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 2022-05-26 00:00:00Z and 2022-05-27 00:00:00Z

[ddnet:master] 4 new commits

05efc25 Add integration test script and coverage map - ChillerDragon 68bcd21 asan+ubsan clean - def- e311621 test -> integration_test - def- 8e3d5c1 Merge #5006 - bors[bot]

00:02

[ddnet/ddnet] Pull request closed: #5006 Add integration test script and coverage map

hmm, I'm not sure how to do that

00:04

if I understand this right it seems to crash when something goes out of scope here

Image attachment

guys, there's this annoying thing, idk if it's intended. so, you press esc, hold the disconnect button and press esc again. now you have set up a trap for whenever you press esc

00:06

this is really annoying

Tater

if I understand this right it seems to crash when something goes out of scope here

lol, mem_zero on class...

00:08

this is a heisenbug waiting to happen

oh wait it's actually in CCharecterCore that the set is placed (edited)

[ddnet/ddnet] Issue opened: #5228 Do not mem_zero classes

Basically title. This is UB as soon as the class is virtual (haven't found any case in our codebase) and for any class member variable as well. Example here: https://github.com/ddnet/ddnet/blob/8e3d5c1b5920e17274e126b8f8d8ad0b49dd6354/src/game/client/gameclient.cpp#L1100-L1104 Member variable that is probably UB (in CCharacterCore): https://github.com/ddnet/ddnet/blob/8e3d5c1b5920e17274e126b8f8d8ad0b49dd6354/src/game/gamecore.h#L225 What happens is: 1. TempCore is initialized, ...

so probably set has no defined size

i created an issue about this UB

00:18

oh, i forgot to credit you

00:18

sorry

oh

00:19

you can edit if you want I think

yup, corrected

00:22

anyway, thanks for the help to improve the game

yeah np :D

00:25

you can credit my gh account if you want @sjrc6

[ddnet/ddnet] Issue opened: #5229 Disconnect "trap" when closing menu while disconnect button is clicked down.

GitHub

Click to see attachment 🖼️

ty ty

theofficialgman

[ddnet/ddnet] Issue opened: #5230 .desktop file preview has no icon

https://github.com/ddnet/ddnet/blob/master/other/ddnet.desktop when compiling the game from source and installing with sudo make install to the default directory of /usr/local/ the icon file doesn't display in the .desktop file. the path could be hardcoded in the .desktop file but that shouldn't be necessary. Does anyone know how to make it so that the Icon=ddnet will pick up on the icons located in /usr/local/share/icons/hicolor/*/apps/

[ddnet-maps:master] 1 new commit

00e44ed M Target, A Merely forth 2 - ddnet-maps

Hey all. I'm a developer from over at pi-apps (an open source raspberry pi app store). We are adding DDNet to the "store" but usually like to have out apps versioned (as most projects do). We have CI setup to automatically update our apps if they post github releases or any other method of obtaining the latest version. With DDNet, there doesn't appear to be any way to get the latest version number. You create tags on the github https://github.com/ddnet/ddnet but they aren't always for versions and you don't have any releases on the page. Using the website isn't really an option as that information can't really be parsed easily or reliably. Thoughts? (edited)

@theofficialgman the https://info2.ddnet.tw/info JSON contains a "version" string, that should always be up-to-date, it's what the client uses to determine whether it's out-of-date

are you referring to the news section at the bottom or something else?

05:32

ah nevermind I see it

05:32

"version": "16.1", is this guaranteed to match the github tag? not a big deal if it doesn't, our updater will just throw a warning. but its good to know anyway (edited)

Chairn

lol, mem_zero on class...

And I doubted the basics of C++ only because I had not figured out what is wrong and you see the mem_zero and think directly the right xD nice

imo good style is to use struct for PODs and classes for complex types

06:10

tho generally mem_zero is more c style than c++ style

yes, I also do not understand why we use mem_zero so often instead of using value initialisation

yes and since c++11 u can also member initialize e.g. class CTest{ int m_A = 5; }; this cannot be skipped by the compiler

06:13

(in case u want to use a different value than 0)

06:13

its imo also nicer than to use constructors

1

[ddnet/ddnet] Pull request opened: #5231 Fix windows crash

Fixes #5226

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-addresss...

@Tater can you test https://github.com/ddnet/ddnet/pull/5231 (edited)

[ddnet/ddnet] Issue opened: #5232 Improve Integration Test

@ChillerDragon A few suggestions to further expand the integration tests 1. Run the tests also on windows 2. Can we maybe output the output from the client/server somehow? currently nothing is output !grafik 3. can we do more complex processes? can you research if you could somehow simulate inputs, e.g. test if you can hook (other tees) and run.

ou i think im onto something :D

08:11

are there any known client crashes on current master?

you mean except when loading just about any malformed map or demo?

yes xd

08:17

something new

08:17

maybe the new hud

last hud issue i reported was known

08:17

ill investigate

Stun lets the client sometimes crash (edited)

08:21

but I only get it on release build

[quakenet] ChillerDragon

maybe the new hud

last hud issue i reported was known

what was that again xD

08:22

flickering Weapon display?

[quakenet] ChillerDragon

are there any known client crashes on current master?

do you have https://github.com/ddnet/ddnet/pull/5222 included in your build?

yes i merged like 1h ago

08:28

oh wow ./DDNet > foo.log keeps colors o.O

08:28

isnt it common that programs detect such things

08:28

for example ls is colored but ls > ls.txt isnt

[ddnet/ddnet] Issue opened: #5233 Suppress log colors when redirecting stdout to a file

[ddnet:master] 3 new commits

cf03ada revert to SetHookedPlayer in ReadCharacterCore - C0D3D3V 62cad6b use value initialization in Evolve lambda - C0D3D3V 3b81157 Merge #5231 - bors[bot]

09:17

[ddnet/ddnet] Pull request closed: #5231 Fix windows crash

09:17

[ddnet/ddnet] Issue closed: #5226 Nightly build crashs on windows if you join a server

[ddnet/ddnet] Pull request opened: #5234 Enable cppcoreguidelines-pro-type-member-init

Strictly enforces initialization of members, could be considered too extreme, but at least we'd end up with fewer uninitialized member accesses: https://clang.llvm.org/extra/clang-tidy/checks/cppcoreguidelines-pro-type-member-init.html

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
[ ] ...

wew i get plenty of client crashes

report all

I WILL REPORT DA CRASHES TO THE AUTHORITIES IN A MINUTE

09:58

but first i gotta sell some 0d to US government

i dont get the crash

[quakenet] ChillerDragon

oh wow ./DDNet > foo.log keeps colors o.O

You can set NO_COLOR env variable to disable colors

10:12

https://no-color.org/

@c0d3d3v did u fix crash?

[ddnet/ddnet] Pull request opened: #5235 Enable less controversial cppcoreguidelines clang-tidy checks

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...

c0d3d3v

but I only get it on release build

RelWithDebInfo too? Better for debugging

deen

RelWithDebInfo too? Better for debugging

Next time I play I will use that.

Ясно Понятно

@c0d3d3v did u fix crash?

yes should be fixed, and merged

Ясно Понятно

@c0d3d3v did u fix crash?

see https://github.com/ddnet/ddnet/pull/5231

[ddnet/ddnet] Pull request opened: #5236 Don't enable colors when stdout is redirected to a file (fixes #5233)

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...

c0d3d3v

see https://github.com/ddnet/ddnet/pull/5231

Ok, ty later i check

@deen yea sure but its just bad style isnt it? As I used i expect the cli colors to behave as in EVERY other application

10:42

also who of u trols is slidin in ma irc dms?

10:43

Embedded image

10:44

OK scratch my crashbugs they are only in my fork

why is the client so insanly unstable rn xD what prs should i test

10:54

heap-use-after-free CGameClient::UpdatePrediction() /src/game/client/gameclient.cpp:2425:74

[ddnet/ddnet] Issue opened: #5237 Client crashes on Linux when no X11 is present

``` Program terminated with signal SIGSEGV, Segmentation fault. #0 0x00007f833c0ec804 in pthread_mutex_lock () from /usr/lib/libc.so.6 [Current thread is 1 (Thread 0x7f832cf75000 (LWP 2543964))] (gdb) bt #0 0x00007f833c0ec804 in pthread_mutex_lock () at /usr/lib/libc.so.6 #1 0x000055afa324028c in gthread_mutex_lock(gthread_mutex_t*) (__mutex=0x10) at /usr/include/c++/12.1.0/x86_64-pc-linux-gnu/bits/gthr-default.h:749 #2 0x000055afa3240323 in std::mutex::lock() (this=0x10) at /u...

10:59

[ddnet/ddnet] Issue opened: #5238 Crash on client connect

``` ==50506==ERROR: AddressSanitizer: heap-use-after-free on address 0x61900009f848 at pc 0x5593e52787b7 bp 0x7ffe0a0eac70 sp 0x7ffe0a0ea440 READ of size 8 at 0x61900009f848 thread T0 #0 0x5593e52787b6 in __asan_memcpy (/build_all/DDNet+0xc2a7b6) (BuildId: 901a71f18e0a233fe9284c4602d7754f132418f1) #1 0x5593e649eadd in CGameClient::UpdatePrediction() /src/game/client/gameclient.cpp:2425:74 #2 0x5593e648aba1 in CGameClient::OnNewSnapshot() /src/game/client/gameclient.cpp:1710:3...

11:00

[ddnet/ddnet] Issue opened: #5239 Crash on client while just standing around

``` ==50953==ERROR: AddressSanitizer: heap-use-after-free on address 0x61900431288b at pc 0x559fbd021cf2 bp 0x7ffdd999fe70 sp 0x7ffdd999fe68 READ of size 1 at 0x61900431288b thread T0 #0 0x559fbd021cf1 in CCharacterCore::Move() /src/game/gamecore.cpp:502:23 #1 0x559fbe01f148 in CCharacter::TickDefered() /src/game/client/prediction/entities/character.cpp:577:9 #2 0x559fbe0758e9 in CGameWorld::Tick() /src/game/client/prediction/gameworld.cpp:199:10 #3 0x559fbdfa71fa in CGa...

11:02

[ddnet/ddnet] Pull request opened: #5240 Fix crash when graphics init failed (fixes #5237)

Yet another data structure that was uninitialized btw

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](http...

c0d3d3v

yes should be fixed, and merged

All work, ty

11:04

@c0d3d3v some time when i kill, i crash

11:05

Image attachment

that smells like someone subtracted from a nullpointer

deen

that smells like someone subtracted from a nullpointer

i dont change anything

11:08

just pull , latest master, and start game

yes i also have lot of crashes

11:09

basically unusable rn the client

@Chairn u love small executables, remove dynamic_casts from our code and disable RTTI fno-rtti

[ddnet/ddnet] Pull request opened: #5241 Ignore our weird vptr issue (fixes #5238, fixes #5239)

See for example https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94325

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...

Ясно Понятно

Click to see attachment 🖼️

what is the exception, and why do you have diffrent code numbers

11:46

ah read access violation, I was blind

[ddnet/ddnet] Pull request opened: #5242 Add valgrind to integration test

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-web:master] 1 new commit

ebf16c1 Frontpage video - def-

happy

1

Ясно Понятно

@c0d3d3v some time when i kill, i crash

also got a crash on kill right now. Maybe related to you get hooked + kill

c0d3d3v

also got a crash on kill right now. Maybe related to you get hooked + kill

#5241

https://github.com/ddnet/ddnet/pull/5241

Ignore our weird vptr issue (fixes #5238, fixes #5239) by def- · Pu...

See for example https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94325 Checklist Tested the change ingame Provided screenshots if it is a visual change Tested in combination with possibly related ...

Jupstar ✪

#5241

strange that this change causes the delete function to crash

14:09

Image attachment

Jupstar ✪

#5241

so I let bors run on it? are you ok with it

no lets wait until the supress is gone

k

c0d3d3v

strange that this change causes the delete function to crash

The code currently relies on UB, and the code that was added hid that

14:11

and changed behavior

I see

c0d3d3v

also got a crash on kill right now. Maybe related to you get hooked + kill

I correct myself xD It basically crashes if any tee that the client knows of gets killed, but that should be clear by the issue causing it (edited)

@Ryozuki do you use any language server in vim?

YMC masterrace

15:04

YCM

15:05

a

15:05

@Learath2 as a true deen fanboi i use https://github.com/ycm-core/YouCompleteMe

GitHub - ycm-core/YouCompleteMe: A code-completion engine for Vim

A code-completion engine for Vim. Contribute to ycm-core/YouCompleteMe development by creating an account on GitHub.

15:05

thats all you need

I'm guessing YCM with clangd?

15:05

Or just YCM?

15:06

Or does it come with clangd nowadays?

i think just YCM

15:06

im not aware of any clangd

15:06

never heard of it

15:06

but it works pretty feature complete to me

15:06

but i barley touch any c++ from vim

15:07

Y M C A

[ddnet/ddnet] Issue opened: #5243 Crash on client CGameClient::UpdatePrediction() pLocalChar->Destroy();

``` Thread 1 "DDNet" received signal SIGSEGV, Segmentation fault. 0x00005555557f9df0 in CGameClient::UpdatePrediction (this=0x7fffec3a8010) at /home/chiller/Desktop/git/ddnet/src/game/client/gameclient.cpp:2263 2263 pLocalChar->Destroy(); (gdb) bt #0 0x00005555557f9df0 in CGameClient::UpdatePrediction() (this=0x7fffec3a8010) at /home/chiller/Desktop/git/ddnet/src/game/client/gameclient.cpp:2263 #1 0x00005555557f65d1 in CGameClient::OnNewSnapshot() (this=0x7f...

15:17

[ddnet/ddnet] Issue closed: #5243 Crash on client CGameClient::UpdatePrediction() pLocalChar->Destroy();

so since we have the freeze progress bar, give us the option to disable the freeze stars

there was discussion here: https://discord.com/channels/252358080522747904/293493549758939136/977252263280144464 don't know what's next.

[ddnet-web:master] 1 new commit

4dc13ed Add donation by Steinchen99 - def-

https://www.broadcom.com/company/news/financial-releases/60271

Broadcom Inc. | Connecting Everything

16:18

Insane

16:18

Broadcom to Acquire VMware for Approximately $61 Billion in Cash and Stock

16:18

The last large buy was microsoft deal to buy blizzard for 69 bn

peanuts

wow VMware is worth 61 billion? (edited)

if anyone is interested: my 1 hour demo with out of sound sync (at 1:00:00), and the demo mp4: https://btp.uber.space/ddnet/stronghold_2_bugdemo.mp4 demo: https://btp.uber.space/ddnet/stronghold_2.demo

cauldron

if anyone is interested: my 1 hour demo with out of sound sync (at 1:00:00), and the demo mp4: https://btp.uber.space/ddnet/stronghold_2_bugdemo.mp4 demo: https://btp.uber.space/ddnet/stronghold_2.demo

pm me your config

16:37

i wanna try out, u are on windows?

no, on linux, and it was rendered a few months ago, i used the client from website back then. must be right after finish time. what config? opengl? is it a file?

settings_ddnet.cfg in ~/.teeworlds

about the removement of the freeze stars, because of freeze bar: what about deepfreeze? permanent bar? no stars at all? well i guess no stars?

Image attachment

[ddnet/ddnet] Pull request opened: #5244 Remove predicted character from gameworld before calling base class d…

…estructor cc #5241

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/#u...

cauldron

if anyone is interested: my 1 hour demo with out of sound sync (at 1:00:00), and the demo mp4: https://btp.uber.space/ddnet/stronghold_2_bugdemo.mp4 demo: https://btp.uber.space/ddnet/stronghold_2.demo

ok it finished and its perfectly in sync for me what is your processor CPU & GPU?

intel i5, intel hd3000 (edited)

is this some intel hardware issue

17:37

i mean i used your config and can absolutly not reproduce it

17:37

perfectly in sync

17:38

https://discord.com/channels/252358080522747904/757720336274948198/979125287222640701 some other person with really modern hardware has this too

17:39

maybe smone of the devs with an intel CPU can test it

i also have this issue: https://discord.com/channels/252358080522747904/757720336274948198/964252317094383736 maybe it is related. I rendered with mesa driver for gpu. i tried to render with LIBGL_ALWAYS_SOFTWARE=true, but then it was just black screen i think. maybe i have to select another renderer then in the settings? i would want to try to render without gpu and look if it works then. (if it's in sync then) (edited)

the software renderer can also just be buggy if it gives u black screen

17:41

but can test it quickly

17:43

yeah black for me too

could it work with another option/choice in graphic settings>renderer? (edited)

maybe

[ddnet/ddnet] Pull request opened: #5245 disable dragger beams soon, not only every 150ms

The check whether a dragger is active, I somehow got lost in the dragger teams. Sorry! Actually, I should have noticed this during testing. But when I tested Fall into the Future today, it seemed strange to me (already at the first part). I should have noticed it when I tested it. I have compared that in any case with 15.9.1 and looked at the code again and saw that there the check is made every tick.

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a v...

i think software renderers mostly focus on opengl 4.5 nowadays

17:45

GLES 3 seems to work

17:45

but just to be clear, i rendered the 1h demo with OpenGL 3.0 just as your config

i don't know anything about this, but i have the suspicion that it is related to my overall mesa freezes / crashes like in the short clip in watch_your_time. i have also issues with gpu / hardware acceleration for java guis.

i mean you have a intel GPU, what do you expect xd

17:49

but i also have a laptop with intel hd 3000, i think everything ran fine there ^^

yeah, i'm planning to buy a new laptop, anyway demos render with software and gles for me too, so i gonna use that.

cauldron

could it work with another option/choice in graphic settings>renderer? (edited)

so just to sum some stuff up the video rednerer works on the CPU, thats one of the weirdest things about this bug normally CPU bugs are rather ez to reproduce bencie tested it with vulkan and had the problem with video rendering too, so probs not even opengl related

cauldron

yeah, i'm planning to buy a new laptop, anyway demos render with software and gles for me too, so i gonna use that.

but with software rendering u wont have the delay? or just not tested it yet

i tested a 28 second demo. but will try the 1 hour demo tonight when i sleep

ok

[ddnet:master] 2 new commits

2402edc Remove predicted character from gameworld before calling base class destructor - trml aa5471f Merge #5244 - bors[bot]

17:58

[ddnet/ddnet] Issue closed: #5238 Crash on client connect

17:58

[ddnet/ddnet] Issue closed: #5239 Crash on client while just standing around

i might be getting nvidia in the near future: https://www.mediamarkt.de/de/product/_msi-gf63-thin-10scxr-1449-2710234.html xD

MSI GF63 THIN 10SCXR-1449, Gaming-Laptop mit 15,6 Zoll Display, Int...

MSI GF63 THIN 10SCXR-1449, Gaming-Laptop mit 15,6 Zoll Display, Intel® Core™ i5 Prozessor, 8 GB RAM, 512 GB SSD, NVIDIA GeForce

[ddnet/ddnet] Pull request closed: #5244 Remove predicted character from gameworld before calling base class d…

cauldron

i might be getting nvidia in the near future: https://www.mediamarkt.de/de/product/_msi-gf63-thin-10scxr-1449-2710234.html xD

gl on linux xd

i'll make it work

Susge

make sure u can disable nvidia optimus, if the laptop has that

18:00

i warned you

laptops suck

18:01

beantyping

[ddnet/ddnet] Pull request closed: #5241 Ignore our weird vptr issue (fixes #5238, fixes #5239)

vulcan graphic renderer problem button in the hud have no text on rx 580

HappyFray

vulcan graphic renderer problem button in the hud have no text on rx 580

update your drivers

18:25

this happens on outdated amd drivers

18:26

use ones from amd.com

18:26

not from windows

maybe someone wants to review https://github.com/ddnet/ddnet/pull/5183

@noby

19:46

aw someone deleted troll

troll

[ddnet-web:master] 1 new commit

d3bacca Add 20 € donation by Blade - def-

[ddnet:master] 2 new commits

62c9951 disable dragger beams soon, not only every 150ms - C0D3D3V 74b94b3 Merge #5245 - bors[bot]

21:06

[ddnet/ddnet] Pull request closed: #5245 disable dragger beams soon, not only every 150ms

@Chairn shouldn't stuff like crashes be high priority, and not some general code cleanup stuff or https://github.com/ddnet/ddnet/issues/3533 where we don't even know the impact?

after 14 hours some bullets stopped moving and my grenades don't ro...

As reported by Im 'corneum: Well maybe you can recreate it. I used cheat engine to speed hack and after 14 hours some bullets stopped moving and my grenades don't rotate. Even after...

highprio imo is either crash or potential cheat

[ddnet:master] 2 new commits

ad06a05 render allways the nameplate of a specchar - C0D3D3V e63b8e2 Merge #5183 - bors[bot]

21:24

[ddnet/ddnet] Pull request closed: #5183 render allways the nameplate of a specchar (or the real player)

which code cleanup stuff?

21:25

my issue is about specifically a crash and UB that may or may not enable cheats or simply crashes in general

ah, so the title is wrong? https://github.com/ddnet/ddnet/issues/5228

Do not mem_zero classes · Issue #5228 · ddnet/ddnet

Basically title. This is UB as soon as the class is virtual (haven't found any case in our codebase) and for any class member variable as well. Example here: ddnet/src/game/client/gameclien...

21:27

don't mem_zero CCharacterCore in that line?

the issue does cover multiple things indeed, do you want me to split it in two?

I guess, I was confused by the "high priority" of getting rid of all mem_zeros. maybe move that to a different issue

it's just that i found this special case given in the first comment that was indeed crashing with MSVC due to their implementation of std::set, but the general rule applies to all mem_zero applied on classes instances

only those that are not POD

well, now that the special case has been solved, we can keep the current issue for the general case

21:30

yeah, but as you can see, the class CCharacterCore was POD until C0ded3v added std::set member

21:30

i'll try to move on on my initall branch this weekend

maybe describe your approach in the issue first

21:34

https://github.com/Chairn/ddnet/commit/78e5e5778af127f2e0880b47fdbd6196d5405c1c#diff-29ffe32916052e1066eed938021aadfeb29f855d7c63d9c1ca3350aceffb79f3 this creates quite a large diff. how do I verify that this doesn't change anything gameplay-related?

Init all data in constructor, disable some default one · Chairn/ddn...

you don't, because it relied on UB anyway

21:36

non initialized value inherited random value from memory

21:36

this value could be non random if it used same space as a previous that always had the same value

21:36

but it 100% depends on the compiler

I understand what UB is

21:44

sometimes UB results in consistent actual behavior

21:44

e.g. with the shotgun bug

21:45

also where are you picking the values from?

no where especially, i zero initialize most of them in a safe way

21:45

compiler probably optimize most of it i hope

[ddnet/ddnet] Issue closed: #5233 Suppress log colors when redirecting stdout to a file

21:45

[ddnet:master] 2 new commits

d2e9a78 Don't enable colors when stdout is redirected to a file (fixes #5233) - def- 0648f22 Merge #5236 - bors[bot]

21:45

[ddnet/ddnet] Pull request closed: #5236 Don't enable colors when stdout is redirected to a file (fixes #5233)

no, I see quite some non-zero values

21:46

e.g. m_Jumps, some weapon stuff, etc.

those are default gameplay value i think, 2 jumps, infinite ammo for hammer and gun

21:48

i'll have to check those obviously

(edited)

I don't know. currently, I don't see the appeal of this solution, over, say letting the compiler initialize everything with a zero-value (I think you can do that in C++?)

heinrich5991

@Chairn shouldn't stuff like crashes be high priority, and not some general code cleanup stuff or https://github.com/ddnet/ddnet/issues/3533 where we don't even know the impact?

I will try to reproduce this, and make a video tomorrow... I still think this is just some visual bugs, caused by the external speed hack that forced the client to desync a lot from the server gametick

heinrich5991

I don't know. currently, I don't see the appeal of this solution, over, say letting the compiler initialize everything with a zero-value (I think you can do that in C++?)

i looked for compiler flags for this, haven't found anything that actually works

c0d3d3v

I will try to reproduce this, and make a video tomorrow... I still think this is just some visual bugs, caused by the external speed hack that forced the client to desync a lot from the server gametick

you can probably just initialize the right timer to a very big value

the bug is not about huge timer!

it's not?

the issue is basically about he crying his speed hack is not working

uh no? it's about bullets not moving and visual bug about grenade not rotating

Image attachment

I think it's a bug because we think that this might also happen after a long time on the server

I do not think its a bug. Speed hacks produce UB .. they can change what ever they want and make the client desync

21:59

... I will test it tomorrow

22:00

Also the description is pretty poor

22:01

He does not mention if he used the hack on client or server, and what he mean with "Even after reload"

I don't think it's fair to call that UB (unless you mean something different from the UB for the C++ standard)

[ddnet:master] 3 new commits

e3f27b6 Enable less controversial cppcoreguidelines clang-tidy checks - def- f312308 Add portability-* clang-tidy checks - def- 551275a Merge #5235 - bors[bot]

22:03

[ddnet/ddnet] Pull request closed: #5235 Enable less controversial cppcoreguidelines clang-tidy checks

22:04

[ddnet/ddnet] Pull request opened: #5246 Use std::vector instead of sorted_array, remove base/tl/sorted_array.h

This replaces all usages of sorted_array with std::vector. This requires the following general changes:

add_unsorted is replaced with push_back.
add is replaced with push_back and subsequent sort or stable_sort must be ensured.

In some cases, immediately sorting the entire list after adding an item was unavoidable. Previously items were added at the correct position, which was O(N) because all items after the inserted one had to be moved in any case.
`sort_rang...

heinrich5991

I don't think it's fair to call that UB (unless you mean something different from the UB for the C++ standard)

we could also detect the speedhack server side, We already send the client ticks to the server? We just could compare them to the server tick, and if the margin is to big, we can kick the player

there is no "speed hack" that influences the server from the client

22:07

the client only ever sends inputs

22:07

hack detections never work, the only way to defend against them is on a systematic level

22:07

which we do, we don't trust the client

heinrich5991

there is no "speed hack" that influences the server from the client

that what I say the whole time... thats what I want to explain you

are you interested in my opinion on why this is an issue on github?

if the client would use a speed hack, its ticks would envolve faster, resulting in faster sending of inputs -> detectable -> kickable

heinrich5991

are you interested in my opinion on why this is an issue on github?

sure

okay, it seems that deen heard about this behavior of the ddnet client or server after it ran for a while with "speedhacking" enabled

22:10

i.e. likely something with faster time

22:11

deen thinks that this might have exposed a bug that our servers run into even when normally running for a long time

22:11

so he wants to see if that can be reproduced

22:12

so you think this is an issue that the servers cannot run into when running normally for a long time?

what would be the purpose of an artificial input margin? to circumvent some kind of input rate limit? (I dont see any)

22:14

oh, wait, I see, nvm (edited)

it is only intended to increase user experience. A player should not turn on a speedhack and then complain that nothing moves. instead, you could kick the player with speedhack so that he also has nothing to complain xD

22:16

but you don't need to kick them, as I said, they only desync their client.

I see

heinrich5991

so you think this is an issue that the servers cannot run into when running normally for a long time?

I'll just try that tomorrow and see where the server despairs xD

i think i'll open a discussion about constructor/destructor and class initialization

thanks. you're doing much better than me on this, I was only armchair-discussing the issue and you're verifying it. nice!

Chairn

i think i'll open a discussion about constructor/destructor and class initialization

thanks

speed hack produces UB even on the server?

c0d3d3v

we could also detect the speedhack server side, We already send the client ticks to the server? We just could compare them to the server tick, and if the margin is to big, we can kick the player

It would be nice to still keep a high max value, like 5-10 seconds or so, just in case

Tater

speed hack produces UB even on the server?

"speed hack" in the client doesn't affect the server

22:23

"speed hack" in the server affects the server

Oh ok

22:24

So what's the benefit of kicking the "speed hack" player?

maaaybe helping the user realize that it doesn't work

22:24

but I guess it might also just make them waste time trying to get around the check

But speed hack should be useless right?

it is useless

22:25

at least on official servers

the server doesn't trust the client. it just accepts inputs

22:26

speedhacking like in minecraft or cs:go is literally impossible in ddnet

I don't see why we should care other than to improve the user experience of someone who is trying to cheat?

this was just a side note, I think

we do not have to... I also do not want to implement it. I just wanted to point out it would be possible xD

[ddnet-web:master] 1 new commit

2ed6213 Add 20 € donation by corneum - def-

don't fret about it too long ^^

oh ok, sorry ill stop talking about it now xd

[ddnet/ddnet] Issue opened: #5247 Initialisation of classes

Recently, we had some bug due to badly initialized classes or overriding whole class with mem_zero. An easy would be to initialise all class members everywhere in the code. One way is to declare a default constructor and to zero initialize each members. Other way is to use C++11 assignment in the class declaration. 1. Default constructor with initialisation in initialiser list (either in header or source file) ```c++ class A { public: A() : m_i(0) {} private: int m_i; };...

if it only happens with time skip software, just close the issue. if it happens on real server with 14 hours, fix it. easy

1

you didn't specify about bullets being player's gun bullet, or game freezing bullet

I can ask Im 'corneum maybe he remembers xD

22:31

but is probably not important...

22:31

After testing it, we will have more data

[ddnet/ddnet] Issue closed: #5237 Client crashes on Linux when no X11 is present

22:34

[ddnet/ddnet] Pull request closed: #5240 Fix crash when graphics init failed (fixes #5237)

[ddnet/ddnet] Pull request opened: #5248 Add back compiler warnings on macOS (oops)

Judged the scope wrong.

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/#us...

[ddnet/ddnet] Pull request opened: #5249 RFC? Extract server main() to its own file, do some small cleanup

I need to use CServer class from some custom main(). It turned out to be a good reason to start cleaning this up. The first step is extracting server main() to its own file. I'm open to all ideas on how to handle InterruptSignaled (I've added an external getter function, somewhy it was the first that came to mind).

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration o...

Exported 297 message(s)