00:48 <+bridge> <jxsl13> with a pitch fork and a dead microsoft logo on the ground
10:09 <+bridge> <chillerdragon> @patiga do you have a use case for detecting maps from images? or did you just build it to build it?
10:27 <+bridge> <0xdeen> I had the same thought when I read about the layoffs. I think https://www.machinegames.com/, which is next to id Software, not below it; but also belongs to Xbox.
10:35 <+bridge> <reitw> He does work at machine games yes
10:41 <+bridge> <chillerdragon> matricks the goat
10:57 <+bridge> <patiga> chillerdragon: I saw too many times that people ask for help on a map part with a screenshot and don't care to mention which map they are on. Also, sometimes people want to know which map a clip from a video is. This was in the back of my mind for some time, and then I had image classification as a subtopic in a uni course. That's where I first took this as a challenge (and there is also a sub-channel on this discord where people can post images and
11:00 <+bridge> <patiga> the repo I use for those two project rn is this one: https://gitlab.com/ddnet-rs/twml
11:00 <+bridge> <patiga> (and if you are interested in the weights I trained last time, I also freely shared them here)
11:49 <+bridge> <cellegenrih> Ooooh, so if we get the discord bot to run your algorithm to identify what segment of what map it is, the bot can just immediately tell us based on the screenshot where exactly it is
11:49 <+bridge> <cellegenrih> Sounds actually useful
13:16 <+bridge> <zavzagali> :kek:
13:33 <+bridge> <cellegenrih> Interestingly there are a dozen of duplicate functionality in map editor code with slight differences that can be combined into one thing
13:34 <+bridge> <cellegenrih> I'll investigate that once I'm done with my own stuff I guess
13:48 <+bridge> <chillerdragon> I think I managed to place 7 segmentation faults in a single line of code gg
13:48 <+bridge> <chillerdragon> `str_copy(pClient->m_aInput + Idx - 1, aRight, sizeof(pClient->m_aInput) - Idx);`
14:10 <+bridge> <boxymaloxy> when is a custom engine applicable in game development just wondering like is it easier is it more free
14:10 <+bridge> <boxymaloxy> (and if you have double the time maybe give tiny opinion on raylib vs custom engine)
14:31 <+bridge> <cellegenrih> the closest game engine I could give a fair comparison with other custom game engines (when it comes to Teeworlds / DDNet) is Godot Engine. Mainly C++ oriented, scalable and probably more stable than DDNet's current engine
14:31 <+bridge> <cellegenrih> To implement a fully functional client however... eh, I would rather try implementing it from scratch instead of copying everything DDNet has.
14:33 <+bridge> <ryozuki> - performance
14:33 <+bridge> <ryozuki> - niche things like the physics engine not fitting or wanting to support weird platforms
14:33 <+bridge> <ryozuki> - its more fun
14:36 <+bridge> <heinrich5991> if you want to build a game engine, build a game engine. if you want to build a game, use an existing game engine, otherwise you're likely never going to get a game
14:40 <+bridge> <ryozuki> i think if i were to make a game like Noita would make my own engine, but if i make a game like dead cells not
14:40 <+bridge> <ryozuki> https://store.steampowered.com/app/881100/Noita/
15:25 <+bridge> <Assa> Noita mentioned :deen_star: 
15:25 <+bridge> <Assa> I had enldess fun with this game
15:26 <+bridge> <Assa> To go into your conversation: It really depends on what game you want to make. Don't create a custom engine if you don't need it - in the case of Noita there was no engine that would support the vision of the game so they had to create their own
15:28 <+bridge> <Assa> raylib is like a good middleground - you can still create your own engine on top of raylib and it makes a lot of stuff easier - but it's not as feature complete as the unity engine for example
15:36 <+bridge> <cellegenrih> To go with a large game on an unknown engine, idk how to feel about that
15:38 <+bridge> <Assa> that's because you checkout a lot of gameengines before you choose one that fits your needs - like make a small audit
15:38 <+bridge> <Assa> that's why you checkout a lot of gameengines before you choose one that fits your needs - like make a small audit
15:39 <+bridge> <Assa> then it's not unknown anymore
16:01 <+bridge> <jxsl13> never :saddo:
16:29 <+bridge> <cellegenrih> :pepeW: But I will
16:29 <+bridge> <cellegenrih> It's horrible
16:29 <+bridge> <cellegenrih> Not the implementations themselves, just the process of adding new features
18:22 <+bridge> <cellegenrih> Can someone check out the branch for this half-baked WebP support prototype? I ain't expecting this to be a release-ready change, but at least we can have a moment to experience using WebP textures in DDNet
18:22 <+bridge> <cellegenrih> https://github.com/VertaireStudio/ddnet/tree/webp-support
18:22 <+bridge> <cellegenrih> I tried giving my fairest assumption of the change in the generated files.
18:23 <+bridge> <cellegenrih> I tried giving my fairest assumption of the change in the generated files. And yes, I did read the code 😄
18:26 <+bridge> <milkeeycat> > And yes, I did read the code
18:26 <+bridge> <milkeeycat> Did you write the code?
18:27 <+bridge> <cellegenrih> No, that's why it's a prototype, a concept that it's possible to do
18:27 <+bridge> <cellegenrih> If we need to change things, we can argue about it and change it together, without AI
18:28 <+bridge> <milkeeycat> I'm sure there will be tons of people happy to read the LLM generated code
18:28 <+bridge> <cellegenrih> :pepeW: That's why I don't PR it bruh
18:28 <+bridge> <cellegenrih> Pushing it as is will definitely make me the asshole
18:29 <+bridge> <cellegenrih> So, that's why I want your opinion on my public branch instead, easier to mess around that way
18:46 <+bridge> <cellegenrih> :justatest: robyt noticed it
18:49 <+bridge> <milkeeycat> I'm a weblate noob, how do I add my translations? Do I have to suggest them?
18:49 <+bridge> <milkeeycat> https://cdn.discordapp.com/attachments/293493549758939136/1525182142928392273/image.png?ex=6a52740f&is=6a51228f&hm=60e5e9a026c18d4d7f5304f8d235c4c144a551d09ddb045a4ed04add6c730003&
18:58 <+bridge> <cellegenrih> @robyt3 Turns out it is useful in terms of file size compression, as in we don't really care about the encoding / decoding speed different between PNG and this.
18:58 <+bridge> <cellegenrih> With a lossless WebP picture that retains the same quality of a lossless PNG picture, I can determine the following:
18:58 <+bridge> <cellegenrih> - for low resolution files (max. 1024x1024), file decreased by ~5-20%
18:58 <+bridge> <cellegenrih> - for high resolution files (max. 4096x4096), file decreased by ~40-60%
18:58 <+bridge> <cellegenrih> 
18:58 <+bridge> <cellegenrih> Given a simple test of converting PNG textures of described sizes into WebP textures of the equilavent size and quality, it proves a better file compression.
18:58 <+bridge> <cellegenrih> 
18:58 <+bridge> <cellegenrih> Assuming that the bare minimum of 10-20% size reduction in https://maps.ddnet.org/ is to be expected, it is a good deal.
18:58 <+bridge> <cellegenrih> Reducing a map size from 1MB to <800KBs in terms of the download speed you get from DDNet servers, it's a huge W
18:58 <+bridge> <cellegenrih> @robyt3 Turns out it is useful in terms of file size compression, as in we don't really care about the encoding / decoding speed different between PNG and WebP.
18:58 <+bridge> <cellegenrih> With a lossless WebP picture that retains the same quality of a lossless PNG picture, I can determine the following:
18:58 <+bridge> <cellegenrih> - for low resolution files (max. 1024x1024), file decreased by ~5-20%
18:59 <+bridge> <cellegenrih> - for high resolution files (max. 4096x4096), file decreased by ~40-60%
18:59 <+bridge> <cellegenrih> 
18:59 <+bridge> <cellegenrih> Given a simple test of converting PNG textures of described sizes into WebP textures of the equilavent size and quality, it proves a better file compression.
18:59 <+bridge> <cellegenrih> 
18:59 <+bridge> <cellegenrih> Assuming that the bare minimum of 10-20% size reduction in https://maps.ddnet.org/ is to be expected, it is a good deal.
18:59 <+bridge> <cellegenrih> Reducing a map size from 1MB to <800KBs in terms of the download speed you get from DDNet servers, it's a huge W
18:59 <+bridge> <cellegenrih> I think I once mentioned the same thing (but with less detail and proof) of the same concept years ago
19:00 <+bridge> <melon> roby closed it with dependency/maintenance cost as a reason, not that its a bad format
19:01 <+bridge> <cellegenrih> Well, maintenance is responsibility from whoever implements it, I guess they don't want to do that for years.
19:01 <+bridge> <melon> we could theoretically propose WebP solely for maps.ddnet.org
19:01 <+bridge> <cellegenrih> I can take responsibility
19:02 <+bridge> <cellegenrih> I understood half of the implementations related to file searches and just loading in the texture to memory, however Idk how saving / loading textures from maps work yet
19:02 <+bridge> <melon> we could theoretically propose WebP solely for maps.ddnet.org (not client embedded)
19:02 <+bridge> <cellegenrih> wouldn't there need to be a converter.... plus a way to use WebP files anyway?
19:03 <+bridge> <cellegenrih> I wouldn't be mad if we just keep maps with the PNGs, but when downloading DDNet maps, we use a special WebP version of it
19:04 <+bridge> <cellegenrih> where, maybe for backwards compat, converting to PNGs for unsupported clients are advantagious
19:04 <+bridge> <cellegenrih> actually, idk how old clients could interact with WebP based formats without conversion
19:06 <+bridge> <robyt3> In Arabic you can only suggest translation after the last incident where someone broke most translations (strings need to be reverse manually because the client doesn't support proper RTL rendering).
19:07 <+bridge> <robyt3> Suggestions should get accepted after 3 positive votes.
19:11 <+bridge> <milkeeycat> And for other languages I can set translations and they don't require explicit approval by owners?
19:12 <+bridge> <robyt3> Yes, you only need to save the string. They should get synced with the git repository every 24 hours or so.
19:18 <+bridge> <robyt3> The size reduction doesn't really seem that useful for how much complexity it adds. Downloading via HTTP is much faster than via game server anyway.
19:19 <+bridge> <cellegenrih> complexity? Nobody bothered with the PNG format for years, and I have essentially vibecoded my way to a working prototype in less than an hour
19:19 <+bridge> <robyt3> It's not only about all the glue code but also about adding a dependency on libwebp itself.
19:20 <+bridge> <cellegenrih> Well, is there a problem with webp lib's license?
19:21 <+bridge> <robyt3> In general adding more dependencies should be carefully evaluated. Can't have any vulnerabilities in the webp format by not supporting it.
19:22 <+bridge> <cellegenrih> Yeah, which is why I need people to review my branch first, then once every bit of necessity is applied to the branch, I could then confidently PR it
19:23 <+bridge> <cellegenrih> At least you could look at the branch for a moment, I tried to give my best for the commits
19:24 <+bridge> <cellegenrih> The WebP formatted textures load into memory as WebPs for now, but the moment you try to save it to a map, it converts it to a PNG
19:24 <+bridge> <cellegenrih> The WebP formatted textures load into memory as WebPs for now, but the moment you try to save it to a map, it converts it to a PNG
19:24 <+bridge> <cellegenrih> that's the only note worthy bug I noticed yet
19:24 <+bridge> <cellegenrih> (And the map extract only detecting .png formats, as expected)
19:27 <+bridge> <cellegenrih> I'll try to make it work
19:28 <+bridge> <robyt3> I did look at the branch. It looked like AI code to me --- correct on the surface but also with various issues. You disabled the CI on your repository but I assume it wouldn't pass.
19:29 <+bridge> <cellegenrih> Yes, as the commits also say on some files only checking for linux distributions
19:29 <+bridge> <cellegenrih> Obviously I won't forget about that cross-platform compatibiltiy
19:29 <+bridge> <cellegenrih> Obviously I won't forget about that cross-platform compatibility
19:32 <+bridge> <louis> Hashed images will probably get u like 90% filesize reduction
19:32 <+bridge> <louis> png is also a lot more familiar and historically worked fine so i cant see the benefits of supporting more extensions
19:32 <+bridge> <louis> (as in mapres decoupled from map files)
19:33 <+bridge> <louis> decoupling images from mapfiles will probably get u like 90% filesize reduction
19:34 <+bridge> <cellegenrih> webp is pretty much used on the internet as one of the most recognizable image formats
19:34 <+bridge> <cellegenrih> It would make sense, given the fact that ddnet was once built to a web application to run, to have some sort of WebP support
19:34 <+bridge> <cellegenrih> And the benefits are just a nice bonus
19:50 <+bridge> <heinrich5991> there's little reason to introduce a second image format to ddnet
19:50 <+bridge> <heinrich5991> we already have one that works, png, and I doubt there are significant savings justifying the introduction of a second image format to ddnet
19:51 <+bridge> <heinrich5991> for this, decoupling images from maps would net much better savings. they wouldn't have to be downloaded for each map individually
19:51 <+bridge> <heinrich5991> we have an issue for this: https://github.com/ddnet/ddnet/issues/5165
19:55 <+bridge> <cellegenrih> I don't know if you do this on purpose, or you truly believe that based on what I just mentioned, you don't even consider ever implementing one if it's inferior
19:56 <+bridge> <cellegenrih> I'll try to make my implementation work with as least amount of errors as I can detect, then we can discuss this properly, no?
19:57 <+bridge> <heinrich5991> I don't think we should implement webp if it's merely a bit superior, no
19:58 <+bridge> <heinrich5991> that doesn't depend on whether you provide an implementation or not
19:58 <+bridge> <cellegenrih> good to know
19:59 <+bridge> <heinrich5991> (if you look above, you can see that @robyt3 also thinks so)
19:59 <+bridge> <cellegenrih> If I look above?
20:00 <+bridge> <cellegenrih> I spoke to him 😄
20:00 <+bridge> <cellegenrih> And I understand that he might not want it, especially in it's current state. But at least he's open about it and he's willing to give me advices
20:13 <+bridge> <cellegenrih> Out of curiousity, is there someone implementing that? Cuz that means reconstructing and exporting maps on the server
20:14 <+bridge> <heinrich5991> I don't think anyone is implementing that right now
20:14 <+bridge> <cellegenrih> https://github.com/ddnet/ddnet/issues/5165
20:15 <+bridge> <cellegenrih> damn, a lot of discussion there
20:15 <+bridge> <heinrich5991> he's open about saying that he doesn't want it
20:15 <+bridge> <heinrich5991> he's not open about having it AFAICT
20:15 <+bridge> <cellegenrih> xd, ok now you are doing it on purpose
20:15 <+bridge> <cellegenrih> you win I guess
20:27 <+bridge> <robyt3> Apparently GitHub finally fixed their search index so that all PRs/issues are shown again. Now the update time of a bunch of PRs/issues appears to have been bumped to roughly the same time though.
21:54 <+bridge> <milkeeycat> If I want to add previously used string, can I copy it from .txt file or does that string have to be reversed?
21:56 <+bridge> <robyt3> Yeah, the strings in the txt should have been reversed properly already.
22:01 <+bridge> <milkeeycat> @robyt3 do you know by any change what's the Ewe language? I can't find `.txt` file for it
22:10 <+bridge> <robyt3> Looks like some kind of mistake with the language code being interpreted incorrectly. It looks like Estonian.
22:31 <+bridge> <robyt3> ^ Fixed. You can adjust the `Add friend` string assuming it's Estonian. I'll rename the file/translation after the changes from Weblate are pushed to the repository.