Time Nick Message 00:03 [MatrxMT] Not a warning, it crashes :P but huh, I had no idea about how that really worked 00:14 MTDiscord Zughy: ah right, misremembered that. it was part of the strict non-nil checks which iirc were relaxed to warnings at least for vectors at some point (?), guess that didn't happen for colorspecs. generally we try to have a warning first when making things stricter :p 00:14 MTDiscord "The job was not started because your account is locked due to a billing issue." github is really pissing me off now 00:15 MTDiscord (i've already filed a support ticket) 00:16 MTDiscord somehow i have the feeling this is a farce to get users to divulge billing info. there's no way they're unable to fix such an egregious, known bug. 00:23 MTDiscord <.zenonseth> on the plus side of the github issue: if github truly is starting to require devs submit card details to operate, it might be the kick luanti needs to do some self-hosting solution or something ? 00:28 [MatrxMT] I don't think they're this stupid 00:30 rubenwardy I could fully see this happening - requiring a card to use CI to combat abuse, even if not charged. Would push many away though 01:58 MTDiscord Github doesn't have to be especially stupid. They can just ride the wave of the rest of the world getting stupid. A lot of places are already requiring credit cards just to meet those new age verification laws. I wouldn't discount the possibility. 12:08 MTDiscord update: GitHub support fixed it, yay 13:03 MTDiscord i feel like luanti-org already should plan on migrating to some self-hosted instance of forgejo or gitea since github is getting down more and more often 13:21 MTDiscord Because the forums have such a great track record 13:22 MTDiscord Or formerly the wikis 13:22 MTDiscord While miles better, cdb as well technically 13:26 celeron55 github is doing a better job than any self hosted instance would do long term 13:27 [MatrxMT] And Codeberg is always an option 13:30 celeron55 codeberg is a better proposition and still the current escape hatch from github. but if uptime is what you're going by, i don't think you can expect any more from codeberg than github 13:31 celeron55 it's not a good argument. better ones are e.g. ethics, or a preference in UI, or seeking of a different contributor base 13:35 [MatrxMT] I'd say that not being on a platform belonging to a platform accused of genocide by the UN is a strong point regarding ethics. And we know there is a big list of other ethical reasons 13:41 ZenonSeth doesn't codeberg have a strict "No AI" policy now? unless maybe you get grandfathered-in ? 13:42 celeron55 well, that's one potential downside or upside, depending on who you ask. it's a big sad that things are so divisive these days. back when i decided to move minetest from bitbucket to github, there were absolutely no politics involved, it was simply a "those guys have a more fleshed out system" question 13:42 celeron55 a bit* 13:44 celeron55 (the fleshed out system being git overall - years after that bitbucket added git support, but they never managed to add the amount of polish and freebies github has) 13:45 ZenonSeth I'm not debating right or wrong and def not politics :) I was thinking just the consideration that Luanti is already on some "ethical foss" list in codeberg as "using AI" - so I was just thinking if that move would be allowed, should it happen 13:45 ZenonSeth be allowed by codeberg, I mean 13:46 celeron55 yes, as a pure allow/not allow question it is still a question and i don't know the answer 13:46 sfan5 funnily enough I remember Flathub also banning apps made using AI but there was a clause that this didn't apply to existing apps 13:48 celeron55 i really don't like such clauses. it piles up power onto people sitting on old projects instead of letting new projects grow. it comes out of practical issues of course, new projects spending resources without a userbase (like codeberg wrote in their blog) 13:49 celeron55 if FOSS becomes a "no AI here" museum, the future for it is weak 13:49 celeron55 interesting in its own sense, but useless for regular users 14:52 MTDiscord Codeberg unfortunately has adopted changes to their Terms of Use a couple months ago which I don't agree with. 14:53 MTDiscord "Content that harms the reputation of Codeberg" was extended to "such as cryptocurrency related projects". Now, I don't have much to do with crypto, I think a lot of the stuff surrounding it is sketchy, sure, but for a software hosting platform to outright ban everything related to crypto shows a very limited tolerance for software that might be considered "inappropriate". 14:55 MTDiscord The other thing being the "You must not share projects that mostly consist of code written by "generative AI"-tools (including services such as Claude, OpenAI Codex).", justified with unclear copyright status and lacking safeguards. 14:55 MTDiscord Content-based bans that are not imposed by external legal regimes are very much against the individual liberty and self-determination spirit of FOSS. If you want to find "content that harms the reputation of codeberg" I guess now you don't even have to look outside of their terms themselves. 15:02 [MatrxMT] I'm impressed by how people are annoyed by the Codeberg policy whilst being on an oligopoly-driven platform and being fine with it 15:03 [MatrxMT] Like, pick your favourite https://en.wikipedia.org/wiki/Criticism_of_Microsoft 15:07 sfan5 I think the difference is that Codeberg policies can likely actually impact us (project or individual devs) while things Microsoft does elsewhere are easy to ignore 15:11 [MatrxMT] I'd say that using the platform of our main "competitor" as a platform to build our version can impact us as well 15:13 [MatrxMT] I mean, we had NSA agents looking at random nudes, why not having some minions checking private discussions and alerts in case we become too big? We've been removed twice by Google Play 15:15 MTDiscord The platform I'm on is git, and it's not oligopoly-driven. My code being available on github or gitlab or whatever is basically a non-issue. However, apparently I can't include a codeberg mirror because they might turn their nose up at arbitrary bits of my work. 15:21 [MatrxMT] Nice twisting of words. Anyway, it's quite easy to host whatever slop you'd like when 1) your company can buy how much space it needs 2) profits from the very AI 15:22 [MatrxMT] Because let's be honest here: you might produce good code with AI, but the majority of people won't 15:23 MTDiscord It would've been reasonable for them to restrict compute for AI-generated repos. But they chose to go for a full on ban, and that they're doing the same for crypto while including a vague "stuff that might harm our reputation" clause shows where they stand. 15:30 celeron55 luatic: i definitely agree. hosting git takes very little resources, and you don't need to hand CI to AI repos or anyone at all to begin with. of course, you'll get more users if you do, but that's the cost vs. benefit comes into play 15:30 celeron55 that's where* 15:30 celeron55 it seems like an emotional over response 15:31 celeron55 and emotional over responses are not what i want from any kind of hosting 15:35 celeron55 and i don't want to be part of any elitist clubs. i don't want to be in a room where people categorically laugh at those who use AI, and neither do i want to be in a room where people categorically laugh at those who don't use AI 15:36 celeron55 offering free CI to vetted FOSS projects is fine and i'd argue absolutely expected. but also being able to pay for CI for you AI-created projects is expected 15:41 rubenwardy I think the crypto/AI rules are to combat a specific kind of spam and vibecoded projects, which Luanti is neither 15:41 rubenwardy and Codeberg has also said that they're doing this because these projects use a lot of their resources in CI 15:46 MTDiscord I think most people understand why they do it and agree that spam and resource consumption is a problem to be solved. What we don't agree with is their chosen solution 15:47 rubenwardy another thing is that we have a good relationship with codeberg so should consider raising this concerns directly with them 15:51 celeron55 i think the main thing is that people build on luanti using different amounts of AI all the way up to pure vibe coding. luanti isn't merging any of that (without proper review at least), but it would be bad if those projects couldn't exist on the same code hosting platform luanti is using 15:54 celeron55 and it's also not good if the platform actually allows those but is still openly talking against them 16:25 rubenwardy I think that the good thing about git is that we don't all need to use the same platform. GitHub is being suffocated with slop, it's fair that Codeberg doesn't want to be 17:26 MTDiscord rubenwardy: we can bring it up, sure. i don't think it'll change much however: this policy was decided by large votes among the Codeberg members iirc. 17:27 MTDiscord Luanti itself could tolerate their policy as it is, sure. But can the projects and people surrounding Luanti? I have some doubts. 17:28 MTDiscord I think it is becoming apparent that our stances towards AI for example are not really compatible. 17:29 MTDiscord Projects which ContentDB permits would not be allowed on Codeberg, for example. 17:44 [MatrxMT] Personally I think that accepting vibecoded stuff on CDB was a mistake in the first place 17:46 [MatrxMT] It's mostly a way to contribute to the slop. As a friend said, is it really open if people don't collaborate anymore because they can vibecode anything? Who's gonna ever read such code? Not even the authors 18:07 ZenonSeth to be pedantic, the answer is 'yes' - it's still open source if generated by llm. good usage of LLMs are readable too, but strictly speaking 'open source' means the code is available so you can fix, repurpose, or integrate it as you need to. The community is nice ofc but not what 'open source' means technically 18:09 ZenonSeth also re: cdb accepting LLM work: first it's near impossible to prove something IS or ISN'T LLM generated. If people lie about it, you'd spend more time arguing. Second, if the policy was that strict, people would start publishing other ways, and someone would make a good competitor platform. 18:19 [MatrxMT] So the reasoning is, we should accept slop because if we don't, somebody else will? 18:23 ZenonSeth That was only half my argument. But maybe we also have different ideas of what 'slop' means. 'slop' to me is a subset of LLM generated content - aka bad quality code. and I'm talking about CDB only, not engine development. 18:24 [MatrxMT] The other half is about don't understanding what is AI-made and what's not. I disagree, I'd say it's usually pretty straightforward 18:24 [MatrxMT] *not understanding 18:26 Noisytoot moving away from github is an ethical necessity 18:27 ZenonSeth re Zughy: we do indeed disagree on that. But even if you believe it's easy to tell, are you willing to waste time arguing about it with authors? Or is the solution to iron-fist block CDB content because one or multiple people may agree it's LLM generated? 18:28 Noisytoot moving luanti to codeberg doesn't mean everything on CDB has to be on codeberg too 18:28 Noisytoot > first it's near impossible to prove something IS or ISN'T LLM generated. If people lie about it, you'd spend more time arguing. 18:28 Noisytoot this is a dumb argument. do you think we should allow plagiarism just because it's impossible to prove it? 18:30 Noisytoot apart from all the ethical issues, github is still (in 2026!) inaccessible from the modern internet (it doesn't support ipv6) 18:30 [MatrxMT] ZenonSeth: people can report it and CDB staff can check, I don't see the problem 18:30 [MatrxMT] Just erring on the side of caution 18:30 MTDiscord Noisytoot: Not the point I'm making. When choosing a Git host, it matters where the people are, and where the people want to be. And a Git host with a policy incompatible with CDB policy isn't a good fit for that. 18:31 [MatrxMT] Literally the CDB developer is on GitHub only because of Luanti 18:32 [MatrxMT] Same applies here 18:32 ZenonSeth Zughy: that's fine in principle, but in practice it takes a lot of effort, and generates a lot of friction. I recently tried to enforce a "AI-generated" label on one package and has lead to long lasting discussion and ongoing issues, sadly wasting time of everyone. 18:32 Noisytoot I don't think a git host with a policy incompatible with software freedom and that geoblocks several countries (e.g. iran) is a good fit for that either 18:33 MTDiscord The geoblocking is required for compliance with US law, it's not really fair to blame GitHub for that 18:34 ZenonSeth that geoblocking is also arguably ethical 18:34 ZenonSeth since we're making such a big fuss about ethics 18:35 [MatrxMT] Wait, are we painting the picture Iran bad US good? 18:35 ZenonSeth no 18:36 [MatrxMT] ZenonSeth RE moderation: I can imagine but moderation is.. Not a fun task, yet a very important one 18:43 ZenonSeth > this is a dumb argument. do you think we should allow plagiarism just because it's impossible to prove it? 18:44 ZenonSeth Noisytoot, no i don't think that. I also don't think we should allow murders because they're hard to prove in a court system. But I do think much less effort should go into solving plariarism than solving murders. We're comparing different things. I don't equate LLM usage to plagiarism 18:45 ZenonSeth i've trained my brain on other people's code, I've been copy-pasting and then modifying solutions from SO for nearly a decade before LLMs were around. Does that count as plagiarism ? 18:48 MTDiscord legally, SO content is under CC-BY-SA. oftentimes however, if the code is the "obvious" way of doing something, copyright doesn't really apply. 18:49 ZenonSeth yeah i don't remember a developer who has ever cared or cited SO content in their life 19:05 Noisytoot The geoblocking is required for compliance with US law, it's not really fair to blame GitHub for that 19:05 Noisytoot it's avoidable by using a service not based in the US 19:05 Noisytoot ... such as codeberg 19:17 MTDiscord it is a legitimate minor pro of codeberg, though contingent on european and german policy. 21:01 [MatrxMT] codeberg's own blog post justified their ToS changes in terms of them having limited resources and wanting to shed what they didn't like. It's their prerogative as an operator to be choosy and that's fine by me, not every git forge needs to just let everyone post all kinds of projects 21:03 ZenonSeth it's funny cause when i say we don't have the resources to police all submissions for AI content, the reaction is that it's a hard but necessary job 21:03 [MatrxMT] there's no reason for luanti-org to be concerned about breaking codeberg's ToS 21:04 ZenonSeth to be clear, i have nothing against codeberg doing whatever they want with their policy. I'm not morally upset at them 21:04 [MatrxMT] That's what I think gets to the crux of how Codeberg's ban will look too: They just find the most egregious commit spammers and dubious crypto projects and yank them to save space 21:06 [MatrxMT] "serving git is cheap" - but Codeberg now (as all git web forges) has to mitigate a load of bots looking at every version of every file 21:24 Krock Hm... dug a bit in why RAII does not work with MinGW on Windows. For some reason, the (SEH) exception handler is only added after linking the program. Hence, the checks in LuaJIT turn out negative 21:26 Krock relevant https://github.com/LuaJIT/LuaJIT/blob/master/src/lj_err.c#L37-L41 21:32 Krock -DLUAJIT_UNWIND_EXTERNAL does cause LJ_UNWIND_EXT to be set, yet something still seems to be missing.. 21:54 MTDiscord Blockhead256: Codeberg is free to make their terms as they please, and we are free not to use it, no hard feelings. 21:55 MTDiscord That's what terms are for, to create an understanding of what to expect, and I think Codeberg has communicated that fairly clearly. 21:56 MTDiscord As I said earlier, their resources are not really a sufficient explanation for their policy. 21:57 MTDiscord The rationale for their crypto policy is clearly given, in their own Terms of Use, as it being potentially harmful to Codeberg's reputation, not as it being stressful on their servers. 21:58 MTDiscord The rationale for their AI policy explicitly includes software quality concerns. 22:00 MTDiscord The way I see it, if you look at the numbers, GitHub has a much larger active community than Codeberg. If you look at the services they can and do provide for free, GitHub again has massive advantages (very unsurprising since it's donations vs a hyperscaler). 22:03 MTDiscord The main thing Codeberg had going for them is being the "nice" alternative; their ToU changes have reduced my trust in their notion of "nice". 22:20 sfan5 if we decide to switch away from github the question becomes: what is the most "nice" github alternative? 22:29 [MatrxMT] Codeberg 22:31 [MatrxMT] We also get ~€800 every month, so we could donate some of it to Codeberg infrastructure. Just spitballing