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 2018-12-07 00:00:00Z and 2018-12-08 00:00:00Z

[ddnet/ddnet] Pull request opened: #1395 Fix laser/rifle bounce delay, when server time(server ticks) are high/huge

Caused by unprecise floating point arithmetic (m_EvalTick can be high and is casted to float). Generelly a cast to double wouldn't hurt, e.g. if the tick speed will be changed in future.

I must say, banks have one of the most broken corporate structures that I've ever seen

Why?

You can complain all you want, the branch manager can be apologetic all he wants, but if he can't communicate with the internal department responsible for your problem your problem won't get fixed

12:30

No way you can directly talk to the internal department either

12:31

So complaining is pointless, you just submit a request and wait until the responsible person decides to read complaints

be prepared to wait at least 6 months to get what you need

luckily I know a guy who knows a guy

12:48

It's either that or you need to have a ridiculous amount of money in your account (which sadly I don't)

that helps

Oh well, luckily I never had problems with my bank and they even correctly detected and reimbursed me when someone stole my credit card info (might be DDNet server related, tried way too many hosters)

13:16

Not sure if that's just your bank or all though

yo guiis

13:20

can some1 explain snapshots?

13:20

I wonder how ofthen the CGameClient::OnNewSnapshop() function gets called

hey, no idea i'm not familiar with that, you could either ++ a static var or use tools like hmm i don't remember is it gprof ? gcov/lcov lvoc something like that

You get one snapshot per ticks that could be processed before the next ticks start time

13:37

That is unless you are connecting or recovering and your snap rate is limited

lol all this code gets executed every tick?

13:42

It iterates over all skin names and compares them for evey player?

No, it gets executed every couple ticks

13:42

but it could potentially be executed every tick yes

how is the client still that fast haha

In the runloop the server gets the current time, then until it's time for the next tick to start it'll keep ticking. As soon as it's time for the next tick to start, it'll snap. And unless you are a rate-limited client you'll get a snap

So the client should get slower the more skins you have i guess

Because languages like C/C++/Rust are no joke, they have a ridiculous amount of optimisation that goes into them and they are blazingly fast

I thought the skin graphics are saved on join and on change

A computer can do a lot more then one might imagine after using a language like Java or something interpreted like python

Because i want to add something similar to a skin system to my client without slowing it down :c

13:46

its like a second skin

13:48

So i am save with iterating over a big array of strings every tick and compare it agianst all the playernames

Where in OnNewSnapshot do we iterate every skin?

m_aClients[ClientID].m_SkinID = g_GameClient.m_pSkins->Find(m_aClients[ClientID].m_aSkinName);

13:49

i think here

13:49

if i understood it correctly

13:49

find does stuff and then calls FindImpl

13:49

int CSkins::FindImpl(const char *pName) const
{
    for(int i = 0; i < m_aSkins.size(); i++)
    {
        if(str_comp(m_aSkins[i].m_aName, pName) == 0)
        {
            return i;
        }
    }
    return -1;
}

Yeah i see

13:49

It's for each snapped player mind you

13:50

so only players close to you

oh so snapped players are only close players?

13:50

yea so all visible players i guess

13:51

i play a lot whit showall on full ddnet servers

Also only for players that changed Clientinfo

wait only if they changed it

13:51

where do i see that

13:52

if(Item.m_Type == NETOBJTYPE_CLIENTINFO)

13:52

ah here?

Snapshots are deltas

13:52

they indicate change

what ever that means xd

Which word do you not understand in that sentence?

delete but i get it now

13:54

delete is change i rememberd something from school

13:54

delta*

13:54

not delete xd

The server sends delta snapshots but the client reconstructs the full snap before processing it. But well... calling this code 25 times per second isn't that much

k thanks @redix

What the actual fuck is wrong with windows? .....

@Learath2 what do you mean

19:49

everything is fcked

19:49

october up, edge -> chromium, ....

I'm trying to get the stupid utf8 output working on console

Not possible I think

19:59

maybe we can just create our own console-like window in sdl2?

good luck, from what i saw it might be done but with big caveats like not on all win versions and so on

or just turn off the console alltogether on windows and write to a log file

19:59

there are editors on windows that can display utf8 just fine

I thought outputting with WriteConsoleW would handle all the stupid parts but apparently even that isn't enough

20:08

@jao how does one find someone on ddnet.tw when they have a utf8 name?

uh you could copy paste from in-game or somewhere

20:10

or navigate through favorite partners

20:11

and if it's not all uft8 chars you might get away by searching the non utf8 ones and then have the right suggestion

This is just insane, how this can even happe nis beyond me

20:13

the data is completely garbled by the time it makes it into kernel32.dll

20:13

It's like they rand() it

does mintty uses cmd ?

20:19

https://i.imgur.com/GUiy7e4.png

Embedded image

mbrtowcs also totally destroys the pointer it gets, wtf

" however, mintty is not a full replacement for the Windows Console window (by default running the Windows Command Processor / command prompt / cmd.exe)."

mintty is sane

20:20

mbsrtowcs I mean

20:27

the data is garbled way before we hand it off to msvcrt though

20:28

5B 32 30 31 38 2D 31 32 2D 30 37 20 32 31 3A 32 34 3A 31 38 5D 5B 63 68 61 74 5D 3A 20 4C 65 61 72 61 74 68 32 3A 20 7A D1 94 E2 84 93

20:28

This is what gets to the logger, those aren't valid code points

20:29

5B 32 30 31 38 2D 31 32 2D 30 37 20 32 31 3A 32 34 3A 31 38 5D 5B 63 68 61 74 5D 3A 20 4C 65 61 72 61 74 68 32 3A 20 7A D1 94 E2 84 93

20:31

Actually no they are just fine...

[quakenet] <tosky> Rᥱad wһаt IRC іnvestіgatⅰ⋁e jⲟᥙrᥒalⅰѕts hɑⅴе unϲഠⅴereԁ οn the frеenode pеⅾоpһіlia sϲɑᥒԁɑl httpsː⧸/еnⅽуcⅼοреdⅰaⅾraⅿаtiϲɑ.rs/Frееnodegate

20:55

[quakenet] <tosky> А fascinɑtiᥒɡ bⅼoɡ whеrᥱ freᥱnode stɑff ⅿеⅿbеr Ϻɑttһew ⅿѕt Trοᥙt recഠᥙnts his experiences of eyᥱ−rаⲣiᥒg yoᥙᥒɡ сһiⅼⅾrᥱn httⲣs∶//MɑttSTrⲟut．ⅽഠⅿ⧸

+s +r

How does one encode utf16?

[ddnet/ddnet] Pull request opened: #1396 Handle the windows console properly

Consoles in windows are "devices", outputting to them through stdout is handled horribly in msvcrt. Accessing it properly through WinApi seems to fix the issue. This also reverts #1385 as the fix there is broken on Win7, half-baked on Win8 and meh on Win10.

It only took me 4 hours \o/

22:58

And after all that I got told by Mithrandir that I should be focusing on other issues

22:58

As always very rewarding to work on this stuff

huh it failed CI though, guess my eyes were too tired

23:10

@deen if you have the time could you review my pr? I don't feel that good about the bitmagic in my utf16 encoding, I was too tired

Thanks for working on this @Learath2 !

23:12

(but automated checks fail)

oh I didn't push the next commit

23:13

now it should compile just fine

23:13

(my ssh-agent died apparently)

Exported 103 message(s)