setup-build-server.sh

#!/bin/bash

# Setup build server for DDA development
# Installs dependencies and configures the build environment

set -e

VERBOSE="${VERBOSE:-0}"
LOG_FILE="/tmp/setup-build-$(date +%s).log"
ERRORS=()
WARNINGS=()

# Helper function to run commands with logging
run_cmd() {
    local cmd="$@"
    if [ "$VERBOSE" = "1" ]; then
        echo "[RUN] $cmd"
        eval "$cmd" | tee -a "$LOG_FILE"
    else
        echo "[RUN] $cmd"
        if ! eval "$cmd" >> "$LOG_FILE" 2>&1; then
            ERRORS+=("Failed: $cmd")
            return 1
        fi
    fi
}

# Helper to add errors
add_error() {
    ERRORS+=("$1")
    echo "❌ $1" | tee -a "$LOG_FILE"
}

# Helper to add warnings
add_warning() {
    WARNINGS+=("$1")
    echo "⚠️  $1" | tee -a "$LOG_FILE"
}

# Trap errors and show summary
trap 'show_summary' EXIT

show_summary() {
    echo ""
    echo "=========================================="
    if [ ${#ERRORS[@]} -eq 0 ] && [ ${#WARNINGS[@]} -eq 0 ]; then
        echo "✅ Build server setup complete!"
    elif [ ${#ERRORS[@]} -eq 0 ]; then
        echo "✅ Build server setup complete (with warnings)"
    else
        echo "❌ Setup completed with ERRORS"
    fi
    echo "=========================================="
    
    if [ ${#ERRORS[@]} -gt 0 ]; then
        echo ""
        echo "ERRORS ENCOUNTERED:"
        printf '%s\n' "${ERRORS[@]}"
    fi
    
    if [ ${#WARNINGS[@]} -gt 0 ]; then
        echo ""
        echo "WARNINGS:"
        printf '%s\n' "${WARNINGS[@]}"
    fi
    
    echo ""
    echo "📋 Full log: $LOG_FILE"
    if [ "$VERBOSE" != "1" ] && [ ${#ERRORS[@]} -gt 0 ]; then
        echo "Run with VERBOSE=1 to see detailed output:"
        echo "  VERBOSE=1 $0"
    fi
    
    if [ ${#ERRORS[@]} -eq 0 ] && [ ${#WARNINGS[@]} -eq 0 ]; then
        echo ""
        echo "Next steps:"
        echo "  1. Verify Docker access: docker ps"
        echo "  2. Build component: ./gdk-component-build-and-publish.sh"
    fi
    
    echo ""
    
    if [ ${#ERRORS[@]} -gt 0 ]; then
        return 1
    fi
}

echo "Setting up build server..."
echo "Log file: $LOG_FILE"
echo ""

echo "▶ Detecting host Ubuntu release..."
# Detect the host release once: /etc/os-release first, with the
# lsb_release fallback the Python 3.11 block below already uses.
OS_RELEASE=$(. /etc/os-release 2>/dev/null && echo "$VERSION_ID") || OS_RELEASE=""
if [ -z "$OS_RELEASE" ] && command -v lsb_release >/dev/null 2>&1; then
    OS_RELEASE=$(lsb_release -rs 2>/dev/null) || OS_RELEASE=""
fi
echo "✓ Detected Ubuntu release: ${OS_RELEASE:-unknown}"

# Ubuntu 24.04 (noble) marks the system Python as externally managed
# (PEP 668), so pip installs need --break-system-packages there. Jammy's
# pip rejects that option as unknown, so the flag is release-keyed: it
# expands to nothing on 22.04 and earlier.
PIP_BREAK_FLAG=''
if [ "$OS_RELEASE" = "24.04" ]; then
    PIP_BREAK_FLAG='--break-system-packages'
fi

echo "▶ Hardening apt against IPv6-only regional mirrors..."
# Fresh EC2 instances get sources pointing at the regional mirror
# (e.g. us-east-1.ec2.ports.ubuntu.com / us-east-1.ec2.archive.ubuntu.com).
# Those hostnames can resolve to IPv6 addresses only; on a subnet without
# IPv6 egress apt-get update then hangs indefinitely and the whole
# bootstrap stalls before any tool is installed (observed on arm64/JP6
# build servers). Repoint to the canonical hosts and force IPv4 so apt
# can never block on an unreachable AAAA record.
echo 'Acquire::ForceIPv4 "true";' | sudo tee /etc/apt/apt.conf.d/99dda-force-ipv4 >/dev/null
for f in /etc/apt/sources.list /etc/apt/sources.list.d/*.list /etc/apt/sources.list.d/*.sources; do
    [ -f "$f" ] || continue
    sudo sed -i -E 's@[a-z0-9-]+\.ec2\.(ports|archive)\.ubuntu\.com@\1.ubuntu.com@g' "$f"
done
echo "✓ apt sources repointed to canonical mirrors, IPv4 forced"

echo "▶ Updating package manager..."
# Every apt-get carries DPkg::Lock::Timeout=600: on a fresh instance an
# SSM patch run can hold the dpkg lock, and install/remove then wait up
# to 600 s for it instead of failing at once. (apt does not apply the
# option to the lists lock `apt-get update` takes; that failure stays
# tolerated, as before.)
run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 update" || true

echo "▶ Installing system dependencies..."
run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 install -y python3 python3-pip nodejs npm git curl snapd zip" || add_warning "Some system packages failed to install"

echo "▶ Removing old Docker packages from apt..."
run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 remove -y docker.io docker-compose" || true

echo "▶ Installing Docker via snap..."
if ! command -v docker >/dev/null 2>&1; then
    run_cmd "sudo snap install docker" || add_error "Failed to install Docker via snap"
    echo "✓ Docker installed via snap"
else
    echo "✓ Docker already installed"
fi

echo "▶ Installing docker-compose..."
if ! command -v docker-compose >/dev/null 2>&1; then
    if [ "$OS_RELEASE" = "24.04" ]; then
        # Noble ships only the `docker compose` plugin; write a
        # docker-compose shim delegating to it instead of the snap.
        if run_cmd "printf '#!/bin/sh\nexec docker compose \"\$@\"\n' | sudo tee /usr/local/bin/docker-compose > /dev/null" \
            && run_cmd "sudo chmod +x /usr/local/bin/docker-compose"; then
            echo "✓ docker-compose shim installed (delegates to the docker compose plugin)"
        else
            add_error "Failed to install docker-compose shim"
        fi
    else
        run_cmd "sudo snap install docker-compose" || add_error "Failed to install docker-compose via snap"
        echo "✓ docker-compose installed via snap"
    fi
else
    echo "✓ docker-compose already installed"
fi

echo "▶ Starting Docker daemon..."
if run_cmd "sudo systemctl start snap.docker.dockerd.service"; then
    echo "✓ Docker daemon started"
else
    add_error "Failed to start Docker daemon"
fi

echo "▶ Configuring Docker permissions..."
# Create docker group if it doesn't exist
if ! getent group docker > /dev/null; then
    if run_cmd "sudo groupadd docker"; then
        echo "✓ Created docker group"
    else
        add_error "Failed to create docker group"
    fi
else
    echo "✓ Docker group already exists"
fi

# Add user to docker group
if run_cmd "sudo usermod -aG docker $USER"; then
    echo "✓ Added $USER to docker group"
else
    add_error "Failed to add user to docker group"
fi

# Fix Docker socket permissions
if run_cmd "sudo chmod 666 /var/run/docker.sock"; then
    echo "✓ Fixed Docker socket permissions"
else
    add_warning "Could not fix Docker socket permissions"
fi

# Reload systemd daemon
if run_cmd "sudo systemctl daemon-reload"; then
    echo "✓ Reloaded systemd daemon"
else
    add_warning "Could not reload systemd daemon"
fi

echo "▶ Verifying Docker installation..."
if sudo docker ps > /dev/null 2>&1; then
    echo "✓ Docker is accessible"
else
    add_warning "Docker may not be fully accessible (may need to restart or log in again)"
fi

echo "▶ Installing Python 3.11..."
if ! command -v python3.11 >/dev/null 2>&1; then
    UBUNTU_VERSION=$(lsb_release -rs)
    if [ "$UBUNTU_VERSION" = "18.04" ]; then
        echo "  Building Python 3.11 from source (Ubuntu 18.04)..."
        run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 install -y build-essential zlib1g-dev libncurses5-dev libgdbm-dev libnss3-dev libssl-dev libreadline-dev libffi-dev wget" || add_warning "Failed to install build dependencies"
        
        if [ ! -d /tmp/Python-3.11.9 ]; then
            run_cmd "cd /tmp && wget --tries=5 --waitretry=30 --retry-on-http-error=429,500,502,503 https://www.python.org/ftp/python/3.11.9/Python-3.11.9.tgz" || add_error "Failed to download Python 3.11"
            run_cmd "cd /tmp && tar -xf Python-3.11.9.tgz" || add_error "Failed to extract Python 3.11"
        fi
        
        run_cmd "cd /tmp/Python-3.11.9 && ./configure --enable-optimizations" || add_error "Failed to configure Python 3.11"
        run_cmd "cd /tmp/Python-3.11.9 && make -j 8" || add_error "Failed to build Python 3.11"
        run_cmd "cd /tmp/Python-3.11.9 && sudo make altinstall" || add_error "Failed to install Python 3.11"
    else
        echo "  Installing Python 3.11 from PPA..."
        run_cmd "sudo add-apt-repository ppa:deadsnakes/ppa -y" || add_warning "Failed to add deadsnakes PPA"
        run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 update" || add_warning "Failed to update package manager"
        run_cmd "sudo apt-get -o DPkg::Lock::Timeout=600 install -y python3.11 python3.11-venv python3.11-dev" || add_error "Failed to install Python 3.11"
    fi
    
    # Set as default
    run_cmd "sudo update-alternatives --install /usr/local/bin/python3 python3 /usr/local/bin/python3.11 1" || add_warning "Failed to set Python 3.11 as default"
    echo "✓ Python 3.11 installed"
else
    echo "✓ Python 3.11 already installed"
fi

echo "▶ Configuring PATH..."
if ! grep -q 'export PATH="$HOME/.local/bin:$PATH"' ~/.bashrc; then
    echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
    echo "✓ Added ~/.local/bin to PATH"
else
    echo "✓ PATH already configured"
fi
export PATH="$HOME/.local/bin:$PATH"

echo "▶ Installing AWS CLI..."
if ! command -v aws >/dev/null 2>&1; then
    run_cmd "sudo pip3 install $PIP_BREAK_FLAG awscli" || add_warning "Failed to install AWS CLI"
else
    echo "✓ AWS CLI already installed"
fi

echo "▶ Installing botocore[crt] (required by GDK publish for credential resolution)..."
run_cmd "sudo pip3 install $PIP_BREAK_FLAG --no-compile 'botocore[crt]'" || add_warning "Failed to install botocore[crt] - 'gdk component publish' may fail to resolve credentials"

echo "▶ Installing AWS CDK (optional for build server)..."
if ! command -v cdk >/dev/null 2>&1; then
    run_cmd "sudo npm install -g aws-cdk" || add_warning "Failed to install AWS CDK (optional - not required for component builds)"
else
    echo "✓ AWS CDK already installed"
fi

echo "▶ Installing AWS Greengrass Development Kit (GDK)..."
if ! command -v gdk >/dev/null 2>&1; then
    run_cmd "pip3 install --user $PIP_BREAK_FLAG git+https://github.com/aws-greengrass/aws-greengrass-gdk-cli.git@v1.6.2" || add_warning "Failed to install GDK CLI from GitHub"
else
    echo "✓ GDK CLI already installed"
fi

echo ""
echo "✓ Build server setup complete!"