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 2021-10-27 00:00:00Z and 2021-10-28 00:00:00Z

[ddnet/ddnet] Pull request opened: #4256 Update spanish.txt

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...

@deen any plans to keep the current site theme as a third option after Halloween? i really like the contrast feelsamazingman

feelsamazingman

[ddnet-maps:master] 1 new commit

b5e0bb7 M SolarStrike - ddnet-maps

lynn

@deen any plans to keep the current site theme as a third option after Halloween? i really like the contrast feelsamazingman

feelsamazingman

the current site theme is the dark theme

hmm im not sure what you mean, the link colors changed

ye cuz i changed the old theme

05:42

there is no new theme

05:42

u probs never knew about the dark theme before right

05:42

it always existed

i guess not? I thought the halloween theme was just adapted as the "dark theme"

the dark theme was originally made for halloween years back too

05:43

we just keep it cuz its a cool dark theme

well, im happy with this so im fine to not have the lower contrast "halloween" theme

05:43

:D (edited)

ye it lacked lot of contrast

looks great

05:44

does the color match the brand orange? cant tell

the logo orange

05:44

ye

okay cool

05:44

sorry web dev term

05:44

xd

well im also a web dev lol

05:44

monkaS

"the atc cta is not flush with the bottom of the fold"

05:45

ill start talking like that xD

wat

atc = add to cart

05:45

cta = call to action (dumb way to say button)

this looks more specific things to u than in general xd

fold = visible area without scrolling

05:45

yea i mean maybe it's more specific to e-commerce. (edited)

lynn

fold = visible area without scrolling

the way i remember this one is like a newspaper, without opening it, you can only see the front "fold"

After Halloween we just switch the default back from dark to bright theme, but anyone that explicitly set the theme by using the "Switch Theme" button will stay on their choice

There is some script to recompute points right? I migrated all my .dtb files to the new sqlite (poggers) and now it would be nice to add the points after the fact.

for ddnet we do it with scripts/ranks.py

09:25

but that also generates the ranks websites

whats this https://github.com/ddnet/ddnet-scripts/blob/master/servers/scripts/recalculate-points.py

ddnet-scripts/recalculate-points.py at master · ddnet/ddnet-scripts

The magic scripts that make DDNet run. Contribute to ddnet/ddnet-scripts development by creating an account on GitHub.

09:25

ah its all mysql connect scripts

well, it's called recalculate, but it just reads the values calculated by ranks.py and fills the db with them

ok so i gotta get ranks.py running

09:28

lemme see what i have to edit

09:28

I wish i could pr against scripts did i mention this already? :D

09:29

  with Cache('/home/teeworlds/servers/players-cache', eviction_policy='none', sqlite_auto_vacuum=0, sqlite_journal_mode='off') as cache:

for example using a relative home path would be a start to not force using the username "teeworlds"

I don't think you want to get it running

09:30

it does so many things

09:30

better code your own points recalculation

creating a bit of html shouldnt be so bad right?

09:33

So what do you suggest? Rewriting my own script to recalculate points? :c

yes

oky

09:36

thanks :)

Hm, this crash is so bizarre

10:21

m_pGameWorld seems to have ended up null in the entity, however that should never really be possible

10:22

I wonder if this is somehow a stale entity wrongly considered a CCharacter due to ub

monkaS

Well, asan didn't catch anything weird

10:34

so that excludes most of the error sources I would have expected

Okay that script seems to work fine :) https://paste.zillyhuhn.com/kQ can/should i pr that against some ddnet repo?

12:04

Can not only be useful for .dtb migrators but also for users that run the new sqlite servers for a while before inserting the map records that define the points

What is insane is pEnt can't possibly be null, which means the entity is still inserted in the gameworld, but such an entity cant possibly have a null gameworld pointer, but it does

Seems the entire character object we are ticking for is nulled for some reason

one could certainly test valgrind locally, which might find uninitialized values Else its probs a critical section issue or a logic error

Zwelf: Am i allowed to touch the ddnet-server.sqlite file? I replaced it with a new version but points where not picked up until i restarted the server. Are points chached or was it an issue with me replacing the file?

You're invited to talk on Matrix

You're invited to talk on Matrix

the file is only opened once when starting the server and kept open as long as the server is open. The server is still using the file from the startup, even if you removed/moved it if I understand the linux file systems correctly

13:40

if you are touching it with the sqlite executeable, you are fine (or the sqlite python library)

13:43

so just replacing it while the server is running doesn't work

Oh okay so that was the issue then nice. But when two servers write via sql connection it syncs live. Ye linux somehow keeps some zombie files :D

14:05

But yea good to know one shoud not replace the dtb file

https://wiki.ddnet.tw/docs/classCComponent__coll__graph_org.svg

15:12

looks cool

https://github.blog/changelog/2021-10-26-updates-to-our-issue-status-icons-and-colors/

Updates to our issue status icons and colors | GitHub Changelog

Updates to our issue status icons and colors

16:37

As part of our roadmap issue today we are starting a rollout that updates the issue closed icon from red to purple.

16:37

we did it, communism is no more

[ddnet/ddnet] Pull request closed: #4256 Update spanish.txt

16:43

[ddnet:master] 1 new commit

1e16c90 Update spanish.txt (#4256) - n0Ketchp

hi! my friend have some problem: when he set cl_refresh_rate 0 it's ~4k fps but when he sets 2000, he have ~1000 fps how to explain that? (edited)

gfx_refresh_rate at 0?

i'll ask

17:08

yep

17:08

it's 0

17:08

exactly, it was 1500, but he set 0 and it's the same

graphics thread might have to sometimes wait for cpu thread, so it's not that unexpected

he says when he set cl_refresh_rate to 1000 it's like V-Sync is on

https://github.com/stong/how-to-exploit-a-double-free

GitHub - stong/how-to-exploit-a-double-free: How to exploit a doubl...

How to exploit a double free vulnerability in 2021. 'Use-After-Free for Dummies' - GitHub - stong/how-to-exploit-a-double-free: How to exploit a double free vulnerability in 2021. &...

17:47

ARM only bug ^_^

17:48

https://discord.com/channels/252358080522747904/757720336274948198/902976757051850814

Is it arm-only though? x86(-64) is usually the odd one out, IBM's Power CPUs have a similar memory model to ARM

Is there a way to go past the zoom out level in the editor?

i don't think so

feelsbadman

we could make it respect the cl_limit_max_zoom_level

would be nice

18:04

i cant select the whole map at this zoom rate

18:04

and i cant move my mouse while selecting

[ddnet/ddnet] Pull request opened: #4257 Allow zooming any distance in editor

as requested by Knuski

Checklist

[x] 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/...

Ryozuki

As part of our roadmap issue today we are starting a rollout that updates the issue closed icon from red to purple.

lol why tho

further encourages repo maintainers to just close everything prematurely now smh (edited)

@deen concern about zooming in editor: it'll enable people to crash their client by zooming out too far or too fast. I guess you considered that?

i mean they have to change config for that

ah right

Um, I think I'm out of ideas for this crash

19:27

I just dont see it happening ever, let alone frequently :/

@Learath2: do you now the commit before these crashes happened?

Nope

ok

I've been digging into the coredump

19:29

But it makes no sense to me

but i remember the bug i introduced once

19:29

and from there i read all commits

19:30

most suspecious are the ones i dont fully understand, the removale of volatile and the teehistorian stuff The rest looks pretty harmless, tho i only read the history of the src/game/server history

The volatile removal should change nothing where a single thread is involved

yeah

19:32

thats what i wasnt sure about, if something like the database is async somewhere and writes to the log

Hm, that is unless ub is involved actually

but i also didnt find anything

19:33

was antibot updated?

I remember some ubsan warnings about ccharacter alignment

alignment or about using it even tho the destructor was called

Jupstar ✪

was antibot updated?

I dont think we got the API change in

19:35

I wonder if it has sth to do with the virtual destructor

its bcs of the order how destructors are called

19:35

it uses a character feature in the destructor call of ientity (in the world call) but at this point its not a character object anymore

19:36

atleast the character part could be considered destructed already

19:37

src/game/client/prediction/gameworld.cpp:151:23: runtime error: downcast of address 0x55c0d1095db0 which does not point to an object of type 'CCharacter'
0x55c0d1095db0: note: object is of type 'CEntity'
 00 00 00 00  98 15 ec cc c0 55 00 00  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  e0 60 c8 cd
              ^~~~~~~~~~~~~~~~~~~~~~~
              vptr for 'CEntity'
    #0 0x55c0ccb3e4d4 in CGameWorld::RemoveEntity(CEntity*) /media/ddnet/src/game/client/prediction/gameworld.cpp:151:23
    #1 0x55c0ccb38fd2 in CEntity::~CEntity() /media/ddnet/src/game/client/prediction/entity.cpp:33:16
    #2 0x55c0ccaedef6 in CCharacter::~CCharacter() /media/ddnet/src/game/client/prediction/entities/character.h:31:7
    #3 0x55c0ccb21846 in CCharacter::~CCharacter() /media/ddnet/src/game/client/prediction/entities/character.h:31:7
    #4 0x55c0ccb4901a in CGameWorld::CopyWorld(CGameWorld*) /media/ddnet/src/game/client/prediction/gameworld.cpp:532:4
    #5 0x55c0ccacc291 in CGameClient::OnPredict() /media/ddnet/src/game/client/gameclient.cpp:1710:19
    #6 0x55c0cc577bf0 in CClient::Update() /media/ddnet/src/engine/client/client.cpp:2812:20
    #7 0x55c0cc583ede in CClient::Run() /media/ddnet/src/engine/client/client.cpp:3237:4
    #8 0x55c0cc5a5b0e in main /media/ddnet/src/engine/client/client.cpp:4341:11
    #9 0x7ff8e66a6151 in __libc_start_main (/usr/lib/libc.so.6+0x28151)
    #10 0x55c0cc2f2e0d in _start (/media/ddnet/DDNet+0x705e0d)

19:38

this is also the last issue of https://github.com/ddnet/ddnet/issues/3050 and the fix is rather hard

Go through ubsan and clean up · Issue #3050 · ddnet/ddnet

CC=clang CXX=clang++ CXXFLAGS="-fsanitize=undefined" CFLAGS="-fsanitize=undefined" cmake -GNinja -DWEBSOCKETS=OFF -DMYSQL=ON -DCMAKE_BUILD_TYPE=Debug -DVIDEORECO...

19:38

i tried it, but it relys on so many things, kinda annoying design

19:40

(ok in this case its the prediction code, but i think its pretty similar to the server code) cant remember if that was fixed in the server code or even happened

Mh, thats clientside

did the server bug always happen on the same line (Collision()->IsSwitch)? Found that a bit strange since there are many other calls to GameContext/Collision before this one

19:47

but don't really know how to debug these kind of things (edited)

https://github.com/ddnet/ddnet/blob/master/src/game/server/entity.cpp#L28 (@Learath2)

Mh, thats clientside

19:48

server uses similar code tho

19:49

but i doubt that this causes the bug

19:51

but if this was smth different can you post the bt? or add it in the issue (@Learath2)

I remember some ubsan warnings about ccharacter alignment

19:57

sad that the thread sanitizer is so hard to use :D

monkalaugh

oh nice with TSAN_OPTIONS="ignore_noninstrumented_modules=1" it works quite well nowadays

20:50

no idea how much it disables, but i tested it on a small program with a datarace and it detected it :D

possibly the volatile buffer or smth, so maybe safe under x86 but not all hardware, but didnt dig deeper into it

[ddnet/ddnet] Pull request opened: #4258 Fix data race in backend_sdl

Checklist

[x] 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--und...

22:56

[ddnet/ddnet] Issue opened: #4259 Remaining TSan warnings

``` WARNING: ThreadSanitizer: data race (pid=1133109) Read of size 4 at 0x000004e0db44 by thread T29 (mutexes: write M4012, write M3986): #0 Mix(short, unsigned int) src/engine/client/sound.cpp:148:26 (DDNet+0xa76897) #1 SdlCallback(void, unsigned char*, int) src/engine/client/sound.cpp:290:2 (DDNet+0xa6a905) #2 SDL_RunAudio SDL/src/audio/SDL_audio.c:752:13 (libSDL2-2.0.so.0+0x269ba) Previous write of size 4 at 0x000004e0db44 by main thread: #0 CSound::SetListen...

Exported 144 message(s)