Announcement

Collapse
No announcement yet.

A Brief Look At Fedora 24

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

  • #21
    Originally posted by Ericg View Post

    Stability? probably worse, though there are things in progress that try to remedy it.

    Up-to-dateness? Even more than Arch, most likely. Rawhide packages the release candidate kernels, mesa, and in-development Gnome snapshots.
    I haven't actually run Arch myself, but this sounds fairly accurate. We are definitely trying to make Rawhide much more reliable than it has been before; we've been continuously beefing up the openQA automated tests that are run nightly, and I try to be pretty proactive in getting things fixed when the tests find breakage. Current Rawhide nightlies are working pretty well, as you can see from https://openqa.fedoraproject.org/tes...dora&groupid=1 - note a couple of the test fails are test problems rather than real bugs.

    Comment


    • #22
      Originally posted by AdamW View Post

      I haven't actually run Arch myself, but this sounds fairly accurate. We are definitely trying to make Rawhide much more reliable than it has been before; we've been continuously beefing up the openQA automated tests that are run nightly, and I try to be pretty proactive in getting things fixed when the tests find breakage. Current Rawhide nightlies are working pretty well, as you can see from https://openqa.fedoraproject.org/tes...dora&groupid=1 - note a couple of the test fails are test problems rather than real bugs.
      Hey, Adam, question for you in regards to the OpenQA tests that get run. The other day I snagged a Rawhide ISO from your /nightlies page, and I noticed something odd.

      Fedora Rawhide x86_64 has been broken since March (like we talked about), and is still broken as of today. Yet, Fedora Rawhide 32-bit is now passing tests. If the problem was just related to that accountservices daemon and SELinux.. why would 64-bit fail, but 32-bit pass?
      All opinions are my own not those of my employer if you know who they are.

      Comment


      • #23
        Originally posted by Ericg View Post

        Hey, Adam, question for you in regards to the OpenQA tests that get run. The other day I snagged a Rawhide ISO from your /nightlies page, and I noticed something odd.

        Fedora Rawhide x86_64 has been broken since March (like we talked about), and is still broken as of today. Yet, Fedora Rawhide 32-bit is now passing tests. If the problem was just related to that accountservices daemon and SELinux.. why would 64-bit fail, but 32-bit pass?
        Heh, good timing - it's actually due to a very bizarre bug that I'm investigating right now:



        The only test that fails for the 64-bit Workstation live is the UEFI install test. It fails right at the end, because openQA expects to see that new user guide screen you complained about in the review, but it doesn't render properly.

        Turns out the reason it doesn't render properly is because there is no write access to /dev/shm, and that causes problems for Webkit, which yelp uses as its content renderer.

        There are a couple of other consequences of this, but nothing world ending, and it's only relevant if you're doing a UEFI install from the live image. All the other tests pass.

        This obviously doesn't affect the i686 image because there's no UEFI test there. The i686 image in fact only has a single test run for it, the basic boot-and-install test. We kinda cut down on the number of tests run on the i686 images to save openQA resources, because i686 is no longer release blocking, so a green result on nightlies is a bit weaker for i686 than for x86_64.

        Comment


        • #24
          Originally posted by AdamW View Post

          Heh, good timing - it's actually due to a very bizarre bug that I'm investigating right now:



          The only test that fails for the 64-bit Workstation live is the UEFI install test. It fails right at the end, because openQA expects to see that new user guide screen you complained about in the review, but it doesn't render properly.

          Turns out the reason it doesn't render properly is because there is no write access to /dev/shm, and that causes problems for Webkit, which yelp uses as its content renderer.

          There are a couple of other consequences of this, but nothing world ending, and it's only relevant if you're doing a UEFI install from the live image. All the other tests pass.

          This obviously doesn't affect the i686 image because there's no UEFI test there. The i686 image in fact only has a single test run for it, the basic boot-and-install test. We kinda cut down on the number of tests run on the i686 images to save openQA resources, because i686 is no longer release blocking, so a green result on nightlies is a bit weaker for i686 than for x86_64.
          Huh.. I wonder if that's why Steam was broken for me on my rawhide tester. I could load up the steam app, but it couldn't display the store. Steam's got chromium embedded in it, which probably still uses a similar system to what Webkit2 is running.
          All opinions are my own not those of my employer if you know who they are.

          Comment


          • #25
            Originally posted by Ericg View Post

            Huh.. I wonder if that's why Steam was broken for me on my rawhide tester. I could load up the steam app, but it couldn't display the store. Steam's got chromium embedded in it, which probably still uses a similar system to what Webkit2 is running.
            Actually sounds plausible. Try 'chmod go+w /dev/shm' and see if it helps...you'd have to do it at every boot, as it's mounted on boot as a tmpfs.

            Comment


            • #26
              Nice writeup Ericg! I certainly appreciate your writing Can you do KDE Neon next?

              Comment


              • #27
                Originally posted by boxie View Post
                Nice writeup Ericg! I certainly appreciate your writing Can you do KDE Neon next?
                I don't know if the KDE folks have forgiven me for last year's debacle But I'll look into it, I've got a spare 16GB ssd in this laptop that I keep for testing different distros.
                All opinions are my own not those of my employer if you know who they are.

                Comment


                • #28
                  Does the font rendering engine in Fedora 24 support subpixel rendering? Even when using hintslight+rgb, fonts look slightly poorer unless using subpixel rendering, and from what I remember, that is patented. Ubuntu comes with a rendering engine that supports this but Fedora has not previously (last time I checked).

                  Comment


                  • #29
                    Originally posted by Ericg View Post
                    ==Fonts==

                    Anti-aliasing: slight
                    Hinting: RGBA

                    ==Other changes==

                    dconf changes: org.gnome.desktop.interface.show-battery-percentage = true
                    Thanks a bunch for the extra suggestions! Loving the fonts

                    Comment


                    • #30
                      Originally posted by finite9 View Post
                      Does the font rendering engine in Fedora 24 support subpixel rendering? Even when using hintslight+rgb, fonts look slightly poorer unless using subpixel rendering, and from what I remember, that is patented. Ubuntu comes with a rendering engine that supports this but Fedora has not previously (last time I checked).
                      I think that patent is still in effect, therefore it's restricted to RPMFusion's freetype-freeworld package
                      All opinions are my own not those of my employer if you know who they are.

                      Comment

                      Working...
                      X