DDraceNetwork - bugs

DDraceNetwork

Development / bugs

Report client & server bugs, especially but not limited to release candidates.

Between 2022-12-05 00:00:00Z and 2022-12-06 00:00:00Z

run_alboni two bugged tps

Image attachment

3

tela

Click to see attachment 🖼️

When i debugged yesterday i could not imagine it. The name is created on main thread. And i doubt another thread accidentally perfectly rewrites the name but not the rest . This has to be in a higher level already. I assume it's related to some feature i didn't have activated like ghost. Which still uses old data or smth. Can you try a clean config. Back up ur current. Else if u know how to debug look when exactly the invalid name is generated. Since it happens when reloading community skins it's pretty sure related to the skin refresh It's just weird i can't trigger it bcs that all happens sync

Jupstar ✪

When i debugged yesterday i could not imagine it. The name is created on main thread. And i doubt another thread accidentally perfectly rewrites the name but not the rest . This has to be in a higher level already. I assume it's related to some feature i didn't have activated like ghost. Which still uses old data or smth. Can you try a clean config. Back up ur current. Else if u know how to debug look when exactly the invalid name is generated. Since it happens when reloading community skins it's pretty sure related to the skin refresh It's just weird i can't trigger it bcs that all happens sync

tried with a clean config, same results, I don't think I know how to debug, does that mean building it with -DCMAKE_BUILD_TYPE=Debug?

yes and setting breakpoints, or add dbg_msgs but if you never done that its probably not so ez^^

ya i never have. If there's any docs for it online or if you feel like explaining I am willing to try

yes i think event driven debugging can already help indetify the problem

18:05

i'd start with adding

c++
    dbg_msg("debugging", "%s", pName);

src/game/client/components/skins.cpp:444 and see what the output is

18:05

then we can already identify if that problem really exists on the main thread

does :444 mean at line 444

exactly

kk, and then compile with that flag i mentioned?

cant hurt to test that too yeah

18:11

ah ok there is a EscapeUrlwhich indeed is a curl call

18:11

this might be the issue all along

18:12

c++
dbg_msg("debugging2", "%s", aEscapedName);

maybe add this directly too

18:12

after the EscapeUrl a few lines down

Jupstar ✪

after the EscapeUrl a few lines down

so, basically the very end of the file?

18:18

im seeing escapeurl at line 448

yes

18:18

after that line

oh right after that line, ok

18:24

alright, there's some debugging info in console surrounding these errors now, not sure if that's what is useful

18:26

https://pastebin.com/raw/3pdzW0n7

18:26

this doesn't seem useful to me but idk lol

at least we now know that these invalid strings are already created earlier

ooo hell yea progress

gimme a second

18:32

i hoipe this works under macos above ::FindImpl (its the same function we already added the dbg_msg to) add

c++


#include <execinfo.h>
void dbg_print_stack_trace()
{
    int j, nptrs;
    void *buffer[100];
    char **strings;

    nptrs = backtrace(buffer, 100);
    dbg_msg("System", "backtrace of %d addresses", nptrs);

    strings = backtrace_symbols(buffer, nptrs);
    if(strings != nullptr)
    {
        for(j = 0; j < nptrs; j++)
            dbg_msg("System", "%s", strings[j]);

        free(strings);
    }
}

then add

c++

    dbg_print_stack_trace();

after

c++

    dbg_msg("debugging", "%s", pName);

18:34

mhh wait the symbols probably require a special compile flag gimme a second

no prob

% cmake .. -DCMAKE_EXE_LINKER_FLAGS="-rdynamic" invoke cmake with this additional parameter

18:38

i hope it works bcs then we have a stack trace already ^^

-g should be enough to have symbols

Jupstar ✪

i hoipe this works under macos above ::FindImpl (its the same function we already added the dbg_msg to) add

c++


#include <execinfo.h>
void dbg_print_stack_trace()
{
    int j, nptrs;
    void *buffer[100];
    char **strings;

    nptrs = backtrace(buffer, 100);
    dbg_msg("System", "backtrace of %d addresses", nptrs);

    strings = backtrace_symbols(buffer, nptrs);
    if(strings != nullptr)
    {
        for(j = 0; j < nptrs; j++)
            dbg_msg("System", "%s", strings[j]);

        free(strings);
    }
}

then add

c++

    dbg_print_stack_trace();

after

c++

    dbg_msg("debugging", "%s", pName);

i'm seeing 3 instances of "FindImpl", 398, 404, 407, that's quite a bit above the 444 line, not sure where to add this

Learath2

-g should be enough to have symbols

i think in this case it doesnt load debug symbols

18:52

but just the mangled symbols

tela

i'm seeing 3 instances of "FindImpl", 398, 404, 407, that's quite a bit above the 444 line, not sure where to add this

const CSkin *CSkins::FindImpl(const char *pName)

idk what you are doing anyway, feel free to ignore me

tela

i'm seeing 3 instances of "FindImpl", 398, 404, 407, that's quite a bit above the 444 line, not sure where to add this

tbh just put it somewhere in the file xD

18:53

doesnt really matter

18:53

just not inside a function

lol ok, I found that line ill add it above it

You can run it under lldb and get the backtraces that way too btw, I do it sometimes

sadly the problem doesnt always occur

18:54

so im too lazy to explain how to debug it

break set -> break command add -> bt -> continue -> DONE

if i know where the problem comes from it should be findable by logic

18:55

or this is some weird unrelated bug

it'll spew out backtraces everywhere

the whole time?

What is the problem btw?

for some reason on macos the skin names can be broken

Jupstar ✪

the whole time?

every time the breakpoint you set is hit, it'll do a backtrace and resume

for no apparent reason

18:55

valgrind doesnt find anything, asan, ubsan not

Learath2

every time the breakpoint you set is hit, it'll do a backtrace and resume

ah yeah makes sense^

Learath2

What is the problem btw?

also probably on the main thread

18:56

which makes it even weirder

Jupstar ✪

for some reason on macos the skin names can be broken

where is it broken? like through what codepath? is the skin name from the server broken or is the name we get from the fs broken?

i am not sure yet

18:57

but he said he toggled community skins

18:57

so i assume its not from server

The curl code is actually pretty broken fwiw, I can see it having a lot of UB

18:57

My curl multi pr fixes a lot of it but it's just not finished

https://pastebin.com/raw/pGqYWibk

ah yeah ghost

18:58

as if i called it

18:58

well ofc i cant repro, if i dont have a ghost file xDD

19:00

mh yeah i see it now too

19:00

should have tested more properly than i probably would have found it

19:00

thanks for the help tho

ya of course, lmk if u need me to do more

â•°â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€â”€

looks like we broke the console output on windows again somehow

its macos xd

pepeW

@Robyt3 help, is it normal that IntsToStr returns weird string if all ints are 0?

19:04

is that expected behavior?

lol i dont like macos either, but unfortunately i'm stuck with it because I can't be fucked to learn WSL or dual boot

I never really tried to fix and test that method, it currently uses undefined behaviour (edited)

ok that sucks in this case xD does somebody know if CGhostItem::Emptyis the function to check if a ghost is used or not?

19:09

i'll just assume yes, its used a few times

19:13

seems like this is not the correct way to check it

it only checks for bool Empty() const { return m_Path.Size() == 0; }

mhh or the problem is simply that it only happens once

so not related to the skin itself

i think the ghost stuff is simply not designed to work with reloads in mind

19:14

for database skins which are delayed

19:15

guess i call it a day and just fix it so it at least doesnt call invalid stuff

someone mentioned UB?

21:14

aPES_Crab

hey, i have trouble connecting to ddnet ger servers. is this the right channel for these kind of stuff ? can someone help me please ?

Ryozuki

aPES_Crab

lmao u really have emojis for everything

Iman

hey, i have trouble connecting to ddnet ger servers. is this the right channel for these kind of stuff ? can someone help me please ?

whats the error

Too weak connection (not acked for 10 seconds)

try LAN cable instead of wifi for example

my connection is fine and it works fine on other servers

where are you from?

france

uh, weird routing maybe? but losing packets sounds weird

might be a routing problem between u & the server

21:21

if so u cant really do anything

i cant even enter the server, the error message appear before i enter

21:21

i cant do anything at all ?

that message shouldn't even appear before entering the server

mhm

what ddnet version do you use?

i use f-client for 0.7

well, then you're on your own

21:22

or ask @fokkonaut

Iman

i use f-client for 0.7

can u simply try ddnet?

or use ddnet client

i dont want to move to 0.6 client all my playtime is on 0.7

u can replace all files with ddnet client files and rename DDNet to teeworlds

you can use tw 0.7 client on ddnet servers i think

what

we have a translation layer i believe

21:24

(to be confirmed)

i used 0.7 client to play on ddnet servers multiple times in the past

21:25

and it worked fine

does it work on other non ger server, like NLD or POL ?

i would like to keep using a 0.7 client, and i think ddnet isnt a 0.7 client ? maybe im misunderstanding something

21:25

i tried on china it worked

21:26

im gonna try on other servers

21:26

same issue on POL

21:27

same issue on NLD

21:27

same on RUS

21:28

same brazil

0.7 is basically unmaintained

21:28

do you need 0.7 for any specific reason?

Iman

i would like to keep using a 0.7 client, and i think ddnet isnt a 0.7 client ? maybe im misunderstanding something

DDNet client is a 0.6 client. DDNet server also supports 0.7 clients, but not well

yes, i need 0.7 because i spent thousand hours playing 0.7 and i believe that 0.6 is slightly different and i dont want to screw up my muscle memory. also id rather play on newer versions of tw, and the servers i like and my friends use to play on 0.

Iman

i use f-client for 0.7

@fokkonaut will be happy to hear

7

21:31

ok

Iman

Too weak connection (not acked for 10 seconds)

@fokkonaut Any idea why 0.7 connections to DDNet server are broken?

Iman

yes, i need 0.7 because i spent thousand hours playing 0.7 and i believe that 0.6 is slightly different and i dont want to screw up my muscle memory. also id rather play on newer versions of tw, and the servers i like and my friends use to play on 0.

well ddnet is the newer version of tw tbh i dunno about muscle memory, but input delay should be better with ddnet with a modern graphics backend also ddnet will probably shift more and more into an own implementation network and other stuff wise

We also ported many of the changes in TW 0.7 to DDNet client which we considered good

as much as i love ddnet and i think its a great project id rather not move to the ddnet client and keep playing teeworlds :I

ok u'll have to accept tho that this isnt mainstream and will probably cause more and more problems in future

21:36

most devs arent testing with 0.7 in mind

Jupstar ✪

ok u'll have to accept tho that this isnt mainstream and will probably cause more and more problems in future

for reference: https://steamdb.info/graph/?compare=380840,412220

Comparing 2 apps

also its not related to ddnet but it looks like i have the same issue on kog servers

KoG server switched to recent DDNet server version

22:04

so it could still be a DDNet server issue. You can try if Vanilla Teeworlds 0.7 client works fine, if yes, then f-client issue. If not, then DDNet server issue

i tried with an other 0.7 client and it didnt worked

22:15

im gonna try with vanilla client

In terms of what is the most and least maintained is like this: TW 0.6 -> TW 0.7 -> DDNet

22:18

Meaning ddnet even if it's based on teeworlds 0.6, it is still the one that will have the newest features and be the most played

i know that already

22:18

tw vanilla client doesnt work either

Well in my opinion ddnet shouldn't be forced to not do new stuff if it meant breaking compatibility with older teeworlds

22:20

So if you really like ddnet then you should even if it would be hard, switch to ddnet (edited)

Hm, we try to keep 0.7 working atleast, I'll take a look at what broke

22:24

But honestly, it's not the greatest experience as it uses the old tune hacks and stuff instead of proper prediction

ok, thank you for your support

Exported 155 message(s)