mirror of
https://github.com/espressif/esp-idf.git
synced 2024-10-05 20:47:46 -04:00
104 lines
5.4 KiB
ReStructuredText
104 lines
5.4 KiB
ReStructuredText
****************
|
|
IDF Docker Image
|
|
****************
|
|
|
|
..
|
|
When changing this page, please keep tools/docker/README.md in sync.
|
|
|
|
.. highlight:: bash
|
|
|
|
IDF Docker image (``espressif/idf``) is intended for building applications and libraries with specific versions of ESP-IDF, when doing automated builds.
|
|
|
|
The image contains:
|
|
|
|
- Common utilities such as git, wget, curl, zip.
|
|
- Python 3.7 or newer.
|
|
- A copy of a specific version of ESP-IDF (see below for information about versions). ``IDF_PATH`` environment variable is set, and points to ESP-IDF location in the container.
|
|
- All the build tools required for the specific version of ESP-IDF: CMake, ninja, cross-compiler toolchains, etc.
|
|
- All Python packages required by ESP-IDF are installed in a virtual environment.
|
|
|
|
The image entrypoint sets up ``PATH`` environment variable to point to the correct version of tools, and activates the Python virtual environment. As a result, the environment is ready to use the ESP-IDF build system.
|
|
|
|
The image can also be used as a base for custom images, if additional utilities are required.
|
|
|
|
Tags
|
|
====
|
|
|
|
Multiple tags of this image are maintained:
|
|
|
|
- ``latest``: tracks ``master`` branch of ESP-IDF
|
|
- ``vX.Y``: corresponds to ESP-IDF release ``vX.Y``
|
|
- ``release-vX.Y``: tracks ``release/vX.Y`` branch of ESP-IDF
|
|
|
|
.. note::
|
|
|
|
Versions of ESP-IDF released before this feature was introduced do not have corresponding Docker image versions. You can check the up-to-date list of available tags at https://hub.docker.com/r/espressif/idf/tags.
|
|
|
|
Usage
|
|
=====
|
|
|
|
Setting up Docker
|
|
~~~~~~~~~~~~~~~~~
|
|
|
|
Before using the ``espressif/idf`` Docker image locally, make sure you have Docker installed. Follow the instructions at https://docs.docker.com/install/, if it is not installed yet.
|
|
|
|
If using the image in CI environment, consult the documentation of your CI service on how to specify the image used for the build process.
|
|
|
|
Building a project with CMake
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
In the project directory, run::
|
|
|
|
docker run --rm -v $PWD:/project -w /project espressif/idf idf.py build
|
|
|
|
|
|
The above command explained:
|
|
|
|
- ``docker run``: runs a Docker image. It is a shorter form of the command ``docker container run``.
|
|
- ``--rm``: removes the container when the build is finished
|
|
- ``-v $PWD:/project``: mounts the current directory on the host (``$PWD``) as ``/project`` directory in the container
|
|
- ``espressif/idf``: uses Docker image ``espressif/idf`` with tag ``latest`` (implicitly added by Docker when no tag is specified)
|
|
- ``idf.py build``: runs this command inside the container
|
|
|
|
To build with a specific docker image tag, specify it as ``espressif/idf:TAG``, for example::
|
|
|
|
docker run --rm -v $PWD:/project -w /project espressif/idf:release-v4.4 idf.py build
|
|
|
|
You can check the up-to-date list of available tags at https://hub.docker.com/r/espressif/idf/tags.
|
|
|
|
Using the image interactively
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
It is also possible to do builds interactively, to debug build issues or test the automated build scripts. Start the container with `-i -t` flags::
|
|
|
|
docker run --rm -v $PWD:/project -w /project -it espressif/idf
|
|
|
|
|
|
Then inside the container, use ``idf.py`` as usual::
|
|
|
|
idf.py menuconfig
|
|
idf.py build
|
|
|
|
.. note::
|
|
|
|
Commands which communicate with the development board, such as ``idf.py flash`` and ``idf.py monitor`` will not work in the container unless the serial port is passed through into the container. However currently this is not possible with Docker for Windows (https://github.com/docker/for-win/issues/1018) and Docker for Mac (https://github.com/docker/for-mac/issues/900).
|
|
|
|
Building custom images
|
|
======================
|
|
|
|
The Dockerfile in ESP-IDF repository provides several build arguments which can be used to customize the Docker image:
|
|
|
|
- ``IDF_CLONE_URL``: URL of the repository to clone ESP-IDF from. Can be set to a custom URL when working with a fork of ESP-IDF. Default is ``https://github.com/espressif/esp-idf.git``.
|
|
- ``IDF_CLONE_BRANCH_OR_TAG``: Name of a git branch or tag use when cloning ESP-IDF. This value is passed to ``git clone`` command using the ``--branch`` argument. Default is ``master``.
|
|
- ``IDF_CHECKOUT_REF``: If this argument is set to a non-empty value, ``git checkout $IDF_CHECKOUT_REF`` command will be performed after cloning. This argument can be set to the SHA of the specific commit to check out, for example if some specific commit on a release branch is desired.
|
|
- ``IDF_CLONE_SHALLOW``: If this argument is set to a non-empty value, ``--depth=1 --shallow-submodules`` arguments will be used when performing ``git clone``. This significantly reduces the amount of data downloaded and the size of the resulting Docker image. However, if switching to a different branch in such a "shallow" repository is necessary, an additional ``git fetch origin <branch>`` command must be executed first.
|
|
- ``IDF_INSTALL_TARGETS``: Comma-separated list of IDF targets to install toolchains for, or ``all`` to install toolchains for all targets. Selecting specific targets reduces the amount of data downloaded and the size of the resulting Docker image. Default is ``all``.
|
|
|
|
To use these arguments, pass them via the ``--build-arg`` command line option. For example, the following command will build a Docker image with a shallow clone of ESP-IDF v4.4.1 and tools for ESP32-C3, only::
|
|
|
|
docker build -t idf-custom:v4.4.1-esp32c3 \
|
|
--build-arg IDF_CLONE_BRANCH_OR_TAG=v4.4.1 \
|
|
--build-arg IDF_CLONE_SHALLOW=1 \
|
|
--build-arg IDF_INSTALL_TARGETS=esp32c3 \
|
|
tools/docker
|