{ stdenv, config, lib, fetchFromGitHub, cmake, libusb1, ninja, pkg-config, gcc , cudaSupport ? config.cudaSupport or false, cudatoolkit , enablePython ? false, pythonPackages ? null }: assert cudaSupport -> cudatoolkit != null; assert enablePython -> pythonPackages != null; stdenv.mkDerivation rec { pname = "librealsense"; version = "2.41.0"; outputs = [ "out" "dev" ]; src = fetchFromGitHub { owner = "IntelRealSense"; repo = pname; rev = "v${version}"; sha256 = "0ngv9fgja72vg7hq1aiwpa7x4dhniawhpd8mqm85pqkjxiph8s1k"; }; buildInputs = [ libusb1 gcc.cc.lib ] ++ lib.optional cudaSupport cudatoolkit ++ lib.optional enablePython pythonPackages.python; patches = lib.optionals enablePython [ ./py_sitepackage_dir.patch ]; nativeBuildInputs = [ cmake ninja pkg-config ]; cmakeFlags = [ "-DBUILD_EXAMPLES=ON" "-DBUILD_GRAPHICAL_EXAMPLES=OFF" "-DBUILD_GLSL_EXTENSIONS=OFF" ] ++ lib.optionals enablePython [ "-DBUILD_PYTHON_BINDINGS:bool=true" "-DXXNIX_PYTHON_SITEPACKAGES=${placeholder "out"}/${pythonPackages.python.sitePackages}" ] ++ lib.optional cudaSupport "-DBUILD_WITH_CUDA:bool=true"; # ensure python package contains its __init__.py. for some reason the install # script does not do this, and it's questionable if intel knows it should be # done # ( https://github.com/IntelRealSense/meta-intel-realsense/issues/20 ) postInstall = lib.optionalString enablePython '' cp ../wrappers/python/pyrealsense2/__init__.py $out/${pythonPackages.python.sitePackages}/pyrealsense2 ''; meta = with lib; { description = "A cross-platform library for IntelĀ® RealSenseā„¢ depth cameras (D400 series and the SR300)"; homepage = "https://github.com/IntelRealSense/librealsense"; license = licenses.asl20; maintainers = with maintainers; [ brian-dawn ]; platforms = [ "i686-linux" "x86_64-linux" "x86_64-darwin" ]; }; }