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 2020-10-19 00:00:00Z and 2020-10-20 00:00:00Z

[ddnet-maps:master] 1 new commit

8e95a12 A Eliteenator - ddnet-maps

[ddnet/ddnet] Issue opened: #3144 heap-buffer-overflow

``` ================================================================= ==18021==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x61e001fb3a6c at pc 0x00000046e7fe bp 0x7fff0fb4c7e0 sp 0x7fff0fb4bf58 READ of size 42 at 0x61e001fb3a6c thread T0 #0 0x46e7fd in printf_common(void, char const, va_list_tag*) (/home/teeworlds/servers/DDNet-Server-asan+0x46e7fd) #1 0x46ed3c in vsnprintf_chk (/home/teeworlds/servers/DDNet-Server-asan+0x46ed3c) #2 0x6e7296 in vsnprintf ...

@Ryozuki tested now, lld is indeed faster

@Deleted User poggers

poggers

@Deleted User adding -flto to cxxflags makes some golang packages fail to build monkalaugh

monkalaugh

07:20

cannot load DWARF output from $WORK/b036//_cgo_.o: decoding dwarf section info at offset 0x0: too short

07:20

stupid go

https://github.com/settings/repositories

Build software better, together

GitHub is where people build software. More than 50 million people use GitHub to discover, fork, and contribute to over 100 million projects.

08:55

u can change main to master

Image attachment

08:55

monkalaugh

The asan/ubsan on real servers is a great success already. I want to expand it to run more servers with it. I'm wondering if we can also do something like that for client.

10:20

Maybe I should start by running asan client myself all the time

it doesnt slow down server?

10:21

thats great

it does

10:21

but GER2 is strong enough, doesn't matter much

10:21

on others I wouldn't do it

How do I run DDNet-Server with asan? I'm trying to get the same reports to check if I fix them.

CC=clang CXX=clang++ CXXFLAGS="-fsanitize=address -fno-omit-frame-pointer" CFLAGS="-fsanitize=address -fno-omit-frame-pointer" cmake ../master -DMYSQL=ON -DCMAKE_BUILD_TYPE=Debug
ninja
ASAN_OPTIONS=log_path=./ASAN:check_initialization_order=1:detect_leaks=1 ni -15 2 ./DDNet-Server

need help with the client

ultimatum: what's wrong?

when i open the client, the loading screen pops up then suddenly in crashes

10:41

this is only happening right now

@Deleted User @Learath2 want to help with debugging ^?

Since you write on IRC, can you compile your own Debug client and run it in gdb? OS?

I need help it wont open to the interface just the loading screen then it crash without any crash message

10:43

windows 10

10:43

gdb?

ah, you're not a developer, too bad :D

10:43

Visual Studio Debugger would also work

10:44

Try writing gfx_major_version 1 in your settings_ddnet.cfg (in config_dir.bat) and then starting client

ok ill do the gfx

10:45

its still doesnt open until the interface

10:46

i can put the application error of windows event if you like that?

Eh, we don't have symbols to read the windows error report

10:47

ultimatum if you have some free time, it would be great if we could build it on your computer and see why it crashes

10:48

It doesn't happen on any of our machines so we can't debug it on ours

10:51

If you don't have time, just leave the windows report, maybe I can reverse engineer a bit to figure out the addresses

ultimate: can you try a clean version: https://ddnet.tw/downloads/DDNet-nightly-win64.zip and maybe press Windows + r -> %APPDATA%/Teeworlds and rename the config

11:02

ultimatum: also, are you using the auto update(the updater in the client) or steam version

i tried everything XD cant seem to complete the opening process of the client...

11:17

is there an application that visualizes all the content data? maybe that file is corrupted because i can open teeworld client just fine..

well just try a complete clean client

11:22

extract the whole zip, rename the Teeworlds folder in APPDATA to Teeworlds_

that works

the rename or the clean client?

but how hmmmm something is wrong with my teeworld file :(

11:24

yeah

yeah = 1 or 2?

yeah 1

11:24

rename

11:24

it works

11:25

i rename the file teeworlds to teeworlds_

ok, if you want, zip the whole directory and upload it somewhere if it doesnt contain personal data, or try to find out what causes it

roger that XD

11:25

thx Jupstar

@deen that was the 2nd report that the config directory was broken btw

ok it seems i narrowed it down

now im curious

I don't understand why it breaks tbh

neither

ok the reason might be the two cfg files masters or settings_ddnet

so deleting one of them works?

putting my old setting_ddnet or masters seems to crash the client

if its settings_ddnet.cfg, it might be a specific setting

11:32

you can upload it if u want

where should i upload?

dunno github

ok

or just a PM to us

hes not discord

but i'm on irc

ok

I am also on discord btw

11:33

name is mike_

I can reproduce a crash with @mike_ 's config

[ddnet/ddnet] Issue opened: #3145 Crash in CTextRender::UploadEntityLayerText

``` ================================================================= ==296452==ERROR: AddressSanitizer: heap-use-after-free on address 0x7f8e7064b90c at pc 0x555ba89ab926 bp 0x7ffc30e3ebf0 sp 0x7ffc30e3ebe8 WRITE of size 1 at 0x7f8e7064b90c thread T0 #0 0x555ba89ab925 in CTextRender::UploadEntityLayerText(void, int, int, int, int, int, char const, int, float, float, int) /media/ddnet/src/engine/client/text.cpp:1848:41 #1 0x555ba8a8d2ac in CMapImages::UpdateEntityLayerText(void...

@mike_ you can remove the cl_text_entities_size 0 from your config to fix it

thx @deen

so thats the problem @deen ?

yes

then tag bannzay, its his config

11:48

just make the lower limit to 1

11:48

not 0

[ddnet/ddnet] Pull request opened: #3146 Disallow cl_text_entities_size 0 (fixes #3145)

@BannZay

the json parser is really strange, it now crashed at client close bcs it deleted m_pDDNetInfo

[ddnet/ddnet] Pull request opened: #3147 Clamp font size to atleast 2 (fixes #3145)

We dont need to close #3146 i think setting it to 0 makes no sense anyway the config to disable entities text is cl_text_entities not this one

@Deleted User any more info about the json parser crash?

Should we look into a new maybe better tested json library?

If there is one and it's easy enough to replace ours, sure

Well a drop-in replacement is unlikely, but there is cJSON which looks like its a small header like ours and still seems to be getting updated

12:27

If we don't mind a new dependency rapidjson and nlohmann/json look like modern c++ gotos for json

Ideally there's one that won't change interfaces in the next 10 years and we don't have to touch anything

Well idk anyone that can make such a promise but the standards commitee and I don't see json being part of stl for a while

[ddnet/ddnet] Issue closed: #3145 Crash in CTextRender::UploadEntityLayerText

12:30

[ddnet:master] 2 new commits

90db24c clamp font size to atleast 2 (fixes #3145) - Jupeyy e002ab1 Merge #3147 - bors[bot]

12:30

[ddnet/ddnet] Pull request closed: #3147 Clamp font size to atleast 2 (fixes #3145)

i dunno @deen already closed, but the name "nameless tee" makes alot of trouble appearently

I guess anyone could make the promise, but I'd only trust such a promise from them :P

trouble?

12:31

because the info.json comes slowly?

no, bcs this null pointer i showed yesterday

12:33

and now it appearntly tried to free something invalidly

I guess we are just missing a null check somewhere where we use m_pDDNetInfo

[ddnet/ddnet] Pull request opened: #3148 Fix out of bounds array access in DoEditBox

Introduced in https://github.com/ddnet/ddnet/commit/186206d40c2fe65193832bf94389351515c179b8

12:38

[ddnet:master] 2 new commits

5ba8d8e Disallow cl_text_entities_size 0 (fixes #3145) - def- 51b5858 Merge #3146 - bors[bot]

12:38

[ddnet/ddnet] Pull request closed: #3146 Disallow cl_text_entities_size 0 (fixes #3145)

[ddnet:master] 2 new commits

8e6b9f0 Fix out of bounds array access in DoEditBox - def- 4360d78 Merge #3148 - bors[bot]

12:46

[ddnet/ddnet] Pull request closed: #3148 Fix out of bounds array access in DoEditBox

[ddnet:master] 4 new commits

343e482 Editor: Fix selection of None value in auto rules (fixes #3013) - def- debf263 Fix out of bounds access in GetSolo - def- 709d700 Ignore ddnet-info.json dl failure again - def- aca143c Merge #3129 #3138 #3143 - bors[bot]

12:54

[ddnet/ddnet] Issue closed: #3013 Editor Auto Rules improvements

12:54

[ddnet/ddnet] Pull request closed: #3143 Ignore ddnet-info.json dl failure again

12:54

[ddnet/ddnet] Pull request closed: #3138 Fix out of bounds access in GetSolo

[ddnet/ddnet] Pull request opened: #3149 Fix memory leak on team finish and read of uninitialized data

Fixes #2522, #3144 and second leak from #3134.

Imagine teaching a computer networks course in 2020 but your first 4 lectures are all about obsolete stuff like ATM ISDN BISDN. Referring to ITU-T as CCITT. It’s like this guy is stuck in the 70s still pissed that virtual channels never took off

15:10

If it was up to him we’d have dialup netflix

monkalaugh

15:30

@Learath2 who doesnt love being capped at 64kb/s

We could probably have higher network speeds, it’d be very funny to only be able to receive as many services as there are phone lines at your house

15:31

Operator, connect line 4 at my house to netflix, and start playing narcos

poggers

Oh or maybe electomagnetic switching, like you call the station, press 4 for 4th line, 3 for netflix followed by the showid

You can't call and switch simultaneously

You accidentally switch your last telephone line to netflix and they have to manually switch it back

[ddnet:master] 2 new commits

f1bb5b5 Use NUM_DUMMIES for dummy array length - ChillerDragon dfe3bb4 Merge #3139 - bors[bot]

16:45

[ddnet/ddnet] Pull request closed: #3139 Use NUM_DUMMYS for dummy array length

@Learath2 the prof could just rename the course to "computer networks history" and keep his curriculum

lol

1

Heh, that’d indeed be more accurate

17:08

His magnum opus is the slide titled “More Formally” which proceeds to write the throughput of the sliding window protocol in variables that he doesn’t define

17:10

I know it’s T_tx/(T_tx + 2t_p) yet I have no idea what T or t or p are

[ddnet/ddnet] Issue closed: #2522 Game UUID is discarded on load

17:17

[ddnet/ddnet] Issue closed: #3144 heap-buffer-overflow in CSaveTee

17:17

[ddnet:master] 4 new commits

aa53f28 Fix memory leak in MySQL getBlob function - Zwelf 6d11c34 Load game UUID on /load - Zwelf 81131c7 Fix grammar mistake - Zwelf c1dea54 Merge #3149 - bors[bot]

17:17

[ddnet/ddnet] Pull request closed: #3149 Fix memory leak on team finish and read of uninitialized data

Making ddnet SAFE again

18:39

18:40

Lets build a wall to contain the leaks

18:40

And make them pay it with memory

18:40

poggers

is there any way to use custom x_ninja skins thru assets folder thingie

photoshop

4.93 MB

19:13

i tried delete part little and different of every (blocks entities) to be able seen top each other

19:14

but seems me still not know how make entities if any one see this good idea you can do it

19:16

Image attachment

add rcon command switch [ind] [state]

19:25

or idk

19:25

Can u save switches with /r maybe

It gives me immense pleasure that we are finally tackling the UB and the memory leaks

[ddnet/ddnet] Issue opened: #3150 Mac: stuck in in ingame chat.

On Mac, when typing one of those characters ^ or ~ I get stuck in ingame chat and there is no way to legitimately close it or Teeworlds. The only way for me is to unplug my router so I get timeout protection and reboot my pc.

HMM see tee in /spec but were in solo :O showothers is off

Image attachment

[ddnet/ddnet] Issue closed: #3080 Make the ddnet-info.json warning less fatal looking

@ᶰ°Konͧsti that sounds expect. when you're in pause/spec you see all tees

im not in pause as u can see

20:31

ok u dont see much in The screen but there is the outline of a tee xd

ok

20:32

can reproduce

Hm, no, actually can't reproduce

20:50

code looks correct too

20:50

maybe you have dummy connected too?

@ᶰ°Konͧsti latest client right

20:54

xd

it is determined by server whether you see speced tees, not client I htink

@deen

previously allocated by thread T2 here:
    #0 0x55f93f94f579 in malloc (/media/ddnet/DDNet+0x8ec579)
    #1 0x7fd6231802ba in CRYPTO_zalloc (/usr/lib/libcrypto.so.1.1+0x1812ba)

is that like a trace?

https://github.com/ddnet/ddnet/issues/1

Show time of player you're spectating · Issue #1 · ddnet/ddnet

Probably difficult since spectators don't have a character?

https://git-scm.com/docs/hash-function-transition/

21:08

did u know git is trying to transition from sha-1 to sha-256

21:08

found about this here https://github.blog/2020-10-19-git-2-29-released/

Highlights from Git 2.29 - The GitHub Blog

The open source Git project just released Git 2.29 with features and bug fixes from over 89 contributors, 24 of them new. Last time we caught up with you, Git 2.28 had just been released. One version later, let’s

i have a suggestion: add "Show downloaded skins" in Tee menu

@gerdoe why?

why should i manually replace whole folder to get skins from db?

just copy and paste

if we load them all into memory it's not so great for loading time and memory usage

21:35

so i'd prefer not to

it's joke, nvm, it would be pretty cool to see skins ingame instead of manual checking

[ddnet/ddnet] Issue opened: #3151 Server search by multiple keywords

Today I saw the feature with excluding servers by keywords. Also it would be nice if it was possible to search for multiple keywords like kog, hard which would list all the kog servers which also contain hard in their name. Another example ddnet, ger, solo Maybe there is a more convenient operator but you get the idea.

[ddnet/ddnet] Issue opened: #3152 Mac: Window shift on resolution change.

My screen has the following resolution: 2560 x 1600 If I select this resolution in tw the window gets displaced by half of the screen. !window_resolution_bug Maybe the gif helps. The window is simply shifted down the screen on the Y axis.

is there a way for see the time in minutes and not in hours?

Exported 176 message(s)