docker-compose.dev.yml

# Development Docker Compose — database + optional containerized dev server.
# Used by bin/setup to start PostgreSQL for native Rails development.
#
# Usage (via setup script — recommended):
#   bin/setup
#
# Usage (database only — manual):
#   docker compose -f docker-compose.dev.yml up db -d
#   bin/rails db:prepare
#   foreman start -f Procfile.dev
#
# Usage (full containerized dev):
#   docker compose -f docker-compose.dev.yml up -d
#
# For deployment: use docker-compose.yml (the default).
name: vulcan

services:
  db:
    image: postgres:18-alpine
    restart: unless-stopped
    ports:
      - "${POSTGRES_PORT:-5435}:5432"
    environment:
      POSTGRES_USER: ${POSTGRES_USER:-postgres}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-postgres}
      POSTGRES_DB: ${POSTGRES_DB:-vulcan_development}
      # Trust auth for local dev. Corporate VPNs (GlobalProtect, etc.) interfere
      # with scram-sha-256 challenge-response handshakes through container runtime
      # port forwarding proxies (OrbStack, Docker Desktop). This is a known issue
      # with no upstream fix: https://github.com/orbstack/orbstack/issues/2267
      # Mastodon and GitLab CI use the same approach for dev/CI environments.
      # Production uses docker-compose.yml which does NOT set this.
      POSTGRES_HOST_AUTH_METHOD: trust
    volumes:
      - vulcan_dev_dbdata:/var/lib/postgresql
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-postgres}"]
      interval: 10s
      timeout: 5s
      retries: 5

  web:
    build:
      context: .
      dockerfile: Dockerfile
      target: development
    environment:
      DATABASE_HOST: db
      DATABASE_PORT: "5432"
      POSTGRES_USER: ${POSTGRES_USER:-postgres}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-postgres}
      POSTGRES_DB: ${POSTGRES_DB:-vulcan_development}
      RAILS_ENV: development
      VULCAN_FIRST_USER_ADMIN: "true"
      VULCAN_ENABLE_LOCAL_LOGIN: "true"
      VULCAN_ENABLE_USER_REGISTRATION: "true"
    ports:
      - "${VULCAN_PORT:-3000}:3000"
    volumes:
      - .:/rails
      - /rails/node_modules
    depends_on:
      db:
        condition: service_healthy
    # Setup steps run sequentially in the foreground; `bash -e` aborts the
    # container if any of them fail, so a failed db:prepare surfaces as a dead
    # container instead of a booted server that raises PendingMigrationError on
    # every request. Only the asset watcher is backgrounded — `;` separators
    # scope the trailing `&` to `yarn build:watch` alone. `exec` hands PID 1 to
    # Puma so docker stop/Ctrl-C signals reach it directly.
    command: >
      bash -ec "rm -f tmp/pids/server.pid;
      bin/rails db:prepare;
      yarn install --frozen-lockfile;
      yarn build;
      yarn build:watch &
      exec bundle exec rails server -b 0.0.0.0 -p 3000"

volumes:
  vulcan_dev_dbdata: