This document explains the implementation of the ESP-IDF build system and the concept of "components". Read this document if you want to know how to organize and build a new ESP-IDF project or component.
..note:: This document describes the CMake-based build system, which is the default since ESP-IDF V4.0. ESP-IDF also supports a :doc:`legacy build system based on GNU Make <build-system-legacy>`, which was the default before ESP-IDF V4.0.
An ESP-IDF project can be seen as an amalgamation of a number of components. For example, for a webserver that shows the current humidity, there could be:
ESP-IDF makes these components explicit and configurable. To do that, when a project is compiled, the build system will look up all the components in the ESP-IDF directories, the project directories and (optionally) in additional custom component directories. It then allows the user to configure the ESP-IDF project using a text-based menu system to customize each component. After the components in the project are configured, the build system will compile the project.
- A "project" is a directory that contains all the files and configuration to build a single "app" (executable), as well as additional supporting elements such as a partition table, data/filesystem partitions, and a bootloader.
- "Project configuration" is held in a single file called ``sdkconfig`` in the root directory of the project. This configuration file is modified via ``idf.py menuconfig`` to customise the configuration of the project. A single project contains exactly one project configuration.
- An "app" is an executable which is built by ESP-IDF. A single project will usually build two apps - a "project app" (the main executable, ie your custom firmware) and a "bootloader app" (the initial bootloader program which launches the project app).
- "components" are modular pieces of standalone code which are compiled into static libraries (.a files) and linked into an app. Some are provided by ESP-IDF itself, others may be sourced from other places.
- "ESP-IDF" is not part of the project. Instead it is standalone, and linked to the project via the ``IDF_PATH`` environment variable which holds the path of the ``esp-idf`` directory. This allows the IDF framework to be decoupled from your project.
The :ref:`getting started guide <get-started-configure>` contains a brief introduction to how to set up ``idf.py`` to configure, build, and flash projects.
``idf.py`` should be run in an ESP-IDF "project" directory, ie one containing a ``CMakeLists.txt`` file. Older style projects with a Makefile will not work with ``idf.py``.
-``idf.py menuconfig`` runs the "menuconfig" tool to configure the project.
-``idf.py build`` will build the project found in the current directory. This can involve multiple steps:
- Create the build directory if needed. The sub-directory ``build`` is used to hold build output, although this can be changed with the ``-B`` option.
- Run CMake_ as necessary to configure the project and generate build files for the main build tool.
- Run the main build tool (Ninja_ or `GNU Make`). By default, the build tool is automatically detected but it can be explicitly set by passing the ``-G`` option to ``idf.py``.
Building is incremental so if no source files or configuration has changed since the last build, nothing will be done.
-``idf.py clean`` will "clean" the project by deleting build output files from the build directory, forcing a "full rebuild" the next time the project is built. Cleaning doesn't delete CMake configuration output and some other files.
-``idf.py fullclean`` will delete the entire "build" directory contents. This includes all CMake configuration output. The next time the project is built, CMake will configure it from scratch. Note that this option recursively deletes *all* files in the build directory, so use with care. Project configuration is not deleted.
-``idf.py flash`` will automatically build the project if necessary, and then flash it to the target. The ``-p`` and ``-b`` options can be used to set serial port name and flasher baud rate, respectively.
-``idf.py monitor`` will display serial output from the target. The ``-p`` option can be used to set the serial port name. Type ``Ctrl-]`` to exit the monitor. See :doc:`tools/idf-monitor` for more details about using the monitor.
Multiple ``idf.py`` commands can be combined into one. For example, ``idf.py -p COM4 clean flash monitor`` will clean the source tree, then build the project and flash it to the target before running the serial monitor.
In order to make `shell autocompletion <https://click.palletsprojects.com/shell-completion/>`_ supported, please make sure you have at least Python 3.5 and `click <https://click.palletsprojects.com/>`_ 7.1 or newer (:ref:`see also <get-started-get-prerequisites>`).
To enable autocompletion for ``idf.py`` use the ``export`` command (:ref:`see this <get-started-export>`). Autocompletion is initiated by pressing the TAB key. Type "idf.py -" and press the TAB key to autocomplete options.
The autocomplete support for PowerShell is planned in the future.
..note:: The environment variables ``ESPPORT`` and ``ESPBAUD`` can be used to set default values for the ``-p`` and ``-b`` options, respectively. Providing these options on the command line overrides the default.
Advanced Commands
^^^^^^^^^^^^^^^^^
-``idf.py app``, ``idf.py bootloader``, ``idf.py partition_table`` can be used to build only the app, bootloader, or partition table from the project as applicable.
-``idf.py size`` prints some size information about the app. ``size-components`` and ``size-files`` are similar commands which print more detailed per-component or per-source-file information, respectively. If you define variable ``-DOUTPUT_JSON=1`` when running CMake (or ``idf.py``), the output will be formatted as JSON not as human readable text. See ``idf.py-size`` for more information.
-``idf.py reconfigure`` re-runs CMake_ even if it doesn't seem to need re-running. This isn't necessary during normal usage, but can be useful after adding/removing files from the source tree, or when modifying CMake cache variables. For example, ``idf.py -DNAME='VALUE' reconfigure`` can be used to set variable ``NAME`` in CMake cache to value ``VALUE``.
-``idf.py python-clean`` deletes generated Python byte code from the IDF directory which may cause issues when switching between IDF and Python versions. It is advised to run this target after switching versions of Python.
-``idf.py docs`` will open direct link to documentation for project's chip target and version in browser. To see all options use ``idf.py docs --help``
The order of multiple ``idf.py`` commands on the same invocation is not important, they will automatically be executed in the correct order for everything to take effect (ie building before flashing, erasing before flashing, etc.).
To list all available root level options, run ``idf.py --help``. To list options that are specific for a subcommand, run ``idf.py <command> --help``, for example ``idf.py monitor --help``. Here is a list of some useful options:
-``-C <dir>`` allows overriding the project directory from the default current working directory.
-``-B <dir>`` allows overriding the build directory from the default ``build`` subdirectory of the project directory.
-``--ccache`` flag can be used to enable CCache_ when compiling source files, if the CCache_ tool is installed. This can dramatically reduce some build times.
Note that some older versions of CCache may exhibit bugs on some platforms, so if files are not rebuilt as expected then try disabling CCache and build again. CCache can be enabled by default by setting the ``IDF_CCACHE_ENABLE`` environment variable to a non-zero value.
-``--cmake-warn-uninitialized`` (or ``-w``) will cause CMake to print uninitialized variable warnings inside the project directory (not for directories not found inside the project directory). This only controls CMake variable warnings inside CMake itself, not other types of build warnings. This option can also be set permanently by setting the ``IDF_CMAKE_WARN_UNINITIALIZED`` environment variable to a non-zero value.
:ref:`idf.py` is a wrapper around CMake_ for convenience. However, you can also invoke CMake directly if you prefer.
..highlight:: bash
When ``idf.py`` does something, it prints each command that it runs for easy reference. For example, the ``idf.py build`` command is the same as running these commands in a bash shell (or similar commands for Windows Command Prompt)::
mkdir -p build
cd build
cmake .. -G Ninja # or 'Unix Makefiles'
ninja
In the above list, the ``cmake`` command configures the project and generates build files for use with the final build tool. In this case the final build tool is Ninja_: running ``ninja`` actually builds the project.
It's not necessary to run ``cmake`` more than once. After the first build, you only need to run ``ninja`` each time. ``ninja`` will automatically re-invoke ``cmake`` if the project needs reconfiguration.
If using CMake with ``ninja`` or ``make``, there are also targets for more of the ``idf.py`` sub-commands - for example running ``make menuconfig`` or ``ninja menuconfig`` in the build directory will work the same as ``idf.py menuconfig``.
..note::
If you're already familiar with CMake_, you may find the ESP-IDF CMake-based build system unusual because it wraps a lot of CMake's functionality to reduce boilerplate. See `writing pure CMake components`_ for some information about writing more "CMake style" components.
.._flash-with-ninja-or-make:
Flashing with ninja or make
^^^^^^^^^^^^^^^^^^^^^^^^^^^
It's possible to build and flash directly from ninja or make by running a target like::
Available targets are: ``flash``, ``app-flash`` (app only), ``bootloader-flash`` (bootloader only).
When flashing this way, optionally set the ``ESPPORT`` and ``ESPBAUD`` environment variables to specify the serial port and baud rate. You can set environment variables in your operating system or IDE project. Alternatively, set them directly on the command line::
ESPPORT=/dev/ttyUSB0 ninja flash
..note:: Providing environment variables at the start of the command like this is Bash shell Syntax. It will work on Linux and macOS. It won't work when using Windows Command Prompt, but it will work when using Bash-like shells on Windows.
Or::
make -j3 app-flash ESPPORT=COM4 ESPBAUD=2000000
..note:: Providing variables at the end of the command line is ``make`` syntax, and works for ``make`` on all platforms.
Using CMake in an IDE
---------------------
You can also use an IDE with CMake integration. The IDE will want to know the path to the project's ``CMakeLists.txt`` file. IDEs with CMake integration often provide their own build tools (CMake calls these "generators") to build the source files as part of the IDE.
When adding custom non-build steps like "flash" to the IDE, it is recommended to execute ``idf.py`` for these "special" commands.
For more detailed information about integrating ESP-IDF with CMake into an IDE, see `Build System Metadata`_.
ESP-IDF works well with all supported Python versions. It should work out-of-box even if you have a legacy system where the default ``python`` interpreter is still Python 2.7, however, it is advised to switch to Python 3 if possible.
``idf.py`` and other Python scripts will run with the default Python interpreter, i.e. ``python``. You can switch to a different one like ``python3 $IDF_PATH/tools/idf.py ...``, or you can set up a shell alias or another script to simplify the command.
To manage the Python version more generally via the command line, check out the tools pyenv_ or virtualenv_. These let you change the default Python version.
- A top-level project CMakeLists.txt file. This is the primary file which CMake uses to learn how to build the project; and may set project-wide CMake variables. It includes the file :idf_file:`/tools/cmake/project.cmake` which implements the rest of the build system. Finally, it sets the project name and defines the project.
- "sdkconfig" project configuration file. This file is created/updated when ``idf.py menuconfig`` runs, and holds configuration for all of the components in the project (including ESP-IDF itself). The "sdkconfig" file may or may not be added to the source control system of the project.
- Optional "components" directory contains components that are part of the project. A project does not have to contain custom components of this kind, but it can be useful for structuring reusable code or including third party components that aren't part of ESP-IDF. Alternatively, ``EXTRA_COMPONENT_DIRS`` can be set in the top-level CMakeLists.txt to look for components in other places. See the :ref:`renaming main <rename-main>` section for more info. If you have a lot of source files in your project, we recommend grouping most into components instead of putting them all in "main".
- "main" directory is a special component that contains source code for the project itself. "main" is a default name, the CMake variable ``COMPONENT_DIRS`` includes this component but you can modify this variable.
- "build" directory is where build output is created. This directory is created by ``idf.py`` if it doesn't already exist. CMake configures the project and generates interim build files in this directory. Then, after the main build process is run, this directory will also contain interim object files and libraries as well as final binary output files. This directory is usually not added to source control or distributed with the project source code.
Component directories each contain a component ``CMakeLists.txt`` file. This file contains variable definitions to control the build process of the component, and its integration into the overall project. See `Component CMakeLists Files`_ for more details.
Each component may also include a ``Kconfig`` file defining the `component configuration`_ options that can be set via ``menuconfig``. Some components may also include ``Kconfig.projbuild`` and ``project_include.cmake`` files, which are special files for `overriding parts of the project`_.
Each project has a single top-level ``CMakeLists.txt`` file that contains build settings for the entire project. By default, the project CMakeLists can be quite minimal.
The inclusion of these three lines, in the order shown above, is necessary for every project:
-``cmake_minimum_required(VERSION 3.5)`` tells CMake the minimum version that is required to build the project. ESP-IDF is designed to work with CMake 3.5 or newer. This line must be the first line in the CMakeLists.txt file.
-``include($ENV{IDF_PATH}/tools/cmake/project.cmake)`` pulls in the rest of the CMake functionality to configure the project, discover all the components, etc.
-``project(myProject)`` creates the project itself, and specifies the project name. The project name is used for the final binary output files of the app - ie ``myProject.elf``, ``myProject.bin``. Only one project can be defined per CMakeLists file.
These variables all have default values that can be overridden for custom behaviour. Look in :idf_file:`/tools/cmake/project.cmake` for all of the implementation details.
-``COMPONENT_DIRS``: Directories to search for components. Defaults to ``IDF_PATH/components``, ``PROJECT_DIR/components``, and ``EXTRA_COMPONENT_DIRS``. Override this variable if you don't want to search for components in these places.
-``EXTRA_COMPONENT_DIRS``: Optional list of additional directories to search for components. Paths can be relative to the project directory, or absolute.
-``COMPONENTS``: A list of component names to build into the project. Defaults to all components found in the ``COMPONENT_DIRS`` directories. Use this variable to "trim down" the project for faster build times. Note that any component which "requires" another component via the REQUIRES or PRIV_REQUIRES arguments on component registration will automatically have it added to this list, so the ``COMPONENTS`` list can be very short.
To set these variables, use the `cmake set command <cmake set_>`_ ie ``set(VARIABLE "VALUE")``. The ``set()`` commands should be placed after the ``cmake_minimum(...)`` line but before the ``include(...)`` line.
The build system provides special treatment to the ``main`` component. It is a component that gets automatically added to the build provided that it is in the expected location, PROJECT_DIR/main. All other components in the build are also added as its dependencies, saving the user from hunting down dependencies and providing a build that works right out of the box. Renaming the ``main`` component causes the loss of these behind-the-scenes heavy lifting, requiring the user to specify the location of the newly renamed component and manually specifying its dependencies. Specifically, the steps to renaming ``main`` are as follows:
2. Set ``EXTRA_COMPONENT_DIRS`` in the project CMakeLists.txt to include the renamed ``main`` directory.
3. Specify the dependencies in the renamed component's CMakeLists.txt file via REQUIRES or PRIV_REQUIRES arguments :ref:`on component registration<cmake_minimal_component_cmakelists>`.
The build sets some global build specifications (compile flags, definitions, etc.) that gets used in compiling all sources from all components.
..highlight:: cmake
For example, one of the default build specifications set is the compile option ``-Wextra``. Suppose a user wants to use override this with ``-Wno-extra``,
This ensures that the compile options set by the user won't be overriden by the default build specifications, since the latter are set inside ``project()``.
Each project contains one or more components. Components can be part of ESP-IDF, part of the project's own components directory, or added from custom component directories (:ref:`see above <component-directories>`).
The list of directories in ``COMPONENT_DIRS`` is searched for the project's components. Directories in this list can either be components themselves (ie they contain a `CMakeLists.txt` file), or they can be top-level directories whose sub-directories are components.
When CMake runs to configure the project, it logs the components included in the build. This list can be useful for debugging the inclusion/exclusion of certain components.
When ESP-IDF is collecting all the components to compile, it will do this in the order specified by ``COMPONENT_DIRS``; by default, this means ESP-IDF's internal components first (``IDF_PATH/components``), then any components in directories specified in ``EXTRA_COMPONENT_DIRS``, and finally the project's components (``PROJECT_DIR/components``). If two or more of these directories contain component sub-directories with the same name, the component in the last place searched is used. This allows, for example, overriding ESP-IDF components with a modified version by copying that component from the ESP-IDF components directory to the project components directory and then modifying it there. If used in this way, the ESP-IDF directory itself can remain untouched.
..note:: If a component is overridden in an existing project by moving it to a new location, the project will not automatically see the new component path. Run ``idf.py reconfigure`` (or delete the project build folder) and then build again.
-``SRCS`` is a list of source files (``*.c``, ``*.cpp``, ``*.cc``, ``*.S``). These source files will be compiled into the component library.
-``INCLUDE_DIRS`` is a list of directories to add to the global include search path for any component which requires this component, and also the main source files.
-``REQUIRES`` is not actually required, but it is very often required to declare what other components this component will use. See :ref:`component requirements`.
Use the command ``idf.py create-component`` for creating a new component. The new component will contain set of files necessary for building a component. You may include the component's header file into your project and use its functionality. For more information execute ``idf.py create-component --help``.
The example will create a new component in the subdirectory `components` under the current working directory. For more information about components follow the documentation page :ref:`see above <component-directories>`.
-``COMPONENT_DIR``: The component directory. Evaluates to the absolute path of the directory containing ``CMakeLists.txt``. The component path cannot contain spaces. This is the same as the ``CMAKE_CURRENT_SOURCE_DIR`` variable.
-``COMPONENT_NAME``: Name of the component. Same as the name of the component directory.
-``COMPONENT_ALIAS``: Alias of the library created internally by the build system for the component.
-``COMPONENT_LIB``: Name of the library created internally by the build system for the component.
-``CONFIG_*``: Each value in the project configuration has a corresponding variable available in cmake. All names begin with ``CONFIG_``. :doc:`More information here </api-reference/kconfig>`.
-``ESP_PLATFORM``: Set to 1 when the CMake file is processed within ESP-IDF build system.
The following are some project/build variables that are available as build properties and whose values can be queried using ``idf_build_get_property``
from the component CMakeLists.txt:
-``PROJECT_NAME``: Name of the project, as set in project CMakeLists.txt file.
-``PROJECT_DIR``: Absolute path of the project directory containing the project CMakeLists. Same as the ``CMAKE_SOURCE_DIR`` variable.
-``COMPONENTS``: Names of all components that are included in this build, formatted as a semicolon-delimited CMake list.
-``IDF_VER``: Git version of ESP-IDF (produced by ``git describe``)
-``IDF_VERSION_MAJOR``, ``IDF_VERSION_MINOR``, ``IDF_VERSION_PATCH``: Components of ESP-IDF version, to be used in conditional expressions. Note that this information is less precise than that provided by ``IDF_VER`` variable. ``v4.0-dev-*``, ``v4.0-beta1``, ``v4.0-rc1`` and ``v4.0`` will all have the same values of ``IDF_VERSION_*`` variables, but different ``IDF_VER`` values.
-``IDF_TARGET``: Name of the target for which the project is being built.
Each component can also have a ``Kconfig`` file, alongside ``CMakeLists.txt``. This contains configuration settings to add to the configuration menu for this component.
When compiling each component, the ESP-IDF build system recursively evaluates its dependencies. This means each component needs to declare the components that it depends on ("requires").
-``PRIV_REQUIRES`` should be set to all components whose header files are #included from *any source files* in this component, unless already listed in ``REQUIRES``. Also any component which is required to be linked in order for this component to function correctly.
- The values of ``REQUIRES`` and ``PRIV_REQUIRES`` should not depend on any configuration choices (``CONFIG_xxx`` macros). This is because requirements are expanded before configuration is loaded. Other component variables (like include paths or source files) can depend on configuration choices.
- Not setting either or both ``REQUIRES`` variables is fine. If the component has no requirements except for the `Common component requirements`_ needed for RTOS, libc, etc.
If a components only supports some target chips (values of ``IDF_TARGET``) then it can specify ``REQUIRED_IDF_TARGETS`` in the ``idf_component_register`` call to express these requirements. In this case the build system will generate an error if the component is included into the build, but does not support the selected target.
..note:: In CMake terms, ``REQUIRES`` & ``PRIV_REQUIRES`` are approximate wrappers around the CMake functions ``target_link_libraries(... PUBLIC ...)`` and ``target_link_libraries(... PRIVATE ...)``.
.._example component requirements:
Example of component requirements
---------------------------------
Imagine there is a ``car`` component, which uses the ``engine`` component, which uses the ``spark_plug`` component:
..code-block:: none
- autoProject/
- CMakeLists.txt
- components/ - car/ - CMakeLists.txt
- car.c
- car.h
- engine/ - CMakeLists.txt
- engine.c
- include/ - engine.h
- spark_plug/ - CMakeLists.txt
- plug.c
- plug.h
Car component
^^^^^^^^^^^^^
..highlight:: c
The ``car.h`` header file is the public interface for the ``car`` component. This header includes ``engine.h`` directly because it uses some declarations from this header::
/* car.h */
#include "engine.h"
#ifdef ENGINE_IS_HYBRID
#define CAR_MODEL "Hybrid"
#endif
And car.c includes ``car.h`` as well::
/* car.c */
#include "car.h"
This means the ``car/CMakeLists.txt`` file needs to declare that ``car`` requires ``engine``:
..code-block:: cmake
idf_component_register(SRCS "car.c"
INCLUDE_DIRS "."
REQUIRES engine)
-``SRCS`` gives the list of source files in the ``car`` component.
-``INCLUDE_DIRS`` gives the list of public include directories for this component. Because the public interface is ``car.h``, the directory containing ``car.h`` is listed here.
-``REQUIRES`` gives the list of components required by the public interface of this component. Because ``car.h`` is a public header and includes a header from ``engine``, we include ``engine`` here. This makes sure that any other component which includes ``car.h`` will be able to recursively include the required ``engine.h`` also.
Engine component
^^^^^^^^^^^^^^^^
..highlight:: c
The ``engine`` component also has a public header file ``include/engine.h``, but this header is simpler::
/* engine.h */
#define ENGINE_IS_HYBRID
void engine_start(void);
The implementation is in ``engine.c``::
/* engine.c */
#include "engine.h"
#include "spark_plug.h"
...
In this component, ``engine`` depends on ``spark_plug`` but this is a private dependency. ``spark_plug.h`` is needed to compile ``engine.c``, but not needed to include ``engine.h``.
This means that the ``engine/CMakeLists.txt`` file can use ``PRIV_REQUIRES``:
..code-block:: cmake
idf_component_register(SRCS "engine.c"
INCLUDE_DIRS "include"
PRIV_REQUIRES spark_plug)
As a result, source files in the ``car`` component don't need the ``spark_plug`` include directories added to their compiler search path. This can speed up compilation, and stops compiler command lines from becoming longer than necessary.
Spark Plug Component
^^^^^^^^^^^^^^^^^^^^
The ``spark_plug`` component doesn't depend on anything else. It has a public header file ``spark_plug.h``, but this doesn't include headers from any other components.
This means that the ``spark_plug/CMakeLists.txt`` file doesn't need any ``REQUIRES`` or ``PRIV_REQUIRES`` clauses:
..code-block:: cmake
idf_component_register(SRCS "spark_plug.c"
INCLUDE_DIRS ".")
Source File Include Directories
-------------------------------
Each component's source file is compiled with these include path directories, as specified in the passed arguments to ``idf_component_register``:
..code-block:: cmake
idf_component_register(..
INCLUDE_DIRS "include"
PRIV_INCLUDE_DIRS "other")
- The current component's ``INCLUDE_DIRS`` and ``PRIV_INCLUDE_DIRS``.
- The ``INCLUDE_DIRS`` belonging to all other components listed in the ``REQUIRES`` and ``PRIV_REQUIRES`` parameters (ie all the current component's public and private dependencies).
- Recursively, all of the ``INCLUDE_DIRS`` of those components ``REQUIRES`` lists (ie all public dependencies of this component's dependencies, recursively expanded).
Main component requirements
---------------------------
The component named ``main`` is special because it automatically requires all other components in the build. So it's not necessary to pass ``REQUIRES`` or ``PRIV_REQUIRES`` to this component. See :ref:`renaming main <rename-main>` for a description of what needs to be changed if no longer using the ``main`` component.
Common component requirements
-----------------------------
To avoid duplication, every component automatically requires some "common" IDF components even if they are not mentioned explicitly. Headers from these components can always be included.
- If you set the ``COMPONENTS`` variable to a minimal list of components used directly by your project, then the build will expand to also include required components. The full list of components will be:
It's possible for a project to contain Component A that requires (``REQUIRES`` or ``PRIV_REQUIRES``) Component B, and Component B that requires Component A. This is known as a dependency cycle or a circular dependency.
CMake will usually handle circular dependencies automatically by repeating the component library names twice on the linker command line. However this strategy doesn't always work, and it's possible the build will fail with a linker error about "Undefined reference to ...", referencing a symbol defined by one of the components inside the circular dependency. This is particularly likely if there is a large circular dependency, i.e. A->B->C->D->A.
The best solution is to restructure the components to remove the circular dependency. In most cases, a software architecture without circular dependencies has desirable properties of modularity and clean layering and will be more maintainable in the long term. However, removing circular dependencies is not always possible.
To bypass a linker error caused by a circular dependency, the simplest workaround is to increase the CMake `LINK_INTERFACE_MULTIPLICITY`_ property of one of the component libraries. This causes CMake to repeat this library and its dependencies more than two times on the linker command line.
- This line should be placed after ``idf_component_register`` in the component CMakeLists.txt file.
- If possible, place this line in the component that creates the circular dependency by depending on a lot of other components. However, the line can be placed inside any component that is part of the cycle. Choosing the component that owns the source file shown in the linker error message, or the component that defines the symbol(s) mentioned in the linker error message, is a good place to start.
- Usually increasing the value to 3 (default is 2) is enough, but if this doesn't work then try increasing the number further.
- Adding this option will make the linker command line longer, and the linking stage slower.
Advanced Workaround: Undefined Symbols
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
If only one or two symbols is causing a circular dependency, and all other dependencies are linear, then there is an alternative method to avoid linker errors: Specify the specific symbols required for the "reverse" dependency as undefined symbols at link time.
For example, if component A depends on component B but component B also needs to reference ``reverse_ops`` from component A (but nothing else), then you can add a line like the following to the component B CMakeLists.txt to resolve the cycle at link time:
..code-block:: cmake
# This symbol is provided by 'Component A' at link time
- The ``-u`` argument means that the linker will always include this symbol in the link, regardless of dependency ordering.
- This line should be placed after ``idf_component_register`` in the component CMakeLists.txt file.
- If 'Component B' doesn't need to access any headers of 'Component A', only link to a few symbol(s), then this line can be used instead of any ``REQUIRES`` from B to A. This further simplifies the component structure in the build system.
See the `target_link_libraries`_ documentation for more information about this CMake function.
- Very early in the CMake configuration process, the script ``expand_requirements.cmake`` is run. This script does a partial evaluation of all component CMakeLists.txt files and builds a graph of component requirements (this :ref:`graph may have cycles <component-circular-dependencies>`). The graph is used to generate a file ``component_depends.cmake`` in the build directory.
- The main CMake process then includes this file and uses it to determine the list of components to include in the build (internal ``BUILD_COMPONENTS`` variable). The ``BUILD_COMPONENTS`` variable is sorted so dependencies are listed first, however as the component dependency graph has cycles this cannot be guaranteed for all components. The order should be deterministic given the same set of components and component dependencies.
- The value of ``BUILD_COMPONENTS`` is logged by CMake as "Component names: "
- Configuration is then evaluated for the components included in the build.
- Each component is included in the build normally and the CMakeLists.txt file is evaluated again to add the component libraries to the build.
- Order that :ref:`project_include.cmake` files are included into the project.
- Order that the list of header paths is generated for compilation (via ``-I`` argument). (Note that for a given component's source files, only that component's dependency's header paths are passed to the compiler.)
For components that have build requirements which must be evaluated before any component CMakeLists files are evaluated, you can create a file called ``project_include.cmake`` in the component directory. This CMake file is included when ``project.cmake`` is evaluating the entire project.
``project_include.cmake`` files are used inside ESP-IDF, for defining project-wide build features such as ``esptool.py`` command line arguments and the ``bootloader`` "special app".
Unlike component ``CMakeLists.txt`` files, when including a ``project_include.cmake`` file the current source directory (``CMAKE_CURRENT_SOURCE_DIR`` and working directory) is the project directory. Use the variable ``COMPONENT_DIR`` for the absolute directory of the component.
Note that ``project_include.cmake`` isn't necessary for the most common component uses - such as adding include directories to the project, or ``LDFLAGS`` to the final linking step. These values can be customised via the ``CMakeLists.txt`` file itself. See `Optional Project Variables`_ for details.
``project_include.cmake`` files are included in the order given in ``BUILD_COMPONENTS`` variable (as logged by CMake). This means that a component's ``project_include.cmake`` file will be included after it's all dependencies' ``project_include.cmake`` files, unless both components are part of a dependency cycle. This is important if a ``project_include.cmake`` file relies on variables set by another component. See also :ref:`above<component-requirements-implementation>`.
Take great care when setting variables or targets in a ``project_include.cmake`` file. As the values are included into the top-level project CMake pass, they can influence or break functionality across all components!
This is an equivalent to ``project_include.cmake`` for :ref:`component-configuration` KConfig files. If you want to include configuration options at the top-level of menuconfig, rather than inside the "Component Configuration" sub-menu, then these can be defined in the KConfig.projbuild file alongside the ``CMakeLists.txt`` file.
Take care when adding configuration values in this file, as they will be included across the entire project configuration. Where possible, it's generally better to create a KConfig file for :ref:`component-configuration`.
``project_include.cmake`` files are used inside ESP-IDF, for defining project-wide build features such as ``esptool.py`` command line arguments and the ``bootloader`` "special app".
Special components which contain no source files, only ``Kconfig.projbuild`` and ``KConfig``, can have a one-line ``CMakeLists.txt`` file which calls the function ``idf_component_register()`` with no arguments specified. This function will include the component in the project build, but no library will be built *and* no header files will be added to any include paths.
For full details about CMake_ and CMake commands, see the `CMake v3.5 documentation`_.
Some tips for debugging the ESP-IDF CMake-based build system:
- When CMake runs, it prints quite a lot of diagnostic information including lists of components and component paths.
- Running ``cmake -DDEBUG=1`` will produce more verbose diagnostic output from the IDF build system.
- Running ``cmake`` with the ``--trace`` or ``--trace-expand`` options will give a lot of information about control flow. See the `cmake command line documentation`_.
When included from a project CMakeLists file, the ``project.cmake`` file defines some utility modules and global variables and then sets ``IDF_PATH`` if it was not set in the system environment.
It also defines an overridden custom version of the built-in CMake_ ``project`` function. This function is overridden to add all of the ESP-IDF specific project functionality.
.._warn-undefined-variables:
Warning On Undefined Variables
------------------------------
By default, ``idf.py`` passes the ``--warn-uninitialized`` flag to CMake_ so it will print a warning if an undefined variable is referenced in the build. This can be very useful to find buggy CMake files.
If you don't want this behaviour, it can be disabled by passing ``--no-warnings`` to ``idf.py``.
Because the build environment tries to set reasonable defaults that will work most
of the time, component ``CMakeLists.txt`` can be very small or even empty (see `Minimal Component CMakeLists`_). However, overriding `component variables`_ is usually required for some functionality.
Note that component dependencies may depend on ``IDF_TARGET`` variable, but not on Kconfig variables. Also one can not use Kconfig variables in ``include`` statements in CMake files, but ``IDF_TARGET`` can be used in such context.
Some components will have a situation where a source file isn't supplied with the component itself but has to be generated from another file. Say our component has a header file that consists of the converted binary data of a BMP file, converted using a hypothetical tool called bmp2h. The header file is then included in as C source file called graphics_lib.c::
This answer is adapted from the `CMake FAQ entry <cmake faq generated files_>`_, which contains some other examples that will also work with ESP-IDF builds.
In this example, logo.h will be generated in the current directory (the build directory) while logo.bmp comes with the component and resides under the component path. Because logo.h is a generated file, it should be cleaned when the project is cleaned. For this reason it is added to the `ADDITIONAL_MAKE_CLEAN_FILES`_ property.
If generating files as part of the project CMakeLists.txt file, not a component CMakeLists.txt, then use build property ``PROJECT_DIR`` instead of ``${COMPONENT_DIR}`` and ``${PROJECT_NAME}.elf`` instead of ``${COMPONENT_LIB}``.)
If a a source file from another component included ``logo.h``, then ``add_dependencies`` would need to be called to add a dependency between the two components, to ensure that the component source files were always compiled in the correct order.
Sometimes you have a file with some binary or text data that you'd like to make available to your component - but you don't want to reformat the file as C source.
The names are generated from the full name of the file, as given in ``EMBED_FILES``. Characters /, ., etc. are replaced with underscores. The _binary prefix in the symbol name is added by objcopy and is the same for both text and binary files.
Place this line after the ``project()`` line in your project CMakeLists.txt file. Replace ``myproject.elf`` with your project name. The final argument can be ``TEXT`` to embed a null-terminated string, or ``BINARY`` to embed the content as-is.
For an example of using this technique, see the "main" component of the file_serving example :example_file:`protocols/http_server/file_serving/main/CMakeLists.txt` - two files are loaded at build time and linked into the firmware.
In the example above, ``my_processed_file.bin`` is generated from ``my_unprocessed_file.bin`` through some command ``my_process_file_cmd``, then embedded into the target.
To specify a dependence on a target, use the ``DEPENDS`` argument::
ESP-IDF has a feature called linker script generation that enables components to define where its code and data will be placed in memory through linker fragment files. These files are processed by the build system, and is used to augment the linker script used for linking app binary. See :doc:`Linker Script Generation <linker-script-generation>` for a quick start guide as well as a detailed discussion of the mechanism.
Obviously, there are cases where all these recipes are insufficient for a certain component, for example when the component is basically a wrapper around another third-party component not originally intended to be compiled under this build system. In that case, it's possible to forego the ESP-IDF build system entirely by using a CMake feature called ExternalProject_. Example component CMakeLists::
(The above CMakeLists.txt can be used to create a component named ``quirc`` that builds the quirc_ project using its own Makefile.)
-``externalproject_add`` defines an external build system.
-``SOURCE_DIR``, ``CONFIGURE_COMMAND``, ``BUILD_COMMAND`` and ``INSTALL_COMMAND`` should always be set. ``CONFIGURE_COMMAND`` can be set to an empty string if the build system has no "configure" step. ``INSTALL_COMMAND`` will generally be empty for ESP-IDF builds.
- Setting ``BUILD_IN_SOURCE`` means the build directory is the same as the source directory. Otherwise you can set ``BUILD_DIR``.
- Consult the ExternalProject_ documentation for more details about ``externalproject_add()``
- The second set of commands adds a library target, which points to the "imported" library file built by the external system. Some properties need to be set in order to add include directories and tell CMake where this file is.
- Finally, the generated library is added to `ADDITIONAL_MAKE_CLEAN_FILES`_. This means ``make clean`` will delete this library. (Note that the other object files from the build won't be deleted.)
..note:: When using an external build process with PSRAM, remember to add ``-mfix-esp32-psram-cache-issue`` to the C compiler arguments. See :ref:`CONFIG_SPIRAM_CACHE_WORKAROUND` for details of this flag.
CMake has some unusual behaviour around external project builds:
-`ADDITIONAL_MAKE_CLEAN_FILES`_ only works when "make" is used as the build system. If Ninja_ or an IDE build system is used, it won't delete these files when cleaning.
- However, the ExternalProject_ configure & build commands will *always* be re-run after a clean is run.
- Therefore, there are two alternative recommended ways to configure the external build command:
1. Have the external ``BUILD_COMMAND`` run a full clean compile of all sources. The build command will be run if any of the dependencies passed to ``externalproject_add`` with ``DEPENDS`` have changed, or if this is a clean build (ie any of ``idf.py clean``, ``ninja clean``, or ``make clean`` was run.)
2. Have the external ``BUILD_COMMAND`` be an incremental build command. Pass the parameter ``BUILD_ALWAYS 1`` to ``externalproject_add``. This means the external project will be built each time a build is run, regardless of dependencies. This is only recommended if the external project has correct incremental build behaviour, and doesn't take too long to run.
The best of these approaches for building an external project will depend on the project itself, its build system, and whether you anticipate needing to frequently recompile the project.
For example projects or other projects where you don't want to specify a full sdkconfig configuration, but you do want to override some key values from the ESP-IDF defaults, it is possible to create a file ``sdkconfig.defaults`` in the project directory. This file will be used when creating a new config from scratch, or when any new config value hasn't yet been set in the ``sdkconfig`` file.
To override the name of this file or to specify multiple files, set the ``SDKCONFIG_DEFAULTS`` environment variable or set ``SDKCONFIG_DEFAULTS`` in top-level CMakeLists.txt. If specifying multiple files, use semicolon as the list separator. File names not specified as full paths are resolved relative to current project.
Some of the IDF examples include a ``sdkconfig.ci`` file. This is part of the continuous integration (CI) test framework and is ignored by the normal build process.
In addition to ``sdkconfig.defaults`` file, build system will also load defaults from ``sdkconfig.defaults.TARGET_NAME`` file, where ``TARGET_NAME`` is the value of ``IDF_TARGET``. For example, for ``esp32`` target, default settings will be taken from ``sdkconfig.defaults`` first, and then from ``sdkconfig.defaults.esp32``.
If ``SDKCONFIG_DEFAULTS`` is used to override the name of defaults file/files, the name of target-specific defaults file will be derived from ``SDKCONFIG_DEFAULTS`` value/values using the rule above.
There are some scenarios that we want to flash the target board without IDF. For this case we want to save the built binaries, esptool.py and esptool write_flash arguments. It's simple to write a script to save binaries and esptool.py.
After running a project build, the build directory contains binary output files (``.bin`` files) for the project and also the following flashing data files:
Alternatively, it is possible to manually copy the parameters from the argument file and pass them on the command line.
The build directory also contains a generated file ``flasher_args.json`` which contains project flash information, in JSON format. This file is used by ``idf.py`` and can also be used by other tools which need information about the project build.
The bootloader is built by default as part of ``idf.py build``, or can be built standalone via ``idf.py bootloader``.
The bootloader is a special "subproject" inside :idf:`/components/bootloader/subproject`. It has its own project CMakeLists.txt file and builds separate .ELF and .BIN files to the main project. However it shares its configuration and build directory with the main project.
The subproject is inserted as an external project from the top-level project, by the file :idf_file:`/components/bootloader/project_include.cmake`. The main build process runs CMake for the subproject, which includes discovering components (a subset of the main components) and generating a bootloader-specific config (derived from the main ``sdkconfig``).
To select the target before building the project, use ``idf.py set-target <target>`` command, for example::
idf.py set-target esp32s2
..important::
``idf.py set-target`` will clear the build directory and re-generate the ``sdkconfig`` file from scratch. The old ``sdkconfig`` file will be saved as ``sdkconfig.old``.
..note::
The behavior of ``idf.py set-target`` command is equivalent to:
1. clearing the build directory (``idf.py fullclean``)
2. removing the sdkconfig file (``mv sdkconfig sdkconfig.old``)
3. configuring the project with the new target (``idf.py -DIDF_TARGET=esp32 reconfigure``)
It is also possible to pass the desired ``IDF_TARGET`` as an environment variable (e.g. ``export IDF_TARGET=esp32s2``) or as a CMake variable (e.g. ``-DIDF_TARGET=esp32s2`` argument to CMake or idf.py). Setting the environment variable is a convenient method if you mostly work with one type of the chip.
To specify the _default_ value of ``IDF_TARGET`` for a given project, add ``CONFIG_IDF_TARGET`` value to ``sdkconfig.defaults``. For example, ``CONFIG_IDF_TARGET="esp32s2"``. This value will be used if ``IDF_TARGET`` is not specified by other method: using an environment variable, CMake variable, or ``idf.py set-target`` command.
The ESP-IDF build system "wraps" CMake with the concept of "components", and helper functions to automatically integrate these components into a project build.
However, underneath the concept of "components" is a full CMake build system. It is also possible to make a component which is pure CMake.
..highlight:: cmake
Here is an example minimal "pure CMake" component CMakeLists file for a component named ``json``::
add_library(json STATIC
cJSON/cJSON.c
cJSON/cJSON_Utils.c)
target_include_directories(json PUBLIC cJSON)
- This is actually an equivalent declaration to the IDF ``json`` component :idf_file:`/components/json/CMakeLists.txt`.
- This file is quite simple as there are not a lot of source files. For components with a large number of files, the globbing behaviour of ESP-IDF's component logic can make the component CMakeLists style simpler.)
- Any time a component adds a library target with the component name, the ESP-IDF build system will automatically add this to the build, expose public include directories, etc. If a component wants to add a library target with a different name, dependencies will need to be added manually via CMake commands.
CMake is used for a lot of open-source C and C++ projects — code that users can tap into for their applications. One of the benefits of having a CMake build system is the ability to import these third-party projects, sometimes even without modification! This allows for users to be able to get functionality that may not yet be provided by a component, or use another library for the same functionality.
For an actual example, take a look at :example:`build_system/cmake/import_lib`. Take note that what needs to be done in order to import the library may vary. It is recommended to read up on the library's documentation for instructions on how to import it from other projects. Studying the library's CMakeLists.txt and build structure can also be helpful.
It is also possible to wrap a third-party library to be used as a component in this manner. For example, the :component:`mbedtls` component is a wrapper for Espressif's fork of `mbedtls <https://github.com/ARMmbed/mbedtls>`_. See its :component_file:`component CMakeLists.txt <mbedtls/CMakeLists.txt>`.
The CMake variable ``ESP_PLATFORM`` is set to 1 whenever the ESP-IDF build system is being used. Tests such as ``if (ESP_PLATFORM)`` can be used in generic CMake code if special IDF-specific logic is required.
The above example assumes that the external library ``foo`` (or ``tinyxml`` in the case of the ``import_lib`` example) doesn't need to use any ESP-IDF APIs apart from common APIs such as libc, libstdc++, etc. If the external library needs to use APIs provided by other ESP-IDF components, this needs to be specified in the external CMakeLists.txt file by adding a dependency on the library target ``idf::<componentname>``.
-``target_name``- name that can be used to reference the imported library, such as when linking to other targets
-``lib_path``- path to prebuilt library; may be an absolute or relative path to the component directory
Optional arguments ``REQUIRES`` and ``PRIV_REQUIRES`` specify dependency on other components. These have the same meaning as the arguments for ``idf_component_register``.
Take note that the prebuilt library must have been compiled for the same target as the consuming project. Configuration relevant to the prebuilt library must also match. If not paid attention to, these two factors may contribute to subtle bugs in the app.
ESP-IDF provides a template CMake project for easily creating an application. However, in some instances the user might already have an existing CMake project or may want to create a custom one. In these cases it is desirable to be able to consume IDF components as libraries to be linked to the user's targets (libraries/ executables).
The example in :example:`build_system/cmake/idf_as_lib` demonstrates the creation of an application equivalent to :example:`hello world application <get-started/hello_world>` using a custom CMake project.
..note:: The IDF build system can only set compiler flags for source files that it builds. When an external CMakeLists.txt file is used and PSRAM is enabled, remember to add ``-mfix-esp32-psram-cache-issue`` to the C compiler arguments. See :ref:`CONFIG_SPIRAM_CACHE_WORKAROUND` for details of this flag.
Retrieve a :ref:`build property<cmake-build-properties>`*property* and store it in *var* accessible from the current scope. Specifying *GENERATOR_EXPRESSION* will retrieve the generator expression string for that property, instead of the actual value, which can be used with CMake commands that support generator expressions.
Set a :ref:`build property<cmake-build-properties>`*property* with value *val*. Specifying *APPEND* will append the specified value to the current value of the property. If the property does not previously exist or it is currently empty, the specified value becomes the first element/member instead.
Present a directory *component_dir* that contains a component to the build system. Relative paths are converted to absolute paths with respect to current directory.
All calls to this command must be performed before `idf_build_process`.
This command does not guarantee that the component will be processed during build (see the `COMPONENTS` argument description for `idf_build_process`)
Performs the bulk of the behind-the-scenes magic for including ESP-IDF components such as component configuration, libraries creation, dependency expansion and resolution. Among these functions, perhaps the most important from a user's perspective is the libraries creation by calling each component's ``idf_component_register``. This command creates the libraries for each component, which are accessible using aliases in the form idf::*component_name*.
These aliases can be used to link the components to the user's own targets, either libraries or executables.
- SDKCONFIG_DEFAULTS - list of files containing default config to use in the build (list must contain full paths); defaults to empty. For each value *filename* in the list, the config from file *filename.target*, if it exists, is also loaded.
- BUILD_DIR - directory to place ESP-IDF build-related artifacts, such as generated binaries, text files, components; defaults to CMAKE_BINARY_DIR
- COMPONENTS - select components to process among the components known by the build system (added via `idf_build_component`). This argument is used to trim the build.
Other components are automatically added if they are required in the dependency chain, i.e. the public and private requirements of the components in this list are automatically added, and in turn the public and private requirements of those requirements, so on and so forth. If not specified, all components known to the build system are processed.
Specify the executable *executable* for ESP-IDF build. This attaches additional targets such as dependencies related to flashing, generating additional binary files, etc. Should be called after ``idf_build_process``.
Get the value of the specified config. Much like build properties, specifying *GENERATOR_EXPRESSION* will retrieve the generator expression string for that config, instead of the actual value, which can be used with CMake commands that support generator expressions. Actual config values are only known after call to ``idf_build_process``, however.
These are properties that describe the build. Values of build properties can be retrieved by using the build command ``idf_build_get_property``. For example, to get the Python interpreter used for the build:
Retrieve a specified *component*'s :ref:`component property<cmake-component-properties>`, *property* and store it in *var* accessible from the current scope. Specifying *GENERATOR_EXPRESSION* will retrieve the generator expression string for that property, instead of the actual value, which can be used with CMake commands that support generator expressions.
Set a specified *component*'s :ref:`component property<cmake-component-properties>`, *property* with value *val*. Specifying *APPEND* will append the specified value to the current value of the property. If the property does not previously exist or it is currently empty, the specified value becomes the first element/member instead.
Register a component to the build system. Much like the ``project()`` CMake command, this should be called from the component's CMakeLists.txt directly (not through a function or macro) and is recommended to be called before any other command. Here are some guidelines on what commands can **not** be called before ``idf_component_register``:
- SRCS - component source files used for creating a static library for the component; if not specified, component is a treated as a config-only component and an interface library is created instead.
- SRC_DIRS, EXCLUDE_SRCS - used to glob source files (.c, .cpp, .S) by specifying directories, instead of specifying source files manually via SRCS. Note that this is subject to the :ref:`limitations of globbing in CMake<cmake-file-globbing>`. Source files specified in EXCLUDE_SRCS are removed from the globbed files.
- INCLUDE_DIRS - paths, relative to the component directory, which will be added to the include search path for all other components which require the current component
- PRIV_INCLUDE_DIRS - directory paths, must be relative to the component directory, which will be added to the include search path for this component's source files only
- REQUIRES - public component requirements for the component
- PRIV_REQUIRES - private component requirements for the component; ignored on config-only components
- LDFRAGMENTS - component linker fragment files
- REQUIRED_IDF_TARGETS - specify the only target the component supports
The following are used for :ref:`embedding data into the component<cmake_embed_data>`, and is considered as source files when determining if a component is config-only. This means that even if the component does not specify source files, a static library is still created internally for the component if it specifies either:
These are properties that describe a component. Values of component properties can be retrieved by using the build command ``idf_component_get_property``. For example, to get the directory of the ``freertos`` component:
- COMPONENT_ALIAS - alias for COMPONENT_LIB used for linking the component to external targets; set by ``idf_build_component`` and alias library itself is created by ``idf_component_register``
- COMPONENT_OVERRIDEN_DIR - contains the directory of the original component if :ref:`this component overrides another component<cmake-components-same-name>`
- COMPONENT_LIB - name for created component static/interface library; set by ``idf_build_component`` and library itself is created by ``idf_component_register``
- EMBED_FILES - list of files to embed in component; set from ``idf_component_register`` EMBED_FILES argument
- EMBED_TXTFILES - list of text files to embed in component; set from ``idf_component_register`` EMBED_TXTFILES argument
- INCLUDE_DIRS - list of component include directories; set from ``idf_component_register`` INCLUDE_DIRS argument
- KCONFIG - component Kconfig file; set by ``idf_build_component``
- KCONFIG_PROJBUILD - component Kconfig.projbuild; set by ``idf_build_component``
- LDFRAGMENTS - list of component linker fragment files; set from ``idf_component_register`` LDFRAGMENTS argument
- PRIV_INCLUDE_DIRS - list of component private include directories; set from ``idf_component_register`` PRIV_INCLUDE_DIRS on components of type LIBRARY
- PRIV_REQUIRES - list of private component dependentices; set from ``idf_component_register`` PRIV_REQUIRES argument
- REQUIRED_IDF_TARGETS - list of targets the component supports; set from ``idf_component_register`` EMBED_TXTFILES argument
- REQUIRES - list of public component dependencies; set from ``idf_component_register`` REQUIRES argument
- SRCS - list of component source files; set from SRCS or SRC_DIRS/EXCLUDE_SRCS argument of ``idf_component_register``
.._cmake-file-globbing:
File Globbing & Incremental Builds
==================================
..highlight:: cmake
The preferred way to include source files in an ESP-IDF component is to list them manually via SRCS argument to ``idf_component_register``::
This preference reflects the `CMake best practice <https://gist.github.com/mbinna/c61dbb39bca0e4fb7d1f73b0d66a4fd1/>`_ of manually listing source files. This could, however, be inconvenient when there are lots of source files to add to the build. The ESP-IDF build system provides an alternative way for specifying source files using ``SRC_DIRS``::
idf_component_register(SRC_DIRS library platform
...)
This uses globbing behind the scenes to find source files in the specified directories. Be aware, however, that if a new source file is added and this method is used, then CMake won't know to automatically re-run and this file won't be added to the build.
The trade-off is acceptable when you're adding the file yourself, because you can trigger a clean build or run ``idf.py reconfigure`` to manually re-run CMake_. However, the problem gets harder when you share your project with others who may check out a new version using a source control tool like Git...
For components which are part of ESP-IDF, we use a third party Git CMake integration module (:idf_file:`/tools/cmake/third_party/GetGitRevisionDescription.cmake`) which automatically re-runs CMake any time the repository commit changes. This means if you check out a new ESP-IDF version, CMake will automatically rerun.
For project components (not part of ESP-IDF), there are a few different options:
- If keeping your project file in Git, ESP-IDF will automatically track the Git revision and re-run CMake if the revision changes.
- If some components are kept in a third git repository (not the project repository or ESP-IDF repository), you can add a call to the ``git_describe`` function in a component CMakeLists file in order to automatically trigger re-runs of CMake when the Git revision changes.
- If not using Git, remember to manually run ``idf.py reconfigure`` whenever a source file may change.
For integration into IDEs and other build systems, when CMake runs the build process generates a number of metadata files in the ``build/`` directory. To regenerate these files, run ``cmake`` or ``idf.py reconfigure`` (or any other ``idf.py`` build command).
-``compile_commands.json`` is a standard format JSON file which describes every source file which is compiled in the project. A CMake feature generates this file, and many IDEs know how to parse it.
-``project_description.json`` contains some general information about the ESP-IDF project, configured paths, etc.
-``flasher_args.json`` contains esptool.py arguments to flash the project's binary files. There are also ``flash_*_args`` files which can be used directly with esptool.py. See `Flash arguments`_.
-``CMakeCache.txt`` is the CMake cache file which contains other information about the CMake process, toolchain, etc.
-``config/sdkconfig.json`` is a JSON-formatted version of the project configuration values.
-``config/kconfig_menus.json`` is a JSON-formatted version of the menus shown in menuconfig, for use in external IDE UIs.
JSON Configuration Server
-------------------------
.. highlight :: json
A tool called ``confserver.py`` is provided to allow IDEs to easily integrate with the configuration system logic. ``confserver.py`` is designed to run in the background and interact with a calling process by reading and writing JSON over process stdin & stdout.
You can run ``confserver.py`` from a project via ``idf.py confserver`` or ``ninja confserver``, or a similar target triggered from a different build generator.
Aside from these files, there are two other important CMake scripts in :idf:`/tools/cmake`:
- idf.cmake - Sets up the build and includes the core modules listed above. Included in CMake projects in order to access ESP-IDF build system functionality.
- project.cmake - Includes ``idf.cmake`` and provides a custom ``project()`` command that takes care of all the heavy lifting of building an executable. Included in the top-level CMakeLists.txt of standard ESP-IDF projects.
The rest of the files in :idf:`/tools/cmake` are support or third-party scripts used in the build process.
Build Process
-------------
This section describes the standard ESP-IDF application build process. The build process can be broken down roughly into four phases:
- Get ESP-IDF git revision and store as ``IDF_VER``.
- Set global build specifications i.e. compile options, compile definitions, include directories for all components in the build.
- Add components in :idf:`components` to the build.
- The initial part of the custom ``project()`` command performs the following steps:
- Set ``IDF_TARGET`` from environment variable or CMake cache and the corresponding ``CMAKE_TOOLCHAIN_FILE`` to be used.
- Add components in ``EXTRA_COMPONENTS_DIRS`` to the build.
- Prepare arguments for calling command ``idf_build_process()`` from variables such as ``COMPONENTS``/``EXCLUDE_COMPONENTS``, ``SDKCONFIG``, ``SDKCONFIG_DEFAULTS``.
The call to ``idf_build_process()`` command marks the end of this phase.
Enumeration
^^^^^^^^^^^
This phase builds a final list of components to be processed in the build, and is performed in the first half of ``idf_build_process()``.
- Retrieve each component's public and private requirements. A child process is created which executes each component's CMakeLists.txt in script mode. The values of ``idf_component_register`` REQUIRES and PRIV_REQUIRES argument is returned to the parent build process. This is called early expansion. The variable ``CMAKE_BUILD_EARLY_EXPANSION`` is defined during this step.
- Recursively include components based on public and private requirements.
This phase processes the components in the build, and is the second half of ``idf_build_process()``.
- Load project configuration from sdkconfig file and generate an sdkconfig.cmake and sdkconfig.h header. These define configuration variables/macros that are accessible from the build scripts and C/C++ source/header files, respectively.
- Include each component's ``project_include.cmake``.
- Add each component as a subdirectory, processing its CMakeLists.txt. The component CMakeLists.txt calls the registration command, ``idf_component_register`` which adds source files, include directories, creates component library, links dependencies, etc.
Finalization
^^^^^^^^^^^^
This phase is everything after ``idf_build_process()``.
- Create executable and link the component libraries to it.
- Generate project metadata files such as project_description.json and display relevant information about the project built.
Browse :idf_file:`/tools/cmake/project.cmake` for more details.
Some aspects of the CMake-based ESP-IDF build system are very similar to the older GNU Make-based system. The developer needs to provide values the include directories, source files etc. There is a syntactical difference, however, as the developer needs to pass these as arguments to the registration command, ``idf_component_register``.
An automatic project conversion tool is available in :idf_file:`/tools/cmake/convert_to_cmake.py`. Run this command line tool with the path to a project like this::
The project directory must contain a Makefile, and GNU Make (``make``) must be installed and available on the PATH.
The tool will convert the project Makefile and any component ``component.mk`` files to their equivalent ``CMakeLists.txt`` files.
It does so by running ``make`` to expand the ESP-IDF build system variables which are set by the build, and then producing equivalent CMakelists files to set the same variables.
..important:: When the conversion tool converts a ``component.mk`` file, it doesn't determine what other components that component depends on. This information needs to be added manually by editing the new component ``CMakeLists.txt`` file and adding ``REQUIRES`` and/or ``PRIV_REQUIRES`` clauses. Otherwise, source files in the component will fail to compile as headers from other components are not found. See :ref:`component requirements`.
The conversion tool is not capable of dealing with complex Makefile logic or unusual targets. These will need to be converted by hand.
No Longer Available in CMake
----------------------------
Some features are significantly different or removed in the CMake-based system. The following variables no longer exist in the CMake-based build system:
-``COMPONENT_BUILD_DIR``: Use ``CMAKE_CURRENT_BINARY_DIR`` instead.
-``COMPONENT_LIBRARY``: Defaulted to ``$(COMPONENT_NAME).a``, but the library name could be overriden by the component. The name of the component library can no longer be overriden by the component.
-``CC``, ``LD``, ``AR``, ``OBJCOPY``: Full paths to each tool from the gcc xtensa cross-toolchain. Use ``CMAKE_C_COMPILER``, ``CMAKE_C_LINK_EXECUTABLE``, ``CMAKE_OBJCOPY``, etc instead. `Full list here <cmake language variables_>`_.
-``HOSTCC``, ``HOSTLD``, ``HOSTAR``: Full names of each tool from the host native toolchain. These are no longer provided, external projects should detect any required host toolchain manually.
-``COMPONENT_ADD_LDFLAGS``: Used to override linker flags. Use the CMake `target_link_libraries`_ command instead.
-``COMPONENT_ADD_LINKER_DEPS``: List of files that linking should depend on. `target_link_libraries`_ will usually infer these dependencies automatically. For linker scripts, use the provided custom CMake function ``target_linker_scripts``.
-``COMPONENT_SUBMODULES``: No longer used, the build system will automatically enumerate all submodules in the ESP-IDF repository.
-``COMPONENT_EXTRA_INCLUDES``: Used to be an alternative to ``COMPONENT_PRIV_INCLUDEDIRS`` for absolute paths. Use ``PRIV_INCLUDE_DIRS`` argument to ``idf_component_register`` for all cases now (can be relative or absolute).
-``COMPONENT_OBJS``: Previously, component sources could be specified as a list of object files. Now they can be specified as a list of source files via ``SRCS`` argument to `idf_component_register`.
-``COMPONENT_OBJEXCLUDE``: Has been replaced with ``EXCLUDE_SRCS`` argument to ``idf_component_register``. Specify source files (as absolute paths or relative to component directory), instead.
-``COMPONENT_EXTRA_CLEAN``: Set property ``ADDITIONAL_MAKE_CLEAN_FILES`` instead but note :ref:`CMake has some restrictions around this functionality <ADDITIONAL_MAKE_CLEAN_FILES_note>`.
-``COMPONENT_OWNBUILDTARGET`` & ``COMPONENT_OWNCLEANTARGET``: Use CMake `ExternalProject`_ instead. See :ref:`component-build-full-override` for full details.
- In the legacy Make-based build system, it is required to also set ``COMPONENT_SRCDIRS`` if ``COMPONENT_SRCS`` is set. In CMake, the equivalent is not necessary i.e. specifying ``SRC_DIRS`` to ``idf_component_register`` if ``SRCS`` is also specified (in fact, ``SRCS`` is ignored if ``SRC_DIRS`` is specified).
``make flash`` and similar targets still work to build and flash. However, project ``sdkconfig`` no longer specifies serial port and baud rate. Environment variables can be used to override these. See :ref:`flash-with-ninja-or-make` for more details.
