Posts by vpeter
-
-
It would be easier to include required libs in image (need to build your own).
Use this post_makeinstall_target() function in packages/lang/Python2/package.mk
Code
Display Morepost_makeinstall_target() { rm -fr $PKG_BUILD/.$TARGET_NAME/build/temp.* for dir in bsddb idlelib lib-tk lib2to3 msilib pydoc_data test unittest; do # added if [ "$dir" = "lib2to3" ]; then rm -rf $INSTALL/usr/lib/python*/$dir/tests continue fi rm -rf $INSTALL/usr/lib/python*/$dir done rm -rf $INSTALL/usr/lib/python*/config # removed # rm -rf $INSTALL/usr/bin/2to3 rm -rf $INSTALL/usr/bin/idle rm -rf $INSTALL/usr/bin/pydoc rm -rf $INSTALL/usr/bin/smtpd.py rm -rf $INSTALL/usr/bin/python*-config cd $INSTALL/usr/lib/$PKG_PYTHON_VERSION $TOOLCHAIN/bin/python -Wi -t -B $PKG_BUILD/Lib/compileall.py -d /usr/lib/$PKG_PYTHON_VERSION -f . find $INSTALL/usr/lib/$PKG_PYTHON_VERSION -name "*.py" -exec rm -f {} \; &>/dev/null # strip chmod u+w $INSTALL/usr/lib/libpython*.so.* debug_strip $INSTALL/usr }Clean Python2 package and make image
install addon which should work (for me it ask for credentials).
-
-
My fix is working because LD_LIBRARY_PATH is set with LE environment libreelec-8.0/config/path · GitHub
Because host doesn't set it (it is empty) you will get the path with colon at the end. And this means "use current folder". I assume this change started to be problematic sometime with Ubuntu updates (didn't investigate where, why/...)
LE 9 unsets this path that's why there is no this issue and will not pop up any time libreelec-9.0/config/path - GitHub
-
-
I am surprised this was not caught earlier.
I assume some change come with host system updates. Because 8.2.5 was build without a problem when it was current.
Interesting read: How Torch broke ls and made me vulnerable (or the hidden dangers of LD_LIBRARY_PATH)
-
-
And the winner is
Codemakeinstall_target() { make INSTALL_PREFIX=$INSTALL install_sw LD_LIBRARY_PATH= make INSTALL_PREFIX=$SYSROOT_PREFIX install_sw chmod 755 $INSTALL/usr/lib/*.so* chmod 755 $INSTALL/usr/lib/engines/*.so }After closer examination I noticed that LD_LIBRARY_PATH is set to
and (seems) it means current folder is also set because nothing exists after ":" separator. And make and every other command fails in the folder where libraries are stored
Code/home/le/LibreELEC.tv.8.2.5/build.LibreELEC-Generic.x86_64-8.2.5/toolchain/x86_64-libreelec-linux-gnu/sysroot/usr/libCodecd /home/le/LibreELEC.tv.8.2.5/build.LibreELEC-Generic.x86_64-8.2.5 le@here:/home/le/LibreELEC.tv.8.2.5/build.LibreELEC-Generic.x86_64-8.2.5/toolchain/x86_64-libreelec-linux-gnu/sysroot/usr/lib$ LD_LIBRARY_PATH="/dummy" which make /usr/bin/make le@here:/home/le/LibreELEC.tv.8.2.5/build.LibreELEC-Generic.x86_64-8.2.5/toolchain/x86_64-libreelec-linux-gnu/sysroot/usr/lib$ LD_LIBRARY_PATH="/dummy:" which make Segmentation fault (core dumped) -
-
Marked as resolved

Happy viewing.
-
Try to use this makeinstall_target() function in packages/security/openssl/package.mk
Codemakeinstall_target() { make INSTALL_PREFIX=$INSTALL install_sw #make INSTALL_PREFIX=$SYSROOT_PREFIX install_sw chmod 755 $INSTALL/usr/lib/*.so* chmod 755 $INSTALL/usr/lib/engines/*.so cp -a $INSTALL/* $SYSROOT_PREFIX }Didn't check why make crashes when installing to SYSROOT.
This also works
Code
Display Moremakeinstall_target() { make INSTALL_PREFIX=$INSTALL install_sw find . -type f -name "Makefile*" -exec sed -i "s|chmod 555|chmod 755|g" {} \; find . -type f -name "Makefile*" -exec sed -i "s|\$(MAKE) -f \$\$here|env -i \$(MAKE) -f \$\$here|g" {} \; touch Makefile Makefile.org Configure config make INSTALL_PREFIX=$SYSROOT_PREFIX install_sw chmod 755 $INSTALL/usr/lib/*.so* chmod 755 $INSTALL/usr/lib/engines/*.so } -
I don't see crash with this version

-
Do I have to worry about the sha256 if it is using the commit number?Just use empty
-
-
Just open file packages/mediacenter/kodi-binary-addons/pvr.nextpvr/package.mk in LibreELEC folder and change three lines I mentioned above.
Version d6bf9a3b4bf22702f907d00adb8618af161e1f23 is from git and no tag is needed.
And Welcome here Martin

-
-
Let me rebuild it and see what the issue is. Last time I build it successfully.
-
Tell Martin that delay with sleep() was just to show that thread was not created yet when it was destroyed. Removing false argument will fix the issue.
When there is no server there is almost no delay.