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 2020-09-21 00:00:00Z and 2020-09-22 00:00:00Z

I updated client since i can bind Ctrl shift q again thx acqwerty

does it work in menu

[ddnet/ddnet] Pull request opened: #2894 Add new renderer(as streamed vertices) for editor

Does not improve performance in editor!

1

@Ravie ^

04:17

try to remember what u did if u find bugs

04:17

helps for reproducing later

monkalaugh

Image attachment

@murpi could you send me the map (background) aswell?

1

04:42

glad you guys solved the issue with skins :p

04:42

the custom map BG in menu is already in 15.0.4 or latest beta needed?

nah, i just tested the pull request

@plsplsplslol no, only F-binds work in menu

[ddnet/ddnet] Pull request opened: #2895 Reload skin menu when skin count changed

fixes graphics bug when new skins are added, skins array is reallocated and we accessed freed memory Thanks to Jupeyy & Learath2 for also looking into this Screenshot of the bug for reference, reproduced by removing all skins in data/skins, opening skin menu, joining a full server and waiting a bit: !screenshot-20200921@005129

[ddnet/ddnet] Pull request opened: #2896 Add edit box to write skin name, Add link to Skin Database

@Deleted User If you manage to add to DDNet the same renderer that Vanilla 0.7+ uses, dilate would no longer be needed, right?

Ddnet renderer should be better

Let's make console renderer

@Learath2 any ideas about mac input problems? https://steamcommunity.com/id/parriseinstein/recommended/412220/

Steam Community :: Parwiz :: Review for DDraceNetwork

Wtf am I playing. I'm using macbook pro and funny how the display is god awful that i can't even see the chat since the dock blocked it. Also except WASD, all other button not responding, mouse click not responding couldn't even quit the game. Worst experience ever

then differently asked: can we get rid of the damn dilate at some point soon? xd

You just need to switch workspaces once after launch and the dock goes away and everything works

That sounds unintuitive, any way we can fix that in code?

Not much we can do about any of the issues as SDL support on macOS is shit

09:26

I couldn't debug it at all

09:26

You are welcome to try if you want

ok

09:26

so it's the same problem with other SDL2 apps?

I don't have any other SDL2 app to check actually. Let me find one

maybe we just need to grab focus more agressively?

09:27

I also noticed once on Windows it wasn't focussed

09:27

and I had to click in the window

let me see the current state when I launch from steam, see if it regressed

09:31

hum, my steam library isn't loading at all on mac

09:33

Now that I think about it, this guy seems to be lying, how did he join a server to see that chat is blocked when no buttons even responded

maybe mouse worked a bit?

The current state is that you need to switch workspaces and it starts working, before that mouse clicks don't seem to be caught

09:35

Sometimes pressing F1 twice works too

09:35

How do we regrab the mouse after f1?

09:35

Maybe we can do that when the game is fully launched

Input()->MouseModeRelative(); I think

09:36

which does: SDL_SetRelativeMouseMode(SDL_TRUE); Graphics()->SetWindowGrab(true);

09:37

So let's just add another SetWindowGrab(true) after the initialization of window is done?

This seems to be some race condition somewhere. Sometimes I can see the OS cursor too, at that point I can't grab the mouse with just double f1

09:38

Yeah we can try a grab after the initialization

I guess we don't have to set relative mouse mode again

Oh btw can we let the mouse escape the window if esc is pressed too?

I guess we could, not sureif we want to

09:39

the mouse sensitivity feels a bit different then

09:40

and might have muscle memory and accidentally click the wrong button

Ah, since we get the os cursor. Yeah not the greatest idea

[ddnet/ddnet] Pull request opened: #2897 After initialization: Agressively try to grab window again

since on Mac OSX we don't have focus otherwise @Learath2 Please try if this works.

@Learath2 u got factorio,made with sdl2

09:53

Well not open source tho

Mh the github artifact crashes instantly

also other versions or my change makes it crash?

10:00

I can build you a nightly or you can build yourself?

Ah, it's just my bug that sdl still hasn't merged the fix for

10:00

I'll just build it myself, with my own sdl

10:03

mh, there is a different grabbing issue when I launch it

10:03

let me try to replace the steam one and launch from steam

10:06

If I build it myself I get a completely different grabbing issue

10:07

The cursor starts at the very top left, seems it grabbed fine, when you go to click something it clicks the apple logo at the top bar instead

10:07

@deen would be nice if you could build a nightly so I can test that

Number of skins of DDNet Skin Database is reducing, may I know why? I'm just curious

@Rafael Fontenelle Soreu, Ravie, louis are cleaning it up to remove non-teeish and ugly skins

Nice

since they are auto-downloaded now quality is more important

Talking about skin auto-download, is this resource working right now? I mean, is ddnet-skins package not needed anymore?

There was a bug with it so we disabled it, but it should work in the next patch again

we didn't disable it

Oh, then it sort of works right now, it will work much better next patch

10:11

@deen why add_unsorted?

because the original array is sorted already, so if we just filter out elements we don't need to resort

Ah, makes sense

10:13

I think there is still a bug here, you better compare the filter properly, or use a flag

10:15

Oh, nvm. This should work

@Learath2 can you try https://ddnet.tw/downloads/DDNet-nightly-osx.dmg ?

Mh, the same thing happens with this build too

10:52

The current one on steam doesn't behave like this, do you remember what commit we built that from?

Nightly on steam or regular?

10:53

It's tagged in git

regular on steam

c2f388a52ba1458b5dd10557826c42e1b89f351b

On the regular, clicks just won't work until I double f1 or switch workspaces

10:54

With the new builds it does the top left corner thing

10:55

Hm, I wonder if it's the bundled SDL the steam version links to

11:02

Yep, it's SDL 2.0.13

11:02

I'll bisect for it in SDL later. #2897 doesn't help sadly @deen

https://github.com/ddnet/ddnet/pull/2897

After initialization: Agressively try to grab window again by def- ...

since on Mac OSX we don't have focus otherwise @Learath2 Please try if this works.

How do I download skin? Didn't find a "Download skin option.

@Rafael Fontenelle it downloads automatically

11:11

Well for other people

11:11

You can set your own skin to something you don't have with player_skin xyz in f1 for now

@Rafael Fontenelle settings -> skins -> download skins

Is there some padding before chunks?

what chunks?

I know the decompression works fine because I can see the token at the end, but the chunk header being all 0s makes no sense

11:21

Net chunks

no idea, sorry

Does a net chunk with 00 00 header make sense to anyone?

maybe name it "auto-download skins"?

Thanks @deen

Ah a chunk header with 00 00 makes no sense, I'm using the wrong offset

[ddnet-scripts:master] 1 new commit

0cd5dcc scripts update - def-

@Soreu is auto dilate not enough?

yes and no

12:36

new dilate makes the image file size bigger

12:36

with 0.7 dilate is not needed and so the map can weight less

12:36

info from @mind though

Yes, but 0.7 converts images on runtime.. That's the trade

12:37

But idc. Somebody can add ofc

Image attachment

12:40

@mind hope you don't mind sharing part of our conversation

Strange in my pm he gave green light, seems like he is not happy

Because we both are, but due to the different rendering ways in DDNet & Vanilla, one could have same file without dilate being lighter

12:41

and that's the annoying part

12:41

not your fault tho

I mean the file size thing can be easily removed. It's only a few KB tho

whoever cares more, always tires to reduce map size as much as possible

Or lowered

12:43

Ah BTW, maybe dilate has less aggressive PNG options than gimp or what u use.. I'll look into this later

No idea what mind is using, Ravie uses Inkscape & Gimp, Louis uses Gimp too I think, and I'm stuck on Photoshop and one lil crap that is pretty much advanced paint (depending on the needs ^^')

12:46

feel free to ping when you need any info

i use inkscape and paint.net kek

12:47

maybe i should learn gimp

if its more about drawing than editing for you, then you could maybe check out https://krita.org/en/ <3 gimp too tho

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

and https://vectr.com/ ^^

12:55

Actually, maybe I'd switch from ocassionally launching AI to full-time using Vectr... tee_thinking

tee_thinking

13:00

@Patiga Krita looks far better then the software I have used long time ago for ~this kind of stuff

awesome :)

ArtRage, reminded myself the name xd

13:01

I can relate to the second part of its name

I spent an hour debugging because matricks got the bit shifting the wrong way around...

13:07

If anyone else wants to chop up teeworlds packets, it's not size_seq, it's seq_size, which makes it a PITA to read in wireshark because it's not contiguous

13:08

flags_size seq_size

https://github.com/heinrich5991/libtw2/blob/master/doc/packet.md

heinrich5991/libtw2

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

13:10

is this info here?

Yep, should have checked that instead of relying on ancient documentation in comments

Q: How many projectiles and/or powerups I cand give within one snapshot before it will cause troubles?

13:14

*with

13:14

*within? :panic: :send_help:

You should ask that to @fokkonaut

There is a serverside limit only which can be increased to your wishes iirc

But a snapshot is still have to fit into a UDP packet? or not?

no idea

or is it getting split/reassembled by a game proto?

13:17

so... it's never a problem, right?

A packet can't be larger than NET_MAX_PAYLOAD

13:17

A chunk can't be split

I knew that! It should fit into IP packet.

13:18

But what it is a chunk? Snapshot is being sends in chunks so it reassembles on arrival by a client?

So if you only manage to fit one snapshot within the packet, you'll have NET_MAX_PAYLOAD - <packet header size> - <chunk header size> - <message header>

Did window focus options change between 15.0.1 and 15.0.4?

13:18

because after launch I can't move mouse unless I click first

But there is a problem there, compression helps this a bit, so you can't really know if it fit or not

By a compression you mean delta?

Both delta and huffman compression will help there

all that "full snapshots" vs "delta snapshots" thing?

13:20

So if you only manage to fit one snapshot within the packet

So my real question is do I have to fit everything in one packet?

13:20

And what will happen when my snapshot will get larger than 1392 bytes including headers etc after all compression applied.

ok, it's not happening all the times, weird

@Deleted User packets will get dropped, you'll have a bad time

I have some weird sporadic lags in my mod after addition of some new pickups icons.

13:22

Oh... so it must be it.

I think you can safely ignore the effects of delta compression, as being unable to send full snapshots will break it anyway

Shouldn't a server complain about this in server logs or somewhere else?

@Deleted User it does

@Soreu hasn't changed, but we want to fix it in next release

dbg_msg("netserver", "packet payload too big. %d. dropping packet", pChunk->m_DataSize);

Well, weird then because I haven't had that problem before xD

13:23

idk, anyway got more weird problem with server

@Soreu my DDNet client started to freeze on close for 5 seconds ~2 weeks ago after an update. Very annoying.

13:24

Everyone is telling that it's just me.

if I launch server from its own exe, it launches autoexec_server.cfg which then loads myServerconfig.cfg pointing to & loading server\config.cfg but... if I launch server via client, with the Start Server button, it doesn't load myServerconfig.cfg because settings in server\config.cfg are not applied at all

13:25

how the hell is that possible?

TIL europe is trying to make processors? https://www.european-processor-initiative.eu/

Home - European Processor Initiative

@Deleted User Can't relate with your problem, sorry justatest

justatest

13:27

oh lol, found what was causing the difference... kid of

13:28

changing from exec server/config.cfg to exec server\config.cfg made it execute the command properly regardless of whether server was started directly or via client ._.

13:30

whatever

@Ryozuki I wonder what will they do now when NVidia bought ARM.

13:35

They will use RISC perhaps. Anyway I don't think we will see those processors any soon.

13:36

Or that hey will be interesting for regular consumers like us.

13:36

We have Elbrus processor in Russia. And nobody can buy it except military.

I know closing server on client close was requested, but is there option to disable this feature? . . .

[ddnet/ddnet] Pull request opened: #2898 Inform players that they are in team already

@Soreu nope, just run it outside of client

13:38

it's a bit ugly because then the original client process can't quit

13:39

and we have too many options already, guarantees that there are combinations of those options that have bugs when used together

yea, k

13:40

that's not a big problem

I think this is good enough to release for now

Image attachment

13:44

It currently only does 0.6 and ddnet because I kinda nuked the 0.7 part when the stupid chunk size kept being wrong

nice

Such amazing work

13:50

Is this dissector available already?

13:51

I want to finally look into teeworlds traffic to finally understand what's going on

13:51

and to estimate snapshot entities sizes and limit too

13:52

I've spend a week relocating my server from Moscow to Frankfurt then back and then to Amsterdam to finally get today that it's related to a small addition of an extra pickup in a snapshot.

I will put it up today, but just because @heinrich5991 asked to take a look, it's not ready yet

No haste ofc.

It still needs decoding of the actual chunks and wrapping up the conversation tracking feature so I can dissect 0.7 traffic too

Is it in C++ or a python?

C

13:53

wireshark only allows C or lua dissectors

13:53

I didn't want to port the huffman code to lua, so C it is

13:53

(and I kinda love C so it's a perfect fit)

1

idk if u answered me already @Learath2 but are you planning to get it into official wireshark?

It needs much work before it's ready for that, but sure

[ddnet/ddnet] Issue closed: #2599 Alt tabbing hang issue

sure and nice

btw, I like how wireshark has memory pools, it's a cool idea

13:59

You allocate memory in specific pools that has specific lifetimes, so it's ghetto garbage collection

I know that opinion will not be popular here, but I hope one day every language will have a GC

GC costs waaay too much

14:01

It's unacceptable for high performance applications

There are many high-performance applicatoins in Java and C#.

Java keeps it's entire memory pool hot, so the kernel can't swap it away, ever, even if you rarely ever use the data. So it can have decent performance, at the cost of horrible memory usage

But didn't meant to start a holywar. Just it's hard to get from interpreted languages like Ruby or GC ones like C# to C++ and get all kinds of obscure and hard to detect/track and debug memory corruption and access errors.

It also throws off prefetch

I can write application in ruby that will preallocate a memory that will be reused but never GC. Just I rarely need it.

The only implementation of GC that I like is in go-lang

Yeah, and golang.

It has great performance, without much extra overhead

It's just I never know why my app will segfault next time. And when it does sometimes there is nothing useful in coredump.

14:04

I've learned about ASAN but still it's hard. So much troubles on memory management.

14:05

There is teesharp which is performant enough. I would love to use sharp. Machine resources are cheaper than my time.

@Deleted User there are ways of achieving memory sanity without any performance penalty

Ofc, it's just have a steep learning curve.

Wiresharks pool approach is really cute imho

I've started to make Unity project with grown very big and never hard any memory-related problems.

14:06

Last month I am just debugging them and learn about yet another way a memory could get corrupted without being noticed in C++.

Memory can't get corrupted in C or C++ unless you touch the memory

14:07

One of the beauties of C/C++/Rust is that they don't do things behind your back

I will learn everything eventually ofc, smart pointers and such. But for now language is getting on my way.

14:08

Yeah, I wish that memory management patterns would get to my programmers subconcious already to this moment.

14:08

This is my first C++ project, so I get all problems a noob could get in an old enough C++ project.

14:09

In infCroya there are no smart pointers, not even nullptr. Just integers.

14:09

I will rewrite this as soon I will grasp this. Just with things could be easier for me.

14:09

Being a C++ programmer for me always was a macho dream

14:10

It's just too lowlevel for leisure.

I was looking into redoing teeworlds with smart pointers, but who owns what isnt really easy to figure out

14:10

Leisure is subjective

Ofc, it's depends on your goals too.

I love how everything is predictable in these low level languages

If one's goal is to make an experience then framework and language are just tools and one could complain if they're having a steep learning curve. (edited)

14:11

If your goal to challenge yourself into learning something hard then it's a leisure already and small result and imrovements are nice.

14:13

Don't mind me. I will get used to it or will just switch to something more rewarding in time.

14:13

At least C++ syntax wasn't hard. Just memory management side effects are.

14:13

It's a chad language after all

14:15

Ah, I am editing my messages here again on reflexes. Time to leave, thanks

@Learath2 too large snapshots aren't a problem due to the thing you specified

14:21

they get broken up into more packets

@Deleted User @Soreu My dilemma was if I should predilate the images or not, since a lot of people are still mapping with old client versions (so predilate is usefull) while on the other hand there actually is a 0.7 mapping scene and the filesize would be up to 40% bigger for them. Removing dilate for newer versions won't solve this at all. In the end I decided to just predilate the images and I'm super happy with the new autodilate.

@Learath2 see SNAP, SNAPSINGLE, SNAPEMPTY net messages

1

Is there a command line utility to see if map is compatible with 0.6/ddnet BTW?

14:29

I keep accidentally overwriting my 0.6 maps with 0.7 editor. Would love to have this check in my deployment pipeline

or finally add 0.7 map support to ddnet

0.7 map incompatibility is my fault, sorry

14:31

and retrospectively, not even for a good reason

fix it then :p

can never fix it, will always have to support the new format

yea thats what i ment support the new stuff in ddnet is a fix for me

14:32

problem solved

We have convert to 07 utility. Just convert_to_06 and which_version tools are missing. (edited)

14:34

Does 0.7 map format have something new that ddnet doesn't have or just or just 0.7 strips some compatibility data blocks?

it has my awesome compression

Nice

14:35

Any simple ways of knowing if the map 0.6 compatible?

14:35

Like some keyword or a header flag?

checking rn

@heinrich5991 really? I thought snaps needed to fit into chunks too and we can't split chunks

snaps get split into multiple chunks

14:38

*can get split into multiple chunks

Oh yep, didn't know that

14:40

So I guess no limit

Can a snapshot of individual entity within a server snapshot be split too?

snaps have an inherent limit though

Or there are no big entities and they cannot grow?

64KiB I think

Oh

14:41

Is it to limit bandwidth and guarantee a rate?

14:41

Update rate

14:42

Also if snapshot splits then I suppose the client has to wait before it could reassemble

yep, for all packets that contain the snap

If a client will get a packet with a chunk for a newer snapshot while it is aggregating older one's chunk to assemble will those chunk be just discarded?

14:44

I mean older ones

14:44

That would lean to ordering problem in UDP

I'd guess the old chunks will be discarded

14:45

but I'd have to read the source to be sure

14:46

ye, discards all old partial data if new data arrives

So when snapshot grows there are more chances it will have some sort of error... racing condition

14:47

No... that it will have higher chances of not to finish assuming when getting newer snapshot packet out of order

wdym, out of order

Up does not guarantee order of packets

yes

14:47

we wait 50ms between two snapshots

14:48

packets don't delay randomly, they usually take the same route, and usually even arrive in the same order AFAIK

So older snapshot could consist of so many packets that server will start generating and sending a newer snapshot before client will assemble old one

not really, to my understanding

Yes, mostly they will be in order ofc

I mean the packet would have to get delayed by 50ms to get a chance to be interleaved with the next snapshot

Or have a lag spike for 50 ms + have multiparty routing anywhere on its way, right?

14:56

*multipart

14:56

Multipath

14:56

Which is common for the internet

is it common that packets take different paths?

14:58

AFAIK no

Anyway it works surprisingly well irl

14:58

Yes

got a citation for that (don't have one for my gut feeling)?

I worked in ISP, channels get aggregated all the time

k

14:59

i.e. that packets send in time proximity routinely take different paths

Could take yes

wdym could take?

14:59

if they don't actually routinely take different paths, then there's no issue, no?

Depends on channel aggregation algo

15:01

Sometimes its just every consequent packet gets a new route

15:01

But some try to maintain a route

15:02

For a stream (ip src+dest combination)

https://www.openmymind.net/How-Unreliable-Is-UDP/

How unreliable is UDP?

Unlike TCP, UDP doesn't provide any delivery or ordering guarantees. Put into practice, just how bad, or good, is UDP?

Otherwise you would get into troubles with up sooner I guess

apparently quite bad about the ordering

15:02

check table # of ordered packets

Yes, but irl for internet backbone router it is cheaper to emulate streams of ip packets

15:07

That's why situation is not that horrible for us

15:07

Also 50 ms delay helps a lot

15:07

But the more congested the bandwidth for a server or a client the worse situation will get anyway

15:08

Maybe thats why there is magic 64 kb limit

15:09

I could emulate out of order delivery with some network tools and test if u curious

64k is just the size of the buffer, which is used for building the snapshots. What the server sends is usually way less, thanks to delta snaps and compression.

15:19

Ordering isn't a problem. Chunks within one tick are numbered. And chunks for older ticks are just dropped.

So it's just all chunks should arrive in a 50ms time window.

Yeah. Otherwise the client will just use the next one

https://github.com/Learath2/teedissect/blob/master/packet-teeworlds.c

Learath2/teedissect

Wireshark teeworlds dissector. Contribute to Learath2/teedissect development by creating an account on GitHub.

So it means that server will get problems with internet bandwidth of 10 Mbps if 64KB will be sent every 50ms

It's quite messy right now, but I need to take a break. Might aswell leave it as is for now

2

So I assume it's not gets full in general

Sending 64kb raw UDP data every 50ms to all clients isn't the best idea :D

Java keeps it's entire memory pool hot, so the kernel can't swap it away, ever, even if you rarely ever use the data. So it can have decent performance, at the cost of horrible memory usage

@Learath2 i think linux works around this with overcommit iirc

My server sends that easily with a lot of custom entities

15:25

But it's a sum for all players

Don't think so. The server only sends what the tee can see. And it only sends what changed since last tick

15:26

Debug mode has a field that shows the transfer rate I think

A lot of thing could change in my mod. I have circles drawn via particles moving in circles

15:26

And they update their position every snapshot

15:26

Though it was never a problem

15:27

I think I have a lot of headroom yet

Still... Delta snaps, int pack and Huffman can reduce that a lot

But would like to check it

@Ryozuki overcommit helps with reducing the usage of memory, it doesn't help with the memory never being able to get swapped out. So you still need a lot of ram

Delta cannot pack anything if every entity have different coordinates I think

Delta can only help if an object hasn't changed

But yes, it's not causing problems yet

15:28

Well before adding 20+ more entities

But it compares the old entity with the new one and only sends the difference.

15:29

Using int pack and Huffman makes that rather small

Projectiles essentially have nothing but a vel and pos which are not constant in sfx particles

15:30

I update them every snapshot to format a circle

15:30

And it looks nice

15:30

To form

15:31

And circle is rotating, I will show you how it look later

Still the coordinates don't change that much. It only sends the difference since last snap and not the absolute coordinates

Then compressions will help ofc

15:32

I am curious now to see actual size of my snap )

Most of the time int pack can reduce each of these integer values to one byte, since the difference since last snap is kinda small

It can bee seen in a debug hud, right?

Not the actual size of the snap but the average send rate I think

Server tick is 1/50 right?

15:34

So I can just divide it by 5o and get snapshot size.

Yeah but by default it only sends 25 times per second

Hmm... TIL

15:35

Ok thanks

You have a server running?

Protocol issues were a nightmare for such amateur modder as I am

15:35

Yes

15:35

Just I am not at PC atm

15:37

BTW if I am sending a projectile 1 snapshot but not on consequent, will it be removed by a client or will it be handled by a prediction?

15:38

I want to switch from pickup free taxi indicator to temporary projectile ones

projectiles don't even have prediction

Hmm... but they have curves? Or are they server-side only?

and the client only predicts what is included in the last snapshot

15:39

the curve is calculated based on the tick and the start pos+vel

15:40

has the advantage that the snaps do not change

Hm... so proj snap have start tick and start vel and pos and it does not change in later snapshots?

15:41

I see why it is compressed nicely

15:41

That means that client calculate ballistic isn't it?

15:42

(Server too to get hits)

https://www.reddit.com/r/C_Programming/comments/hkd0ze/question_about_what_to_do_when_malloc_fails/ interesting threas

15:42

Thread

@Deleted User infNext? with 2 rotating circles and a few tees on screen it's ~14 kB/s for me

15:46

You are also sending lasers and projectiles when there is nothing on my screen

15:48

the table in debug mode shows what is updated in the snaps

Yes, I don't know why it sends every reserved entity every tick

15:51

Thats a good find I've thought that wa normal in tw )

15:51

Thanks for testing

15:53

14 kBps on almost empty server would become much larger when special weapons will be used

15:53

But thanks for the starting point

15:53

I will investigate it further

15:54

When I add katana pickup above character head as a sign of activated taxi function it starts to lag... sometimes on 10+ players

15:54

So I was thinking I am getting some limits

15:55

Maybe I still get them, I'll check it

15:56

Thanks gods I have now chillerbot pentest tools so I can reproduce a load locally and not to wait for players

15:57

ChillerDragon thanks

Hover seems broken for the clear buttons of the filter lines

it should be fixed

16:35

i remember reporting it on a pr

16:36

it prob isnt merged yet

16:37

https://github.com/ddnet/ddnet/pull/2880#issuecomment-695760260

16:37

@Learath2 did u update ur fork

ah I was on a branch

https://www.zdnet.com/article/github-to-replace-master-with-main-starting-next-month/

GitHub to replace 'master' with 'main' starting next month | ZDNet

All new Git repositories on GitHub will be named "main" instead of "master" starting October 1, 2020.

YAY we finally did it boys, the ceo of racism is dead

@ᶰ°Konͧsti topkek

16:53

Existing repositories that have "master" set as the default branch will be left as is.

16:54

ill change my new repos to use master

16:54

fuck the system

16:54

monkalaugh

they should just donate to blm charities instead of changing stuff that's been like that for forever

16:54

if they want to help

its sjw public relations

sjw? xd

haha look a corporation cares about my rights haha

16:55

ill now proceed to love it

16:55

@louis sjw means social justice warrior

i know

social warrior of justice

what github is doing is purely Virtue signalling

16:55

i love this word

16:55

https://en.wikipedia.org/wiki/Virtue_signalling

Virtue signalling

Virtue signalling is a pejorative neologism for the conspicuous and disingenuous expression of moral values with the intent to enhance one's own image.

i like it

16:55

master sounds stupid xd

BLM isnt against racism, its literal a racist movement

16:56

White sjws like nouis dont check that

trollsti

what github is oding is like twitter people changing their pfp to the french flag when a terror attack happens

16:56

Virtue signalling

i think main sounds better than master as well but i heard it messed some things up?

the most cringy ppl

ok konsti send evidence as to why blm is racist then maybe i'll listen to u

https://github.com/openzfs/zfs/issues/10458

request: to not continue on this virtue signalling path · Issue #10...

hello, as a man of non-white background (a mix of brown, black, and white), the removal of "slave" and "blacklist" from the source code itself feels racist. i&am...

16:58

hello, as a man of non-white background (a mix of brown, black, and white), the removal of "slave" and "blacklist" from the source code itself feels racist.

16:58

lmao

kek

for me its not about racism or smth, master just sounds like the wrong word

why

16:59

i find it normal

i dunno main branch sounds more fitting imo

https://www.zdnet.com/article/linux-team-approves-new-terminology-bans-terms-like-blacklist-and-slave/

Linux team approves new terminology, bans terms like 'blacklist' an...

Linux creator Linus Torvalds puts stamp on proposal to use neutral language in kernel code and documentation.

17:00

ah

17:00

2020

17:00

a lovely year

i mean whats wrong with it, they rename stuff, bcs the words are stupid anyway

yeah lets use requester/responder

17:01

instead of master/slave

17:02

monkalaugh

17:02

"director/performer"

17:02

we are in a circus

but master slave is better? XD

Agree with @Deleted User , it seems a bit like fanboys and haters. I guess it's easier to care about such small things as naming than the more important things

strange argument

master slave feels normal

yes u are used to it

u can give any word a bad meaning

primary/secondary

17:03

sounds logical for me

master/slave too sounds logical

well idc about naming, aslong its intuitive

its like they pretend bits and bytes have feelings or something

Considering that "git" means idiot, all the other weird names fit in well

in what language does it mean that?

17:06

oh

17:06

english

17:06

TIL

ironical, bcs without git, our software projects were handled by gits xd

@Deleted User

17:11

Torvalds sarcastically quipped about the name git (which means "unpleasant person" in British English slang): "I'm an egotistical bastard, and I name all my projects after myself. First 'Linux', now 'git'."[24][25] The man page describes Git as "the stupid content tracker

17:12

Image attachment

17:12

monkalaugh

gitos

if you have 8 gits, do you have 1 gyte?

sendhelp

5

17:15

@Soreu btw have u tested the new editor

17:15

i need ppl to test, if i broke smth

what new editor, what am I missing again lately? xD

no not new editor sry, just the new rendering in it

17:15

https://github.com/ddnet/ddnet/pull/2894

Add new renderer(as streamed vertices) for editor by Jupeyy · Pull ...

Does not improve performance in editor!

17:16

the game layer looks funny

17:16

so clear xd

17:22

Do you think there are still ppl with 0.6 vanilla playing on ddnet servers?

a few, yes

ok, bcs today we had the debate with dilate again

17:25

0.7 uses premultiplied alpha colors

17:25

that saves a bit file size

a few KB probably don't matter that much

yeah, thats also why i added dilate to the editor

17:26

so nobody has to care

well, skin creators, game skin creators still have to care

true

17:28

@fokkonaut have u tried #2874 btw?

https://github.com/ddnet/ddnet/pull/2874

Add unmasked entities for all mods by Jupeyy · Pull Request #2874 ·...

Fixes switch tile for unmasked ddnet And ddnet mods send that they are a race and ddrace mod appearently? Not very intuitive that ddnet is a superset of them, bcs tiles changed. Waiting for a revie...

have you noticed that the game stays at pos 0,0 twice longer after you connected to the server for the first time (seems like initialization stuff going on). Comparing to 14.7.1 (edited)

17:30

it takes almost a second for me. Probably we should show loading screen litle bit longer

" that the game is about twice longer"

17:30

what does this mean

is there a loading bar for connecting to server?

17:31

or what do u mean by loading screen

17:31

or just uses more ticks before rendering

Image attachment

if that takes longer is probs bcs its building the map when onmapload is called

When I connect to the server my camera shows game at position 0.0 for a second

ah so after the loading screen?

yes

oh ok

17:33

have u tried on older ddnet servers? maybe its not client side

but it is "so slow" only for the first connect. When I reconnect to the server everything is fine

oh ok

yes I tried 14.7.1

17:34

the same but for 14.7.1 it shows about for a 0.3 sec

yeah nvm thought server side, buts its only once, so nvm

yeah I need to restart the game to reproduce (edited)

[ddnet/ddnet] Pull request opened: #2899 Editor: Select game layer by default

Since it's the most important layer

[ddnet:master] 2 new commits

3276f29 Reload skin menu when skin count changed - def- adf4103 Merge #2895 - bors[bot]

17:50

[ddnet/ddnet] Pull request closed: #2895 Reload skin menu when skin count changed

[ddnet:master] 4 new commits

f9e66f0 Add edit box to write skin name - def- d17e0e4 Add link to Skin Database - def- a273969 Fix clang-format - def- d61c973 Merge #2896 - bors[bot]

17:58

[ddnet/ddnet] Pull request closed: #2896 Add edit box to write skin name, Add link to Skin Database

@Deleted User not yet, but from the code i read on the train i think its good

18:06

i will try it rn

ty

does anyone have png exports of all ddnet client-shipped skins

https://github.com/ddnet/ddnet/tree/master/data/skins

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/ddnet

like, rendered

18:15

like, rendered

what

@Deleted User works, thanks

alright

@plsplsplslol like how it would look ingame

you could download client and just look at skin list

@louis

18:22

Image attachment

18:22

like this?

18:23

if yes: https://ddnet.tw/skins/

18:23

that page uses the skin aseembler iirc

18:23

its somewhere open src i believe

18:23

Image attachment

18:23

actually pretty nice to use as "emojis" in discord

oh neat

np

but then you have to find which ones in the skin database are in ddnet client

true

if they are shipped with DDNet client by default, they are in database too

18:35

Image attachment

ye but it also includes extra

[ddnet/ddnet] Issue closed: #2834 Select game layer on map load in editor

20:30

[ddnet:master] 2 new commits

ddb88c6 Editor: Select game layer by default - def- 481a56d Merge #2899 - bors[bot]

20:30

[ddnet/ddnet] Pull request closed: #2899 Editor: Select game layer by default (fixes #2834)

Jupstar: Your theory-patch worked. Thanks. :)

nice

Jupstar: The patch fixes 1.1.0 internally, but it uses a 3.0 context. Should it use more features or did it not really anyway?

[ddnet:master] 2 new commits

5208a58 Add unmasked entities for all mods - Jupeyy 9e23623 Merge #2874 - bors[bot]

20:56

[ddnet/ddnet] Pull request closed: #2874 Add unmasked entities for all mods

i think sdl just defaults to EGL 3.0 or smth, and ddnet just shows it as opengl 3.0

20:59

or did u not change the config?

20:59

then it just creates EGL 3.0 with compability

20:59

but that should be fine too, internally the patch should only call 1.1 functions (edited)

[ddnet/ddnet] Pull request closed: #2897 After initialization: Agressively try to grab window again

21:17

[ddnet/ddnet] Pull request opened: #2900 Version 15.0.5

Still needs to go through Release Candidate testing.

i see a month passed monkalaugh

monkalaugh

21:22

joke xD

[ddnet-web:pr-15.0.5] 1 new commit

c11079a Version 15.0.5 - def-

21:22

[ddnet/ddnet-web] New branch created: pr-15.0.5

21:23

[ddnet/ddnet-web] Pull request opened: #100 Version 15.0.5

Still needs to go through Release Candidate testing.

we just need the skin fix, thats why xd

21:24

@deen want the home screen config in? on kog there was another person complaining bcs he didnt find the setting (edited)

@Ryozuki it's supposed to be a bug fix release, i don't think the big changes you talked about made it in

ah

And I didn't see any super-high-risk changes, so I didn't want to branch off

https://github.com/ddnet/ddnet/compare/15.0.4...master

Not sure what you guys meant by "not rushing". I can also delay the bugfix release if that makes the community happy

honestly most ppl dont even notice when on steam

21:30

its always 10 ppl complaining

I considered not doing the bugfix release since the skin bug is not too bad

well maybe a bug fix can be merged fast, but big features should definitly be tested more i guess

might crash, but haven't heard anyone say so

the best would be to have someone with a old pc able to test too tho

@Ryozuki even with testing edge cases arent always seen easliy

i know

21:32

but the skin download bug

21:32

it being static

e.g. for the skinprefix it had to exceed a specific length

was it rly that hidden?

i'd say not very intuitive

21:32

u could have found it sure

@Ryozuki let's see how many bugs the release candidates will find in the future

21:33

hard to predict and depends on how many people participate and how thoroughly they search

okay

21:33

i will definitly try to test

21:33

but for example if jupstar makes a graphic change we should definitly get someone with a old pc to test it

... i tried

21:34

but most ppl prefer not to answer

21:34

only this brazilian guy did

justatest

and then it was quickly fixed

u cant simulate a old pc right?

21:34

forcing openg 1 would not be the same

i cant simulate broken drivers, no xD

or smth like that

21:34

xD

well, most players are 12-16 year old kids, so it's no surprise they are not the best bug reporters

https://github.com/ddnet/ddnet/pull/2894/files this definitly looks like a big change that may need to be tested xd

well i valgrind and test few cases always

21:36

and tried all opengl contexts

not in the release

yeah i know

@Deleted User i just got a weird error starting client

4.73 KB

i what version

hm, rebuilding client fixed it. Maybe I just mixed some old client version with new shaders

21:38

nevermind then

oh ok

21:38

yeah makes sense

21:38

i use the shader for opengl 3.3 now too

this x has never been centered right? xD

Image attachment

btw i added text render flags for such things

21:39

wait

21:39

TEXT_RENDER_FLAG_NO_X_BEARING = 1<<0, TEXT_RENDER_FLAG_NO_Y_BEARING = 1<<1, TEXT_RENDER_FLAG_ONLY_ADVANCE_WIDTH = 1<<2, TEXT_RENDER_FLAG_NO_PIXEL_ALIGMENT = 1<<3, TEXT_RENDER_FLAG_KERNING = 1<<4, TEXT_RENDER_FLAG_NO_OVERSIZE = 1<<5, TEXT_RENDER_FLAG_NO_FIRST_CHARACTER_X_BEARING = 1<<6, TEXT_RENDER_FLAG_NO_LAST_CHARACTER_ADVANCE = 1<<7,

21:39

u can basically enable almost all

21:39

and it should be centered more

21:40

ok noit the first two xd

21:40

and maybe also not the 4th xd

monkalaugh

but they can controll things for single characters

21:40

i added them for the gui icon font

always when i set fsaa to 16 and restart it resets to 8

21:42

not related to this release tho

probs ur driver doesnt allow more

21:42

mine just completly ignores it for some reason

@Deleted User is multisampling enabled?

u mean in shaders?

src/engine/client/backend_sdl.cpp 4262: SDL_GL_SetAttribute(SDL_GL_MULTISAMPLESAMPLES, FsaaSamples); 4267: SDL_GL_SetAttribute(SDL_GL_MULTISAMPLESAMPLES, 0);

21:45

justatest

yes

21:46

xd

is this really fsaa?

21:46

it looks to me like its MSAA

Should we have a separate channel for reporting bugs (for the release candidate)?

but i dont know much about this

21:47

yeah why not

i think MSAA is just one possibility

21:47

you can also do super sampling

21:48

e.g. for the ffmpeg it might be nice to have framebuffers at higher resolution, so we can export in any format

Most modern GPUs support 2×, 4×, and 8× MSAA samples.

21:48

so the 8 cap makes sense

21:50

put it under #developer

21:50

i guess

the only problem i see is, that announcements will be flooded by releases, while its more for actual announcements

you can create another category, "development" which can have 3 channels, announcements, discussion, bugs, if u do that u should move this channel and rename it to discussion i guess, to keep the webhook

21:59

or maybe its not needed

yeah sounds good

22:00

well dev discussion xd

22:00

but that would split it right?

22:00

from chat

yeah

ok

thats what a category means

22:01

its this thingy

Image attachment

yes

22:01

i hoped that discord allows subcategories or smth

it doesnt

22:02

xd

22:03

@Deleted User did someone test 8?

22:03

i dont even know what that mask means xD

fokko

22:03

its just that a mod can decide to show invalid tiles

i see

22:08

any idea how to trigger Reload skin menu when skin count changed

it does auto

the interesting case was when the size of skins doubled

22:09

so you could remove all skins, start client, open menu, connect to a server with many players and it would bug

ok ill try

but u have todo smth different than us xd

skins are saved on .teeworlds right?

something we didnt have thought of (edited)

not in the data folder

both are taken, same as for all

downloaded in .teeworlds

looks like it works

22:11

btw before i deleted downloaded skins i had some interesting files there

Image attachment

22:12

br bro brow brownb brownbe

22:12

funny evolution

the tmp files should also get deleted with the new update now

yeah i dont see any now

[ddnet/ddnet] Pull request opened: #2901 Remove failed downloaded files immediately

Especially skins were only deleted after they were requested again

I should code a graphics driver, it would be buggy asf so we can test this stuff

feature request: if a player has dummy connected before a server restart, dummy autoconnects after the restart

23:09

if the map is the smae

23:09

same*

doesnt server restart when the server is empty

is there such a thing as a moderator in game that can punish people that see as fit

@plsplsplslol for mapping i do f2 restart to update the map

23:14

@Siya yes, you'll have to ping them in discord using @ moderator and give them the names of the people who are misbehaving (or the server region and map name)

Well is there a way i can be a moderator i wanna help people because i see trolling 24/7

there's some info here: https://forum.ddnet.tw/viewforum.php?f=69

23:16

if you're new you might want to get acquainted with the game+community a bit before you apply

ive been in the game for about a month maybe

23:19

Is there an age requirement? @louis

i don't think so

There's a little bug on the players profile page: When i click on the difficulty rating of one of the last finished maps, it tries to redirect me to a none existing page. e.g. https://ddnet.tw/players/Wollwolke/novice/

Exported 706 message(s)