Note: The Docker image generated by this repository is being used by DuckDB's CI, and mention in DuckDB PR#16549

DuckDB RISC-V CI

A containerized environment for building and testing DuckDB on RISC-V architecture using QEMU emulation.

Overview

This project provides the infrastructure to:

1. Set up a RISC-V emulation environment using QEMU
2. Build DuckDB for RISC-V architecture
3. Run tests to verify DuckDB functionality on RISC-V

Components

• Docker Container: Debian-based environment with QEMU(Base on DQIB) and RISC-V tools
• QEMU Script: Configures and launches a RISC-V virtual machine
• CI/CD Pipeline: Automates building and publishing the container image

Requirements

• Docker
• GitHub account (for using the pre-built container)

Usage

Using the pre-built Docker image

# Pull the image
docker pull ghcr.io/mocusez/duckdb-riscv-ci/duckdb-riscv-ci:latest

# Run the container on DuckDB repo
docker run -i --rm -v $(pwd):/duckdb --workdir /duckdb ghcr.io/mocusez/duckdb-riscv-ci/duckdb-riscv-ci <<< "apt-get install -y cmake ninja-build libssl-dev g++-riscv64-linux-gnu && GEN=ninja CC='riscv64-linux-gnu-gcc -march=rv64gcv_zicsr_zifencei_zihintpause_zvl256b' CXX='riscv64-linux-gnu-g++ -march=rv64gcv_zicsr_zifencei_zihintpause_zvl256b' DUCKDB_PLATFORM=linux_riscv make && cd / && ./start_qemu.sh && cd /duckdb && make clean && echo 'DOCKER TEST RESULT: SUCCESS' || (echo 'DOCKER TEST RESULT: FAILURE' && make clean)" 2>&1

Building locally

# Clone the repository
git clone https://github.com/mocusez/duckdb-riscv-ci.git
cd duckdb-riscv-ci

# Build the Docker image
docker build -t duckdb-riscv-ci .

# Run the container
docker run -it --rm duckdb-riscv-ci

# Inside container, run the QEMU script
./start_qemu.sh

RISC-V Configuration

The QEMU environment is configured with:

• RV64GCV CPU with Vector extensions (RVV)
• 4GB RAM
• 4 CPU cores
• Network access with port forwarding for SSH

CI/CD Pipeline

GitHub Actions automatically builds and pushes the Docker image on commits to the main branch. The workflow:

1. Checks out the code
2. Logs in to GitHub Container Registry
3. Builds the Docker image
4. Tags and pushes the image to ghcr.io