Posts by chewitt

    Your download request hits our mirror redirector (running mirrorbits) which we would expect to route you to a mirror geographically close or at least in the same continent. The geo data isn't always accurate though, and with ISPs trading IPv4 address space ASNs that used to belong to someone else can give occasional random results. Or maybe the wrong number of jelly beans were in the jar today.. we honestly have no idea (and no real influence) on how it works.

    I think calling Generic "broken" because you can't use the Chrome browser add-on is a little over-dramatic. For the majority of users it works great. For a minority that need specific features currently tied to Xorg the GBM image is a regressive step; but they are the minority.

    We stopped formally supporting ARM devices with < 1GB RAM with LE10 as our feeling was the codebase using the newer video pipeline has outgrown 512MB devices and 1GB was the minimum needed for a good user experience. There's been no attempt to actively prevent installs on 512MB devices and stats show some LE10 users have been using it, but it sounds like things finally tipped over the edge.

    The issue is likely to be RAM footprint. You can save memory in a custom image rebuilding the kernel with fewer things enabled (and more use of =m than =y) and disabling services that might not be required; WiFi, BT, Samba, etc. and perhaps experimenting with ZRAM (there's a Pull Requst for that on GitHub). All of them are marginal gains but collectively they might accumulate enough to be meaningful.

    The easier option is repurposing a Model B board with 1GB RAM.

    External Content www.youtube.com
    Content embedded from external sources will not be displayed without your consent.
    Through the activation of external content, you agree that personal data may be transferred to third party platforms. We have provided more information on this in our privacy policy.

    Code
    2023-03-09 14:15:11.682 T:1194  warning <general>: Failed to get an OpenGL context supporting core profile 3.2, using legacy mode with reduced feature set
    2023-03-09 14:15:11.765 T:1194     info <general>: Using visual 0x20
    2023-03-09 14:15:11.796 T:1194     info <general>: VDPAU::CreateContext - creating decoder context
    2023-03-09 14:15:11.799 T:1194     info <general>: vdp_device = 0x000000 vdp_st = 0x000001
    2023-03-09 14:15:11.799 T:1194    error <general>: (VDPAU) unable to init VDPAU - vdp_st = 0x1.  Falling back.
    2023-03-09 14:15:11.800 T:1194     info <general>: CRenderSystemGL::InitRenderSystem - Version: 2.1 Mesa 22.3.5, Major: 2, Minor: 1
    2023-03-09 14:15:11.800 T:1194     info <general>: GL_VENDOR = Mesa Project
    2023-03-09 14:15:11.800 T:1194     info <general>: GL_RENDERER = i915 (chipset: Pineview M)
    2023-03-09 14:15:11.800 T:1194     info <general>: GL_VERSION = 2.1 Mesa 22.3.5
    2023-03-09 14:15:11.800 T:1194     info <general>: GL_SHADING_LANGUAGE_VERSION = 1.20

    The GPU fails to create the OpenGL surface needed for Kodi ^

    CvH I'm wondering if this device requires the older i965 driver?

    I added that ^ to the release announcement for LE 11.x final (and all beta releases). In the same text I also tell people to read the instructions in the wiki as boot processes have changed from older releases (things like needing to set dtb names). I also state that these images might not meet everyone's requirements, and it seems you are one of those users, so please continue to use LE 9.x .. or Android .. or anything else.

    I'm not sure there will be much add-on breakage with Omega. Kodi devs are generally running their own code so anything from the Kodi repo normally works fine, and there are no major API differences between Nexus/Omega (so far) to cause disruption with other add-ons; the Py2 to Py3 changes are long done and settled now. One change that will happen in LE12 is a switch from arm to aarch64 userspace as 64-bit widevine libs have finally appeared.

    On the kernel side, if there's anything notable to backport from newer kernels to 6.1.y then it can be done. I plan to send at least one more bump for LE 11.0.1 to resolve the "pink line" seen on S912 boards when scaling SD media to 1080p. I'm not expecting much progress to happen on hardware decoders in the near future though.

    NB: I will probably close this thread and fork a new one for K21 images soon. I do need to move the kernel codebase forwards to drop patches and ensure I'm not sitting on too much technical debt.

    Why does this download-file exist? And why does a SD-card with that file (for SD-card) not boot?

    Running LE11 from the internal emmc requires upstream u-boot; which that image contains. It cannot boot the box from SD card because vendor u-boot doesn't support the extlinux boot files needed for upstream u-boot. so you boot the box image (so emmc is not in-use) and then download and install the board image (that image) to emmc. Others have grasped that and succesfully installed. You haven't, but that's not really something I can solve. Your weird screen issue is not reported anywhere before so it's probably rare or something odd and unique about your environment. And the AMLGX image is not perfect, nor advertised as being perfect. But it works good-enough for me and my very typical media collection, and likely others have a similar experience (active install numbers are increasing). So sorry it's not for you, but we can't please everyone. I suggest you stick with 9.0.2 or bin the device or run Android or whatever you prefer.

    Your "serious flaw" is known because it's been seen before; maybe 4-5 times since public beta testing started. In that period there are ~20,000 installs on a wide range of hardware so it's definitely niche and not worth holding up the 11.0 release for. It's not been seen by anyone on staff or anyone in our ecosystem with a clue about triage/test .. and since we aren't really sure where the issue resides it's not clear what info might be required (but we do know the usual Kodi and boot logs will not show anything interesting). In time there'll probably be a Eureka moment and it'll be resolved. Until then .. /shrug