Linux is the name of a computer operating system kernel and a group of systems built on it, like Ubuntu or Mint. Linux is the basic infrastructure of the Internet. It’s important. A report ju…
  • self@awful.systemsEnglish
    30·
    1 day ago

    update: the patch set from Linus Torvalds, Boy Genius, went into review and the reviewers immediately found a race condition: https://woof.tech/@techokami/117446164348546588

    wow! it’s really easy to write fast kernel code when it’s wrong

    and now he’s apparently just dumping it on someone else because he’s “going on vacation”, a move that post-LLM commercial developers on most teams will be extremely familiar with

    • diz@awful.systemsEnglish
      6·
      7 hours ago

      By the way, his argument about “you wouldn’t mention tools” is just plain wrong.

      A few years back, at work, I found a security bug introduced by a commit that said something like “fix Coverity warnings”. I ended up reviewing every commit that mentioned Coverity, because that legacy codebase had a lot of rather bad tricks that triggered Coverity to have a false positive.

      And when a large number of false positives were getting resolved by editing code, sometimes that resulted in a false negative. E.g. Coverity complains about a memory leak, you end up with double free. Or vice versa, it complains about double free, and you end up with a leak. (The exact example was even more on the nose; Coverity complained about use-after-free, and the fix for it introduced a real use-after-free).

      Bottom line is, they do not want to have those labels because that would enable someone to actually measure how good or bad LLMs are at code.

      This is another thing that is extremely familiar to most commercial developers: there’s an LLM-powered code review bot, but unlike any tool ever before, nobody’s even trying to capture statistics whether it works or not. Why? Because whoever rolls out AI tooling isn’t sure that the tooling actually works for its stated purpose, and wants to for sure get their recognition and promotion for rolling out AI tooling, without jeopardizing it with the risk of adverse KPIs.

    • diz@awful.systemsEnglish
      41·
      7 hours ago

      Oh another thing, purely anekdotally, coding LLMs still suck at multithreading.

      I feel that’s one part where they could legitimately make actual “AI” for, basically a tool that either proves that nothing depends on exact timing, finds a counter example, or tells you it can’t do either (in which case your code is probably just very bad). But big overfunded AI labs probably feel that such tooling is beneath them.

  • supersquirrel@lemmy.caEnglish
    21·
    2 days ago

    At what point does this become best considered as purposeful or accidental, targeted sabotage?

    • lichtmetzger@discuss.tchncs.deEnglish
      31·
      1 day ago

      linus probably reviewed more code than almost any other dev in the world.

      We should measure Linus by his current actions, not by his actions from the past.

      • self@awful.systemsEnglish
        361·
        1 day ago

        but it’s so much harder to glaze the guy if you pay any attention to anything he’s done in the current decade:

        • receiving a gigantic paycheck from the Linux Foundation, almost all of which comes from AI corporations and their affiliates
        • failing to learn basic frontend Python for a side project and getting overly impressed with Google Antigravity (whatever the fuck that is) slopping it out for him
        • changing the kernel’s LLM policy based entirely on that experience to force engagement with a (reputedly horrible, but that goes without saying) LLM review bot and slop commits and responding to the pushback with “Linux isn’t a social warrior project”, a sentence that’s fractally incorrect and also a callback to the old SJW right-wing conspiracy theory
        • entirely slop-coding core kernel features, tests, and docs because Claude Code rewrote an old patch in a way that impressed him and that sole experience was all the excuse he needed to decide that LLMs are good at kernel code, in spite of all other projects to write less complex systems code with LLMs (Anthropic’s slop browser, Anthropic’s slop C compiler, bun, rsync, the open source firmware for the Xteink X4, MeshOS, that one disaster System 7 clone, probably more) ending in disaster
        • deciding at a closed doors maintainer summit that LLM commits no longer need to be tagged as such, eliminating documentation stating that the tagging was done as a quality control measure and making it impossible to tell if the coming shitstorm of issues for the kernel are due to LLM code (but they absolutely are)
        • declaring at the same summit that the kernel no longer wants inexperienced maintainers; that all maintainers are expected to either have the wealth to pay for Claude Code or to educate themselves (the latter learning path being by bashing their heads against LLM-generated PRs and the LLM CI nagbot that flags human code much more frequently than LLM code until they burn out, from long personal experience with slopified projects)
        • declaring at that same cursed fucking summit that LLMs work great if you set them up right and using the old patch as “proof” (a common blame-shifting move in corporate teams using LLMs) then refusing to share his setup or the steps he took to clean up the old patch (also a common move in corporate teams using LLMs)

        but if you paid any attention to any of these things, you might think he’s an asshole accepting money and tokens (he doesn’t pay for any of his own LLM usage) from some of the most corrupt corporations in the world in order to wrench control over Linux away from its own community

        and that would be crazy!

        • self@awful.systemsEnglish
          27·
          1 day ago

          oh and who can forget:

          at that same fucking summit he talked about how LLMs were irreplaceable tools for modifying the callers of a function whose signature you changed, a thing that IDEs have been able to do much, much more reliably and faster than LLMs since… the 80s I think? possibly the 70s? and they don’t drink several bottles of water doing it. they don’t even spike your CPU. it’s such a basic thing for an IDE or an editor plugin to do. ctags is a command line tool for this that came with fucking BSD 2.0.

          but torvalds famously used to only use a version of microemacs from 1987 to work on the kernel, an ultra cool choice that doesn’t even support syntax highlighting and later forked to become… nano of all god damn things

          so now he’s ultra-impressed by the basic shit you can do, because like a lot of programming dickheads he never learned any of his tools properly and made it everyone else’s problem

          and now like an MBA who thinks he can slop out a billion dollar app, his fried brain thinks it’s so impressive when the LLM does the basic editing tasks he never learned to do himself but much slower and much worse than an editor. and like that same MBA when his project’s burning to the ground, he’s shouting that it’s his underlings’ faults for not learning the LLM properly, for not prompting it right. no he won’t share his setup, because if he did it’d become very obvious he’s out of his fucking depth.

          godlike programmer my ass

          • self@awful.systemsEnglish
            16·
            1 day ago

            weird how all the assholes who insisted on making programming worse for themselves and others are now fully addicted to LLMs

              • self@awful.systemsEnglish
                10·
                1 day ago

                I mean, given that I’m typing this from an X window managed by emacs emulating vim…

                both vim and emacs are really good editors with an ecosystem of modern features (and a bunch of stuff that goes beyond what IDEs usually provide too). unlike an LLM, an editor is a tool that rewards learning. the problem is, guys like Torvalds are so unwilling to learn a new way of working that they turned the horrible way of working they started with into a fetish, and they made that fetish everyone else’s problem. there’s a lot of horrible development practices in the kernel that stem from conservatism and an unwillingness to learn.

                and as it turns out, LLMs are perfect in so many ways if you’re an overwhelmed conservative trying to hold onto power without ever having to learn a damn thing again.

    • self@awful.systemsEnglish
      231·
      1 day ago

      no matter how much you glaze him, torvalds will never know who you are

      • corbin@awful.systemsEnglish
        9·
        1 day ago

        Oh, to have a second upvote. Literally the only two modes of interaction I’ve had with Torvalds, across multiple kernel summits, are either (1) he ignores me, maybe kind of grunting, while I’m trying to say something like “good morning” or “how are you?”, or (2) yelling at the person standing next to me, who I’m trying to collaborate with, and thus yelling at me by proxy. I hope it’s no wonder that I only hack kernel in private, for patches I need, or for close friends.

        • self@awful.systemsEnglish
          8·
          1 day ago

          god, imagine the system we’d have if working on this shit didn’t fucking suck due to guys like torvalds