Posts by HiassofT

I tried playing the linked youtube video but I haven't been able to reproduce the issue here. Video info showed that the 1920x1080 25fps stream was played and tvservice -s showed that video output was 1920x1080p50.

I've tested both with LE 9.2.0 and the upcoming 9.2.1 release (should be out soon) with Estuary and Confluence skins. All overlays (progress bar, info) showed fine and the video signal didn't drop out.

Please post a debug log, that should give some hints if some of your settings are different to what I'm using here.

PS: reboot taking a couple of minutes when you hit the no signal issue is quite normal, I also get that here (the system has to run into a couple of timeouts until it's able to shut down).

so long,

Hias

Make sure you disable 25-30Hz in the whitelist (i.e. only include 24 and 50-60Hz) mode. If you enable 25-30Hz then these will be picked up when playing 50 or 60Hz interlaced content (which need to be output at 50p/60p) and playback will go south.

Which such a setup I didn't have any issues playing 1080i50 and 1080p50 H264 content with Hyperion enabled.

If you still have issue please post a link to a sample video that will trigger this issue - so far I haven't been able to reproduce the no screen issue by playing back a 1080 file, only with Netflix and 4k content.

Not sure why a reboot doesn't solve the no signal issue for you, this works fine here. Did your RPi really come back after a reboot (i.e. could you ssh in)? If that's the case have you tried turning of/on your TV - could also be that the TV somehow got upset by the signal from the RPi.

so long,

Hias

The problem with the dispmanx grabber in the RPi firmware is that it needs a lot of RAM bandwidth. It downscales the full (HD or 4k) image down to 64x64 (or whatever's configured in hyperion) which is then processed by hyperion. If that needs too much RAM bandwidth the HDMI output "starves" (can't get it's data in time) and the screen cuts out.

The downscale processes every pixel on screen which could probably be optimized (eg checking only every 2nd/4th/8th... pixel) which should still give a good enough approximation for hyperion, be a lot quicker and save a lot of RAM bandwidth.

Such a thing might be easier to do with V4L2 decoders, a new grabber could directly hook into kodi (or get the video buffers with decoded images from kodi) and could implement optimized downscaling in userspace.

But that's something for the future and of course a new grabber (which could then work on all platforms using V4L2 decoders) plus some plumbing on kodi side would need to be implemented in hyperion.

so long,

Hias

Probably, as the dispmanx grabber has similar known issues with 4k playback etc.

As this issue here only seems to show up when using the binary-only widevine CDM, eg on netflix, (youtube with inputstream adaptive works fine here) it's tricky to find out what's actually happening.

Dispmanx grabber issues are known for a long time but fixing them is rather low priority. Development focus of RPi kernel+firmware is on reducing power consumption (which already looks quite good) and on getting the non-firmware graphics driver and V4L2 decoders working.

so long,

Hias

Quote from JohnWayne111

Any Updates about this bug ?

No.

First make sure you have the hyperion addon version 9.2.0.112 installed - in earlier versions the screenshot function was broken.

I don't have an external grabber but with the internal (dispmanx) grabber on RPi4 the screenshot function works fine in hypercon. If I select the v4l2 grabber the connection log ("Show Traffic" in hypercon) shows that it tries to use hyperion-v4l2 but can't open /dev/video0 (which is expected).

If you can't get it working please post the output of "Show Traffic" after trying to grab a screenshot.

so long,

Hias

There are no downsides using the LibreELEC addon. We regularly update it to current version (though there isn't much happening ATM - we skipped the opencv grabber as that's not used in LE) and as it's created within our build system you can also be sure that it'll work, even if you update LibreELEC to a new version (in which case the hyperion addon will be automatically updated, too).

so long,

Hias

I'm not really familiar what the hyperion installer is actually doing, you'll have to ask on their forum instead.

The main profit of switching to the LibreELEC addon is that we can answer your questions.

so long,

Hias

I'd strongly recommend switching to the LibreELEC hyperion addon, then you'll also get regular updates.

ISTR the hyperion installer copies stuff to /storage/hyperion and modifies /storage/.config/autostart.sh to start it automatically on boot.

It should be enough to backup hyperion.config.json from your current installation, remove the files and directories (including autostart.sh), install the LibreELEC addon and then use the LibreELEC hypercon version to upload your hyperion.config.json file.

See Hyperion [LibreELEC.wiki] and Hypercon [LibreELEC.wiki] and/or search the forum for Hyperion

so long,

Hias

Quote from keeslampers

Has anyone managed to netboot Libreelec on a RPi4 yet?

Yes, I use this regularly for testing master builds.

Note that LibreELEC 9.2.0 has a couple of issues with netboot. I've fixed a bunch of these issues in our development trees a while ago, they'll be included in 9.2.1.

so long,

Hias

Quite certainly Hyperion is the culprit - I'm getting no signal quite frequently here. See this post for a workaround: Hyperion addon does not work with inputstream.adaptive (Netflix etc.) with Pi4

so long,

Hias

What you'd probably want is zswap, not zram, and I'm not too keen on enabling it.

zswap is still marked as experimental in the Kernel config and the info states it's not fully tested and can cause issues. Also the reports of people who tried it are rather mixed (including actual issues). As the stuff occupying RAM in this case is mainly uncompressed JPEG images I also have doubts how much it can be compressed by LZO/RLE.

LibreELEC 10 will swtich to the vc4-kms driver and V4L2 decoders where only a bit of gpu_mem is needed for H264 decoding but a lot of cma memory (curently testing with gpu_mem=76 and cma=256M/512M - 512M cma needed for 4k video playback) so things are going to change quite a bit. If there are issues with cma they'll need to be resolved anyways.

so long,

Hias

systemctl mask is essentially the same as symlinking a unit to /dev/null

LibreELEC stores the systemd user configuration in /storage/.config/system.d instead of /etc/systemd/system - therefore there's nothing in the (read only) /etc

so long,

Hias

111 is rpcbind. If you don't plan to use NFS you can disabling it by masking rpcbind.service and rpcbind.socket

so long,

Hias

Bug reports, especially with detailed info, are always welcome.

For issues in milhouse builds it's best to post in his threads on the kodi forum so everything's in one place.

The dmesg output helped me track down a bug lingering in the pcm512x driver for a long time, I've sent a fix upstream and it's already scheduled for inclusion in the next 5.6 kernel (and will then find it's way into earlier kernels)

ASoC: pcm512x: Fix unbalanced regulator enable call in probe error path - Patchwork

As posted in the milhouse thread I suspect the actual error is with the clock driver of the Hifiberry card (which Hifiberry folks need to look into, I don't have the hardware here to test myself) which then triggered the bug in the pcm512x driver.

so long,

Hias

Quote from mlac99

The original remote is broken, I was thinking of using a second one with this IR reciever. So, I must get another IR reciever ?

Yes. I'd recommend getting an (HP branded) MCE USB receiver, which is very well supported in Linux (and LibreELEC).

so long,

Hias

The logs look fine,

The Ortek IR receiver shows up as keyboard+mouse input devices and isn't handled by the rc-core subsystem so it won't show up in ir-keytable and you can only use it with the remote it was bundled with.

I did a quick test here on RPi4 with the remote+receiver and it worked out of the box.

so long,

Hias

The default RPi4 image has not been changed, reducing CMA memory is a bit risky (it could break "heavier" skins that otherwise work fine on the 2 and 4 GB models), so if you are affected by that issue please reduce CMA memory manually for now.

To be future proof I'd definitely recommend getting a 2GB (or 4GB) model.

so long,

Hias