The purpose of the chef app is to to:
Chef uses the shell app a starting point, but processes the data model defined on ZAP files during build time. This procedure is handled by its unified build script: chef.py.
When processing ZAP files as part of the build process, Chef places the auto-generated zap artifacts under its out temporary folder. Chef uses artifacts from zzz_generated for CI/CD.
All device types available (.zap files) are found inside the devices folder.
chef.py the first time to create a config.yaml configuration file. If you already have SDK environment variables such as IDF_PATH (esp32) and ZEPHYR_BASE (nrfconnect) it will use those values as default.config.yaml. TTY is the path used by the platform to enumerate its device as a serial port. Typical values are: # ESP32 macOS
TTY: /dev/tty.usbmodemXXXXXXX
# ESP32 Linux
TTY: /dev/ttyACM0
# NRFCONNECT macOS
TTY: /dev/tty.usbserial-XXXXX
# NRFCONNECT Linux
TTY: /dev/ttyUSB0
$ chef.py -u to update zap and the toolchain (on selected platforms).$ chef.py -gzbf -t <platform> -d lighting. This command will run the ZAP GUI opening the devices/lighting.zap file and will allow editing. It will then generate the zap artifacts, place them on the zap-generated folder, run a build and flash the binary in your target.chef.py -h to see all available commands.$ chef.py -g -d <device> to open in the ZAP GUI a device to be used as a starting point.Save As and save the file with the name of your new device type into the devices folder. This device is now available for the script. See chef.py -h for a list of devices available.<platform>: build system and main.cpp file for every supported platform. When porting a new platform, please minimize the source code in this folder, favoring the common folder for code that is not platform related.common: contains code shared between different platforms. It may contain source code that enables specific features such as LightingManager class or LockManager, as long as the application dynamically identify the presence of the relevant cluster configurations and it doesn't break the use cases where chef is built without these clusters.devices: contains the data models that may be used with chef. As of Matter 1.0 the data models are defined using .zap files.out: temporary folder used for placing ZAP generated artifacts.sample_app_util: guidelines and scripts for generating file names for new device types committed to the devices folder.config.yaml: contains general configuration for the chef.py script. As of Matter 1.0 this is used exclusively for toolchain and TTY interface paths.chef.py: main script for generating samples. More info on its help chef.py -h.All CI jobs for chef can be found in .github/workflows/chef.yaml.
These jobs use a platform-specific image with base chip-build. Such images contain the toolchain for the respective platform under /opt.
CI jobs call chef with the options --ci -t $PLATFORM. The --ci option will execute builds for all devices specified in ci_allow_list defined in cicd_config.json (so long as these devices are also in /devices) on the specified platform.
CI jobs also call the function bundle_$PLATFORM at the end of each example build. This function should copy or move build output files from the build output location into _CD_STAGING_DIR. Typically, the set of files touched is the minimal set of files needed to flash a device. See the function bundle_esp32 for reference.
First, implement a bundle_$PLATFORM function.
Next, ensure that the examples in ci_allow_list build in a container using the relevant platform image. You can simulate the workflow locally by mounting your CHIP repo into a container and executing the CI command:
docker run -it --mount source=$(pwd),target=/workspace,type=bind connectedhomeip/chip-build-$PLATFORM:$VERSION
In the container:
chown -R $(whoami) /workspace cd /workspace source ./scripts/bootstrap.sh source ./scripts/activate.sh ./examples/chef/chef.py --ci -t $PLATFORM
Once you are confident the CI examples build and bundle in a container, add a new job to the chef workflow.
Replace all instances of $PLATFORM with the new platform. Replace $VERSION with the image version used in the rest of the workflows, or update the image version for all images in the workflow as needed.
chef_$PLATFORM: name: Chef - $PLATFORM CI Examples runs-on: ubuntu-latest if: github.actor != 'restyled-io[bot]' container: image: connectedhomeip/chip-build-$PLATFORM:$VERSION options: --user root steps: - uses: Wandalen/wretry.action@v1.0.41 name: Checkout with: action: actions/checkout@v3 with: | token: ${{ github.token }} attempt_limit: 3 attempt_delay: 2000 - name: Checkout submodules run: | scripts/checkout_submodules.py --allow-changing-global-git-config --shallow --platform $PLATFORM - name: Bootstrap timeout-minutes: 25 run: bash scripts/bootstrap.sh - name: CI Examples $PLATFORM shell: bash run: | ./scripts/run_in_build_env.sh "./examples/chef/chef.py --ci -t $PLATFORM"
Once CI is enabled for a platform, the platform may also be integrated into integrations/cloudbuild/, where chef builds are defined in chef.yaml. See the README in this path for more information.
Note that the image used in chef.yaml is chip-build-vscode. See docker/images/chip-build-vscode/Dockerfile for the source of this image. This image is a combination of the individual toolchain images. Therefore, before a platform is integrated into chef CD, the toolchain should be copied into chip-build-vscode and chef.yaml should be updated to use the new image version.
Finally, add the new platform to cd_platforms in cicd_config.json. The configuration should follow the following schema:
"$PLATFORM": { "output_archive_prefix_1": ["option_1", "option_2"], "output_archive_prefix_2": [], }
Take note of the configuration for linux:
"linux": { "linux_x86": ["--cpu_type", "x64"], "linux_arm64_ipv6only": ["--cpu_type", "arm64", "--ipv6only"] },
This will produce output archives prefixed linux_x86 and linux_arm_64_ipv6only and will append the respective options to each build command for these targets.
To test your configuration locally, you may employ a similar strategy as in CI:
docker run -it --mount source=$(pwd),target=/workspace,type=bind connectedhomeip/chip-build-vscode:$VERSION
In the container:
chown -R $(whoami) /workspace cd /workspace source ./scripts/bootstrap.sh source ./scripts/activate.sh ./examples/chef/chef.py --build_all --keep_going
You may also use the Google Cloud Build local builder as detailed in the README of integrations/cloudbuild/.
To add new devices for chef:
python sample_app_util.py zap <zap_file> --rename-file to rename the example and place the new file in examples/chef/devices.README in examples/chef/sample_app_util/ for more info.scripts/tools/zap_regen_all.py, commit zzz_generated and examples/chef/devices..github/workflows/zap_templates.yaml.