Posts by adam.h.

Ok, so to complete this journey here is the log with Linux version 6.10.0-rc3 and FFmpeg version 6.1.1, the stream is not playable when hardware acceleration is enabled.

Let me know if you think I can contribute with anything else.

I'm using that setup: <setting id="debug.setextraloglevel">128,2048,32768,262144</setting> so ffmpeg is covered. I didn't try with ffmpeg 6.1.1 though.

I tried that sample file attached to 23699 and it works fine for both software codec and hardware acceleration, see the log. As expected, with vaapi enabled it plays very smoothly, while it slightly stutters with software codec.

So I suppose it is something else im my stream which causes these issues.

Hi chewitt, I checked that all these HEVC samples work well with and without hardware acceleration, so it seems that there is something specific with my tested HEVC stream.

I could provide like 10 sec sample of that failing HEVC stream for further investigations, but since simple hacks with playercorefactory/ffmpeg are not working with the new Kodi and this is prioprietary Disney contents I suppose I would need some guidance (PM?) how to proceed with this.

Thanks for the pointers. I compiled PR for 6.10-rc3 but the latest kernel didn't change anything.

In case you want to look int the log file, first part shows software codec (it works), second part shows hardware support enabled (it breaks).

Hi chewitt, I don't think this is LE issue (perhaps more the problem in the NUC BIOS), but if you want to see the log, here it is.

Maybe this was discussed already here, but I cannot find anything specific. Perhaps my adventures will help someone.

In theory NUC10 should be fine with HEVC (hardware decoding is available in 10th gen.). But not in my case, the HEVC movie was just generating sound and video screen was not refreshing. I could see multiple entries in the log file, like: ffmpeg[0x1c3b8a30]: [hevc] hardware accelerator failed to decode picture.

So it helped to disable hardware acceleration in the Player settings. Apparently hardware driver was not as good as the software driver. Attaching also screenshot for the reference.

Thanks chewitt, just verified that the fix works on x86 as well.

Can someone please trigger nightly build for x86 as well?

btw. I always had impression that nightlies are triggered via some automation if at least one PR is merged during a day? But it seems that they are triggered somehow randomly, otherwise we should see 5 builds during June 11-15 and this would be consistent for all device types:

Changelogs:
    2024-06-11 (cb6b341):	#8978	kernel-firmware: update to 20240610
    2024-06-11 (e762f58):	#8979	alsa update to 1.2.12
    2024-06-12 (1a71ae6):	#8976	Migrate to upstream 6.11 RTL8192DU
    2024-06-13 (5301016):	#8983	package updates
    2024-06-13 (79d4d33):	#8982	systemd: update to 256
    2024-06-13 (18becd8):	#8984	linux (RPi): update to 6.6.33
    2024-06-14 (2216317):	#8988	iwlwifi-firmware: update to githash 1303bad
    2024-06-14 (12989ea):	#8989	rust: update to 1.79.0
    2024-06-15 (09e42f7):	#8991	vulkan: update to 1.3.288
    2024-06-15 (59be5f5):	#8990	eventlircd: bump to fix assert issue with systemd udev

RPi4 Builds:
    LibreELEC-RPi4.aarch64-13.0-nightly-20240611-dcca4e3.img.gz (sha256)	2024-06-11 09:35	146.8M
    LibreELEC-RPi4.aarch64-13.0-nightly-20240613-1a71ae6.img.gz (sha256)	2024-06-13 10:40	146.7M
    LibreELEC-RPi4.aarch64-13.0-nightly-20240614-18becd8.img.gz (sha256)	2024-06-14 07:50	146.8M
    LibreELEC-RPi4.aarch64-13.0-nightly-20240615-59be5f5.img.gz (sha256)	2024-06-15 17:02	146.8M

Generic Builds:
    LibreELEC-Generic.x86_64-13.0-nightly-20240612-1a71ae6.img.gz (sha256)	2024-06-13 03:46	254.1M
    LibreELEC-Generic.x86_64-13.0-nightly-20240613-18becd8.img.gz (sha256)	2024-06-13 22:55	254.3M

nightly-20240613-18becd8 - remote is not working with this one (no reaction to any button)

nightly-20240612-1a71ae6 - remote is back to normal

Perhaps this is the reason:

Jun 15 00:10:24.294870 NUC-ELEC eventlircd[541]: Assertion 'udev_device && key' failed at src/libudev/libudev-device.c:194, function udev_device_get_property_value(). Aborting.
Jun 15 00:10:24.329341 NUC-ELEC systemd[1]: eventlircd.service: Main process exited, code=dumped, status=6/ABRT
Jun 15 00:10:24.329360 NUC-ELEC systemd[1]: eventlircd.service: Failed with result 'core-dump'.

I switched the remote to One For All URC 7935, as I missed play/pause key

It works with TSOP sensor, attaching my keytable file.

It looks like texturecache.py was used for that purpose, some users tested it for Nexus so probably it works for Omega too.

texturecache.py qax movies @qaperiod=9999 @qa.nfo.refresh=1

where "@qa.nfo.refresh=1" means: refresh (remove/reload) all movies with nfo file modified during previous 1 day.

The problem is that it needs Python 2.x which is not available with Omega, so probably the easiest workaround is to bundle it together with old Python into Docker image and run it that way on LE.

Try to run some tests without NFS and see if this is the same or different?

Quote from popcornmix

The information you have linked to is over 7 years old and doesn't apply. Don't try this.

Good to know, thanks. However I would expect also some positive response to this guy, say "nope, no way, just restart your device" or "use that magic-command instead". Unless you are waiting for some ritual of ash and dust begging?

Quote from flintrpi

it would be nice if there was a command line I could enter from a putty session to reenable HDMI if this happens again.

see here

but probably you will need to install binaries via LE plugin

or try this

I wonder if YUV mode can be forced by editing EDID? This blog describes how to do it in the opposite direction (force RGB over YUV), but perhaps it won't work for LE.

Won't boot, no ACT light, no anything - bad SD card in that case...

Also the green led should blink with some pattern to indicate main problem - see the codes table here.

https://forums.raspberrypi.com/viewtopic.php?t=329860