ci.yml

---
#
# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#

name: CI/CD pipeline for rshim - deb and rpm for amd64 and arm64

on:
  push:
    branches:
      - '**'
    tags:
      - '*'

permissions:
  contents: write

jobs:
  build-and-test:
    runs-on: ubuntu-22.04
    steps:
      - name: Checkout code
        uses: actions/checkout@v3
        with:
          fetch-depth: 0
          fetch-tags: true

      - name: Create apt-get update retry script
        run: |
          cat << 'EOF' > retry_apt_update.sh
          #!/bin/bash

          # Retry apt-get update with exponential backoff and built-in retries
          retry_apt_get_update() {
            local attempt=0
            local max_attempts=5
            local delay=10

            while (( attempt < max_attempts )); do
              echo "Running apt-get update (attempt $((attempt+1))/$max_attempts)..."
              if sudo apt-get update -o Acquire::Retries=3; then
                return 0
              fi
              echo "apt-get update failed. Retrying in ${delay}s..."
              sleep $delay
              delay=$(( delay * 2 ))
              attempt=$(( attempt + 1 ))
            done

            echo "apt-get update failed after $max_attempts attempts."
            return 1
          }
          EOF
          chmod +x retry_apt_update.sh

      - name: Install dependencies (for .deb)
        run: |
          . ./retry_apt_update.sh
          if sudo apt-get update || retry_apt_get_update; then
            systemd_dep="systemd"
            if apt-cache show systemd-dev >/dev/null 2>&1; then
              systemd_dep="systemd-dev"
            fi
            sudo apt-get install -y \
              build-essential debhelper devscripts fakeroot git \
              libpci-dev libusb-1.0-0-dev libfuse-dev pkg-config \
              "$systemd_dep"
          else
            echo "Unable to refresh apt cache"
            exit 1
          fi

      - name: Prepare release body
        id: changelog
        run: |
          body=$(./scripts/get-changelog debian/changelog)
          echo "body<<EOF" >> $GITHUB_OUTPUT
          echo "${body}" >> $GITHUB_OUTPUT
          echo "EOF" >> $GITHUB_OUTPUT

      - name: Create source tarball
        run: |
          [ -d dist ] || mkdir -p dist
          version="$(./scripts/get-ver)"
          tarball="rshim-src-${version}.tar.gz"
          git archive --format=tar.gz -o "$tarball" HEAD
          mv -v "$tarball" dist/
          ls -lh dist

      - name: Build & Test x86_64 .deb package
        run: |
          version="$(./scripts/get-ver)"
          sed -i "1s/(\(.*\))/($version)/" debian/changelog
          DEB_BUILD_OPTIONS="nocheck nostrip nolto" \
            CFLAGS="-fno-lto" LDFLAGS="-fno-lto" \
            dpkg-buildpackage -us -uc -a amd64
          mkdir -p dist
          mv -v ../*.deb dist/
          ls -lh dist
          sudo dpkg -i dist/*.deb || sudo apt-get -f install -y
          echo "Running rshim --version:"
          rshim --version

      - name: Capture UID/GID for Docker
        id: uidgid
        run: |
          echo "HOST_UID=$(id -u)" >> $GITHUB_ENV
          echo "HOST_GID=$(id -g)" >> $GITHUB_ENV

      - name: Build & Test ARM64 .deb package in Docker
        run: |
          set -euo pipefail

          docker run --rm --privileged multiarch/qemu-user-static \
            --reset -p yes

          # Retry function for Docker build
          retry_count=0
          max_retries=3
          delay=20

          while true; do
            echo "Docker build attempt $((retry_count + 1))/$max_retries"

            if docker run --rm --platform linux/arm64 \
              -e HOST_UID="${{ env.HOST_UID }}" \
              -e HOST_GID="${{ env.HOST_GID }}" \
              -v "${{ github.workspace }}:/workspace" \
              -w /workspace \
              arm64v8/ubuntu:22.04 \
              /bin/bash -c "
                set -euo pipefail
                . /workspace/retry_apt_update.sh
                apt-get update || retry_apt_get_update
                systemd_dep="systemd"
                if apt-cache show systemd-dev >/dev/null 2>&1; then
                  systemd_dep="systemd-dev"
                fi
                apt-get install -y \
                  autoconf automake libtool build-essential debhelper \
                  devscripts fakeroot gcc git libpci-dev libusb-1.0-0-dev \
                  libfuse-dev libsystemd-dev libpci3 libusb-1.0-0 fuse \
                  pkg-config "\$systemd_dep"
                git config --global --add safe.directory /workspace && \
                version=\$(/workspace/scripts/get-ver)
                  sed -i \"1s/(\(.*\))/(\$version)/\" /workspace/debian/changelog
                mkdir -p /workspace/build
                DEB_BUILD_OPTIONS='nocheck nostrip nolto' \
                  CFLAGS="-fno-lto" LDFLAGS="-fno-lto" \
                  dpkg-buildpackage -us -uc -a arm64 || {
                    echo '--- pkg-config failure diagnostics ---'
                    pkg-config --modversion libpci || true
                    pkg-config --modversion libusb-1.0 || true
                    pkg-config --modversion fuse || true
                    pkg-config --variable=systemdsystemunitdir systemd || true
                    echo '--- config.log tails ---'
                    find . -name config.log -exec sh -c \
                      'echo ==== {}; tail -n 200 {}' \;
                    exit 1
                  }
                mv -v ../*.deb /workspace/build/ && \
                dpkg -i /workspace/build/*.deb || apt-get -f install -y
                echo 'Running rshim --version:'
                rshim --version
                chown -R \$HOST_UID:\$HOST_GID /workspace
              "; then
                echo "Docker build succeeded on attempt $((retry_count + 1))"
              break
            else
              retry_count=$((retry_count + 1))
              if [ $retry_count -lt $max_retries ]; then
                echo "Docker build failed, waiting ${delay}s before retry..."
                sleep $delay
              else
                echo "Docker build failed after $max_retries attempts"
                exit 1
              fi
            fi
          done

          mkdir -p dist
          mv -v build/*.deb dist/
          ls -lh dist

      - name: Build & Test x86_64 RPM using Rocky Linux
        run: |
          docker run --rm \
            -e HOST_UID="${{ env.HOST_UID }}" \
            -e HOST_GID="${{ env.HOST_GID }}" \
            -v "${{ github.workspace }}:/workspace" \
            -w /workspace \
            quay.io/rockylinux/rockylinux:8 \
            /bin/bash -c "
            set -euo pipefail
              yum -y install gcc make rpm-build git autoconf automake \
                libtool pkgconfig pciutils-devel libusb1-devel fuse-devel && \
              git config --global --add safe.directory /workspace && \
              ARCH=x86_64 CFLAGS="-fno-lto" LDFLAGS="-fno-lto" \
                ./scripts/build-rpm && \
              yum localinstall -y /workspace/rpmbuild/RPMS/x86_64/*.rpm && \
              echo 'Running rshim --version:' && \
              rshim --version && \
              chown -R \$HOST_UID:\$HOST_GID /workspace
            "
          mkdir -p dist
          mv -v $(find rpmbuild/RPMS -name '*.rpm') dist/
          ls -lh dist

      - name: Build & Test arm64 RPM using Rocky Linux
        run: |
          set -euo pipefail

          # Prepare QEMU for ARM64 emulation
          docker run --rm --privileged multiarch/qemu-user-static \
            --reset -p yes

          # Retry function for Docker build
          retry_count=0
          max_retries=3
          delay=20

          while true; do
            echo "Docker build attempt $((retry_count + 1))/$max_retries"

            if docker run --rm --platform linux/arm64 \
              -e HOST_UID="${{ env.HOST_UID }}" \
              -e HOST_GID="${{ env.HOST_GID }}" \
              -v "${{ github.workspace }}:/workspace" \
              -w /workspace \
              quay.io/rockylinux/rockylinux:8 \
              /bin/bash -c "
                set -euo pipefail
                yum -y install gcc make rpm-build git autoconf automake \
                  libtool pkgconfig pciutils-devel libusb1-devel fuse-devel && \
                git config --global --add safe.directory /workspace && \
                ARCH=aarch64 CFLAGS=\"-fno-lto\" LDFLAGS=\"-fno-lto\" \
                  ./scripts/build-rpm && \
                yum localinstall -y /workspace/rpmbuild/RPMS/aarch64/*.rpm && \
                echo 'Running rshim --version:' && \
                rshim --version && \
                chown -R \$HOST_UID:\$HOST_GID /workspace
              "; then
              echo "Docker build succeeded on attempt $((retry_count + 1))"
              break
            else
              retry_count=$((retry_count + 1))
              if [ $retry_count -lt $max_retries ]; then
                echo "Docker build failed, waiting ${delay}s before retry..."
                sleep $delay
              else
                echo "Docker build failed after $max_retries attempts"
                exit 1
              fi
            fi
          done

          mkdir -p dist
          mv -v $(find rpmbuild/RPMS -name '*.rpm') dist/
          ls -lh dist

      - name: Upload artifacts
        uses: actions/upload-artifact@v4
        with:
          name: rshim-packages
          path: dist/*

      - name: Create GitHub Release
        if: startsWith(github.ref_name, 'rshim-')
        uses: softprops/action-gh-release@v1
        with:
          token: ${{ secrets.GITHUB_TOKEN }}
          tag_name: ${{ github.ref_name }}
          name: ${{ github.ref_name }}
          body: ${{ steps.changelog.outputs.body }}
          draft: false
          prerelease: false
          files: dist/*