Fedora 27环境下安装版本<6的gcc与g++求助

Hey there, I’ve been in your exact situation before—dealing with compiler version conflicts when building TensorFlow GPU can be a real headache. Let’s walk through how to install and switch to an older gcc/g++ version (below 6, we’ll use gcc 5 as an example) on your Fedora 27 system:

Fedora’s repositories still host older compiler versions, so we can grab them directly with dnf:

• Run this command to install gcc 5 and g++ 5:
  sudo dnf install gcc5 gcc-c++5
• Verify the installation worked by checking the versions:
  gcc5 --version
g++5 --version

Instead of overwriting your system’s default gcc 7 (which might break other tools), we’ll use update-alternatives to switch between versions safely:

• Add gcc 5 and gcc 7 as alternative options:
  sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc5 10
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc7 20
• Do the same for g++:
  sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++5 10
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++7 20
• Switch to gcc 5 as the default:
  sudo update-alternatives --config gcc
  You’ll see a list of options—enter the number corresponding to /usr/bin/gcc5 and hit enter.
• Repeat for g++:
  sudo update-alternatives --config g++

Run these commands to make sure your default compiler is now the older version:
gcc --version
g++ --version
You should see output showing gcc 5.x instead of 7.3.1.

• If you need to switch back to gcc 7 later, just re-run the update-alternatives --config commands and select the gcc7 option.
• To make sure TensorFlow uses the correct compiler during build, set these environment variables before starting the compilation process:
  export CC=/usr/bin/gcc5
export CXX=/usr/bin/g++5

内容的提问来源于stack exchange，提问作者DEVENDRA MEENA