DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.org/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 2025-01-14 00:00 and 2025-01-15 00:00

: when i decide to push before create a pr & sleep

Image attachment

[ddnet/ddnet] Pull request opened: #9528 Don't show add friend (clan) message when everyone is offline

Count offline friends when deciding whether to show add friend tooltip. Remove the useless offline tooltip. Before: !Before After: !After

Checklist

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

Hecta

Is there a way to filter tiles in a CLayerTiles struct? I know that void CLayerTiles::FillGameTiles(EGameTileOp Fill) exists, but is there a more general way to take one CLayerTile and only keep the tiles requested? For example, only keeping every tile that has an index of 1.

Did anyone answer this yet? The thing is, I made a psudeo-filter but it's really jank and leaves artifacts everywhere.

std::copy_if

06:47

is the stl way to do that

06:48

actually you need iterators

06:48

i'd have to look at code

06:48

im going sleep

hmmm alright

github-merge-queue[bot]

[ddnet:master] 3 new commits

39f2d78 Rename CScoreboard::Active function to IsActive - Robyt3 7a4a7f6 Rename CGameConsole::IsClosed function to IsActive and invert - Robyt3 22cde12 Merge pull request #9525 from Robyt3/Client-Component-IsActive - def-

07:47

[ddnet/ddnet] Pull request closed: #9525 Consistently name client component functions IsActive

good morning

07:54

@Ewan sleep well

1

github-merge-queue[bot]

[ddnet:master] 2 new commits

9706000 Fix crash when loading very long localization lines - Robyt3 6b1dfeb Merge pull request #9526 from Robyt3/Memheap-Overflow-Fix - heinrich5991

08:11

[ddnet/ddnet] Pull request closed: #9526 Fix crash when loading very long localization lines

github-merge-queue[bot]

[ddnet:master] 2 new commits

780110a Don't show add friend (clan) message when everyone is offline - Pioooooo a3c6679 Merge pull request #9528 from Pioooooo/empty-friend-msg - heinrich5991

08:18

[ddnet/ddnet] Pull request closed: #9528 Don't show add friend (clan) message when everyone is offline

Ryozuki

good morning

Good morning, rust up?

rust stable

<ChillerDragon> sometimes i end up getting a map from downloadedmaps when in the editor i say load current map but i also have that map in my regular maps folder

09:48

<ChillerDragon> then i dont notice edit the map reload and get really confused

09:48

<ChillerDragon> dont we check the maps/ folder before the downloadedmaps/ folder? I dont recall this kind of stuff happening back in the days

09:50

<ChillerDragon> oh i fighured out the root cause hmm thats fair

09:50

<ChillerDragon> so when you edit the map and the local server did not reload yet you end up in that situation not sure if it is fixable

09:51

<ChillerDragon> but at least we could throw a warning when you try to edit downloadedmaps/ since that really should not be a workflow then the hash in the filename becomes misleading

[ddnet/ddnet] Issue opened: #9529 Editor: throw a warning when trying to edit downloadedmaps/

Editing maps in the downloadedmaps/ folder should not be recommended. Because it causes a wrong checksum in the filename. If someone wants to edit a map they should resave or move it into maps/ When working on maps on a lan server my server sometimes has a older map than my client which causes me to download the map. And the the editors feature "load current map" opens a downloadedmap/ but what I assumed was a map/ and then I happily start editing without ever noticing it until stuff gets...

<ChillerDragon> @Jupstar ✪ why does a speedup layer count 3 and a tune layer count 1 i do not understand the pattern here

10:07

<ChillerDragon> https://github.com/ddnet/ddnet/blob/a3c667974b2aede95e2c1ebac12c38f75651a90c/src/game/client/components/maplayers.cpp#L1312

ddnet/src/game/client/components/maplayers.cpp at a3c667974b2aede95...

DDraceNetwork, a free cooperative platformer game. Contribute to ddnet/ddnet development by creating an account on GitHub.

10:07

<ChillerDragon> https://github.com/ddnet/ddnet/blob/a3c667974b2aede95e2c1ebac12c38f75651a90c/src/game/client/components/maplayers.cpp#L1318

ddnet/src/game/client/components/maplayers.cpp at a3c667974b2aede95...

DDraceNetwork, a free cooperative platformer game. Contribute to ddnet/ddnet development by creating an account on GitHub.

[ddnet/ddnet] Pull request opened: #9530 Simplify is entity layer check

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 (especially base/) or added coverage to integration test
[ ] 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-addresssan...

ws-client

<ChillerDragon> @Jupstar ✪ why does a speedup layer count 3 and a tune layer count 1 i do not understand the pattern here

bcs speedlayer has 3 visible params, and tune only 1

[ddnet/ddnet] Issue opened: #9531 Editor: unused teleporters/speedups can not be deleted with shift fill mode only with a brush

I am pretty sure other layers such as speedups and tune are also affected. Here a quick video explaining the issue. Make sure to unmute it has sound. https://github.com/user-attachments/assets/fdd61ba5-29e2-4dc9-9279-f8d1e19e77ba

<ChillerDragon> nice thanks!

github-merge-queue[bot]

[ddnet:master] 2 new commits

fc3da83 Simplify is entity layer check - ChillerDragon ae91b32 Merge pull request #9530 from ChillerDragon/pr_isent - Jupeyy

10:51

[ddnet/ddnet] Pull request closed: #9530 Simplify is entity layer check

https://discord.com/channels/252358080522747904/295908390956433410/1328712866093862964

Anime.pdf

https://discord.com/channels/252358080522747904/295908390956433410/1328712866093862964

https://discord.com/channels/252358080522747904/1264537509677891584

not going into ddnet client anyway

13:25

just fun feature to share

Anime.pdf

just fun feature to share

i tested it with a few ppl too, and tbh it sounds cooler than it is xD teeworlds is too dynamic and u often don't hear the others. maybe a game mode where team play isn't essential it works (like how it works in among us)

Jupstar ✪

i tested it with a few ppl too, and tbh it sounds cooler than it is xD teeworlds is too dynamic and u often don't hear the others. maybe a game mode where team play isn't essential it works (like how it works in among us)

the proof is u can speedrun with a partner with 0 voice chat

voice chat is good for pvp modes though

14:10

u always have all information in ddrace

louis

voice chat is good for pvp modes though

especially in pvp it was annoying

14:16

u can only scream a single word

14:16

then the player was gone again xd

Not proximity

Replying to @Jupstar ✪ especially in pvp it was annoying

ye with proximity it's probably troll

yeah ok vc only maybe good if u dont play with friends

14:19

ig in pvp the traffic problem wouldn't be as bad either

14:20

bcs broadcasting the voice packets of a single player to up to 64players is not cheap

Jupstar ✪

especially in pvp it was annoying

We will test this on TF servers with block and fng xdd

@Jupstar ✪ Implement a pencil weapon and allow people to draw on map backgrounds

murpi

@Jupstar ✪ Implement a pencil weapon and allow people to draw on map backgrounds

do you want to implement a skribbl io mod or what

nah, better map testers <-> map creators communication

we need wrench weapon that can place all sorts of block with radial selector overlay

zhn

nah, better map testers <-> map creators communication

for this

15:15

and add a pencil mode to the wrench

15:16

i will pay 25 burgers to whoeber adds this

louis

we need wrench weapon that can place all sorts of block with radial selector overlay

Add Redstone

️

Add worldedit

minetee

louis

we need wrench weapon that can place all sorts of block with radial selector overlay

lowkey nodes

[ddnet-rs:main] 2 new commits

6292d98 Change start arg parsing in client to directly change the config after loaded - Jupeyy d6f4957 Merge pull request #61 from Jupeyy/pr_client_config_parse - Jupeyy

[ddnet/ddnet] Issue opened: #9532 Add translation/localization instructions

Contributions are done at data/languages since always, but now there is Android strings at scripts/android/files/res/values/strings.xml. Should we add translation instructions to guide new translations and current ones that might not know about Android strings? If yes, where (the main README or a separate file e.g. TRANSLATIONS) ?

<ChillerDragon> new clang just dropped

17:15

<ChillerDragon> time to add .clangd config to ddnet

17:16

<ChillerDragon> @heinrich5991 if you dislike redirect layers what are my chances of getting it merged? I really think it is a cool feature that should go into ddnet

17:16

<ChillerDragon> i spent 2 days making it work and i am now polishing out the bugs

17:17

<ChillerDragon> a feature that is easy to use and has more code is better than something that is too hard to find or figure out for users with less code

17:18

<ChillerDragon> also i agree more code is always worse but i think you are taking it a bit too far and stalling feature development

honestly it feels like a very niche thing only like, 3 maps are going to use

<ChillerDragon> it is already use a lot in custom servers

17:19

<ChillerDragon> for example ugblocker

17:19

<ChillerDragon> but in a weird way that requires configs and the entities look cursed for players

17:21

<ChillerDragon> but yea for ddnet i only see 1 map that uses it maybe linear too not sure what the blocker situation is there

17:21

<ChillerDragon> and that also depends on the redirect protocol to be finally extended to allow redirecting to different ips

ws-client

<ChillerDragon> a feature that is easy to use and has more code is better than something that is too hard to find or figure out for users with less code

I get this point, it might be fun for some splitted maps, but can you make the redirect with a password?

<ChillerDragon> yea redirect with password would be really cool! but i think heinrich rejected that in favor of accounts

[ddnet/ddnet] Pull request opened: #9533 mastersrv: Add config file and hot-reloading of it

This allows to add bans and port forward exceptions without restarting the mastersrv itself. It also allows reloading the locations database (with the important caveat that it is mmap-ed, and as such must be removed before being overridden in the file system).

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 (especially base/) or added ...

the deen

I didn't know the CTO of Databricks and Apache Spark creator did water rendering physics for 0 A.D.

https://en.wikipedia.org/wiki/Matei_Zaharia

Matei Zaharia (born 1984 or 1985) is a Romanian-Canadian computer scientist, educator and the creator of Apache Spark. As of April 2022, Forbes ranked him and Ion Stoica as the 3rd-richest people in Romania with a net worth of $1.6 billion.

github-merge-queue[bot]

[ddnet:master] 2 new commits

8c01dcf mastersrv: Add config file and hot-reloading of it - heinrich5991 1ec83d5 Merge pull request #9533 from heinrich5991/pr_ddnet_mastersrv_config - def-

18:31

[ddnet/ddnet] Pull request closed: #9533 mastersrv: Add config file and hot-reloading of it

ws-client

<ChillerDragon> yea redirect with password would be really cool! but i think heinrich rejected that in favor of accounts

But what has redirection to do with accounts?

19:07

They are different things

Avolicious

But what has redirection to do with accounts?

for client state to share i thing

19:07

like cookies

19:07

i think*

Mhmmm, it does sound like two things for me

19:08

Accounts are accounts & redirection is redirection

19:08

It only allows to redirect on the same ip, just a different port

have to be something related with security and cookie hijacking and blah blah blah

[ddnet/ddnet] Issue opened: #9534 Client crashes with Vulkan when rendering very long text

Using Vulkan the client crashes when rendering very long text. When rendering a string of length ~20000 it crashes after a few seconds. With a string of length ~65000 it crashes immediately. Can be tested by changing the replacement string of any line in a localization file. Does not crash with OpenGL. I get two different crash logs, read and write: ``` DDNet.exe caused an Access Violation at location 00007FFE6B7E61D9 in module nvoglv64.dll Reading from location 0000000000000110. Ad...

[ddnet-rs:main] 2 new commits

f9ed119 Add fallback mode for cursor grab. - Jupeyy 8313293 Merge pull request #62 from Jupeyy/pr_client_grabbed_cursor_fallback_none - Jupeyy

[ddnet/ddnet] Issue opened: #9535 Buffer overflow in text renderer with OpenGL 1.1 (and presumingly gfx_quad_as_triangle 0)

Tried to reproduce #9534 . Used OpenGL 1 accidentally ``` src/engine/client/graphics_threaded.cpp:853:3: runtime error: index 32768 out of bounds for type 'CCommandBuffer::SVertex[32768]' (aka 'GL_SVertex[32768]') SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior src/engine/client/graphics_threaded.cpp:853:3 ================================================================= ==175288==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x7bac4a16dbfc at pc 0x55a3a1f81792...

@Robyt3 have u ever seen the onetime buffer flag?

Jupstar ✪

@Robyt3 have u ever seen the onetime buffer flag?

You mean related to the crash? I've seen it in the code.

Robyt3

You mean related to the crash? I've seen it in the code.

The crash should be fixed regardless, but I think the idea of the flag was that a text cannot be bigger than how big our MAX_VERTEX_COUNT is

20:39

that assumption is one of the reasons vk is so much faster in text rendering

20:39

bcs it reuses buffers as much as possible

Jupstar ✪

The crash should be fixed regardless, but I think the idea of the flag was that a text cannot be bigger than how big our MAX_VERTEX_COUNT is

ah, but it's being used for all TextEx rendering

Robyt3

ah, but it's being used for all TextEx rendering

Yes

20:39

do we need 65k character support?

20:39

we can always increase the buffer ig

20:40

or the text renderer has to flush (it probs has to anyway for opengl 1.x)

Jupstar ✪

or the text renderer has to flush (it probs has to anyway for opengl 1.x)

yeah, if that's possible that seems optimal, though limiting to some maximum amount of text instead of crashing would also be sane

[ddnet/ddnet] Pull request opened: #9536 Handle vkMapMemory result. Fail if memory exceeds max buffer size for stream buffers.

This fixes #9534 . Additionally this code changes handles the result of vkMapMemory now to make the code safer. The Vulkan backend intentionally does not support such huge buffers, to reuse buffers in the case of streamed buffers such as during text rendering. The actual fix for a text limit must be handled in the text rendering engine

Checklist

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

can i not add a format specifier in dbg_asserts?

Jupstar ✪

can i not add a format specifier in dbg_asserts?

no, it doesn't have vararg support right now

github-merge-queue[bot]

[ddnet:master] 3 new commits

2d1a48a Handle vkMapMemory result. Fail if memory exceeds max buffer size for stream buffers. - Jupeyy 4ec2858 Also add assert to graphics_threaded's CreateBufferObject & RecreateBufferObject - Jupeyy 51b7185 Merge pull request #9536 from Jupeyy/pr_handle_vk_map_result - Robyt3

22:03

[ddnet/ddnet] Pull request closed: #9536 Handle vkMapMemory result. Fail if memory exceeds max buffer size for stream buffers.

22:03

[ddnet/ddnet] Issue closed: #9534 Client crashes with Vulkan when rendering very long text

22:04

[ddnet/ddnet] Issue opened: #9537 Assertion when rendering very long text with Vulkan and OpenGL 3.3

Continuation of #9534. #9536 fixes the invalid reads/writes but the client now crashes with an assertion when rendering very long text with Vulkan and OpenGL 3.3. The text renderer should possibly flush the buffer when its full to ensure correct usage of the one-time-use flag. See also #9535 which is a different crash but may also be fixed by flushing, if I understand this correctly.

Exported 114 message(s)

Timezone: UTC+0