cppglue

customizable clang based tool to extract c++ code for input to language binding generators

Build docker image

docker build -t llvm-runtime .

Run docker image

Key components for secure container execution:

User Permissions and Storage

• --user $(id -u):$(id -g) - Preserves host system permissions by running as current user
• -v /tmp/llvm1:/app - Maps local storage to container for persistent data

Security Measures

• --security-opt=no-new-privileges - Prevents privilege escalation within container
• --cap-drop=ALL - Removes all Linux kernel capabilities, minimizing attack surface

Interactive Access

• -it - Enables interactive terminal access (combines stdin and TTY allocation)

mkdir -p /tmp/llvm1
docker run -it --user $(id -u):$(id -g) -v /tmp/llvm1:/app --security-opt=no-new-privileges --cap-drop=ALL llvm-runtime --install /app
export PATH=$PATH:/tmp/llvm1/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/tmp/llvm1/lib

# You should now be be able to run clang and llvm tools

Python Bindings

# Generate python bindings
# enter llvm-runtime container
# cmake .. -G Ninja -DCMAKE_INSTALL_PREFIX=/app
# ninja install

Testing Python Package Installation

After building the Python bindings, you can test the package installation:

1. Create a virtual environment:

python -m venv venv
source venv/bin/activate

2. Install the package in development mode:

pip install -e .

3. Test importing the module:

import cppglue
# Your module should now be available

4. To build a wheel for distribution:

python -m build

The wheel file will be created in the dist/ directory.