discord.com/channels/252358080522747904/745926398140612678/838576495827222560

we have people that have game launches like that for sure

i am just not sure if it acutally happens for so many ppl

09:34

ok

but i know how to fix it pretty easily so it wasn't a huge issue for us tbh.

so you want to force the "best" resolution?

09:35

or what is the plan

but i still think we should do a countdown revert on resolution change for your case, and a desktop size check for my case.

09:36

i want to force the desktop resolution if the configed resolution is larger than the desktop

ok

09:36

so we can basically also remove the fake modes?

09:36

i mean they basically uselss then

09:36

who has 8k screen xD

i don't know what fake modes are for initially

neither i do

09:36

they just existed

actually

i prefer if we just check it against the highest resolution possible

09:37

i mean u can set uneven resolution in windowed mode

09:37

i am fine with that

VideoModes are your monitors' (or GPU's) supported display mode right?

guess a mix of both

so fake mode might still be useful for getting a windowed mode with a size that isn't available on your monitor

but i guess only fullscreen modes

09:38

but why should that even happen xD

09:38

also u can just resize on windowed mode

can you do it in the last release already?

yeah

i can't

09:39

let me check master lol

well u should note on windows the window doesnt remove fullscreen flags

09:39

rest should work

09:39

i tested it on windows

09:39

ofc not in borderless

09:39

if u tested that

windowed mode, can't drag to resize

real windowed or borderless?

real windowed

press windows key plus up

09:40

should normally work

09:40

or windows key right

nope

09:41

maximum is not even enabled

09:41

Image attachment

wtf

09:41

did u disable that in windos or smth

how do I do that lol

ok u forced me to switch to windows xd

09:41

i'll test

i saw the code handling resizes, don't know why it isn't possible yet

09:42

been wondering that for a while tbh

Deleted User

ok u forced me to switch to windows xd

kek

I'm so sorry

Image attachment

09:44

oh

09:44

works for me

09:44

strange

09:44

is that really a driver thing

09:44

would kill me

09:44

i can casually resize

09:44

its a bit annoying as the mouse always gets into the window once i dragg, but it certainly works

09:45

and if i rem. correctly it was already in the latest version too

can driver disable window resize tho

i've no idea

09:46

i honestly thought this is all handled by windows itself

09:46

but e.g. windowed fullscreen also only worked in the vm

09:46

was already annoying enough

you need to set some flags to disable it explictly

09:46

SDL might have done that by default

really strange

maybe it is a SDL thing?

what sdl do u use, u can see it in the console

actually

09:47

one sec

justatest

[2021-05-06 11:47:28][sdl]: SDL version 2.0.8 (compiled = 2.0.8)

09:47

i use that version

why is gfx_resizeable a config

AA

09:48

yeah that need to be on xD

09:48

doesnt it default to true? xxD

i don't know

maybe u used an old client once?

Image attachment

09:48

no

oh

09:48

rip

09:48

why is it default off xD

old config rip

09:49

we can't easily change that now

we can

i think its cuz resizing was bugged somewhere?

default values are not saved

oh

Ryozuki

i think its cuz resizing was bugged somewhere?

i fixed all

macos too?

09:49

xd

i even respected macos thing yes

09:49

even tho it was never tested

09:49

i respected the canvas

i'll test resize on the size for macos then

yes

09:50

would be good to know

09:50

then we can defs turn it on again

but i think i'll test it after the resolution list thing

@Deleted User it caused troubles on i3 and other wms before

09:51

i always had to disable that

also SDL_GL_GetDrawableSize seems broken

it changes resolution randomly when i switch workspace iirc

on macos with HiDPI mode off (in game), SDL_GL_GetDrawableSize still reports super high resolution.

09:53

and if you changed window size and called glViewport, SDL_GL_GetDrawableSize will report low resolution again.

09:53

it is weird

nori

@Deleted User it caused troubles on i3 and other wms before

can you test that again?

09:55

just to be sure

09:55

i had a bug in KDE too before, but it was gone after i moved some code around between the graphic thread and the main thread

TsFreddie

also SDL_GL_GetDrawableSize seems broken

well sdl2 says u should use it xD

i know

09:56

but it is broken tho

so random

09:56

i hate randomness xD

I checked SDL flag, if HIDPI is not set, i used SDL_GetWindowSize for a fix.

09:57

but i don't know if that's jank

yeah and everytime we change smth in these things we have to test all OS

09:59

anyway, we can also #ifdef it for windows

09:59

it works good enough for windows

09:59

not perfect but ok

i don't think it matters on windows whether you use SDL_GetWindowSize or SDL_GL_GetDrawableSize

yeah i mean the resizable config

oh

10:00

then yeah.

and about the resolution clamp, i'd personally prefer we implement that

why isn't it a settings in Graphics tho

[ddnet/ddnet] Issue closed: #2885 [regression] Segmentation fault with SDL_VIDEODRIVER=wayland

10:00

[ddnet:master] 8 new commits

c0d6ce6 Add GLES support - Jupeyy f5f05f3 make screenshot work - Jupeyy ecc5a7a Make GLES3 only makros - Jupeyy e05abba Remove GL includes from headers - Jupeyy b8f96ce Allow GLES3 coexists with OpenGL on Linux - Jupeyy 6f01de1 Fix modern GL graphic settings for the multi backend - Jupeyy 8be6f17 Make video's pixel read GLES conform - Jupeyy 97ab2c6 Merge #3798 - bors[bot]

10:00

[ddnet/ddnet] Pull request closed: #3798 Add GLES support

better than allowing stuff that doesnt exists

TsFreddie

why isn't it a settings in Graphics tho

guess bcs it doesnt work on all OS XD

#ifdef the menu

10:00

lol

10:01

just let people enable it and see it is broken, they can disable it without checking whether there is a command (maybe)

i mean why not yeah

10:01

probably most ppl dont play in "real" windowed anyway

10:01

either borderless or fullscreen

i know people who do

10:02

a lot actually lol

chinese ppl test every setting, perfect xD

10:02

quality by quantitiy

10:02

also a way

i don't get why they like windowed mode tho

10:02

i guess they watch stuff while playing the game or something

possibly

I also know a lot of people just don't like fullscreen

Because I have 10 things running at once at any time, need to mulitask

10:04

some people use DDNet client more for chatting and watching than playing

ah yeah

and even when playing, you can be stuck in one place for hours

for them its perfect ofc

because your partner can't do the part well, or you're died and waiting for the next team 0 group

i'd tab out probably xD

10:05

then the game is also muted and stuff

TsFreddie

#ifdef the menu

also since Allow High DPI option does nothing on windows, should we ifdef that out of windows too?

i wondered about that too

10:05

but yeah its already in the meta data of the manifest

10:05

we can trade resize vs hdpi xD

10:05

button swap

wtf it is 6pm already. i haven't done anything today yet.

start with breathing

yesterday someone in chn mentioned that we probably have enough players for teeworlds BR during weekends.

@deen should we just default all intel users to gl 2.x?

10:09

on windows ofc

10:09

there is still no official answer

10:09

i doubt we'll get any soon

yes

[ddnet-web:master] 1 new commit

4b195ab Update frontpage video - def-

[ddnet/ddnet] Pull request opened: #3810 av_init_packet is deprecated, use av_packet_alloc instead

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...

10:31

[ddnet/ddnet] Pull request opened: #3811 Add Intel as vendor to the block list on windows

I also added, that the warning doesnt require to be displayed, as it currently won't help anyway to update the driver. Should set all Intel users to OGL 2.0 default, they can ofc still manually change the setting as they want

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 pos...

[ddnet/ddnet] Pull request opened: #3812 Make init/shutdown when GL processor creation fails safer

Especially with GL 3.3, which has a shutdown command implemented, it should only get called, if it actually created the the backend stuff.

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test if it works standalone, system.c especially
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect...

i'm getting crashes on master if friendlist is empty

oh doesnt sound good

11:14

maybe that list fix update

11:14

with up and down key

emm

11:14

it is broken since april?

11:15

i'm trying to find a good commit, when was the list fix

ah

11:15

no

11:15

its "just" an assert

11:15

bcs the list is empty

oh

it wont happen on release mode

justatest

11:15

ok

i'll just patch it

yeah, do that

11:17

also i think i rebased the resolution thing

11:17

maybe test that on linux

[ddnet/ddnet] Pull request opened: #3813 Only sort friends, if you got any

as TsFreddie pointed out, it asserts inside the array function back(), bcs the list is empty.

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 ...

TsFreddie

also i think i rebased the resolution thing

ok i'll try

emm

ah

11:22

not rebased i see

11:22

but seems to work

My resolution list change nuked higher resolution on windows if the per-app setting forced a low dpi mode

ah it is now

11:23

oh

Image attachment

but yeah still works for me

11:24

i'll try in wayland

11:28

seems to work too

11:28

but also only have 2k screen xd

do you have dpi scaling on or something tho

dunno in linux its probs DE related or smth

11:29

cant even tell tbh

11:29

i have a dpi setting

11:29

it also works for the normal apps

11:30

but it desont report any other canvas size

does the list has less options tho

11:30

resize seems fine on macos btw

oh it told me to restart my pc, after changing the dpi settings, didnt even notice xD

11:34

yeah dunno, doesnt care about these settings

11:34

tried turning off HDPI in the client

11:34

on off no diff, just gives me the normal resolution list

figured

11:35

then it is "working" i guess

11:35

at least didn't break anything

atleast kde doesnt care it seems

maybe it is time to finally port text editing stuff to ddnet

11:37

since i'm still kinda in the flow

would be really cool

also kinda tired of people asking why they can't see their IME and the broadcast can't center and motd can't wordwrap

11:41

and stuff

yeah, its sad that sdl has no opengl backend for ime xD

11:42

that works really well (edited)

yeah

11:42

but i found a way to render IME in game anyway

without removing fullscreen?

ye

nice

vanilla has that in a Robert's PR which I submitted

11:43

he was doing the TextEditing stuff, so I just added that on top

11:43

don't know why i wrote text editing like that

[ddnet/ddnet] Pull request closed: #3588 Add color selector for Tee settings (fixes #3576)

TsFreddie

resize seems fine on macos btw

btw have u also tried that without the hdpi setting in the client As it also calls SDL_GL_GetDrawableSize which then reports the wrong resolution i guess?

SDL_GL_GetDrawableSize reports the right size after glViewport is called apparently

wtf xD

12:33

ok

12:33

ah yeah

12:33

probs makes sense

but i'll test it with hidpi off too

12:34

yep, seems fine

yeah ok

12:35

i also see in the sdl backend it sets the viewport later

12:36

but good that u find all these weird stuff

[ddnet:master] 2 new commits

427c012 av_init_packet is deprecated, use av_packet_alloc instead - def- 555adbd Merge #3810 - bors[bot]

12:36

[ddnet/ddnet] Pull request closed: #3810 av_init_packet is deprecated, use av_packet_alloc instead

Deleted User

i also see in the sdl backend it sets the viewport later

how late tho

https://github.com/ddnet/ddnet/pull/3804/files#diff-442d7485882d4f383630c81a7061d34995210d7f694acf5eae9366d85a9f66e9L972

12:38

oh

12:38

didnt work well xD

12:38

search CCommandBuffer::SCommand_Update_Viewport CmdSDL;

12:38

in backend_sdl

oh i see

12:39

i thought you meant somewhere inside SDL itself

i also dunno when this code is called

12:39

only fullscreen as it seems

anytime you resize it is called

12:40

hmm wait. no, i'm probably wrong then. seems like we call GetViewportSize first before issuing SCommand_Update_Viewport

12:41

then I have no idea why SDL_GL_GetDrawableSize reports the correct size after we resize

12:42

I tried SDL_SetWindowSize before SDL_GL_GetDrawableSize during CGraphicsBackend_SDL_OpenGL::Init which didn't help either

i think it can make sense tho, we alter the viewport which basically is the drawable size

12:43

so after we changed it, the getdrawable size is different

okay, we used the result of SDL_GL_GetDrawableSize to call glViewport

12:44

which makes me more confused

12:44

so maybe SDL_GL_GetDrawableSize will be correct after the window has appeared in the OS?

you need to be very careful with these, I spent a good 10 hours trying to get this working properly on macOS when I got it to resize properly

justatest

your best bet seems to be to react to the sdl resize event instead of trying to figure out when it's safe to query sdl

we are discussing why SDL_GL_GetDrawableSize gives us the wrong size during initialization with HIDPI flag not set.

on macOS that is completely broken until some arbitrary time when the WM decides to populate whatever internal fields this is taken from

sounds about right

I know for a fact that you'll get a SDL_WINDOWEVENT_SIZE_CHANGED when it does populate it though

anyway i just grabbed window size instead of draw size when HIDPI is not set during init

What are you trying to fix btw?

do you think it is too hacky.

I think ddnet already works fine with hidpi, atleast on macOS

yeah, but it doesn't with hidpi off lol

you sure? it worked fine for me on 10.15 iirc

at least some clipping is broken for me

12:50

e.g. the resolution list

ah, I think jupeyy never added the reaction on SDL_WINDOWEVENT_SIZE_CHANGED

TsFreddie

at least some clipping is broken for me

until i resize the game

https://github.com/Learath2/teeworlds/commit/7f594afc9be9b402263f18f866a8b59abb08cbe1 with this patch I got it to work properly on macOS

Try runtime resizing again · Learath2/teeworlds@7f594af

that's just a quirk that i found during me changing the resolution list tho

12:52

i can drop that change

The clipping is indeed wrong because of the resolution being wrong at initialization

12:52

you just need to wait for the event and then get the correct size

i'll try that

(one thing idk is whether this works well on windows and linux hidpi since I have no access to a hidpi display except the retina display on my macbook)

no. but win doesn't need hidpi support

why not? some magic code making it transparent to even SDL?

twintri

i think we made hidpi flag into our executable, which just makes windows automatically use hidpi. and SDL doesn' seems to override that.

yes

and if hidpi is forced off (set to "System scaling"), windows will tell the game that we have a smaller desktop, and it'll just work and windows handles the scaling.

its some manifest file i added a few years ago

13:00

black magic

and do we hide the hidpi setting on windows?

currently not

13:01

but we can trade it for resizable button

but we should

so hidpi does nothing on windows?

yepf

13:02

also d

lol, that is a solution

also does nothing on kde i think @Deleted User said that

yeah but thats a different story i guess

13:03

i havent decided if i'd like it more to just get a dpi floating point scale or smth instead of all these canvas stuff

13:03

then u can actually create dpi awareness

I want to cry when I hear of these things, an abstraction layer like sdl is supposed to just handle this all transparently

HA

13:03

best joke

Deleted User

i havent decided if i'd like it more to just get a dpi floating point scale or smth instead of all these canvas stuff

I really don't like the floating point scale idea, floating points are cancer

sdl behaves so different between all OS and linux variants

13:04

nobody knows what works and what not xD

Please, let's just use the actual resolution and not some internal resolution * scale factor bs

Learath2

I really don't like the floating point scale idea, floating points are cancer

yeah doesnt need to be a c++ ´float´

13:04

just some scale

Learath2

Please, let's just use the actual resolution and not some internal resolution * scale factor bs

actual canvas resolution or actual window resolution? XD

macOS doesn't even give you this information, you can calculate it but that's just goofing around with a non guaranteed API

i calculated it for resolution list...

Canvas resolution is what matters for us, no?

i guess tsfreedie's pr makes sense, bcs else u cannot select fullhd easily on full hd window resolution, bcs it might be 4k canvas

we initialize window with window resolution

13:06

at least SDL does

ah wait

13:07

ur pr does scale the window resolution down?

13:07

or up

13:08

canvas resolution kinda is what u care about

my pr uses videomodes as canvas resolution, calculated the window resolution. compare config to window resolution for selected item, shows canvas resolution in UI

13:08

discards canvas resolution that is too big for desktop resolution

that sounds good to me

13:08

but thats also what learath wants

13:08

so we all happy xD

13:08

whats the problem

As soon as I see "calculated" I'm extremely vary

tried SDL_WINDOWEVENT_SIZE_CHANGED didn't trigger the event after the window is up.

13:09

for some reason

now comes SDL version changes? XXD

i might be doing it wrong. i'll keep trying tho

What is the actual issue you are looking to fix btw?

resolution list showing the wrong size for canvas and selected items with hidpi off

13:11

that fix is already done with the "calculated" thing that worries you.

13:11

https://github.com/ddnet/ddnet/pull/3804/

HiDPI-aware Resolution List by TsFreddie · Pull Request #3804 · ddn...

This is a version of teeworlds/teeworlds#2827 Which gives you a resolution list that allows you to select appropriate window size (which is saved in config and used to initialize videos), while sho...

13:12

ye, SDL_WINDOWEVENT_SIZE_CHANGED is not triggering for me without resizing.

has it side effects to just always HDPI in macos?

13:14

or generally

I mean I had it working and even letting me change hidpi without restart, so maybe an AppKit change at some point?

13:14

What version of macOS are you on?

latest

13:15

11.3

13:15

cmake doesn't work for me with this version

13:15

cmake doesn't even work for me with this version*

wym by cmake doesn't work? some include mess again?

pnglite/version is getting compileud

building on macOS has been broken for a year now, I just patch CMakeCache to fix it

and glew/version

13:16

it was fine a few days ago

oh we got the GLES change in too today

13:17

I was going to take a look at the hidpi thing myself, but I honestly don't have it in me to debug another cmake issue

i just commented both version files

13:17

lol

I'm on 10.15, but my XCode is up to date, so I have sdk version 11.0

13:18

CMake will find some include files in the 10.15 sdk and some in the 11.0 one, so it won't compile

13:19

which is a result of homebrew not actually being aware of sdk versions, so some of my libraries downloaded from homebrew are linked to the old sdk because of a broken pkg-config file

13:20

I also have to delete the bundled sdl framework because even if I have -DPREFER_BUNDLED_LIBS=OFF it will link to the damn framework there

on my machine, you can't even tell it is the version file being compiled

13:21

because it being compiled messes up something in macos sdk

13:21

and spits those errors out

@Learath2 is gles also deprecated on macos? i guess their phones probably still use it?

~~when are we dropping macos support~~

Anyway, I'm fairly scared of your PR, but if you feel confident it works fine everywhere I guess I just don't have the energy to look into it

well you can at least test it i guess

13:22

since I only have one mac

Deleted User

has it side effects to just always HDPI in macos?

what about this?

Deleted User

@Learath2 is gles also deprecated on macos? i guess their phones probably still use it?

OpenGL in general is deprecated, I highly doubt ES has some exception

its completly random anyway

-- Found SDL2: /usr/local/lib/libSDL2.dylib
-- Found SQLite3: /Library/Developer/CommandLineTools/SDKs/MacOSX11.0.sdk/usr/include (found version "3.32.3")
-- Checking for module 'sqlite3'
--   Found sqlite3, version 3.28.0
-- Found SQLite3: /Library/Developer/CommandLineTools/SDKs/MacOSX10.15.sdk/usr/include

13:24

Look at this joke

nouis

I really just want to pay someone to fix our cmakefile

13:24

sed -i "" 's/10\.15/11.0/g' CMakeCache.txt here is the only fix I could come up with

13:24

after configuring patching the configuration

@noby test my PR pls

TsFreddie

well you can at least test it i guess

I can't if I can't compile it

hmm

13:26

do we have artifacts for PRs

Oh btw, they merged my patch to SDL

super

The artifacts for macOS usually don't work because of linking issues

justatest

on the release version we patch several paths using otool

13:28

thank god dyld is very very flexible and lets us patch it as we want

13:28

I'd probably cry if we had to patch ELF files on linux

13:28

or PE files on windows

we have to remove macos support anyway if they actually remove opengl

okay so I fixed the damn mess, now you said something about resolutions in the list being broken?

or we use opengl to vulkan to metal xD

at that point i'd probably want to implement a vk backend

did you build my PR?

13:33

or master

we really shouldn't have removed the emergency exit, nearly killed my mac...

13:33

we need to remove the fullscreen option on macs, it's just broken

TsFreddie

or master

this one

then uncheck hidpi then restart and see the resolution list i guess

both when I have hidpi on and off it'll only show me screen resolutions

13:35

not the actual canvas resolution

oh that, then you should just build my PR and see if it is what you wanted

13:35

but i'm saying the resolution list might have clipping issue

13:35

on master

mh, it doesn't for me atleast

13:36

any other gfx settings I need to get it to clip?

13:36

with the resizing support in ddnet I can't imagine the resolution falling out of sync

are you running it in full screen?

you mean desktop fullscreen or fullscreen

desktop fullscreen i guess

fullscreen will just not work at all

did you run it in windowed mode, no hidpi and the resolution list looks fine?

yep, windowed, no hidpi and it clipped properly

let me take a screenshot justatest

justatest

btw whats the reason it doesnt work in fullscreen?

13:39

does nothing wokr in fullscreen on macos? XD

Image attachment

13:39

mine does this

@Deleted User couldn't debug it at all, it's completely impossible to debug with it completely getting your computer stuck

TsFreddie

Click to see attachment 🖼️

can you tell me exactly your steps?

just launch the game

13:40

it just look like that

13:40

for me

13:40

until i resize it

13:40

then it is fine

I was running gl3.3 but even without that it was fine

this is what i did to the resolution list in the PR btw

Image attachment

can you show me cat yourconfig | grep "^gfx_"?

sure

Ohhh wait I'm running bleeding edge SDL

13:42

maybe it's the SDL version

13:42

check during launch what version sdl reports

how do I check sdl version again?

mh, I'm spending time I don't have on this... I really am supposed to be studying

TsFreddie

how do I check sdl version again?

it prints on the console during launch

2.0.8

13:43

should i update my sdl then

ah, hm 2.0.8 will just instantly crash for me and usually for everyone beyond macOS 10.15

miricale

TsFreddie

should i update my sdl then

no, we ship sdl 2.0.8 so it needs to work there

13:44

I'll compile a patched sdl 2.0.8 to test

oops, already brewed it.

cant we use current master sdl somehow?

what version did you brew?

it probs fixes alot of stuff

wait

Deleted User

it probs fixes alot of stuff

There was one massive performance issue on windows that made us roll back

i don't think i'm using homebrew's sdl anyway

13:46

my homebrew already has 2.0.12

yeah, it's pain to get macOS to link to the correct sdl

13:46

2.0.16 is going to be the first version with my patch to fix the appkit issue

poggers

@TsFreddie afaik the only way to get it to link to homebrew sdl is to delete the sdl window in ddnet-libs

Learath2

There was one massive performance issue on windows that made us roll back

yeah i dont mean for windows

13:47

just for macos

i don't want to deal with that tbh

[ddnet:master] 3 new commits

f99f062 Only sort friends, if you got any - Jupeyy 6465e48 Add unit test for empty range - def- 70f79bf Merge #3813 - bors[bot]

13:48

[ddnet/ddnet] Pull request closed: #3813 Only sort friends, if you got any

I'll try to reproduce your issue on 2.0.8

13:50

with my cute sdl monkeypatcher probably

SDL master

13:50

i really don't like reading sdl2's code every time something happened.

yeah, such a pain

13:50

and most of the time it's not even sdl's fault

13:51

on macOS you almost always are in a world of pain trying to disassemble bits of AppKit

~~well what about the IME stuff on windows~~

How does this damn thing even work for you....

13:51

you are supposed to get SIGILL on launch like all other macOS users

poggers

13:52

maybe no one is running big sur?

13:53

or maybe my macos is fresh since every time something happens i just nuke it and restart.

13:53

i don't save anything on my mac

maybe AppKit on big sur actually noticed this is an insane check

maybe

the only way around is to link your application to an old sdk, which is what we do for the official release

gfx_screen_width 1280
gfx_screen_height 800
gfx_borderless 0
gfx_highdpi 0
gfx_vsync 1
gfx_resizable 1
gfx_opengl_major 2
gfx_3d_texture_analysis_done 1

13:53

don't know if you still need it, but i forgot to send this

13:54

do I even have old sdks on my mac?

13:54

how do i check

ls -la /Library/Developer/CommandLineTools/SDKs/

i have 10.15.sdk

There is a SDK in XCode too but I don't remember the exact path

i don't have xcode installed

13:55

i hate that thing anyway

it has a decent profiler, about the only thing I use in it

i use orbit on windows

13:56

since i don't have visual studio either

hm I've been looking for a way to avoid vs too, is orbit any good?

well i don't know how profiler supposed to work, but i can see nanoseconds and stuff so i'm content with it lol

I just love my tools on linux so much, why wouldn't they just mimic linux?

13:58

perf, callgrind, gprof, valgrind, ASan, MSan

13:58

on windows it's either VS or here have this 10 year old windows NT tool

I run valgrind in WSL for server codes which has been working fine for me

Dr. Memory is a piss poor excuse for ASan

TsFreddie

I run valgrind in WSL for server codes which has been working fine for me

how does WSL work? do you link to another set of libraries?

WSL2 is just a vm now

msvcrt doesn't really take well to hooking into malloc

13:59

Ah, a VM makes sense

with some microsoft jank that bridge the file system and stdin/out somehow

14:00

it is kinda magic. i can run windows version of powershell in the WSL2 vm and run WSL2's zsh inside the powershell that is in the zsh to begin with.

okay, 2.0.8 breaks resizing

14:03

but that's all

emmm

14:03

resizing works fine for me tho

14:04

i doubt my pr fix resizes

Image attachment

i'll try master i guess

14:05

that's a huge screen tho

isnt that bcs of the fullscreen flag?

idk what sdl version I should try with your PR

14:05

@Deleted User ?

or is that pure window

14:05

i mean like clicking fullscreen button

14:05

probs the yellow top right

14:05

never used macos xxd

it's the screenshot of a window

Image attachment

14:06

i mean like this

you want me to try hit the button or sth?

are u in a maximized state

hitting the maximize button breaks the game

like clicking maximize

14:07

oh ok

ye, resizing worked for me just fine

anyway, I wasn't maximized

Image attachment

ok, just thought bcs under windosw sdl doesnt remove the flag for some reason

maximize works for me too lol

maybe learath should update his os xD

14:08

does macos not force u todo it?

with SDL 2.0.15 both maximizing and resizing works fine for me

niceeee update

though we seem to be missing an event hook when we are maximizing

14:10

I need to resize twice to get our code to catch the resize

sooo.. what's the plan

I don't know what to do

14:12

I'll try your PR, if it works for me I guess we merge it since it fixes one more version of macOS I guess

always blame sdl or drivers on linux it works perfect

we also seem to confuse concepts a lot, resizing the screen by dragging or hitting maximize actually changes the render resolution

14:13

but we don't change gfx_screen_width and height

yeah

14:13

i am also not sure if we should do that

14:13

or if we do, we should show the current resolution somewhere

14:14

else u cannot click the current resolution in the list

14:14

current selected

i feel like we shouldn't

mixed feelings tbh

14:14

it can also be nice to restore last state

don't know why but that's how i feel

14:14

ye

14:15

https://streamable.com/ly8obr

the first time I change resolution it doesn't work, after that it works perfectly

14:15

idk wtf is up with macOS in general...

Learath2

the first time I change resolution it doesn't work, after that it works perfectly

that actually happened to me a while ago. i thought you fixed it

14:16

forgot which version was i on. but i'm already on big sur for sure

14:16

i was*

I thought I fixed it too, another bug that will take hours to debug

14:16

okay, I would really like to continue this but I really will be mad at myself if I don't spend at least 1 hour studying today

14:17

so I'm leaving right after I compile your PR

what was the pr number?

3804

okay there is an issue here

14:19

the highest supported resolution is not my entire screen

with hidpi off right

with hidpi off it is my entire screen

14:20

with hidpi on it's smaller

hmm

14:21

i might need to swap canvas size and window size around then

3360x2100 is the highest shown in hidpi

14:22

which is the second highest actually, maybe you are overculling the resolutions?

was is supposed to be the highest btfw

14:22

btw*

maybe a < instead of a <= somewhere?

14:22

it's supposed to be 1920x1200 * 2

14:23

so idkx1400

14:23

2400*

maybe it is something wrong with calculated dpi scale

Image attachment

14:24

afterall

see why the word "calculated" scares me?

14:25

well i can swap canvas size and windows size

Oh and wanna hear something even funnier?

so i keep videomodes as windowsize, and calculate canvas size which is only used for displaying in ui

with sdl 2.0.8 your PR actually creates the clipping issue you reported

hmm

14:26

well fuck me then

and it breaks resizing and it still won't show the correct highest resolution when hidpi is on...

how does it break resizing

pffff this is like sooooo sooooo soooo annoying when macOS behaves so differently between versions

i didn't change anything about resizing tho

oh actually resizing was already broken in 2.0.8 for me

14:27

nvm

14:28

maybe I should just pony up the cash and get an M1 macbook to test this on

if you haven't gone yet, see if the new commit works better, at least in the resolution part.

14:32

not sure about clipping. but i guess i'll keep looking for stuff.

Is SDL bundled so that we can update it?

i'm using bundled SDL

Then why struggle with outdated stuff

14:34

If 2.0.15 works fine

works fine for me, I wonder if it works fine for @TsFreddie e.g.

Actually what I'm using is sdl master

let's wait for 2.0.16 maybe?

@TsFreddie can you try something for me?

yes, sdl master

can you do defaults write -g NSLogUnusualAppConfig -bool YES

14:41

then launch ddnet and see if you get anything extra in the console

14:41

it will show the compatibility changes appkit makes to run the program

sure on sec

without 2021-05-06 16:40:49.478 DDNet[19052:1029624] NSOpenGLContextSuppressThreadAssertions=YES in the log I'm fairly certain you have some alien macOS

14:42

sent by the gods to allow you to run legacy ogl applications with ease

poggers

14:44

btw which console

where stdout goes

ok

the terminal where you launch the game

still building, just making sure

14:45

my mac throttled so hard it take 5 minutes to build ddnet

yeah mine chokes too, an intel mac?

ye

M1s are supposed to be much better

i know

14:45

i'm poor now tho

I would probably never spend my own money on a mac

14:46

unless I was like filthy rich and felt like testing something

Image attachment

14:46

i have alien mac i guess

Learath2

I would probably never spend my own money on a mac

did you raid apple store

I got one in exchange for developing a mobile app

14:49

they were considering hiring a developer, I offered to do it if they got me the mac

ic

14:50

so.. what did I do to my mac

TsFreddie

Click to see attachment 🖼️

you sure these were the only unusual configs?

let me grep it then

It should be right after the opengl context creation btw

Image attachment

one last thing, can you try opengl3

will do

it shouldn't differ but maaaybe in big sur ogl3 takes a different codepath

same thing

and it just works?

looks like it

okay can you do an otool -L DDNet?

do you want the output

I'm looking for just the AppKit line

    /System/Library/Frameworks/AppKit.framework/Versions/C/AppKit (compatibility version 45.0.0, current version 2022.44.149)

14:55

okay that is much much newer than mine, can I have the appkit framework? it's a bit large but if you don't mind I want to check out what they changed

emm how

14:55

do i just archive it or something

yep, that works

AppKit.framework right?

yep

14:57

don't send it here btw

how do you get new AppKit in the first place

new macOS version

make sense

AppKit never breaks compatibility though, so functions take different codepaths in newer AppKits if they are linked to an older version of the sdk

14:58

I want to check out [NSOpenGL_Context setView:]

hmm, the archive is 20mb

jesus

14:59

mh, I guess don't bother, no need to waste time to satisfy my curiosity

14:59

I'll just get big sur sometime and check it out

15:00

I wish the SDKs no longer came with flat manifests instead of the actual framework files

is the size wrong

no it's correct, it's just too large to upload, atleast for my internet speed

15:01

if you don't mind I'd still like to have it

i can upload it pretty quickly

if you can't be assed, it's whatever, again just idle curiosity

and it's done

dm me the link, apple doesn't really enjoy people uploading these

i have something like 100mbits up

I wish

but i still need to upload stuff through a proxy so the file can get to onedrive and stuff, so i guess i'm bottlenecked by the proxy

15:03

still plenty fast th

15:03

tho*

hm, you sure you zipped this correctly?

maybe not.

15:05

Image attachment

can you do a ls -la /System/Library/Frameworks/AppKit.framework/Versions/C/AppKit?

15:06

maybe it's a symlink

/C/AppKit doesn't exists

eh?

Image attachment

15:07

maybe they moved it

but your ddnet was linked to it

emm

eeeeh?

huh

15:07

ye right, it was linked to it.

do you have lldb?

hmm

15:08

i do

15:08

forgot how to use it

lldb ./DDNet

15:08

run , wait a bit ^C to interrupt

15:08

then do image list

can you grep within lldb

no, but AppKit should be within the first 20

[ 15] 10AFBC3A-E9A4-3E62-B9F5-97DF579B7A84 0x00007fff22e79000 /System/Library/Frameworks/AppKit.framework/Versions/C/AppKit

15:10

bruh

okay, I'm starting to believe aliens sent this mac to you as a gift

15:11

I guess I'm not going to get to check out what changed until I get my own big sur

maybe apple hide those path in kernel or something

15:11

well do a sudo ls -la instead

15:11

if root can't see it idk, only god himself knows

gg

Image attachment

15:12

i'm out.

15:12

sendhelp

yeah me too, we tried

15:13

New in macOS Big Sur 11 beta, the system ships with a built-in dynamic linker cache of all system-provided libraries. As part of this change, copies of dynamic libraries are no longer present on the filesystem. Code that attempts to check for dynamic library presence by looking for a file at a path or enumerating a directory will fail. Instead, check for library presence by attempting to dlopen() the path, which will correctly check for the library in the cache. (62986286)

15:14

apparently there is a tool to dump it from the cache, I've already wasted enough of your time, I'll just check it out myself when I upgrade some day

15:15

頑張って on your studying then

15:16

if you still going for that

I have to, so I don't have to live under the bridge with the hobos

TsFreddie

頑張って on your studying then

@Deleted User does this patch fix window icon notifications on KDE? I have only tested it on i3 and it seems to work well.

attentionfix.patch

1.24 KB

@Learath2 I guess try the new commit in my PR and see if that fixed it for you when you got time. Since you are the only one who can test that problem.

Deleted User

Also we need a solution that works without x11,bcs Wayland is on the March

I don't think that there's a Wayland protocol for that. See https://gitlab.freedesktop.org/wayland/wayland-protocols/-/issues/31

Add facilities for clients to request toplevel activation (#31) · I...

This is a meta issue for discussing toplevel activation/focus requests. I realized we don't have one yet. Use-cases Let...

https://blog.rust-lang.org/2021/05/06/Rust-1.52.0.html

Announcing Rust 1.52.0 | Rust Blog

Empowering everyone to build reliable and efficient software.

17:51

poggers

bluekitty

every six weeks

i needed that clippy fix so much

17:51

and im 2 lazy to setup nightly just for that

ah, nice bugfix. otherwise it looks like a rather small release

The most significant change in this release is not to the language or standard libraries, but rather an enhancement to tooling support for Clippy.

17:54

@heinrich5991 there is something "big" idk how big but its a regression

17:54

that affects ring

17:54

since rust is the only real world end user of the llvm noalias stuff they find all the bugs it has xd

17:54

https://github.com/rust-lang/rust/issues/84958

Regression: Miscompilation due to bug in "mutable noalias" logic · ...

The ring test suite started failing when I tried to upgrade to untrusted 0.8.0. The regression first shipped in nightly-2021-03-23: We were further able to narrow down the regression by bisecting d...

17:55

Enable mutable noalias by default on LLVM 12, as previously known miscompiles have been resolved. Now it's time to find the next one ;)

17:55

monkalaugh

I wonder if it'll ever work to enable it

this is kind of an open question: have there been attempts / ideas of a complete rewrite (either in cpp or other lang) with goals to make modding and ddnet contribution easier? i know this is not a simple or straightforward process but idk if it's been attempted before

not really, some attempts have been made

i dont think ddnet is a good base for making a "empty" teeworlds

18:01

it has many ddnet specific code everywhere

e.g. mine

https://github.com/heinrich5991/libtw2/tree/master/server

heinrich5991/libtw2

Some Teeworlds stuff in Rust. Contribute to heinrich5991/libtw2 development by creating an account on GitHub.

teeworlds 0.7 is actually a good base for new mods

18:01

it has 64p support

18:02

and cmake

well specific and not, all of teeworlds + ddnet code would be part of the rewrite

18:03

and even treating them as the same code

@heinrich5991 would it be crazy/improve speed to rewrite the server using a async runtime like tokio? https://github.com/tokio-rs/mio

Metal IO library for Rust. Contribute to tokio-rs/mio development by creating an account on GitHub.

ideally it would be organized like some "core" that generally works with any client but has "extensions" for modes

18:05

@heinrich5991 how much of the protocol does this cover?

@lynn the protocol is completely implemented, the server is a very barebones and you can only move around, nothing else

18:09

no client is implemented

18:11

@Ryozuki I don't think tokio is going to improve much… the server is already using mio to get notified about socket events

oh i didnt see it

so you could connect to it and play? just wouldn't support things like races

you can connect to it and move around

18:12

no weapons, teams, etc. are implemented

18:12

no collision

oh wow

18:12

wait if there's no collision what do you mean move around

no collision with other tees, I mean

oh, with other players

18:12

yeah ok

18:13

well, this is good then

(it's just a proof of concept, I did the interesting parts and stopped before the boring parts came)

maybe i will try to learn rust this weekend and try to PR it with some of this

18:13

i need to stop saying "i will learn rust" and actually do it

good luck learning it in 1 weekend monkaS

monkaS

well not all of it obviously, but enough to add logic to it

another programmer converted greenthing

greenthing

idk ive always liked rust but other stuff has had my attention for forever

18:17

i had a lua phase

i never liked it

18:18

and i rly pushed myself to like it due to factorio modding

18:18

but i dont like it

18:18

(lua)

but you can't mod factorio in rust… ryo already send the next message

18:19

sent*

wdym?

18:19

is it some discord desync

18:19

ah

18:23

their api is rly gud https://lua-api.factorio.com/latest/ feelsbadman

feelsbadman

when tw api?

When programming language researcher get bored due to lockdowns

18:37

https://www.youtube.com/watch?v=77999Td20TM

Pure Implementation -- Official Song of PLDI 2021

i admire how hands off the lua library is

18:38

i enjoyed making my own constructs that exist in other langs in lua

how can i turn off the spec tees?

19:14

i dont wanna see those ghosts

you can't

oh yikes that is inconvienient

19:16

any chance we can get an option?

not really, no

19:17

can you explain why you don't want to see the spec tees?

he could make the x_spec.png skin transparent, no?

ah right

it might still show the nametag though

19:20

(I think)

also right

i really dislike it optically i rather not know than look at something visually unpleasing

you can replace the skin maybe. or you could get used to it for a couple of days and see if it still looks out of place

its not that the design is bad its quite good actually. Its just i do not want to see a tee i can not interact with. So i really have to fork for that cfg?

19:21

i still see the name

19:21

yea probably i can get used to it

its not really visually unpleasing tbh, u just have to get used to it

Maiski

@Deleted User does this patch fix window icon notifications on KDE? I have only tested it on i3 and it seems to work well.

yeah that works

Exported 920 message(s)