[WIP] Multi-stage build for Dockerfile.dependencies_humble

scripts/RADI/Dockerfile.dependencies_humble

@@ -1,7 +1,7 @@
-FROM nvidia/opengl:1.2-glvnd-runtime-ubuntu22.04
+FROM nvidia/opengl:1.2-glvnd-runtime-ubuntu22.04 AS deps-builder 
 
 # Make all NVIDIA GPUS visible
-ARG NVIDIA_VISIBLE_DEVICES=all
+ENV NVIDIA_VISIBLE_DEVICES=all
 ENV NVIDIA_DRIVER_CAPABILITIES=all
 ENV VIRTUALGL_VERSION=3.1.4
 ENV TURBOVNC_VERSION=3.0.3
@@ -371,4 +371,238 @@ WORKDIR /home/ws
 RUN /bin/bash -c "source /opt/ros/humble/setup.bash && \
   export GZ_VERSION=harmonic && \
   colcon build --packages-select gz_ros2_control --allow-overriding gz_ros2_control"
+WORKDIR /
+
+# Runtime stage: fresh-installs its own packages instead of copying the
+# builder's rootfs wholesale. Build toolchain (cmake/colcon/etc.) is kept
+# since Dockerfile.humble runs colcon build on top of this image. See
+# audit.md for the full package DROP/KEEP rationale.
+FROM nvidia/opengl:1.2-glvnd-runtime-ubuntu22.04 AS deps-final
+
+# Make all NVIDIA GPUS visible
+ENV NVIDIA_VISIBLE_DEVICES=all
+ENV NVIDIA_DRIVER_CAPABILITIES=all
+ENV VIRTUALGL_VERSION=3.1.4
+ENV TURBOVNC_VERSION=3.0.3
+
+# Setup NON INTERACTIVE ENVIRONMENT
+ENV DEBIAN_FRONTEND=noninteractive
+
+# Locale (locale files themselves come across via the wholesale COPY)
+ENV LANG=en_US.UTF-8
+ENV LANGUAGE=en_US:en
+ENV LC_ALL=en_US.UTF-8
+
+ENV ROS_DISTRO=humble
+ENV AMENT_PREFIX_PATH=/opt/ros/${ROS_DISTRO}
+ENV COLCON_PREFIX_PATH=/opt/ros/${ROS_DISTRO}
+ENV LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/gazebo-11/plugins:/opt/ros/${ROS_DISTRO}/opt/rviz_ogre_vendor/lib:/opt/ros/${ROS_DISTRO}/lib/x86_64-linux-gnu:/opt/ros/${ROS_DISTRO}/lib:/usr/share/gazebo/../../lib/x86_64-linux-gnu/gazebo-11/plugins:
+ENV PATH=/opt/ros/${ROS_DISTRO}/bin:$PATH
+ENV PYTHONPATH=/opt/ros/${ROS_DISTRO}/lib/python3.10/site-packages
+ENV ROS_PYTHON_VERSION=3
+ENV ROS_VERSION=2
+
+# Force pure-Python protobuf parser (see builder stage for rationale)
+ENV PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
+
+# VirtualGL + TurboVNC on PATH
+ENV PATH="$PATH:/opt/VirtualGL/bin:/opt/TurboVNC/bin"
+
+# Install ca-certificates before the PGDG source is copied in below, else the
+# HTTPS apt update against apt.postgresql.org fails on an untrusted cert
+# (base image has no certs). Same order as deps-builder (certs, then pgdg.list).
+RUN apt-get update && apt-get install -y --no-install-recommends ca-certificates \
+  && rm -rf /var/lib/apt/lists/*
+
+# Repo keyrings/sources from the builder so apt-get update can resolve
+# the ROS 2/Gazebo/PGDG repos here too.
+COPY --from=deps-builder /usr/share/keyrings/ /usr/share/keyrings/
+COPY --from=deps-builder /etc/apt/sources.list.d/ /etc/apt/sources.list.d/
+
+# Locale needs its own layer since locale-gen runs after install.
+RUN apt-get update && apt-get install -y --no-install-recommends locales \
+  && locale-gen en_US.UTF-8 \
+  && rm -rf /var/lib/apt/lists/*
+
+# Runtime apt packages (build toolchain kept, see above). i386 enabled for
+# the 32-bit VirtualGL stack. Editors/lint/test-only packages dropped vs
+# the builder — see audit.md for the full list.
+RUN dpkg --add-architecture i386 \
+  && apt-get update && apt-get install -y --no-install-recommends \
+  apt-utils \
+  build-essential \
+  cmake \
+  git curl wget gnupg lsb-release ca-certificates \
+  sudo \
+  net-tools \
+  pciutils \
+  nlohmann-json3-dev \
+  python3-pip \
+  python-is-python3 \
+  ros-dev-tools \
+  python3-colcon-common-extensions \
+  ros-${ROS_DISTRO}-ros-base \
+  ros-${ROS_DISTRO}-xacro \
+  ros-${ROS_DISTRO}-joint-state-publisher \
+  ros-${ROS_DISTRO}-rviz2 \
+  gz-harmonic \
+  ros-${ROS_DISTRO}-ros-gzharmonic \
+  dbus-x11 \
+  libdbus-c++-1-0v5 \
+  xvfb xauth xfonts-base xkb-data x11-xkb-utils \
+  x11vnc \
+  xterm \
+  xserver-xorg-video-dummy \
+  x11-apps \
+  libxtst6:i386 \
+  libxv1:i386 \
+  libglu1-mesa:i386 \
+  libegl1-mesa:i386 \
+  libegl1-mesa:amd64 \
+  libpq5 \
+  postgresql-client-18 \
+  ros-${ROS_DISTRO}-py-trees \
+  ros-${ROS_DISTRO}-py-trees-ros \
+  libbenchmark-dev \
+  libeigen3-dev \
+  libgeographic-dev \
+  libncurses-dev \
+  libyaml-cpp-dev \
+  pybind11-dev \
+  libpoco-dev \
+  libpcl-dev \
+  python3-jinja2 \
+  python3-pydantic \
+  python3-pymap3d \
+  ros-${ROS_DISTRO}-action-msgs \
+  ros-${ROS_DISTRO}-ament-cmake \
+  ros-${ROS_DISTRO}-ament-cmake-python \
+  ros-${ROS_DISTRO}-ament-index-cpp \
+  ros-${ROS_DISTRO}-backward-ros \
+  ros-${ROS_DISTRO}-builtin-interfaces \
+  ros-${ROS_DISTRO}-cv-bridge \
+  ros-${ROS_DISTRO}-geographic-msgs \
+  ros-${ROS_DISTRO}-geometry-msgs \
+  ros-${ROS_DISTRO}-image-transport \
+  ros-${ROS_DISTRO}-mocap4r2-msgs \
+  ros-${ROS_DISTRO}-nav-msgs \
+  ros-${ROS_DISTRO}-pluginlib \
+  ros-${ROS_DISTRO}-rclcpp \
+  ros-${ROS_DISTRO}-rclcpp-action \
+  ros-${ROS_DISTRO}-rclcpp-components \
+  ros-${ROS_DISTRO}-rclcpp-lifecycle \
+  ros-${ROS_DISTRO}-rclpy \
+  ros-${ROS_DISTRO}-robot-state-publisher \
+  ros-${ROS_DISTRO}-rosidl-default-generators \
+  ros-${ROS_DISTRO}-rosidl-default-runtime \
+  ros-${ROS_DISTRO}-sdformat-urdf \
+  ros-${ROS_DISTRO}-sensor-msgs \
+  ros-${ROS_DISTRO}-std-msgs \
+  ros-${ROS_DISTRO}-std-srvs \
+  ros-${ROS_DISTRO}-tf2 \
+  ros-${ROS_DISTRO}-tf2-geometry-msgs \
+  ros-${ROS_DISTRO}-tf2-msgs \
+  ros-${ROS_DISTRO}-tf2-ros \
+  ros-${ROS_DISTRO}-trajectory-msgs \
+  ros-${ROS_DISTRO}-visualization-msgs \
+  ros-${ROS_DISTRO}-moveit \
+  ros-${ROS_DISTRO}-ros2-control \
+  ros-${ROS_DISTRO}-ros2-controllers \
+  ros-${ROS_DISTRO}-gripper-controllers \
+  ros-${ROS_DISTRO}-rmw-cyclonedds-cpp \
+  ros-${ROS_DISTRO}-ur \
+  ros-${ROS_DISTRO}-pcl-ros \
+  ros-${ROS_DISTRO}-tf-transformations \
+  ros-${ROS_DISTRO}-pcl-conversions \
+  ros-${ROS_DISTRO}-pcl-msgs \
+  && rm -rf /var/lib/apt/lists/*
+
+RUN apt-get update && \
+  apt-get install -y lxde-common && \
+  rm -rf /var/lib/apt/lists/*
+
+# Re-install pip packages fresh rather than copying site-packages. Mirrors
+# the builder's pip blocks minus test/lint/firmware-only wheels — see audit.md.
+RUN python3.10 -m pip install --no-cache-dir --upgrade pip==23.3.1 wheel==0.41.3 setuptools==69.0.2
+
+RUN python3.10 -m pip install --no-cache-dir \
+  pylint==3.3.1 transitions==0.9.0 pydantic==2.4.2 websocket-client==1.5.2 \
+  argparse==1.4.0 cerberus==1.3.4 empy==3.3.4 jinja2==3.0.3 \
+  matplotlib==3.0.* numpy==1.24.3 packaging==21.3 pkgconfig==1.5.5 \
+  pyulog==1.0.1 pyyaml==5.4.1 requests==2.31.0 serial==0.0.97 six==1.16.0 toml==0.10.2 psutil==5.9.0 \
+  onnxruntime-gpu==1.22.0 Pillow==9.0.1 opencv-python==4.5.5.64 watchdog==2.1.5 utm==0.7.0 psycopg2 jedi pyapriltags
+
+# CUDA runtime wheels for onnxruntime-gpu — keep versions in lockstep with the builder block.
+RUN python3.10 -m pip install --no-cache-dir \
+  nvidia-cuda-runtime-cu12==12.8.90 \
+  nvidia-cublas-cu12==12.8.4.1 \
+  nvidia-cudnn-cu12==9.10.2.21 \
+  nvidia-nvjitlink-cu12==12.8.93 \
+  nvidia-cuda-nvrtc-cu12==12.8.93 \
+  nvidia-curand-cu12==10.3.10.19 \
+  nvidia-cufft-cu12==11.4.1.4
+
+# Register the pip-installed nvidia .so files with ldconfig (same as builder).
+COPY config-nvidia-libs.py /
+RUN chmod u+x /config-nvidia-libs.py && python3.10 /config-nvidia-libs.py && ldconfig
+
+# monaco editor formatter (RAM runtime dep)
+RUN python3.10 -m pip install --no-cache-dir black==24.10.0
+
+# websocket server + Django stack
+RUN python3.10 -m pip install --no-cache-dir websocket_server==0.6.4 posix-ipc==1.1.1 django==4.1.7 djangorestframework==3.13.1 \
+  django-webpack-loader==1.5.0 django-cors-headers==3.14.0 websockets==11.0.3 asyncio==3.4.3
+
+# BT studio + drone runtime pip (transforms3d/PySimpleGUI for teleop).
+RUN python3.10 -m pip install --no-cache-dir py-trees autopep8 transforms3d PySimpleGUI-4-foss
+
+# Non-apt/pip artifacts from the builder (git clones, .deb binaries, config
+# writes) — replaces the old blanket `COPY / /`.
+# VirtualGL + TurboVNC install trees + setuid faker libs (.deb already removed in the builder).
+COPY --from=deps-builder /opt/VirtualGL /opt/VirtualGL
+COPY --from=deps-builder /opt/TurboVNC /opt/TurboVNC
+COPY --from=deps-builder /usr/lib/libvglfaker.so /usr/lib/libvglfaker.so
+COPY --from=deps-builder /usr/lib/libdlfaker.so /usr/lib/libdlfaker.so
+COPY --from=deps-builder /usr/lib32/libvglfaker.so /usr/lib32/libvglfaker.so
+COPY --from=deps-builder /usr/lib32/libdlfaker.so /usr/lib32/libdlfaker.so
+COPY --from=deps-builder /usr/lib/i386-linux-gnu/libvglfaker.so /usr/lib/i386-linux-gnu/libvglfaker.so
+COPY --from=deps-builder /usr/lib/i386-linux-gnu/libdlfaker.so /usr/lib/i386-linux-gnu/libdlfaker.so
+
+# noVNC + websockify (git clones)
+COPY --from=deps-builder /noVNC /noVNC
+
+# Downloaded X config
+COPY --from=deps-builder /xorg.conf /xorg.conf
+
+# OMPL + onnxruntime C++ API — the only non-pip content left under
+# /usr/local. Scoped explicitly instead of `COPY /usr/local/ /usr/local/`,
+# which duplicated ~3.5GB of already-reinstalled pip packages (see
+# size-regression-rootcause.md).
+COPY --from=deps-builder /usr/local/lib/libompl.so* /usr/local/lib/
+COPY --from=deps-builder /usr/local/lib/libonnxruntime* /usr/local/lib/
+COPY --from=deps-builder /usr/local/lib/cmake/ /usr/local/lib/cmake/
+COPY --from=deps-builder /usr/local/lib/pkgconfig/ /usr/local/lib/pkgconfig/
+COPY --from=deps-builder /usr/local/include/ompl-1.7/ /usr/local/include/ompl-1.7/
+COPY --from=deps-builder /usr/local/include/core/ /usr/local/include/core/
+COPY --from=deps-builder /usr/local/include/*.h /usr/local/include/
+COPY --from=deps-builder /usr/local/share/ompl/ /usr/local/share/ompl/
+COPY --from=deps-builder /usr/local/share/man/ /usr/local/share/man/
+COPY --from=deps-builder /usr/local/share/ament_index/ /usr/local/share/ament_index/
+COPY --from=deps-builder /usr/local/bin/onnxruntime_test /usr/local/bin/onnxruntime_test
+COPY --from=deps-builder /usr/local/bin/ompl_benchmark_statistics.py /usr/local/bin/ompl_benchmark_statistics.py
+# OMPL Python bindings (SWIG-generated .py wrappers + compiled _*.so modules).
+# install-ompl-ubuntu.sh --python installs these into the system dist-packages
+# via `make install` (PYTHON_EXEC=/usr/bin/python3), NOT under /usr/local, so
+# the /usr/local COPY lines above miss them. The modules link libompl.so.18,
+# already provided by the /usr/local/lib/libompl.so* copy + ldconfig below.
+COPY --from=deps-builder /usr/lib/python3/dist-packages/ompl/ /usr/lib/python3/dist-packages/ompl/
+RUN ldconfig
+
+# Shell env writes (RMW impl, ROS/aerostack2 sourcing) — root's .bashrc
+COPY --from=deps-builder /root/.bashrc /root/.bashrc
+
+# ROS workspaces copied wholesale (install/ + src/, no subsetting).
+COPY --from=deps-builder /home/ws /home/ws
+COPY --from=deps-builder /home/drones_ws /home/drones_ws
+
 WORKDIR /

scripts/RADI/Dockerfile.humble

@@ -65,7 +65,6 @@ COPY --from=ra-base /tmp/app/react_frontend/webpack-stats.json /RoboticsAcademy/
 
 # Compiling and sourcing the workspace
 WORKDIR /home/ws
-RUN /bin/bash -c "source /home/drones_ws/install/setup.bash"
 RUN /bin/bash -c "source /opt/ros/humble/setup.bash; colcon build --symlink-install --packages-ignore gz_ros2_control gz_ros2_control_demos"
 
 # Django server