Announcement

Collapse
No announcement yet.

X.Org Server 1.16 ABI Breakage Causes Concern

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

  • X.Org Server 1.16 ABI Breakage Causes Concern

    Phoronix: X.Org Server 1.16 ABI Breakage Causes Concern

    A late breakage to the X.Org Server 1.16 ABI is resulting in (arguably rightful) concern by the proprietary driver developers at NVIDIA...

    http://www.phoronix.com/vr.php?view=MTY3NTI

  • #2
    How big of an impact can an API/ABI change for just cursor behavior be for NVidia to start bitching about it? I mean, the article makes it look like it was only a function or two that changes the way cursors are handled (and doesn't make it sound like a dramatic change). Surely that shouldn't be too much of a big thing for a company like NVidia...

    Also, while a frozen API/ABI is important, I'd much rather not have my cursor randomly disappearing, thank you very much :P

    Comment


    • #3
      Originally posted by Daktyl198 View Post
      Surely that shouldn't be too much of a big thing for a company like NVidia...:P
      If running a test suite for Q&A is taking a few weeks (and I think the nvidia one must be gigantic) tossing one month of work is probably very upsetting for devs and testers.

      On the other hand I'm not sure why one should expect ABI stability from software in development, that would be nice, but you can't always have nice things.

      Comment


      • #4
        Originally posted by Daktyl198 View Post
        Surely that shouldn't be too much of a big thing for a company like NVidia...
        It's a big thing precisely for a large company like NVidia, where a release is more work than commiting a fix, tagging it and uploading a tarball - all that pesky QA gets in the way. That being said, if their QA doesn't have a shortpath for minor but important bugfixes like these, then I'm not sure their complaints are valid.

        Comment


        • #5
          Really, this is what NVIDIA complains about:
          Code:
          -#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(17, 0)
          +#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(18, 0)
          ? Shouldn't that only be a "change one if/else statement and recompile", especially when they aren't affected by the cursor bug? Don't they need to go through the QA again cause of said cursor bug anyway?

          Comment


          • #6
            It is only and only Nvidia problem. They do not want to develop their driver openly, so they have to suffer of all these problems...

            Comment


            • #7
              Um, having the driver open-source does not affect the time QA takes.

              Comment


              • #8
                Originally posted by TAXI View Post
                Really, this is what NVIDIA complains about:
                Code:
                -#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(17, 0)
                +#define ABI_VIDEODRV_VERSION	SET_ABI_VERSION(18, 0)
                ? Shouldn't that only be a "change one if/else statement and recompile", especially when they aren't affected by the cursor bug? Don't they need to go through the QA again cause of said cursor bug anyway?
                Actually, it's more than that. They're breaking API (here's just one of the functions that change API). Patch at: http://cgit.freedesktop.org/xorg/xse...84efcaa7ec873e

                Code:
                - Bool
                + void
                (*load_cursor_image) (xf86CrtcPtr crtc, CARD8 *image);
                + Bool
                + (*load_cursor_image_check) (xf86CrtcPtr crtc, CARD8 *image);

                Comment


                • #9
                  Originally posted by curaga View Post
                  Um, having the driver open-source does not affect the time QA takes.

                  there is no need for qa in that case, cause even if the abi changes, when you compile xorg/mesa it's all ready and done. Open source drivers do not suffer from abi breakage/api breakage, cause the necessary changes are made alongside the api/abi at the same time, instead the binary blob needs constant catching up.

                  Comment


                  • #10
                    hm?

                    Originally posted by pali View Post
                    It is only and only Nvidia problem. They do not want to develop their driver openly, so they have to suffer of all these problems...
                    the problem is the xorg, we need a new display server. always problems with xorg support over the years, to much changes version by version without gains for users...

                    Comment

                    Working...
                    X