Eli Schwartz

@odd@mstdn.social @hailey@hails.org That doesn't mean we have to put up with horrid slop, it just means that blanket forbidding all LLM assisted software is in practice unenforceable outside our community (where we cannot analyze PRs and require assertion of non-use). Inevitably the result of slop is that the software deteriorates and doesn't work, and Gentoo may respond by freezing updates or removing it entirely (see chardet being frozen). We do what we can, and we make our disapproval clear regarding the rest.

@odd@mstdn.social @hailey@hails.org Given a choice between freezing all packages before vibe coding was "discovered" and packaging upstream releases, the former is not fully practical. The Gentoo social contract requires following council mandated policies about how one interacts with the Gentoo project, but doesn't control what other projects merge. The kernel is slopcoded too, but we can't exactly do without that, now can we? Well, okay, we recently added experimental GNU Hurd support. :P

@odd@mstdn.social @hailey@hails.org The fact that Gentoo, unusually among distros, actually supports (unlimited) multiple versions of a package and for users to pick and choose their versions, means there is a *lot* of latitude for isolating the dangers of slopcoding destroying a package. It's generally always possible to discover that the project fell apart and stopped working, and just tell portage to reinstall the old version. Removing slopcoded trash from the distro is "simply" a matter of removing one choice.

@odd@mstdn.social @hailey@hails.org I don't expect non-buggy versions of rsync to be dropped in favor of slopcoded ones.

@ratsnakegames@mastodon.social @david_chisnall@infosec.exchange The point being made wasn't that AI boosters will or will not be convinced, the point being made is that there is now legally un-retractable evidence that Google is liable for fraud if they advertise LLMs as accurate and someone buys into the advertisement and Bad Things Happen. The "someone" can, after the fact, sue them for misrepresenting the capabilities etc.

@umbrella@aus.social

By a fortunate coincidence, the only CI we utilized was a self hosted check (that is also run every ~15 minutes on pushes to https://gitweb.gentoo.org in order to check whether all ebuilds have consistent metadata and can be safely merged master -> stable) and a self hosted maintainer assignment + bugzilla integration.

The only thing necessary to get CI running on codeberg was to port the API calls for custom status checks etc.

@odoruhako@mastodon.social @davidgerard@circumstances.run Hmm... Even better is stuff like this: https://old.reddit.com/r/GithubCopilot/comments/1tzdcis/whats_the_best_read_cheapest_within_reason_way_to/ The alternative isn't even learning to code, the alternative is installing a real LSP instead of asking an LLM to simulate one.

@bentorkington@mastodon.nz @mid_kid@fosstodon.org @mgorny@social.treehouse.systems It's because anyone who even offers the option of using a systemd alternative is assumed to hate it. Either you're for or against it, there is no middle ground. And when most people hear "distro that supports not-systemd" they are only aware of Gentoo, Alpine, and maybe Devuan with the caveat that many systemd haters (or many people in general) would rather die than associate with the Devuan community. Gentoo is "a big name" in not-systemd.

@cocaine_owlbear@retro.pizza @eldersea@expressional.social @Canageek@wandering.shop The Bits from the DPL in question is older than Lucas Nussbaum's proposed GR. The interesting thing about it is, perhaps, more about the fact that the DPL is evidently in support of using LLMs, which may indicate something about the prevailing influence in that community. Of course, it doesn't seem to be happening on any particular timeframe. Of course also, this is Debian so "timeframes" are an... interesting concept. :)

@hugovk@mastodon.social the writing has been on the wall for a long time. Shockingly it turns out that #GitHub uptime isn't really better than #codeberg anymore, so the only really good excuse to keep using GitHub is if you're hugely reliant on free CI minutes. I sympathize and am in the same boat. I have been wondering for a bit, if it might be viable to have some simple automation that mirrors any PRs on codeberg to GitHub just to run CI, then report back on codeberg.

@cocaine_owlbear@retro.pizza @eldersea@expressional.social @Canageek@wandering.shop Is Debian "waiting for things to settle down" before proposing to go beg LLM vendors to sponsor "free LLM credits for Debian developers"? https://bits.debian.org/2025/07/bits-from-the-dpl-june.html