Kodi Community Forum

Full Version: LibreELEC Testbuilds for x86_64 (Kodi 19.0)
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
New LibreELEC.tv Matrix build #0124: Generic
(Supercedes previous build)

SHA256 Checksum: 07d888de62bf334ea84c29cc86c40a78e133508e23628b4f046d0ecab8bf3e71 (Generic)

text:
# uname -a
Linux NUC 5.4.14 #2 SMP Fri Jan 24 21:03:46 GMT 2020 x86_64 GNU/Linux

# lsb_release
LibreELEC (Milhouse): devel-20200124210300-#0124-gd829752 [Build #0124]

# Kodi version
Kodi (19.0-ALPHA1 Git:43b613a). Platform: Linux x86 64-bit

Based on tip of LibreELEC.tv master (d829752, changelog) and tip of XBMC master (43b613a, changelog) with the following modifications: Build Highlights:
  1. wireguard: add initial support for WireGuard VPNs
  2. Python: update to Python3-3.7.6 and various related packages
Build Details:
  1. XBMC:
    • [Estuary] fix brackets in DialogVideoInfo.xml (PR:17248, 1 commit, 1 file changed)
    • [PVR] Context menus: Implement 'Record', 'Stop recording', 'Switch to channel' for EPG gap tags. (PR:17251, 1 commit, 3 files changed)
    • [PVR] Channel Manager: If started via channel's context menu, presele (PR:17250, 1 commit, 3 files changed)
    • [windows] Fix downloading dependencies from mirrors (PR:17231, 1 commit, 1 file changed)
  2. Additional commits/pull requests/changes not yet merged upstream:
    • Added: [env] PR:4139 (perma): wireguard: add initial support for WireGuard VPNs
    • Added: [env] PR:4141 (perma): Python: update to Python3-3.7.6 and various related packages
    • Added: [env] PR:4142 (perma): boost: update to boost-1.72.0
    • Updated: [pkg] PR:17222 (perma): [url] skip parsing username:password@ for udp/rtp streams
(2020-01-25, 01:53)Milhouse Wrote: [ -> ]my suspicion right now is that the GUI updates that show the user the progress information may be quite expensive - a simple Python3 test outside of Kodi (without GUI updates) achieves 99MB/s. We do use fairly small 32KB chunks in the Kodi download so increasing the chunk size may help improve bandwidth. Need to investigate.

Yes it's the GUI (dialog) progress updates here - combine a fast internet connection with a small (32KB) chunk size and we're updating this dialog over 70 times a second, which is slowing down the whole process (and computed speed).

We need to rate limit the dialog updates - I'll have something to test in #0125.
(2020-01-25, 02:36)Milhouse Wrote: [ -> ]
(2020-01-25, 01:53)Milhouse Wrote: [ -> ]my suspicion right now is that the GUI updates that show the user the progress information may be quite expensive - a simple Python3 test outside of Kodi (without GUI updates) achieves 99MB/s. We do use fairly small 32KB chunks in the Kodi download so increasing the chunk size may help improve bandwidth. Need to investigate.

Yes it's the GUI (dialog) progress updates here - combine a fast internet connection with a small (32KB) chunk size and we're updating this dialog over 70 times a second, which is slowing down the whole process (and computed speed).

We need to rate limit the dialog updates - I'll have something to test in #0125. 

Okay, that explains this weird behavior but the GUI shouldn't have any impact on streams like youtube or amazon, right?
I mean, yes these "downloads" are also done via python but there's no GUI visible that gets updated.
Or am I totally wrong on this one?
(2020-01-25, 11:36)UrushibaraRuka Wrote: [ -> ]Okay, that explains this weird behavior but the GUI shouldn't have any impact on streams like youtube or amazon, right?
I mean, yes these "downloads" are also done via python but there's no GUI visible that gets updated.
Or am I totally wrong on this one?

No you're not wrong, this has nothing to do with YouTube, Amazon etc. Post a Kodi log from a YouTube failure.
(2020-01-25, 14:30)Milhouse Wrote: [ -> ]
(2020-01-25, 11:36)UrushibaraRuka Wrote: [ -> ]Okay, that explains this weird behavior but the GUI shouldn't have any impact on streams like youtube or amazon, right?
I mean, yes these "downloads" are also done via python but there's no GUI visible that gets updated.
Or am I totally wrong on this one?

No you're not wrong, this has nothing to do with YouTube, Amazon etc. Post a Kodi log from a YouTube failure. 

That might be very difficult as there's no real failure that's logged.
When I start a YouTube video it all works well for the first 5-10 minutes, then the video freezes for 15-30 seconds and skips ahead as if the video was running normally and is trying to catch up.
Very rarely it's more than 30 seconds and only then I get thrown back into the menu with the message "One or more Video was unable to play", I get back in, start the same video where it stopped and it works normal.
Same with Amazon, Twitch and so on.
I assume it's sort of a network timeout and it only happens 1-5 times when I just booted kodi, after that everything is normal (apart from the GUI download thing).
Also to note: During these "network timeouts/stutters", the "iperf" showed 0kbps for 1-2 "ticks" every other "tick". (I can try to recreate this if it helps)
It's as if something is taxing the NIC or my machine to the point that it dropps the network for a few seconds.
@Milhouse i confirm that I don't have any problem with Amazon Vod, YouTube, Dplay etc on my ftth 1gbit with my asrock beebox N3150

I confirm that the dialog gui of system update also to me show Max 5500 KB/s
@Milhouse sorry do you have any news on my edid issue ?
(2020-01-25, 21:57)Roby77 Wrote: [ -> ]@Milhouse i confirm that I don't have any problem with Amazon Vod, YouTube, Dplay etc on my ftth 1gbit with my asrock beebox N3150

Thanks. Since you are using the same driver as @UrushibaraRuka it seems to be unlikely that the driver is the problem, but without logs or any other evidence it's hard to say for sure. Typically when one person has a problem with a particular piece of hardware, it's not the software that's the problem...

(2020-01-25, 21:57)Roby77 Wrote: [ -> ]I confirm that the dialog gui of system update also to me show Max 5500 KB/s

Yeah, that will be fixed tonight - appreciate everyone that brought it to my attention, I should have twigged there was something wrong a bit sooner! It is funny how it has gone unnoticed for 7 years - I guess it's only fairly recently that broadband speeds have become fast enough to make it an obvious issue (the GUI-induced throttling seems to limit the maximum download to about 60Mbps so if your speed is lower than that you wouldn't realise there is even an issue). Everyone should see a nice boost once #0125 is installed (you can re-install #0125 to test).

(2020-01-25, 21:58)Roby77 Wrote: [ -> ]@Milhouse sorry do you have any news on my edid issue ?

Not me personally, as I have little knowledge of edid (and even less about hex-editing edids). As before, hopefully someone else knows more about it and is able to chip in... I'll put out some feelers (again) but can't promise if anyone will bite.
Thank you @Milhouse

PS: lol speed went from 5500 Kb to 65.000 Big Grin
(2020-01-25, 23:55)Roby77 Wrote: [ -> ]Thank you @Milhouse

PS: lol speed went from 5500 Kb to 65.000 Big Grin

Smile
New LibreELEC.tv Matrix build #0125: Generic
(Supercedes previous build)

SHA256 Checksum: 12a230ffedac74564f4f505959f0a424dad85f6bb7f5a351e7c1d6cda6e43dd8 (Generic)

text:
# uname -a
Linux NUC 5.4.14 #2 SMP Sat Jan 25 21:03:50 GMT 2020 x86_64 GNU/Linux

# lsb_release
LibreELEC (Milhouse): devel-20200125210304-#0125-gd829752 [Build #0125]

# Kodi version
Kodi (19.0-ALPHA1 Git:8454ab3). Platform: Linux x86 64-bit

Based on tip of LibreELEC.tv master (d829752, changelog) and tip of XBMC master (8454ab3, changelog) with the following modifications: Build Highlights:
  1. Fix performance issue when downloading updates (thanks @UrushibaraRuka)
  2. config/graphic: allow using the new intel gallium iris driver
    Quote:This build includes the new Gallium iris driver for Intel Gen8 (and later) GPUs which should offer improved performance compared to the "classic" i965 driver and will become the default with mesa-20.

    To test it now, enable the new Gallium3D iris driver by adding MESA_LOADER_DRIVER_OVERRIDE=iris to /storage/.config/kodi.conf (create if necessary).

    See Phoronix article for details.

    You should see OpenGL vendor: Intel in Settings > System Information > Video once iris is enabled (if you see OpenGL vendor: Intel Open Source Technology Center then iris is not enabled).
Build Details:
  1. XBMC:
    • DVDVideoCodecDRMPRIME: improve error messages and drain codec on reset (PR:16855, 4 commits, 2 files changed)
    • [music][gui] Ensure disc subtitle is in ListItem.Title for skins to display (PR:17257, 1 commit, 1 file changed)
    • [PVR] EPG Memory optimizations/Performance improvements (PR:17195, 9 commits, 28 files changed)
  2. Additional commits/pull requests/changes not yet merged upstream:
    • Added: [env] PR:4143 (perma): config/graphic: allow using the new intel gallium iris driver
    • Added: [env] PR:4144 (perma): mesa-demos: doesn't install anything, fix that
    • Added: [env] PR:4145 (perma): init: remove file before overwriting
    • Updated: [pkg] PR:147 (perma): Add custom regdom option (service.libreelec.settings)
    • Updated: [pkg] PR:17209 (perma): [depends] Bump samba-gplv3 to version 4.11.4 & build static
    • Updated: [pkg] PR:17222 (perma): [url] skip parsing username:password@ for udp/rtp streams
    • Added: [pkg] PR:148 (perma): improve download performance on fast connections (service.libreelec.settings)
Can confirm that the download for updates went back up to 22000kbps where it was somewhere in the kodi 17 nightlies.

I'll post a log if the streaming issue, where it throws me back into the menu, happens again. But as I said, it's pretty rare.

Maybe there's something wrong with my caching settings in the advanced config that's responsible for the issue? I'll post that one too when I'm back home tomorrow evening.
@Roby77 
Code:
Jan 06 23:33:08 LibreELEC kernel: platform HDMI-A-1: Direct firmware load for edid/edid.bin failed with error -2
Jan 06 23:33:08 LibreELEC kernel: [drm:edid_load] *ERROR* Requesting EDID firmware "edid/edid.bin" failed (err=-2)

getedid fails to put a copy of edid.bin into /storage/.config/firmware/edid for intel/amd.

Try this (untested):
Code:
cd /tmp; rm -fr /tmp/lib; cpio -F /flash/edid.cpio -i
mkdir -p /storage/.config/firmware/edid
cp /tmp/lib/firmware/edid/edid.bin /storage/.config/firmware/edid

@Milhouse getedid fix, untested as well:
Code:
diff --git a/packages/sysutils/busybox/scripts/getedid b/packages/sysutils/busybox/scripts/getedid
index c9afcbc608..ab81e4748e 100755
--- a/packages/sysutils/busybox/scripts/getedid
+++ b/packages/sysutils/busybox/scripts/getedid
@@ -142,6 +142,8 @@ create_edid() {
   # create edid
   mkdir -p /tmp/cpio/lib/firmware/edid
   cat "/sys/class/drm/$card/edid" > /tmp/cpio/lib/firmware/edid/edid.bin
+  mkdir -p /storage/.config/firmware/edid
+  cp /tmp/cpio/lib/firmware/edid/edid.bin /storage/.config/firmware/edid
 
   # create cpio archive
   cd /tmp/cpio/
@mglae fixed! Thank you
Thanks @mglae!