Did you include the kodi patches too?
Ooooh, signs of deinterlacing life! YAY! How exciting!
As the owner of a 4yo watching kids UK channels on SD live TV, I can't wait! I don't watch myself, but my eyes hurt every time they look in the TV's direction!
Things are still at the "frequently scratching head" stage of making it work so I wouldn't get too excited just yet.
Things are still at the "frequently scratching head" stage of making it work so I wouldn't get too excited just yet.
"frequently scratching head" is a start. A start is good! It's several steps above "it's on the to-do list, we'll get to that point one day". It's positive movement!
popcornmix: Thanks for your reply! I now replaced the official kodi package with yours but the latest patches don't work with your codebase. Do you have a working patch set or a working libreelec commit?
You need to remove some of the existing patches that conflict.
I have a script that does:
rm -f ~/projects/http://LibreELEC.tv/projects/RPi/devices/RPi4/patches/kodi/000*
rm -f ~/projects/http://LibreELEC.tv/projects/RPi/devices/RPi2/patches/kodi/000*
rm -f ~/projects/http://LibreELEC.tv/packages/mediacenter/kodi/patches/kodi-999.01-pr20210-drmprime-coordinates.patch
and I create the new patch with:
git diff-index $HASH --binary > /tmp/000-popcornmix-gbm.patch
where HASH comes from PKG_VERSION= in ~/projects/http://LibreELEC.tv/packages/mediacenter/kodi/package.mk
FYI: the latest development version of the deinterlace patches are now in LE master, today's RPi4 testbuild (http://test.libreelec.tv/LibreELEC-RPi4…126-8797e08.tar) already includes them. RPi2 testbuild with these patches will follow tomorrow.
so long,
Hias
Amazing! Just did an upgrade from 10.1 to the latest nightly and deinterlacing is running well. Done a few test runs with some 1080i content and so far so good.
downloaded, upgraded.
Tests with UK FTA DVB-T and -T2 TV
BBC 1 HD - 1080i50 H.264 - Looking great!
Tiny POP - 576i50 MPEG2 - Looking mostly good! (well, as good at 576i can... But it is deinterlaced!). There are sometime some weird artefacts, white lines, more or less where there was movement. See video below. Not ideal, but less annoying than interlaced to my eyes. The artefacts may be present in 1080i, but I could not see them.
https://youplala.net/~nico/le10-bad-video/PXL_20211127_112456189.mp4
Here are 2 sample TS snippets for you guys to play with, if that helps:
https://youplala.net/~nico/le10-bad-video/Gigantosaurus-2021-11-27-12-15-.ts
https://youplala.net/~nico/le10-bad-video/JoJo%20&%20Gran%20Gran-2021-11-27-12-15-.ts
Not great content, but cartoons help highlight the artefacts! 
Would a debug log help you?
Out of curiosity, what deinterlacing method is being used?
In any case, the head scratching was very productive, thanks!
camelreef thanks a lot for testing and the samples!
I've noticed the glitches on MPEG2 SD material, too, and popcornmix already has an idea what could be causing that - it might be related to software decoding, H264 hardware decoded videos don't seem to have those glitches (at least we haven't noticed them there yet).
So no need for a debug log ATM 
IIRC SD content will use the "advanced" deinterlacer, HD will use bob, of the videocore chip - pretty much the same as on RPi2/3 with LE9.2
so long,
Hias
Display Morecamelreef thanks a lot for testing and the samples!
I've noticed the glitches on MPEG2 SD material, too, and popcornmix already has an idea what could be causing that - it might be related to software decoding, H264 hardware decoded videos don't seem to have those glitches (at least we haven't noticed them there yet).
So no need for a debug log ATM
IIRC SD content will use the "advanced" deinterlacer, HD will use bob, of the videocore chip - pretty much the same as on RPi2/3 with LE9.2
so long,
Hias
Didn't 1080i end up being "MMAL Advanced" (as opposed to Bob) deinterlaced eventually on Pi 3B/3B+ chips? (Or am I confusing MMAL Advanced being implemented on Pi Zero/A/B single core models eventually for SD stuff, where previously MMAL Advanced only worked on Quad core Pis?)
ISTR that MMAL Advanced was quite similar to YADIF 2x. AIUI Weston 3-field may be a bit less computationally expensive (it's an option on some other playback applications now I think), and is certainly better than Bob.
"MMAL advanced" is borderline/too much for 1080i, therefore it's done using bob which needs fewer resources (memory bandwidth especially).
I'm not 100% sure about RPi0-3 but ISTR they had similar issues - except for test streams I don't have any 1080i content here.
RPis don't have a hardware deinterlace block therefore this is being done by software running on the videocore CPU - and that code is the same on all RPis.
so long,
Hias
"MMAL advanced" is borderline/too much for 1080i, therefore it's done using bob which needs fewer resources (memory bandwidth especially).
I'm not 100% sure about RPi0-3 but ISTR they had similar issues - except for test streams I don't have any 1080i content here.
RPis don't have a hardware deinterlace block therefore this is being done by software running on the videocore CPU - and that code is the same on all RPis.
so long,
Hias
Ah - I thought there was some GPU acceleration for deinterlace on the Pi (a bit like there is for HEVC on the Pi 3B+ and below) - so not a hardware block for deinterlacing, but some assistance from GPU processing. I've clearly been wrong for years!
(The MMAL is what made me think MMAL Advanced had some non-CPU support)
MMAL advanced is perfectly usable for 1080i on RPi3.
noggin IIRC the videocore code uses some hardware acceleration so I think you were about right. I'm not really familiar with all the gory details of the closed-source firmware blob and RPi folks will probably correct me if I got some details wrong
so long,
Hias
We've added a fix for the "white line" artefacts https://github.com/LibreELEC/LibreELEC.tv/pull/5925 which will be in today's nightly builds.
Testing here on RPi4 looked good so far, more feedback is always welcome
so long,
Hias
We've added a fix for the "white line" artefacts https://github.com/LibreELEC/LibreELEC.tv/pull/5925 which will be in today's nightly builds.
Testing here on RPi4 looked good so far, more feedback is always welcome
so long,
Hias
I'm guessing that it will be on the next nightly, as nightly-20211127-e3a7c06 still gives me the white lines.
I'll eagerly wait, test and report back!
