Announcement

Collapse
No announcement yet.

GNOME Shell & Mutter Broke Their Good Faith With Ubuntu

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by intelfx View Post

    When a new GPU comes out, a competent vendor develops and upstreams the drivers well in advance. Linux release model is fine.
    It doesn't matter, it still sucks. Even when you start upstreaming code before you get silicon back it's still hard to get alignment with Linux distros. Hardware schedules rarely align with distro schedules and hardware programs are usually designed to finish by product launch, not 6 months before product launch. On top of that, every distro uses a different kernel version. Everyone ends up having to backport lots of patches to various different kernel versions. Moreover, lots of customers want to use some older distro using an old kernel so even if you manage to get everything you need upstream and into the appropriate kernel releases to align with the next release of the distros, the customer wants to use a distro that was released a year earlier.

    On top of that, a lot of baseline features that customers rely on are not supported upstream due to bikeshedding and developer disagreements. To support them, you need to provide out of tree drivers with that functionality added on. The distros even support it. They go out of their way to make it possible to install out of tree stuff to enable customers even when there are open source solutions available because of these missing features. It's painful for everyone involved.

    Comment


    • #32
      Originally posted by user1
      What a moronic desicion by Canonical.
      Let's keep the experience worse for Nvidia Wayland users when the Wayland session isn't even the default in 24.04 for Nvidia.
      How is extra QA a moronic decision?

      Removal of the exception to get that package's point releases included automatically only means it has to be inspected and tested like any other package's point release.

      Comment


      • #33
        Originally posted by anda_skoa View Post

        How is extra QA a moronic decision?

        Removal of the exception to get that package's point releases included automatically only means it has to be inspected and tested like any other package's point release.
        Wait, so they are going to ship the point releases after all? My understanding is that at least for now they're not going to ship 46.1. Or am I missing something?

        Comment


        • #34
          Gnome... making friends wherever they go.

          Comment


          • #35
            Originally posted by agd5f View Post

            It doesn't matter, it still sucks. Even when you start upstreaming code before you get silicon back it's still hard to get alignment with Linux distros. Hardware schedules rarely align with distro schedules and hardware programs are usually designed to finish by product launch, not 6 months before product launch. On top of that, every distro uses a different kernel version. Everyone ends up having to backport lots of patches to various different kernel versions. Moreover, lots of customers want to use some older distro using an old kernel so even if you manage to get everything you need upstream and into the appropriate kernel releases to align with the next release of the distros, the customer wants to use a distro that was released a year earlier.
            As far as I know, the new kernel arrives quickly in Fedora, some weeks later. So the problem is that the distributions want to simulate some stability. That does not work well in my experience, because you get this custom patch hell.

            I have now changed it to a distribution that you can easily roll back. Thus, a reliable rollback solves the stability problem.
            Last edited by patrick1946; 23 May 2024, 11:35 AM.

            Comment


            • #36
              Originally posted by Brittle2 View Post
              you can argue that explicity sync is a bug fix for nvidia
              i thought that explicit sync is a bug fix for every gpu except nvidia.
              while nvidia had issues with wayland, i thought that wayland protocol was the problem and nvidia actually fix it for every gpu

              Comment


              • #37
                Originally posted by user1 View Post
                Wait, so they are going to ship the point releases after all? My understanding is that at least for now they're not going to ship 46.1. Or am I missing something?
                The announcement was only about revoking the automatic acceptance of Mutter point releases.

                It did not say anything about whether the release in question would or would not accepted.
                It did say, however, that Mutter point release can still be accepted based on the normal micro release process.
                I.e. with additional QA by Ubuntu.

                Apparently other people consider additional diligence to be a moronic decision and only blind believe in upstream efforts is acceptable.



                Comment


                • #38
                  Originally posted by user1
                  What a moronic desicion by Canonical.
                  Let's keep the experience worse for Nvidia Wayland users when the Wayland session isn't even the default in 24.04 for Nvidia.
                  Canonical didn't say they aren't going to update Gnome to the latest point release. In fact, it clearly states in the quote that they still intend to push point updates for Gnome. The only difference is that before, they would just assume each point update of Gnome was a bugfix/performance update and not check it for breaking changes to the desktop before building and shipping it. Now, they are going to have to manually check each point release to see if it conflicts with or breaks other packages before shipping it to users.

                  Comment


                  • #39
                    Originally posted by anda_skoa View Post

                    The announcement was only about revoking the automatic acceptance of Mutter point releases.

                    It did not say anything about whether the release in question would or would not accepted.
                    It did say, however, that Mutter point release can still be accepted based on the normal micro release process.
                    I.e. with additional QA by Ubuntu.

                    Apparently other people consider additional diligence to be a moronic decision and only blind believe in upstream efforts is acceptable.


                    Ok, I see now. Somehow I completely missed that part.

                    Comment


                    • #40
                      Originally posted by patrick1946 View Post
                      As far as I know, the new kernel arrives quickly in Fedora, some weeks later. So the problem is that the distributions want to simulate some stability. That does not work well in my experience, because you get this custom patch hell. I have now changed it to a distribution that you can easily roll back. Thus, a reliable rollback solves the stability problem.
                      In my experience, LTS kernels and using old kernels is both counterproductive and snake oil at same time. Related article: https://lwn.net/Articles/973996/
                      Last edited by timofonic; 23 May 2024, 12:50 PM.

                      Comment

                      Working...
                      X