stm32pio 2.0.0

Small cross-platform Python app that can create and update PlatformIO projects from STM32CubeMX .ioc files. It uses STM32CubeMX to generate a HAL-framework-based code and alongside creates PlatformIO project with compatible parameters to stick them both together

stm32pio

Small cross-platform Python app that can create and update PlatformIO projects from the STM32CubeMX .ioc files.

It uses STM32CubeMX to generate a HAL-framework-based code and alongside creates PlatformIO project with the compatible parameters to stick them both together.

The GUI version is available, too.

Table of contents

• Features
• Requirements
• Installation
• Usage
  • Project patching
  • Embedding
• Example
• Build
• Test
• CI
• Restrictions

Features

• Start the new complete project in a single directory using only an .ioc file
• Update an existing project after changing a hardware options in CubeMX
• Clean-up the project
• Get the status information
• [optional] Automatically run your favorite editor in the end
• [optional] Automatically make an initial build of the project
• [optional] GUI version (see the dedicated README file)

Requirements:

The app presents zero dependencies by itself and requires only this to run:

• macOS, Linux, Windows
• Python 3.6 and above

Of course, you need to have all the necessary tools to be installed in order to work with them:

• STM32CubeMX with the desired downloaded frameworks (F0, F1, etc.). All recent versions are fine, probably something like 5.0+
• Java CLI (JRE, Java runtime environment) for the CubeMX (likely is already installed if the CubeMX is working). In other words, you should be able to launch the Java from your terminal using java command. Which version is appropriate for the CubeMX you can find in its docs
• PlatformIO (4.2.0 and above) CLI (already present if you have installed PlatformIO via some package manager (pip, apt, brew, etc.) or need to be installed as the "command line extension" from IDE (see docs for more information))

If you for some reasons don't want to or cannot install command line versions of these applications in your system you can always specify the direct paths to them using the project configuration file stm32pio.ini or even override the default values in the source code file core/settings.py.

Also, some external dependencies may be required to build, test and pack the app. See the corresponding sections for more information.

A general recommendation there would be to test both CubeMX (code generation) and PlatformIO (project creation, building) at least once before using the stm32pio to make sure all the tools work properly even without any "glue".

2020 Update: in my tests, recent versions of CubeMX had been shipping with no installer of some sort. So I just unpack the distribution archive and place it in ~/cubemx directory so it can be started simply as java -jar ~/cubemx/STM32CubeMX.exe. That's why the default path is as mentioned above. Tell me if your default path is rather another. Also, the default structure of the generated code is significantly different when you invoke the generation from the GUI version of CubeMX or the CLI one. As stm32pio uses the latter, currently the PlatformIO project structure cannot be properly patched to use a code from the GUI version of CubeMX (at least with the default patch config, you can always tweak it in a configuration file stm32pio.ini).

Installation

As a normal Python package the app can be run in a completely portable way by downloading or cloning the snapshot of this repository and invoking the main script (or the Python module):

stm32pio-repo/ $   python3 stm32pio/cli/app.py  # or
stm32pio-repo/ $   python3 -m stm32pio.cli  # or
any-path/ $   python3 path/to/stm32pio-repo/stm32pio/cli/app.py

(we assume python3 and pip3 hereinafter).

However, it's handier to install the utility to be able to run stm32pio from anywhere. The PyPI distribution (starting from v0.95) is available:

$ pip install stm32pio

To uninstall run

$ pip uninstall stm32pio

Usage

Basically, you need to follow such a pattern:

1. Create CubeMX project (.ioc file), set-up your hardware configuration, save with the compatible parameters
2. Run the stm32pio that automatically invokes CubeMX to generate the code, creates PlatformIO project, patches a platformio.ini file and so on
3. Work with your project normally as you wish, compile/upload/debug etc.
4. Come back to the hardware configuration in CubeMX when necessary, then run stm32pio to re-generate the code

Refer to Example section on more detailed steps. If you face off with some error try to enable a verbose output to get more information about a problem:

$ stm32pio -v [command] [options]

On the first run stm32pio will create a config file stm32pio.ini, syntax of which is similar to the platformio.ini. You can also create this config without any following operations by initializing the project:

$ stm32pio init -d path/to/project

It may be useful to tweak some parameters before proceeding. The structure of the config is separated into two sections: app and project. Options of the first one is related to the global settings (such as commands to invoke different instruments) though they can be adjusted on the per-project base while the second section contains of project-related parameters. See comments in the settings.py file for parameters description.

You can always run

$ stm32pio --help

to see help on available commands. Find the copy of its output on the project wiki page, also.

Project patching

Note, that the patch operation (which takes the CubeMX code and PlatformIO project to the compliance) erases all the comments (lines starting with ;) inside the platformio.ini file. They are not required anyway, in general, but if you need them for some reason please consider saving the information somewhere else.

For those who wants to modify the patch (default one is at settings.py, project one in a config file stm32pio.ini): it has a general-form .INI-style content so it's possible to specify several sections and apply composite patches. This works totally fine for the most cases except, perhaps, some really big complex patches involving, say, the parameters interpolation feature. It is turned off for both platformio.ini and user's patch parsing by default. If there are some problems you've met due to a such behavior please modify the source code to match the parameters interpolation kind for the configs you need to. Seems like platformio.ini uses ExtendedInterpolation for its needs, by the way.

Embedding

You can also use stm32pio as an ordinary Python package and embed it in your own application. Find the minimal example at the examples to see some possible ways of implementing this. Basically, you need to import stm32pio.core.lib module (where the main Stm32pio class resides), optionally set up a logger and you are good to go. If you prefer higher-level API similar to the CLI version, use main() function in cli/app.py passing the same CLI arguments to it. Also, take a look at the CLI (app.py) or GUI versions.

Example

1. Run CubeMX, choose MCU/board, do all necessary tweaking
2. Select Project Manager -> Project tab, specify "Project Name", choose "Other Toolchains (GPDSC)". In Code Generator tab check "Copy only the necessary library files" and "Generate periphery initialization as a pair of '.c/.h' files per peripheral" options

3. Back in the first tab (Project) copy the "Toolchain Folder Location" string (you may not be able to copy it in modern CubeMX versions so use a terminal or a file manager to do this). Save the project

4. Use a copied string (project folder) as a -d argument for stm32pio (can be omitted if your current working directory is already a project directory).
5. Run platformio boards (pio boards) or go to boards to list all supported devices. Pick one and use its ID as a -b argument (for example, nucleo_f031k6)
6. All done! You can now run

   $ stm32pio new -d path/to/cubemx/project/ -b nucleo_f031k6 --start-editor=code --with-build
   

   to trigger the code generation, compile the project and start the VSCode editor with opened folder (last 2 options are given as an example and they are not required). Make sure you have all the tools in PATH (java (or set its path in stm32pio.ini), platformio, python, editor). You can use a slightly shorter form if you are already located in the project directory:

   path/to/cubemx/project/ $   stm32pio new -b nucleo_f031k6
   

7. To get the information about the current state of the project use status command.
8. If you will be in need to update the hardware configuration in a future, make all the necessary stuff in CubeMX and run generate command in a similar way:

   $ stm32pio generate -d /path/to/cubemx/project
   

9. To clean-up the folder and keep only the .ioc file run clean command.

Build

Staring from v2.0.0 PEP517-compatible build process is supported. For the build dependencies list see pyproject.toml file:

$ pip install wheel
$ pip install setuptools setuptools_scm

This process yet described and standardized in PEPs is still early and not fully implemented by the different tools (such as pip, twine) though (at the time of this version at least). So the current way is a little bit messy and depends on different instruments. Use the latest Python and build packages versions. To build a wheel setup.py is not even required:

$ pip wheel . --wheel-dir dist

but for the source distribution tarball it is still necessary:

$ python setup.py sdist

Test

There are some tests in tests directory (based on the unittest module). The test stage is a CubeMX project (.ioc file) with an optional platformio.ini.lockfile config specifying the versions of the used ("locked") PlatformIO libraries (see "CI" chapter) (this config is an ordinary .INI-style file which will be merged with the platformio.ini test file). Several such targets can be placed to the fixtures folder to test against. Finally, run this command setting the current case as an environment variable

stm32pio-repo/ $   STM32PIO_TEST_CASE=nucleo_f031k6 python -m unittest -b -v

to test the stm32pio. Tests code automatically create temporary directory (using tempfile Python standard module) where all actions are performed.

To run the specific group of tests or a single test function you can use:

stm32pio-repo/ $   python -m unittest tests.test_integration.TestIntegration
stm32pio-repo/ $   python -m unittest tests.test_cli.TestCLI.test_verbosity

Automated tests against old .ioc files seems not possible because of CubeMX interactive GUI prompts about migrations. I don't find any CLI key or option to disable them (and in general, documentation for the CubeMX CLI doesn't look complete...).

CI

Azure Pipelines is used to automate test, build, and publish tasks. The repo is tested against all 3 major OSes and for the Linux the test coverage is also calculated. For this purposes some additional external dependencies are necessary, such as

• pytest
• coverage
• yaml

There are some elements of the "reproducible builds" approach using several "lockfiles", isolated test fixtures and caching. Actually, due to a number of different tools in use and their nature the truly and fully "canonical" CI seems challenging to implement so the current system is far from ideal. It probably will be improved in the future, see azure-pipelines.yml, CI for more information for now.

Restrictions

• The tool doesn't check for different parameters compatibility, e.g. CPU frequency, memory sizes and so on. It simply eases your workflow with these 2 programs (PlatformIO and STM32CubeMX) a little bit.
• CubeMX middlewares are not supported yet because it's hard to be prepared for every possible configuration. You need to manually adjust them to build appropriately. For example, FreeRTOS can be added via PlatformIO' lib feature or be directly compiled in its own directory using lib_extra_dirs option:

  lib_extra_dirs = Middlewares/Third_Party/FreeRTOS
  

  You also need to move all .c/.h files to the appropriate folders respectively. See PlatformIO documentation for more information.