Kodi Community Forum

Full Version: LibreELEC Testbuilds for RaspberryPi (Kodi 17.0)
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
New LibreELEC.tv Krypton build #1116: RPi / RPi2
(Supercedes previous build)

# uname -a
Linux rpi512 4.8.8 #1 Wed Nov 16 21:45:51 GMT 2016 armv6l GNU/Linux

# vcgencmd version
Nov 14 2016 16:19:30
Copyright (c) 2012 Broadcom
version 4a839311e7b5fc3422a8e6a8623b30b52891e0ea (clean) (release)

# lsb_release
LibreELEC (Milhouse) - Version: devel-20161116214421-#1116-g5d65364 [Build #1116]

# Kodi version
(17.0-BETA6 Git:db545c6). Platform: Linux ARM 32-bit

Based on tip of LibreELEC.tv master (5d653647, changelog) and tip of XBMC master (db545c69, changelog) with the following modifications: Build Highlights:
  1. Bump to gcc-6.2.0, update libressl - this will break some addons (install latest tvheadend42 from addons etc.)
  2. new tzdata and tzcode packages
Build Details:
  1. LibreELEC.tv:
    • gcc: Bump to gcc-6.2.0 (PR:650, 20 commits, 20 files changed)
    • Disable autoreconf when not needed (PR:735, 163 commits, 163 files changed)
    • RTL8812au update: update package to the stable version branch 4.3.20 (PR:923, 2 commits, 9 files changed)
    • LibreSSL: update to 2.4.4 (PR:925, 1 commit, 1 file changed)
    • Package updates: bump various packages after gcc-6 (PR:935, 13 commits, 15 files changed)
      • taglib: update to taglib-1.11
      • xkeyboard-config: cleanup
      • rtmpdump: build statically only
      • openssh: update to openssh-7.3p1
      • binutils: remove old patch
      • binutils: update to binutils-2.27
      • boost: update to boost-1_61_0
      • make: update to make-4.2.1
      • kmod: update to kmod-23
      • gdb: update to gdb-7.11.1
      • libgcrypt: update to libgcrypt-1.7.3
      • util-linux: update to util-linux-2.28.1
      • libamcodec: update to libamcodec-5e23a81
  2. XBMC:
    • Set notification rumble to weak motor for 300ms (PR:10929, 1 commit, 1 file changed)
    • [cmake] Unify usage of CMAKE_CROSSCOMPILE throughout the project (PR:10863, 5 commits, 6 files changed)
  3. pvr.vbox:
    • Bump version to 3.6.0 (#158) (48f43152)
    • Fixed bug found by Coverity (using uninitialized variable) (#159) (d37f8ffb)
    • Fixed Coverity issue (case using uninitialized var) (#161) (f2e5cfa7)
  4. newclock5:
    • New commits in this build:
      • temp: Add spammy logging to track down seek delays (79b89e4e)
      • Revert "temp: Add spammy logging to track down seek delays" (2ac8d1ba)
  5. kernel 4.8.y:
    • New commits in this build:
      • BCM270x_DT: Drop using ARCH_BCM2708 and ARCH_BCM2709 (b36810b1)
      • Drop ARCH_BCM2708 and ARCH_BCM2709 (c26b00cf)
      • mkknlimg: Retain downstream-kernel detection (657f3edc)
      • Merge pull request #1717 from notro/bcm270x-rip-1 (9fce3365)
      • fixup: fb: Use correct bus address for dest of dma memcpy (8e7ca2c5)
  6. Additional commits/pull requests/changes not yet merged upstream:
    • Added: [env] 3a6cf520: RBP: Drop CONFIG_ARCH_BCM270x after rpi/linux/PR1717
    • Added: [env] PR:939: autoreconf fixes
    • Added: [env] PR:940: timezone-data: convert to new tzdata and tzcode packages
(2016-11-16, 02:18)bill_orange Wrote: [ -> ]I notice that on build #1114 I occasionally get the little thermometer symbol in the top right. Scanning the library is a circumstance that causes it to appear. On build #1114 has overclocking been increased or has the threshold that causes the thermometer to appear changed?

Can someone scan their library and reproduce this?

During music library scan temp is 172 F. Should I do anything about it? This started with build #1114.
(2016-11-17, 04:39)bill_orange Wrote: [ -> ]During music library scan temp is 172 F. Should I do anything about it?

172F is about 78C, which is within limits but may be causing the ARM to throttle (reduce frequency). Not ideal, and a small heatsink would help (or a FLIRC case).

(2016-11-17, 04:39)bill_orange Wrote: [ -> ]This started with build #1114.

Did anything else change, such as your skin, perhaps?

When you are running the scan, is there any text that is constantly scrolling (as this can send the GPU temperature through the roof)?

In #1114 there is a firmware and kernel update, maybe this causes the GPU to run a little hotter than usual.
(2016-11-16, 20:20)MikeKL Wrote: [ -> ]
(2016-11-16, 09:10)gendo Wrote: [ -> ]the change in filmon addon is trivial so it should not result in a kodi crash.. so i suspect problem is somewhere else..
Fair enough, but easy to provoke crash with pvr.filmon enabled, here is another debug crashlog this evening after enabling addon, waiting for channels to be loaded before selecting one playing it and the switching between it and another live channel.

Crash occuring at moment of switch https://gist.github.com/Mike-KL/97bafd19...aaa6b810f4
Debug includes verbose libcurl logging enabled.

---edit---With tvheadened enabled and switching between channels not able to provoke a crash
(tvheadened backend/frontend both running on same rpi2)

it seems related to rtmp addon.. tvheadend does not use rtmp so its fine..
I just cannot figure out which change broke it
Well, problems streaming Bloomberg have emerged again.

Bloomberg streams fine on build #1007. Attempting to play Bloomberg on build #1108 and above results in a LibreELEC restart.

Here is a log. Since a LibreELEC restart occurs, I am not sure if the log will contain what you need. I turned FFMPEG logging on.


To make this even odder. I have to media centers. Bloomberg streams fine on the recent builds including the latest. They do you very different SDs but I would hardly expect hardware dependence.


I retrieved the crash log out of the PI:


@bill_orange are you sure it was LE that restarted and not just Kodi that crashed (big difference!)

Are you able to reproduce with the latest #1116 build? There have been a number of changes since #1108 that may have already fixed the issue, mean that a more up to date crash log is required.

There's a debug-enabled version of build #1116 available: RPi2 - if you use this to produce a crash-log it will have more detailed information.
I misspoke (or miswrote in this case). I mean it crashed KODI. Yes, the problem also occurs in the latest version. I loaded the debug version of build #1116. Here is the crash log.

@bill_orange: That doesn't look like the crash log from the debug enabled #1116y - it looks like the crashlog from the latest non-debug enabled #1116.

Can you double check that you have installed #1116y and that the upgrade succeeded - it might fail if your FAT partition is not 512MB or larger.
@Milhouse and @bill_orange i believe this is the same as crash on stopping pvr.filmon streams.. filmon and bloomberg use rtmp
(2016-11-17, 06:13)bill_orange Wrote: [ -> ]I retrieved the crash log out of the PI:


Looks like you've got ADSP enabled. Don't enable it - it crashes:
20:04:26 T:1939780512   DEBUG: ActiveAE DSP - Create - Audio DSP processing id 0 created:

@Milhouse - we really need to make sure this is not possible to enable in final Krypton release.
Yes! I had ADSP on. Since this PI is hooked to the Home Theater (what I used to call a stereo), I turned on everything that looked helpful rather I needed it or not.

Bloomberg streams in fine now. This also explains why Bloomberg was streaming on the other PI. It is hooked to simple 2 channel computer speakers and all fancier audio processing is turned off.

Do you still need me to create a more helpful boot log?

Got a crashlog with a debug enabled build

(2016-11-17, 19:53)ironic_monkey Wrote: [ -> ]please try https://github.com/notspiff/inputstream....b22ecb1077

Thanks, will include this in tonight's build.
(2016-11-17, 13:10)popcornmix Wrote: [ -> ]@Milhouse - we really need to make sure this is not possible to enable in final Krypton release.

Agreed. The ADSP addons will be dropped from the LibreELEC repository with the next 7.90.009 version bump.

@everyone: For the time being, ADSP should be disabled to improve stability.