Login
You're viewing the f.duriansoftware.com public feed.

Replies

  • [object Object]zzt@mas.to
    Oct 15, 2026, 7:53 PM

    e: I edited out the example screenshots that were here til I get some better ones

    💬 2🔄 2⭐ 6
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 15, 2026, 8:45 PM

    @zzt That post mentions diffing against the wrong commit, it's probably mixing in hundreds of unrelated commits. No wonder it doesn't make any sense.

    💬 1🔄 0⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 15, 2026, 9:01 PM

    @zzt Even then, I've looked at all branches in that repository for changes to net/netfilter/xt_HL.c (the file from the first screenshot), and can't find any change like that. It's baffling, where did that come from? It doesn't seem to be from the Linux git repository (including the branch mentioned in the message).

    💬 0🔄 0⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 15, 2026, 9:03 PM

    @zzt ...wait a bit, that's a diff between net/netfilter/xt_HL.c and net/netfilter/xt_hl.c, two unrelated files (the target and the match)! No wonder the diff makes no sense.

    💬 0🔄 0⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 15, 2026, 9:09 PM

    @zzt And the second screenshot is a diff between net/netfilter/xt_TCPMSS.c and net/netfilter/xt_tcpmss.c, again two unrelated files. One is the implementation for "iptables -j TCPMSS", the other is the implementation for "iptables -m tcpmss".

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • @pndcpndc@treehouse.systems
    Oct 15, 2026, 9:08 PM

    @zzt Those screenshots don't demonstrate what you think they do. Interestingly, they show Linus being arrogant and not as smart as he thinks he is in a completely different way.

    That's the sort of output you get when git has decided that a file copy or rename has occurred and picks a "similar" file to diff against instead, resulting in an incoherent mess. Because git just can't track copies and renames, and just guesses, because that's how Linus designed it. Don't you love it when your version control system just makes stuff up? We'd call that a "hallucination" today.

    💬 2🔄 1⭐ 0
  • @pndcpndc@treehouse.systems
    Oct 15, 2026, 9:19 PM

    @zzt I don't criticise Linus for coming up with that design which in retrospect turned out to have those flaws. It's elegant in its way, and works well enough most of the time, so that much is a success to be proud of. It's that he could never admit that these flaws existed and need fixing.

    💬 1🔄 1⭐ 0
  • [object Object]zzt@mas.to
    Oct 15, 2026, 9:23 PM

    @pndc makes sense, thanks! I’ve edited out the screenshots til I get some better ones

    💬 0🔄 0⭐ 0
  • Anthony "Erin" Fishingtonantonius@phpc.social
    Oct 16, 2026, 12:42 AM

    @pndc @zzt this reminds me of a conversation with a coworker where he was annoyed that git wasn't recognizing his file rename even though he explicitly used the cli to do it

    Having just read the docs I said something like "I think it's cause git doesn't actually record rename and it basically does a heuristic to guess"

    To which the only response I got was "i asked Claude and it said a bunch of crap" lol okay

    💬 0🔄 1⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 16, 2026, 1:14 AM

    @zzt So now I looked at the patch series more calmly. I saw none of the usual bad LLM tells (useless comments, strange phrasing on the commit messages). The code style, comment style, and very detailed commit messages are all usual for the kernel; if it weren't for the "Assisted-by: LLM" tag and/or Linus saying he used it, I wouldn't be able to tell. Even the race condition is typical of these kinds of patch series, and it's not the first time Linus posted a series with a bug like that. 1/2

    💬 3🔄 0⭐ 0
  • Cesarcesarb@mastodon.social
    Oct 16, 2026, 1:26 AM

    @zzt So whatever he did, he seems to still be in control for now. Let's hope he doesn't get dragged into letting the LLM think for him.

    Now, a personal opinion: it wouldn't be the first time Linus goes into what could be considered the wrong direction; it already happened before with Bitkeeper, and we all know how it ended. 2/2

    💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 0
  • David Gerarddavidgerard@circumstances.run
    Oct 16, 2026, 4:18 PM

    @cesarb @zzt given Linus chose this patch to put forward as his showcase for LLM quality, I'm not seeing a reason to construct excuses for him on this one

    💬 0🔄 2⭐ 0
  • The Orange Themetheorangetheme@en.osm.town
    Oct 17, 2026, 12:44 AM

    @cesarb @zzt Consider this: it's also boiling the ocean, stealing everyone's work, ruining the economy, emboldening the nazis, losing asstons of money, and it DOESN'T EVEN FUCKING WORK. Hope this helps.

    💬 0🔄 1⭐ 3
  • Catboy Cody :v_cat:​:v_bott:​:v_mlm::v_ace::ms_furry_pride:CatboyCody@tech.lgbt
    Oct 17, 2026, 8:49 AM

    @zzt Now I imagine giving Linus the slopcode review treatment I usually give my coworkers :nkoevil:

    But neither do I have any experience in kernel development nor am I in his new high-society club to get noticed... So even getting ripped a new one by Linus is probably unachievable. :neocat_googly_blep:

    💬 0🔄 1⭐ 0