Posts by chewitt

On ARM SoC devices that support multiple output planes and where the GPU (2D/3D) is separate from DRM (rendering/compositing) the active planes are processed individually and then composited for HDMI output. There is normally a dedicated OSD plane, and on Amlogic hardware (at least the older generations that LE supports) the OSD plane is limited to 1080p resolution while video planes can handle 4K. The Amlogic vendor kernel does some internal hardware juju to scale the 1080p OSD plane to 4K at a later point in the display pipeline but the visual results of that forced scaling (scaling is always a lossy process) often result in visual artefacts and the output can be visually better when you leave the OSD plane at its native 1080p. I have zero code knowledge of what CE are doing but it sounds like they implemented some settings to fiddle with the kernel and stop the OSD scaling from happening. Other ARM SoC types, e.g. RPi, Rockchip, support 4K on all planes which must be better, right?

The problem is that rendering the Kodi interface always requires scaling of 2D/3D elements (can be accelerated on the GPU) to be combined with bitmap elements; artwork in the skin, artwork from thumbs and scraping (cannot be accelerated, so must be scaled on-the-fly, on the CPU). And as most (all?) Kodi skins are currently designed for 720p or 1080p output, running the 'desktop' at 4K means most of the time you must scale a bunch of things from 720p/1080p to 4K and this has a significant compute impact on the HTPC device. Even if you have "4K native" skins, thumbs and artwork are still scaled, and still generate compute load.

The TV also has its own scaling capabilities for handling SD/HD signals. The TV's capabilities are normally visually better, due to having fancier scaling algorithms, and scaling on the TV has zero compute impact on the HTPC. This means most of the time and even on more compute-capable x86_64 hardware; you have a better overall Kodi experience (fewer visual artifacts from scaling and snappier navigation) when you leave the Kodi GUI at 1080p, not 4K, and only switch to 4K output when playing video. Some users are obsessed with 4K settings though

Quote from kardano

I read that unfortunately the Rasp4 does not have hardware support for H265, which the Rasp5 does

RPi4 supports H264 decode to 1080p, and both RPi4 and RPi5 support H265 decode to 4K60. RPi4 and RPi5 software decode other formats (AV1, VP9, etc.). RPi4 should handle anything else to 1080p whereas RPi5 generally has enough grunt to handle up to 4K30.

RPi5 is faster than RPi4 but since you already have an RPi4 you might as experiment before investing more in upgrades. Amazon shows lots of inexpensive BT or 'USB air mouse' wireless remote options. Or a USB flirc receiver allows you to reuse an existing IR remote that you might have? - RPi4 also supports HDMI-CEC which lots of people seem to use.

popcornmix any ideas?

Done

Stop being a drama queen and exercise some patience. We know IA changes can be disruptive but we are dependent on upstream IA developers figuring out the issue and then merging fixes to their codebase - and we've merged changes to LE within an hour of those fixes being published. The project runs on people working for $free in their spare time, alegedly for fun. You might want to recalibrate your expectations.

I have never booted NOOBS/PINN, but we create tar files for them (still) so see if this works:

https://releases.libreelec.tv/LibreELEC-RPi5.aarch64-12.0.2-noobs.tar

Updates for LE12/LE13 have been merged and should appear in the repo over the next 24h.

There's nothing printed on the board that returns anything on Google, the spec info is too generic, and the manufacturer page contains no details. Dump/share the Android system/boot log and we can probably see what SoC/chipset is inside.

This will be somewhat hard to guess at .. but here's some comments and thoughts:

I'll make an educated guess that you can't SSH to the RPi as the college has implemented guest isolation on the WiFi network. It's a pretty standard restriction in "public" networks to prevent a device from interacting with anything except the gateway. It stops the computer science students from hacking everything (school networks are pretty hostile environments).

RPi boards have no RTC chip so date/time defaults to the libc release date until the network is up and connman makes a successful NTP request to the default pool.0.ntp.org servers (there is normally no need to configure any). TLS certs can be invalidated if time is wrong because e.g. the current date/time is seen to be in the past and the valid-from date for the cert is seen to be in the future so although you have IP connectivity the TLS connection is (correctly) rejected.

So there is probably an issue with NTP. If something blocks external access to pool.0.ntp.org or DHCP does not provide an NTP server (or worst, provides an invalid one) the clock is wrong and services using TLS certs (like repos) may not work. The workaround is to SSH to the device over Ethernet (direct cable) and set the clock to current time. You can also try configuring a local NTP server. If the college doesn't publish anything about their in-network NTP servers, try configuring the IP of the gateway as most will routers will respond to an NTP request.

Another possibility is you need to 'authenticate' to access the Internet? - perhaps the college has a captive portal where you need to enter credentials or click 'go online' to proceed (like a Hotel, Airpot, Coffee Shop). If yes, the two suggestions are:

a) Share the WiFi of a laptop over Ethernet and then connect the RPi to the laptop via Ethernet.

b) Share the WiFi of the RPi over Ethernet and then connect a laptop via Ethernet. This cannot be configured in the GUI, but can be done using connmanctl over SSH; so SSH into the RPi over a directly connected Ethernet cable and then use connmanctl to enable 'tethering' over Ethernet. This will allow you to use a browser on the Laptop to visit the captive portal and complete whatever auth challenge is required.

There's probably other things .. that's just what comes to mind.

Quote from kardano

Hi chewitt, i want to buy a TVBOX with an Amlogic chipset for a good 4K vision: can you recommend me a good device?

In short, no. The 'best' or more accurately 'least worst' supported devices are still S905X/D from the 2015/16 era; I wouldn't really use the word 'good' in the same sentence as Amlogic. Newer hardware has zero support so you're forced to deal with the vendor kernel (we choose not to go there) and their recent technical direction is quite hostile to open-source. I see them as a dead-end for Kodi support on Linux (even with vendor kernels) in the long-term.

The device I use daily and the only one I'll recommend to anyone at the moment is an RPi5.

How is this related to LE?

The crash log shows https://tv-server.fritz.box is configured. If you disabled TLS it should be http://tv-server.fritz.box ?

The better option is trusting the self-signed cert: https://wiki.libreelec.tv/configuration/ssl-tls-certificates

imhotep please boot the image in my share with meson-g12b-dreambox-one.dtb configured in uEnv.ini and paste the log.

Quote from wyup

Is it possible with a manual installer (dd command) to install Libreelec in available space on a disk with a previous media partition?

No, because 'dd' is the wrong tool for that job.

If you know what you're doing with Gparted (under Ubuntu) then it's not that hard to rearranage (shrink/move) existing partitions and filesystems to make space for the TWO partitions LE requires and then perform a manual install. However the method will be dependent on the hardware used and there are intentionally no instructions for this in the wiki; as guiding less experienced users through the process (and the inevitable recovery exercise when things don't go right) is not fun, and we don't support it.

Yasai-san thanks for confirming. I've submitted changes:

linux: disable CONFIG_PROVE_LOCKING for Exynos by chewitt · Pull Request #10024 · LibreELEC/LibreELEC.tv

This solves a messy boot splat (see below) visible in kernel logs shared by @ShigeakiAsai after the u-boot defconfig fix was applied. Googling flagged this…

github.com

linux: disable CONFIG_PROVE_LOCKING for Exynos by chewitt · Pull Request #10025 · LibreELEC/LibreELEC.tv

Backport of #10024

github.com

I've generated the patches by doing diff -Naurp against projects/Samsung/linux/linux.arm.conf and the .config file the kernel creates in the buildsystem linux build folder.

Quote from dipswitch

I don't know what llvm en drm means (this is the same drm as in "digital rights management"?

DRM = Direct Rendering Manager. This is a kernel API that allows userspace applications (mesa) to interact with GPU hardware that is managed by the kernel. LLVM = (not an acronymn) provides software rendering capabilities to mesa when hardware accelerated rendering of 2D/3D objects (e.g. the Kodi GUI) is not possible. We compile mesa with LLVM support although AMD cards will use an accelerated code path.

Quote from ubuntuuser

I am after true 4k or as close to true 4k as I can get in Linux.

4K is nothing more than the dimensions of the image. Both GLES and GL support 4K output. Current LE releases use GLES and there is nothing wrong with GLES. Future LE images will probably switch to GL, and there's nothing wrong with that either. Two different code-paths that use slightly different technologies to basically achieve the same thing.

If you mean 'HDR' output? HDR is only supported on GBM images (Generic) not X11 (Generic Legacy); and it is supported whether the image is LE12 (GLES) or LE13 (could be GLES or GL) as the rendering method isn't important.