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 2021-03-16 00:00:00Z and 2021-03-17 00:00:00Z

Deleted User

nice, finally i can see how good newer hardware compares to mine

interesting results already, rtx 2080 ti beats my gpu by factor 2 there is also an AMD laptop benchmark: https://openbenchmarking.org/result/2102211-PTS-XONOTICD79 Opengl 3.3 is slightly slower than 3.0 there For me its the opposite: https://openbenchmarking.org/result/2103154-HA-JVEGA64RY23,2103152-HA-JVEGA64RY92 would love to see GTX 1080 TI(bcs my GPU was released as a competitor against that) and ofc i'd love too see the current top high end cards

@fokkonaut the values only changed to me after trying to do anything with dummy, like changing skin and other things

05:25

then entire config gets messed up

thanks, will look into it

@fokkonaut does the bug occur to you too?

I'll be on my pc soon

06:41

I hope so

06:41

then I could fix it

06:41

And I can imagine it happening with dummy skin, I had a bug there before, fixed it, and if its broken after a big master update again that would make sense

Souly

@fokkonaut does the bug occur to you too?

Thanks that you responded, it does.

fokkonaut

Thanks that you responded, it does.

does it like randomize the skin or something?

07:21

when you are trying to change one skin part

no, im trying to find the issue right now

i remember that they added a skin randomizer recently to vanilla and maybe that could cause it

nop

Hm, cant find the issue that quickly

07:40

need to do something else first. will look into it later again..

deen

@Deleted User btw, phoronix benchmark should support higher fps now: https://openbenchmarking.org/test/pts/ddnet-1.1.0

wow such low fps

09:12

i get 1500 on my rx580 lmfao

Deleted User

i get 1500 on my rx580 lmfao

at that specific benchmark? its pretty GPU intensive

09:45

ah

09:46

why even becnhmark teeworlds tho...

09:46

like

09:46

who tf needs more than 240fps

09:46

cant play with 240 fps xD

me

also its not so much about fps, but also about frame times, which make the game feel very smooth

09:47

I know

09:47

but like let me guess most of the people here dont have more than 144hz

09:47

yes, but if you dont play with gsync or freesync, then you should have more FPS than ur monitor can output

09:47

1600fps and 0.5ms frame times that is what I have

Deleted User

yes, but if you dont play with gsync or freesync, then you should have more FPS than ur monitor can output

i mean sure but 1500 fps and 500 dont make a difference lmfao

09:48

i'd say > 1000 its pretty stable xd

09:48

i will cap my fps and see the difference

09:49

bet its not even noticable

09:49

also unless you play on a 2004 xp machine you will get more than 200fps

09:49

i notice < 1000 fps

do u have a 144hz

nah but I know how this works

09:49

i play on 60hz

09:50

also the sense of benchmarking is more, too compare hardware

with entities and without health HUD, i get 10000fps

monkalaugh

Image attachment

09:50

monkalaugh

Image attachment

Deleted User

also the sense of benchmarking is more, too compare hardware

with entities and without health HUD, i get 10000fps

teeworlds isnt really the best game to compare hardware lol

09:50

depends

09:50

bruh

09:50

no

09:51

in the end usage is usage, just bcs u might not use specific pipeline features, doesnt mean it doesnt mean anything

09:51

my benchmark wouldnt be fricking teeworlds lmfao

09:52

many cpu benchmarks also just test a single thing

09:52

sure id do it for fun but for actual results nah

09:52

openbenchmark is a collection of different benchmarks

09:52

so it makes sense imo

09:52

sure ig, just not worth it if you wanna get actual results on how your pc performs

09:53

again, depends on what u do on ur cpu

09:53

its like u say ARM sucks, because x86 has stronger floating point calculations

09:53

but youtube doesnt care about floating point performance

09:53

dude

09:54

im just saying

09:54

benchmarking teeworlds for actual results

09:54

is not really efficient

09:54

id benchmark something else

09:54

and i say its part of a collection of benchmarks

09:54

which then makes sense

Deleted User

sure id do it for fun but for actual results nah

and i said this

09:54

if u have a GPU that rules everything but sucks in teeworlds, it still shows there is smth wrong

09:55

does such gpu even exist lol

09:55

Embedded image

09:55

why does rtx 2080 super have so bad fps?=

09:55

it could be a bug in the driver, or smth similar

09:55

just the driver, its more than obvious id say

09:56

so it was useful it was tested

09:56

lmfao

09:57

i suppose so but still, there would be something way more convinient than checking/doing teeworlds benchmarks

09:57

well then think what u want

09:57

lol

09:57

its true

09:57

give me a mathematical proof and i believe u

09:58

what do you mean, what equation should I do xD

09:58

that teeworlds benchmark is not useful

09:58

proof that fact based

09:59

1. teeworld is so unpopular no one would even check that first 2. you could do any other opengl benchmark to check for drivers issues

09:59

I am not saying you should not do the benchmarks, you can, I couldn't care less

09:59

if you do it just to collect more info into a database and compare then that's on you

09:59

I am saying in reality it won't be really applicable universally

09:59

unless you want to check for specific issues with the game

[ddnet/ddnet] Pull request opened: #3710 Add variable length arrays to protocol

This PR adds the posibility to send variable length array of the same netmessages (well, the object the message represents) I added a PlayerTimeMessage and Sv_PlayerRanks as example, they can be removed before this pr is merged. I made this because this can solve problems like the m_pDescription0 to 12 limitation and also because I think this would be needed if we add a proper netmessage for querying player ranks for showing them in a proper UI. It uses std::vector, but we are in mod...

i still disagree, i'm ok with you calling it more unefficient then doing other designed benchmarks, but in the end its just an addition, it doesnt need to be used, but it can

10:01

Then you agree with me that is exactly what I said

10:01

it's just an addition you can do it

Deleted User

I am not saying you should not do the benchmarks, you can, I couldn't care less

.

Deleted User

if you do it just to collect more info into a database and compare then that's on you

.

10:01

that

10:02

It's cool, as I said it just won't be as good as other more popular efficient becnhmarks just like you said

10:02

just as you could not use other benchmarks

10:02

hm?

10:02

as i said, it depends on what u want to know

10:03

and I said that too

Deleted User

unless you want to check for specific issues with the game

.

10:03

like this

10:03

then the benchmark is useful

10:03

its nice that u said that, i said that from the beginning, when you said it is not useful

this convo feels worthelss

10:03

less

10:04

monkalaugh

no conversation is worthless, it might be pointless but not worthless

well yea

10:04

pointless then

10:04

but isnt something pointless also worthless?

no lol

worthless adj (without value)

10:05

pointless adj (without use, goal)

working 100 hours for a dollar is pointless but not worthless

10:05

because the dollar is worth something

10:05

your dignity and 100 hours ;D

xD

10:06

@Learath2 are u here

Deleted User

its nice that u said that, i said that from the beginning, when you said it is not useful

can you quote it? I mightve missed it

Ryozuki

@Learath2 are u here

Yes

Deleted User

teeworlds isnt really the best game to compare hardware lol

read from here

@Learath2 i insert the size of the vector beforehand

10:07

and when unpacking i get it first

10:07

is it wrong?

10:07

or what do u mean

Deleted User

read from here

I said no as in it is a useless benchmark outside of super specific areas, as a general benchmark it is useless

10:08

"useless"

Deleted User

I said no as in it is a useless benchmark outside of super specific areas, as a general benchmark it is useless

dont lie, i said "depends" u said "bruh no"

10:08

I am not lying

10:08

also getting a bit passively agressive, bud?

10:08

this discussion is over, i stay to facts.

10:08

xD

Ryozuki

@Learath2 i insert the size of the vector beforehand

There is a max packet size, beyond that you need to chop it up

ah

10:09

well i dont check that

10:09

hmm

Calm down man, no need to get pissed just because someone is saying something else different from your own opinion

So for a proper VLA your pr looks too simple :D

xd

10:09

i mean

10:09

its fine as long as it stays within the max packet size right

and we both share the same facts -- you got what I said wrong

@Learath2 how could i check that?

Um, the packer has a size method

i mean maybe chopping is easy but what about the remaining data xd

Chopping is not trivial

well if the object within has variables with non defined sizes yeah i guess

Even with defined sizes it's not that easy

10:12

You need to modify the client to hold back incomplete chunks

rip

Or you can maybe do template magic to generate a _MORE packet

@Learath2 are cnetmsgs put together in one big buffer or are they sent one by one

They are put together

monkalaugh

What we call messages are in CNetChunks which we pack together to get a CNetPacket

justatest

If your chunk is larger than how large a packet can be - the header overhead, it'll never fit in a packet

10:16

Dunno what happens then :P

i mean, we can make it so that a single netobjects never goes beyond x packet size

10:17

and then if it cant fit in the current packet

10:17

put it at the start

10:17

of the next

10:17

ez

Yes, but then how useful is this feature

yeah not as useful

Anyway, I'm not sure. I'll let other people comment on it too

10:18

I wanted to do something like this too to get rid of the chunking logic we have sprinkled around the code

10:19

Like votes, rcon commands, chat commands

ye

But I never figured out how to chunk it in a pretty way :P

10:24

@heinrich5991 sometimes gets good ideas for this protocol stuff

@Deleted User any idea about this? https://steamcommunity.com/app/412220/discussions/0/4366772972649191150/?tscn=1615868366 game freezes totally, but gfx_opengl_major 1 didn't help

my friend needs help :: DDraceNetwork General Discussions

so my friend can't play ddrace network i mean like he can run it but he gets frozen and it says connection problems i really love this game but it sorta boring by your self and well i wanna play with him and can't anyone know the fix to the problem if yes please help me and him.

Connection problems doesn’t sound like a gfx issue, no?

deen

@Deleted User any idea about this? https://steamcommunity.com/app/412220/discussions/0/4366772972649191150/?tscn=1615868366 game freezes totally, but gfx_opengl_major 1 didn't help

wasnt that the intel driver bug we talked for around half a year now xd

12:38

i'd love to know of older clients work 100% without the crash while, the newer dont work with gfx_opengl_major 1 bcs if so, it must be some patch between these versions. Else it could still be a windows update or smth similar

12:38

but i'm pretty sure the gl 1.x renderer itself didnt add any feature it didnt have before

12:39

i only added the fallback code, but it only does stuff for >= gl 2.x 1.x is the absolute basic, it disables everything and does nothing special

12:40

maybe a SDL thing that only triggers on intel platforms, or GLEW loading something intels dont like or smth

12:40

would be easier to say if older clients actually dont have problems at all :/

12:41

and i also dont even know if the newest intel driver now actually works

deen

@Deleted User any idea about this? https://steamcommunity.com/app/412220/discussions/0/4366772972649191150/?tscn=1615868366 game freezes totally, but gfx_opengl_major 1 didn't help

ah and also, the question is, if he restarted the client after entering the command

12:48

u always have to assume the ppl do it exactly as you dont want them todo xd

ddnet

Suggestion ddnet

ddnet

new system messages (see the image) 1 -hold "u" (show chat) for see it to not coverage the screen when you play 2-putting images for every effects like ,unhammer,endless etc... until the save if you have one

Image attachment

2

1

I like the idea of having status symbols for your current endless hook, etc. state

14:17

we might even have an issue for that already

ravie even had drawn icons (edited)

I don't understand (1). do you mean we should remove system messages from the chat display when pressing u?

u is already binded to show chat

14:20

bound*

heinrich5991

I don't understand (1). do you mean we should remove system messages from the chat display when pressing u?

i mean put system message above the chat no inside

ah. my first instinct tells me that I wouldn't like that, because those messages usually have an ordering related to chat messages

14:24

e.g. bye xyz, *** xyz has left the game

we alr have an issue for the icons

heinrich5991

e.g. bye xyz, *** xyz has left the game

yes i say this message like this leave it in chat

maybe put server messages like ***you are in a solo part under the icoins in the top left

ah, at that point you'd probably remove the solo part messages and only show the icons

heinrich5991

ah, at that point you'd probably remove the solo part messages and only show the icons

i said this too

m but new players might not know what the icons mean

louis

m but new players might not know what the icons mean

for first 5 second

also for ppl who dont want the icons u still need the messages

my plan is make division system message 3 parts 1-2-3

louis

also for ppl who dont want the icons u still need the messages

we can leave the old system message like old chat

14:29

go click f3

no

14:30

having 3 different sections on one side of the screen is messy

14:31

and if the /top5 and /points stuff is changed then might as well make an actual UI for it rather than keep it in the ugly chat form

louis

having 3 different sections on one side of the screen is messy

it's now in one side (in chat) i want fix the messy

louis

and if the /top5 and /points stuff is changed then might as well make an actual UI for it rather than keep it in the ugly chat form

you can do this too not nesscary do what i say Literally

14:36

place for top5,rank ... and place for left game... and place for effect icons

@Learath2 actually, since various netmsgs are packed in a same packet we can do variable arrays in a different way taking that into account:for example if the player requests top x ranks via the protocol, make that netmsg have a response id, then the server sends x netmsg each having the info about 1 rank + the response id (edited)

14:37

+ a way to know the total amount of ranks

14:37

The ranks is an example to understand it easier

14:38

Writing a wall of text on mobile sucks

I was thinking of abusing the chunking too. I'm not sure how to make either side pretty like that though

14:39

Ideally it shouldnt look any different to sending or receiving any single message

[ddnet/ddnet] Issue opened: #3711 Retry hdiutil create if it fails due to "Resource busy"

If CPython doesn't do anything more sensible, we don't need to do that either. :P https://github.com/python/cpython/commit/0133f9fc9e8a35b223341d7b5641ac2f16ff2a19 https://github.com/python/cpython/pull/11333

louis

m but new players might not know what the icons mean

we can do both things putting icons and writing in chat (solo,unhammer...

louis

m but new players might not know what the icons mean

We so need an introduction for the UI

Is it possible to translate system messages to another language?

no. there's an issue about that

15:09

https://github.com/ddnet/ddnet/issues/3091

Client Side translation of Server Messages · Issue #3091 · ddnet/dd...

The server would send a packet instead of a chat message with a id and arguments, then the client would display the translated messages with those values. We would then need our own printf parser a...

15:10

@murpi can you explain what you need it for, right now?

oh ok

Image attachment

I've been playing alot on asia lately and I feel like people don't understund the system messages very well, like jetpack and such

maybe doing /points player could bring up a display like that

huh, I like it

@louis

the only issue i see is that it wouldnt be serverwide anymore

justatest

with a new scoreboard we could show the points count of every player at all times

[ddnet/ddnet] Issue opened: #3712 Suggestion: new system messages

3 places for the Different system messages to make system messages arranging and clear (see the image) !sm 1-(putting system message above the chat not inside) and you have to hold "u" (show chat) for see it to not covering the screen when you play 2-putting icons for every effects like ,unhammer ,endless etc... until the save if you have one 3- this messages usually ha...

2

1

thanks for creating an issue

I wanted to create a second scoreboard with ranks and such

15:41

Hence my pe abour VLA

15:41

Pr

I want to transmit that stuff out of the protocol tbf

15:43

We'd need to add so many new messages and stuff. I wanted to just query https://info.ddnet.tw/mapinfo?name=<ourname> to get a json with all the relevant stuff

this would make it main-ddnet-servers only feature, I wouldn't like that

Then someone else will have to do it. I don't have it in me to do yet another ad hoc implementation of chunking for both top5, points, leaderboard, regional leaderboard and whatever else we decide we want in the new scoreboard page

15:47

+ it's just asking for something to go wrong, I'm okay with not being the one that wrote the scoreboard that broke ddnet by flooding connections or finally made the networking code not performant enough to push out all the chunks

15:47

we had it with votes and rcon before, not sure how much more we can throttle, rcon completions already arrive at a snails pace

15:49

Ah and a mapchange will definitely timeout everyone logged into rcon without throttling for this stuff

Learath2

we had it with votes and rcon before, not sure how much more we can throttle, rcon completions already arrive at a snails pace

oh yeah why do rcon completions appear so slow

Because the protocol is just not made for this kind of bulk information sending

all continents except europe start with a

16:22

(or eurasia)

and end with a too :p

16:25

oh well i guess north/south america technically starts with n/s

Continent

m i was taught that north america and south america are 2 separate continents

according to wikipedia, there are anywhere between 7 and 4 continents

[ddnet/ddnet] Pull request opened: #3713 Retry hdiutil create if it fails (fixes #3711)

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...

[ddnet/ddnet] Issue closed: #3711 Retry hdiutil create if it fails due to "Resource busy"

17:25

[ddnet/ddnet] Pull request closed: #3713 Retry hdiutil create if it fails (fixes #3711)

@deen, why the heartbeats requests is in UDP and not TCP?

17:28

I am having trouble creating a reliable filter for the UDP packets due to the fact that I break heartbeat requests by doing so, would it be possible to use a different port for heartbeat or send it in TCP?

the heartbeats require the same UDP port that connecting players use

17:30

if you can't receive heartbeats, players can likely not connect to you

Yes they can.

in what way do you distinguish between players and the heartbeats?

[2021-03-16 18:35:35][register]: WARNING: No master servers. Retrying in 60 seconds

17:36

IP: 45.141.57.5:27777

17:36

No heartbeats but I can join and play without problems

what kind of filter do you use?

17:37

why does it filter the heartbeat packets specifically?

Because apparently there are types of attacks that use the same traffic as heartbeats

17:37

I'm using my own filters + voxility

so you can't control the filter. maybe contact voxility that they block your packets

17:38

(in the future, we might have HTTPS masters, but even these will likely still check for UDP connectivity with your server)

17:39

I can't connect to your server btw

Voxility cannot "whitelist" those packets because they are similar to an attack that can cause problems

17:40

@heinrich5991 it is strange, I'm in the server currently

heartbeat packets can be distinguished by being outgoing from your server

Voxility doesn't check the outbound traffic unfortunately

I mean it’s completely reasonable that you can join if their automatic rule is fw**

(I can't join)

Good thing we never tried voxillity. Massive links but as with all other providers apparently they’ll protect themselves even if you are within the limits of your link

Its cheaper to get rid of us

17:42

monkaS

Is this not some sort of fraud btw? I guess they bury it somewhere in the contract that they can actually limit the link however they want fuck you?

"[...] reserves the right to filter to protect the network [...]", probably

Such silliness. Here you go a 10g link, but you can only use as much as we feel like

17:44

When are we getting our own backbone link and datacenter? ;P

17:45

Anyway, Davide if you want your server to appear on the masterserver list it needs to reply to heartbeats, no way around it

Okay

17:49

@heinrich5991 can you try here 45.141.57.6:3000 ?

17:49

If you can join now

However yes, an https heartbeats would be best

but that's only for the far future

Rip

17:57

[2021-03-16 18:56:29][register]: WARNING: No master servers. Retrying in 60 seconds

17:57

Connection with players possible but heartbeats no

17:57

xD

No master servers suggests that either your cou2s or the servers siz2s don't make it through

18:01

I don't think it's the heartbeats since you don't even start sending beats until you choose a master

Mmm, that's probably true, what kind of method do you use to choose the server?

18:02

Ping? (I don't think)

18:04

https://sc.lowhosting.org/itAAMzMqwLJr all seems fine with pings

Embedded image

We do cou2 (get count) to all masters, we get siz2 (count) from the masters, after a timeout or if we get all counts we pick the one with the least count, then you start sending bea2 (heartbeats) periodically

in that stage we're happy if we receive any answer at all

The master server will reply with a fw??, the server should reply with fw!!, if your fw!! gets through you'll get a fwok if it doesn't you'll get a fwer

18:07

If you'd gotten to the fw check stage you'd get a port forwarding error or if you never get the fw?? a "switching master"

18:08

lmk the ip of your server and I can tcpdump on the master to see if your cou2s are getting through

Is there a way to force the server "selection" and go with a statically chosen server? (maybe from the config)

(but I'd guess that is highly unlikely to be the issue and it's most likely an issue with "siz2" never making it through voxility's filter)

The outgoing bandwidth is not managed by Voxility but by me, so the packet comes to you for sure, the problem is the answer from you to me

I don't actually know if you can force the server to use a master, masters.cfg is such a stupid file that I ended up just modifying the register code

18:11

@heinrich5991 you've been looking at the register code too, do you know how to force a master?

Embedded image

18:14

This is the packet that Voxility filter

Yeah, that's part of the register process so if you never receive that it won't work

18:15

you can ask them to maybe whitelist it from the ip of the master you choose?

18:16

Or you can try modifying the register code to skip the count check

Unfortunately Voxility has no Whitelist for IP due to the possibility of spoofing

FWIW their filter is breaking, it's their fault. I can only tell you the facts

what's wrong with that packet?

(honestly it's quite silly too, it's like filtering out SYN and saying it mitigates attacks, filtering all legitimate traffic will indeed stop an attack lol)

Voxility is a strange scrubbing provider, I already know that, but it is one of the only ones that can counter truly gigantic attacks

They seem to "counter" it by blocking legit traffic, any hoster can do that. Try Linode they'll null route you even, 100% solution to all attacks

Comrade

what's wrong with that packet?

Voxility's smart firewall or "smart" network engineers think it's an attack

Hahaha, btw the server is playable.

For now. Next time you'll get an attack that's part of the core protocol and it won't work anymore

18:23

Back at nfoservers they had to remove an automated rule like that all the time for us

All the rules are active, if I receive an attack nothing change

I don't get it is this rule against siz2 not something they added because of an attack?

18:29

It's an essential part of teeworlds protocol, it's like blocking all SYNs because people do SYN Flooding. That's silly imo

No, it has always been blocked

18:30

It is not something due to an attack

18:31

It seems that Voxility has always seen this as illegitimate traffic

I'd guess it's a reaction to a reflection attack against someone else. They don't randomly make rules like this

18:32

Letting them know it's legitimate traffic is all you can do

imagine world where edge routers of all autonomous systems are configured correctly so no packets with spoofed source IP can reach the global internet

I would say it is an almost impossible hope: D

maybe IPv6 is the way because many "admins" refuse to use it

Isps*?

@Learath2 you can force a master by making all the other masters point to non-functional addresses

Image attachment

20:47

poggers

Dark theme for TWs?

what's that?

another theme for github

20:49

i dont like it

20:49

i prefer the dark one

20:49

https://github.com/settings/appearance

20:49

(you need to enable feature preview)

https://system76.com/laptops/pangolin?its-the-same-url-but-the-content-has-changed-thank-you

System76 laptops are custom-built to run Linux flawlessly with the latest technology and super fast solid state storage.

21:00

Operating System Pop!_OS 20.10 (64-bit), Pop!_OS 20.04 LTS (64-bit), or Ubuntu 20.04 LTS (64-bit)

21:00

+100 for this

https://jdk.java.net/16/

21:07

java 16 is out

21:07

monkalaugh

Possible to block hooklines from some players? (edited)

21:24

Its annoying if they run always with the hookline

cl_show_hook_coll_other 0 but that affects all (edited)

Possible to add like block emotes ?

@Deleted User not the Intel problem, he has an AMD CPU, so I really doubt he has Intel graphics: https://steamcommunity.com/app/412220/discussions/0/4366772972649191150/?tscn=1615920752

my friend needs help :: DDraceNetwork General Discussions

so my friend can't play ddrace network i mean like he can run it but he gets frozen and it says connection problems i really love this game but it sorta boring by your self and well i wanna play with him and can't anyone know the fix to the problem if yes please help me and him.

Mhh I better wait for a response, he'd be the first

23:07

BTW the console output also shows the GPU vendor

Exported 358 message(s)