Announcement

Collapse
No announcement yet.

Google Begins Upstreaming Fuchsia OS Support Into Mesa 3D

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

  • #11
    Google should be required to maintain out of tree patches. Reject this.

    This doesn't benefit anybody but Google.

    Comment


    • #12
      It's 30 lines of code. If Fuchsia dies, I'm sure it will be easy enough to remove those 30 lines of code.

      Comment


      • #13
        Originally posted by QwertyChouskie View Post
        It's 30 lines of code. If Fuchsia dies, I'm sure it will be easy enough to remove those 30 lines of code.
        Then you've clearly never worked with FOSS software before. That 30 lines of code is considered "platform support". That means going forward, Mesa would be required to support Fuchsia as a platform, and never make any changes that break that platform, even if Fuchsia dies. The FOSS community is extremely anal about backwards compatibility. It's why we still have platform support for 45-50 year old hardware in modern Linux kernels, because there's the slightest chance somebody wants to run a modern kernel on that hardware, and god forbid they can't.

        Comment


        • #14
          Fuschia OS is one of the operating systems I think looks the neatest, I like a lot of the design decisions they made regarding the userland parts and zircon. GFXstream support in mesa is probably one of the best things to come out of the work.

          I don't see why people are against adding fuscia OS into mesa, it shows nothing to hint at it's dying anytime soon, they did bring down the scope of fuschia OS (see workstation being replaced by workbench https://fuchsia.dev/fuchsia-src/cont...l=en#workbench) and personally I still think fuscia based operating systems still have a lot of potential IMO.

          But in the end, mesa should be getting pushed everywhere. because of the fuscia work, we are now getting gfxstream support in mesa, this means better android support (though we do already have venus support it has been less then reliable to say the least), better windows host support for linux vms etc.

          Comment


          • #15
            Looks surprisingly lively: https://fuchsia.googlesource.com/fuchsia/+log

            Comment


            • #16
              Originally posted by Daktyl198 View Post

              Then you've clearly never worked with FOSS software before. That 30 lines of code is considered "platform support". That means going forward, Mesa would be required to support Fuchsia as a platform, and never make any changes that break that platform, even if Fuchsia dies. The FOSS community is extremely anal about backwards compatibility. It's why we still have platform support for 45-50 year old hardware in modern Linux kernels, because there's the slightest chance somebody wants to run a modern kernel on that hardware, and god forbid they can't.
              Fuchsia isn't covered by any sort of CI, so it's not a functional blocker of any MRs, and extremely unlikely to be a release blocker either. (As far as I can tell, even Gallium Nine regressions aren't considered a release blocker, and that certainly sees more use than Fuchsia platform support does.) Besides, the fact that the extent of adding Fuchsia support was a handful of #ifdef lines, it's clear that A. Mesa is already extremely portable, and B. Fuchsia implements enough of POSIX that software designed to run on POSIX systems can trivially be made to run on Fuchsia. All in all, not much to see here.

              Comment


              • #17
                Originally posted by timofonic View Post

                Google is becoming Microsoft, Microsoft tries to look cool and friendly...

                Google hates Android, but they cannot replace it. ChromeOS is used as a cheap laptop for webapps and such, few hack it.
                ​​Chromebooks are used in most schools in the United States and is given to kids as a school laptop, replacing traditional paper books.

                Comment


                • #18
                  Originally posted by Daktyl198 View Post
                  It's why we still have platform support for 45-50 year old hardware in modern Linux kernels, because there's the slightest chance somebody wants to run a modern kernel on that hardware, and god forbid they can't.
                  Citation needed.

                  If you follow this site, Michael is reporting on Linux ripping out support for platforms much newer than that, left and right.

                  Comment


                  • #19
                    Originally posted by coder View Post
                    Uh, if Fuchsia is its own OS, then maybe it already has kernel graphics drivers? Also, do we know they're using an existing windowing system, or is that also proprietary?
                    The only indication here is that they're using lavapipe, a software renderer. If they're using that at all, it's probably pretty unlikely that they have a real vulkan driver working, and that makes perfect sense for an OS as immature as fushsia. It's a simple and lightweight solution compared to redox's VMs [1] or manually reimplementing a ton of hardware drivers, even if it means bad GPU performance.

                    [1] RedoxOS are in a similar boat, being a new microkernel OS, but don't want software rendering and don't want to reimplement all the kernel drivers. They're taking the highly unconventional step of (I think?) building mini linux VMs just to reuse linux's drivers.

                    Comment


                    • #20
                      Originally posted by Daktyl198 View Post
                      I feel like the Mesa project shouldn't accept code to support an OS that is not shipping, or planned to ship, on any actual PC hardware. At that point, it becomes needless bloat that the devs have to maintain for Google, with 0 actual purpose of existing there. With the way FOSS works, that code will be there for 20 years and never removed "just in case" even after the Fuchsia project dies completely.

                      Unless Google can demonstrate that this code is going to ship on a usable product that isn't a Google Home soon, or can demonstrate that these patch sets make Mesa better in a generic way, they should be outright rejected and Google should be forced to maintain their patches themselves.
                      It doesn't really matter if this ships on google home. Shipping in a product is shipping in a product, even if it's not a PC. A lot of the drivers in MESA are for random phone SoCs.

                      That said, I'm not sure if fuchsia has actually shipped yet in something that has a screen.

                      Comment

                      Working...
                      X