• A_norny_mousse@feddit.orgEnglish
    6·
    3 hours ago

    Who could’ve seen that coming.

    PSA: XX.10 versions are beta versions. If you use them you are an unpaid beta tester for the LTS versions and their paid customers.

    edit: since replacement is the implicit and explicit point of Rust coreutils, it borders on arrogance that they aren’t, first and foremost, 100% compatible with GNU coreutils. Are developers touting these as usable?

    • bluGill@fedia.io
      2·
      3 hours ago

      Posix combatibility should be the goal, and breaking GNUisms isia good thing. BSD and GNU have always been slightly different. (Once upon a time I’d talk about sysV but that seems dead now). Either write posix or be clear that you dependion GNU - depending on something without being clear is a bad thing.

      i don’t think rust coreutils are posix compatible either, but that is the bugs they should be working on first (ane they are clear that they have a lot of such bugs so I can’t blame them)

    • mesa@piefed.socialOPEnglish
      4·
      4 hours ago

      I personally dont care if its rust, python, C/C++ etc… what I do are is that its breaking things on essential services. At work we tried and found a couple of install scripts stop working because of the sudo issue. And that got us off Ubuntu server and now on Debian server (containerized but still). Thank god for CI/CD or this would be an issue later on. I hope they get everything rock solid because you cant have core utils deviate all that much given how many services rely on them. Accuracy should be more important than speed.

        • mesa@piefed.socialOPEnglish
          4·
          3 hours ago

          Very good question.

          We test nightly to make sure nothing breaks the app in CI. Its trivial and saves us future work. Some apps cannot go down no matter what (health related stuff) and we tend to deploy as soon as its stable.

          Its not my preference (I would rather just do LTS and stop) but im a small cog in the machine.

  • Baron Von J@lemmy.worldEnglish
    3·
    4 hours ago

    I must misunderstand what people are reporting about the checksum errors in benchmarking tools. Are the benchmarking tools validating the MD5 checksum of the binaries provided by Coreutils? Wouldn’t that be expected for the checksums of the Coreutils binaries to change, even just between OS releases even? Or are they saying that the Coreutils written in Rust are non-deterministic in the checksums they produce? That would be a huge bug indeed.