Posts by smf007

Appears to be an issue in the GLES code, an unintended variable assignment. I'll be taking this over to Kodi.

I have found an issue with GBM, confirmed with both Generic and Amlogic, when configured as limited colour range. The issue being that images lines for dvb subtitles have the range conversion applied twice. In other words, the lines where dvb_subtitles are rendered, not just the text but the whole width of the screen appears lighter then the rest of the screen.

This does not occur with Generic-Legacy (X11) nor Kodi (X11) on Ubuntu 22.04 (x64).

I can provide a kodi debug log, however I'm not sure it would help.

The current nightlies are using CONFIG_DRM_MESON=y, and has been for 4 years

You can also confirm from the currently running kernel by cat /proc/config.gz | gunzip | grep -i DRM_MESON

There has been a continual up and down over the last couple years with regard to the reboot issue.

If you try a google, there is a back and forth breakage between the Kernel and u-boot. General summary seems to be related to the power for attached devices, e.g. SD-Card, USB, EMMC, so what you have attached also seems to impact whether you are impacted or not. Even just changing the speed of SD-Card has an impact.

For any of the Libreelec team, the following might be of some help if you hadn't already seen:

Armbian - meson64: edge: rework to kernel 5.19 which with regard to the reboot issue just references

This discussion of CONFIG_DRM_MESON=m vs. CONFIG_DRM_MESON=y in terms of a broadly acceptable solution.

Armbian's later update to Kernel 6.0.13 continued to follow suit on CONFIG_DRM_MESON=y

Armbian - Update kernel configs for Meson64 CURRENT / EDGE

Using the latest Generic-Legacy nightlies, there are missing library dependencies for the music visualization plugins. Seems to be a case of the plugins being compiled for Generic (GBM) instead of Generic-Legacy (X11). Below is for shadertoy, but the others are similar.

I had cleaned out any archived copies in the packages folder, so I don't think it got a previously cached copies of the plugins.

ldd: $warning: you do not have execution permission for `./visualization.shadertoy.so.20.3.0'
        linux-vdso.so.1 => linux-vdso.so.1 (0x00007fff70fa6000)
        libGLESv2.so.2 => not found
        libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x00007ff15b05b000)
        libm.so.6 => /usr/lib/libm.so.6 (0x00007ff15af7e000)
        libgcc_s.so.1 => /usr/lib/libgcc_s.so.1 (0x00007ff15af5e000)
        libc.so.6 => /usr/lib/libc.so.6 (0x00007ff15ad9e000)
        /usr/lib64/ld-linux-x86-64.so.2 => /usr/lib64/ld-linux-x86-64.so.2 (0x00007ff15b2dc000)

I have gotten it working. Hopefully the following would provide assistance to anyone else encountering the problem.

I was setup using the main lirc configuration file, Lircd.conf. As a test, I switched to using the distributed toml keymap file (see /usr/lib/udev/rc_keymaps/*). To do so, remove the lircd.conf and create a file ~/.config/.rc_maps.cfg containing:

* * hauppauge.toml

So follow these instructions instead of the Lirc documentation.

I'm using an IR remote that is using the Hauppauge codes (PVR-250). Both a Hauppauge remote as well as a Harmony universal programmed as Hauppauge. The IR receiver being the built-in Odroid C2 one, so meson-ir.

The lirc.conf for the remote is the borg standard Hauppauge from the lirc website.

ir-keytable -t shows only one scan code per key. I also note that it immediately responds for each keypress, there is no waiting for the second keypress.

Using the nightlies on an Arm - Odroid C2, there is an issue for my remote following the upgrade to Lirc 0.10.2.

In short, Lirc 0.10.2 does not issue commands on each keypress. Instead, the key presses appear to be queued, with the queue being translated to commands on every 2nd key press. For example:

1. Press down button - No effect on Kodi
2. Press right button - Kodi responds with a down action (and navigation sound) followed by a right action (and navigation sound).

In terms of the nightlies:

• Working LibreELEC-AMLGX.arm-11.0-nightly-20221007-9c1305e-odroid-c2
• Not working LibreELEC-AMLGX.arm-11.0-nightly-20221008-1d2cd1d-odroid-c2

Between those two builds is Pull lirc: update to 0.10.2 #6966

The Kodi debug log does not help since it only records when it receives the commands from Lirc, which is after the issue.

Quote from vpeter

I misread the situation Thanks for pointing it out.

Sorry. Looking back I realize I could have written that clearer.

So to clarify, for x86-64:

• Generic (GBM) - missing libraries

• Generic-Legacy (x11) - works

vpeter thanks. With the rebuild of the binary plugins, everything is good now.

I will note though that generic-legacy is required due to the dependencies on X11. GBM is missing the libraries so Kodi will just sit with the spinner spinning, no recognition that the libraries didn't load unless you check a debug log. For GBM, the missing are:

libXext.so.6 => not found

libX11.so.6 => not found

libxcb.so.1 => not found

vpeter does the "jre for bd-j menus" plugin need an update for the nightlies due to the libbluray version bump? i.e. do the .jar files need to match the libraries?

Thanks.

The builds versus hash in order of commit history are:

In order of commit, the Pulls covering that range are:

3a4dc81 Pull #6717

1182c1a Pull #6721

13bb8f3 Pull #6724

06fd018 Pull #6715

2cc0585 Pull #6726

771b9b9 Pull #6470

f529cc9 Pull #6727

0da8a26 Pull #6729

0008ab5 Pull #6730

None of them are obvious as to breakage. If you were able to build Libreelec yourself, I would probably start by looking at #6727 as that appears to have triggered issues for others as well.

Sorry, issues is not with Libreelec. The issue is related to makemkv not playing nice with the current Libreelec. Removing makemkv or upgrading to a more recent compile resolves the issue.

daniel235 so it worked fine with 2022 June 08 build of b78941a?

If that build worked fine, then the question would be if pull #6673 caused the problem. That was merged 2022 July 08. So try something just after that. Something like

https://test.libreelec.tv/11.0/RPi/RPi4/LibreELEC-RPi4.arm-11.0-nightly-20220709-6050346.img.gz

Then to confirm that pull is the issue, try the build just before

https://test.libreelec.tv/11.0/RPi/RPi4/LibreELEC-RPi4.arm-11.0-nightly-20220707-379df42.img.gz

If it is that pull, then it will need to be dissected some as there are a number of changes included in it, library and Kodi changes.

When trying to play a physical bluray disc using nightlies (x64)(nightly-20221114-9c34b63 11.0), they aren't working.

Debug log

This is a disc that has worked on this system previously using much older Libreelec versions.

chewitt, issue is if one wants to disable the service. It can't be disabled as the symbolic link, from the "WantedBy" dependency, can never be deleted due to the file system being read only.

When ServiceA has a "WantedBy" ServiceB, a symbolic link is created under ServiceB that results in ServiceA being run. When using systemd to disabled ServiceA, systemd would delete the symbolic link.

In this case our filesystem is read only, so systemd cannot delete the symbolic link under kodi.target, thus the ledfix service can never really be disabled as kodi.target is always running it.

Quote from smf007

Hi chewitt.

I was just trying to disable ledfix.service on my C2 when I discovered that service is disabled already, yet it still is being run at boot. It seems ledfix.service currently has an error in that it lists kodi.target as both a "Wants" as well as a "WantedBy". Likely the "WantedBy" should be dropped.

I take that back. Seems the issue is purely that when having a "WantedBy", systemd would be creating (enable) or removing (disable) a symbolic link under the wantedby target, but that fails since the filesystem in read-only.

Hi chewitt.

I was just trying to disable ledfix.service on my C2 when I discovered that service is disabled already, yet it still is being run at boot. It seems ledfix.service currently has an error in that it lists kodi.target as both a "Wants" as well as a "WantedBy". Likely the "WantedBy" should be dropped.