Ubuntu18.04 安装工具包及问题解决记录

Ubuntu18.04 添加一个硬盘分区并挂载
vi /etc/fstab
UUID=c9fc6df3-ad92-4e07-a225-639828535cd8 /home/dennis/sda ext4 defaults 0 2

Ubuntu18.04精简版安装Packages
sudo apt-get install net-tools
sudo apt-get install samba
sudo apt-get install vim
sudo apt-get install git
sudo apt-get install ssh
sudo apt-get install default-jdk
sudo apt install python3-opencv
sudo apt-get install libncurses5-dev libncursesw5-dev

Ubuntu Linux制作Patches + Patch
资料参考：差异文件（diff）和补丁文件（patch）简介

若需要把一方不存在的文件也打包到Patch中，则需要 “-N”
diff -Naur sources-orig/ sources-fixed/ >myfixes.patch

若不需要即只Patch 内容不同的部分，则只需要 “-ur”
diff -ur sources-orig/ sources-fixed/ >myfixes.patch
比如 A501 Driver source files：
diff -ur sources/ a501.sf/ > a501.src.patch

Patch 代码：
进入目录（如A501 sources/)
patch -p1 < ../a501.src.patch

Bug001.Ubuntu 编译Jetson 内核报错：
scripts/kconfig/lxdialog/dialog.h:32:20: fatal error: curses.h: No such file or directory
解决：
sudo apt-get install libncurses5-dev libncursesw5-dev

Bug002.安装CUDA+TensorFlow后: import tensorflow FAIL

dennis@dennis:~$ python3 
Python 3.6.7 (default, Oct 22 2018, 11:32:17) 
[GCC 8.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow 
python3: Relink `/lib/x86_64-linux-gnu/libudev.so.1' with `/lib/x86_64-linux-gnu/librt.so.1' for IFUNC symbol `clock_gettime'

解决：
通过NVIDIA SDKManager 仅安装了CUDA-10.0，但未安装cuDNN，需要从NVIDIA 网站下载对应的cuDNNLibrary并安装即可解决:
cuDNN-Download

dennis@dennis:~$  python3 -c "import tensorflow as tf; print(tf.__version__)"
1.13.1

image.png