ESP-IDF Tests with Pytest Guide

This documentation is a guide that introduces the following aspects:

The basic idea of different test types in ESP-IDF
How to apply the pytest framework to the test python scripts to make sure the apps are working as expected.
ESP-IDF CI target test process
Run ESP-IDF tests with pytest locally
Tips and tricks on pytest

Disclaimer

In ESP-IDF, we use the following plugins by default:

pytest-embedded with default services esp,idf
pytest-rerunfailures

All the introduced concepts and usages are based on the default behavior in ESP-IDF. Not all of them are available in vanilla pytest.

Installation

All dependencies could be installed by running the install script with the --enable-pytest argument, e.g. $ install.sh --enable-pytest.

Common Issues During Installation

No Package 'dbus-1' found

If you're facing an error message like:

configure: error: Package requirements (dbus-1 >= 1.8) were not met:

No package 'dbus-1' found

Consider adjusting the PKG_CONFIG_PATH environment variable if you
installed software in a non-standard prefix.

If you're running a ubuntu system, you may need to run:

sudo apt-get install libdbus-glib-1-dev

or

sudo apt-get install libdbus-1-dev

For other linux distros, you may Google the error message and find the solution. This issue could be solved by installing the related header files.

Invalid command 'bdist_wheel'

If you're facing an error message like:

error: invalid command 'bdist_wheel'

You may need to run:

python -m pip install -U pip

Or

python -m pip install wheel

Before running the pip commands, please make sure you're using the IDF python virtual environment.

Basic Concepts

Component-based Unit Tests

Component-based unit tests are our recommended way to test your component. All the test apps should be located under ${IDF_PATH}/components/<COMPONENT_NAME>/test_apps.

For example:

components/
└── my_component/
    ├── include/
    │   └── ...
    ├── test_apps/
    │   ├── test_app_1
    │   │   ├── main/
    │   │   │   └── ...
    │   │   ├── CMakeLists.txt
    │   │   └── pytest_my_component_app_1.py
    │   ├── test_app_2
    │   │   ├── ...
    │   │   └── pytest_my_component_app_2.py
    │   └── parent_folder
    │       ├── test_app_3
    │       │   ├── ...
    │       │   └── pytest_my_component_app_3.py
    │       └── ...
    ├── my_component.c
    └── CMakeLists.txt

Example Tests

Example Tests are tests for examples that are intended to demonstrate parts of the ESP-IDF functionality to our customers.

All the test apps should be located under ${IDF_PATH}/examples. For more information please refer to the :idf_file:`Examples Readme <examples/README.md>`.

For example:

examples/
└── parent_folder/
    └── example_1/
        ├── main/
        │   └── ...
        ├── CMakeLists.txt
        └── pytest_example_1.py

Custom Tests

Custom Tests are tests that aim to run some arbitrary test internally. They are not intended to demonstrate the ESP-IDF functionality to our customers in any way.

All the test apps should be located under ${IDF_PATH}/tools/test_apps. For more information please refer to the :idf_file:`Custom Test Readme <tools/test_apps/README.md>`.

Pytest in ESP-IDF

Pytest Execution Process

Bootstrapping Phase

Create session-scoped caches:
- port-target cache
- port-app cache
Collection Phase
1. Get all the python files with the prefix pytest_
2. Get all the test functions with the prefix test_
3. Apply the params, and duplicate the test functions.
4. Filter the test cases with CLI options. Introduced detailed usages here
Test Running Phase
1. Construct the fixtures. In ESP-IDF, the common fixtures are initialized in this order:
  1. pexpect_proc: pexpect instance
  2. app: IdfApp instance
    
    The information of the app, like sdkconfig, flash_files, partition_table, etc., would be parsed at this phase.
  3. serial: IdfSerial instance
    
    The port of the host which connected to the target type parsed from the app would be auto-detected. The flash files would be auto flashed.
  4. dut: IdfDut instance
2. Run the real test function
3. Deconstruct the fixtures in this order:
  1. dut
    1. close the serial port
    2. (Only for apps with unity test framework) generate junit report of the unity test cases
  2. serial
  3. app
  4. pexpect_proc: Close the file descriptor
4. (Only for apps with unity test framework)
  
  Raise AssertionError when detected unity test failed if you call dut.expect_from_unity_output() in the test function.
Reporting Phase
1. Generate junit report of the test functions
2. Modify the junit report test case name into ESP-IDF test case ID format: <target>.<config>.<test function name>
Finalizing Phase (Only for apps with unity test framework)

Combine the junit reports if the junit reports of the unity test cases are generated.

Example Code

This code example is taken from :idf_file:`pytest_console_basic.py <examples/system/console/basic/pytest_console_basic.py>`.

@pytest.mark.esp32
@pytest.mark.esp32c3
@pytest.mark.generic
@pytest.mark.parametrize('config', [
    'history',
    'nohistory',
], indirect=True)
def test_console_advanced(config: str, dut: Dut) -> None:
    if config == 'history':
        dut.expect('Command history enabled')
    elif config == 'nohistory':
        dut.expect('Command history disabled')

Note

Using expect_exact is better here. For further reading about the different types of expect functions, please refer to the pytest-embedded Expecting documentation.

Use Markers to Specify the Supported Targets

You can use markers to specify the supported targets and the test env in CI. You can run pytest --markers to get more details about different markers.

@pytest.mark.esp32     # <-- support esp32
@pytest.mark.esp32c3   # <-- support esp32c3
@pytest.mark.generic   # <-- test env `generic, would assign to runner with tag `generic`

Besides, if the test case supports all officially ESP-IDF-supported targets, like esp32, esp32s2, esp32s3, esp32c3 for now (2022.2), you can use a special marker supported_targets to apply them all in one line.

This code example is taken from :idf_file:`pytest_gptimer_example.py <examples/peripherals/timer_group/gptimer/pytest_gptimer_example.py>`.

@pytest.mark.supported_targets
@pytest.mark.generic
def test_gptimer_example(dut: Dut) -> None:
    ...

Use Params to Specify the sdkconfig Files

You can use pytest.mark.parametrize with “config” to apply the same test to different apps with different sdkconfig files. For more information about sdkconfig.ci.xxx files, please refer to the Configuration Files section under :idf_file:`this readme <tools/test_apps/README.md>`.

@pytest.mark.parametrize('config', [
    'history',     # <-- run with app built by sdkconfig.ci.history
    'nohistory',   # <-- run with app built by sdkconfig.ci.nohistory
], indirect=True)  # <-- `indirect=True` is required

Overall, this test function would be replicated to 4 test cases:

esp32.history.test_console_advanced
esp32.nohistory.test_console_advanced
esp32c3.history.test_console_advanced
esp32c3.nohistory.test_console_advanced

Advanced Examples

Multi Dut Tests with the Same App

This code example is taken from :idf_file:`pytest_usb_host.py <tools/test_apps/peripherals/usb/pytest_usb_host.py>`.

@pytest.mark.esp32s2
@pytest.mark.esp32s3
@pytest.mark.usb_host
@pytest.mark.parametrize('count', [
    2,
], indirect=True)
def test_usb_host(dut: Tuple[IdfDut, IdfDut]) -> None:
    device = dut[0]  # <-- assume the first dut is the device
    host = dut[1]    # <-- and the second dut is the host
    ...

After setting the param count to 2, all these fixtures are changed into tuples.

Multi Dut Tests with Different Apps

This code example is taken from :idf_file:`pytest_wifi_getting_started.py <examples/wifi/getting_started/pytest_wifi_getting_started.py>`.

@pytest.mark.esp32
@pytest.mark.multi_dut_generic
@pytest.mark.parametrize(
    'count, app_path', [
        (2,
         f'{os.path.join(os.path.dirname(__file__), "softAP")}|{os.path.join(os.path.dirname(__file__), "station")}'),
    ], indirect=True
)
def test_wifi_getting_started(dut: Tuple[IdfDut, IdfDut]) -> None:
    softap = dut[0]
    station = dut[1]
    ...

Here the first dut was flashed with the app :idf_file:`softap <examples/wifi/getting_started/softAP/main/softap_example_main.c>`, and the second dut was flashed with the app :idf_file:`station <examples/wifi/getting_started/station/main/station_example_main.c>`.

Note

Here the app_path should be set with absolute path. the __file__ macro in python would return the absolute path of the test script itself.

Multi Dut Tests with Different Apps, and Targets

This code example is taken from :idf_file:`pytest_wifi_getting_started.py <examples/wifi/getting_started/pytest_wifi_getting_started.py>`. As the comment says, for now it's not running in the ESP-IDF CI.

@pytest.mark.parametrize(
    'count, app_path, target', [
        (2,
         f'{os.path.join(os.path.dirname(__file__), "softAP")}|{os.path.join(os.path.dirname(__file__), "station")}',
         'esp32|esp32s2'),
        (2,
         f'{os.path.join(os.path.dirname(__file__), "softAP")}|{os.path.join(os.path.dirname(__file__), "station")}',
         'esp32s2|esp32'),
    ],
    indirect=True,
)
def test_wifi_getting_started(dut: Tuple[IdfDut, IdfDut]) -> None:
    softap = dut[0]
    station = dut[1]
    ...

Overall, this test function would be replicated to 2 test cases:

softap with esp32 target, and station with esp32s2 target
softap with esp32s2 target, and station with esp32 target

Support different targets with different sdkconfig files

This code example is taken from :idf_file:`pytest_panic.py <tools/test_apps/system/panic/pytest_panic.py>` as an advanced example.

CONFIGS = [
    pytest.param('coredump_flash_bin_crc', marks=[pytest.mark.esp32, pytest.mark.esp32s2]),
    pytest.param('coredump_flash_elf_sha', marks=[pytest.mark.esp32]),  # sha256 only supported on esp32
    pytest.param('coredump_uart_bin_crc', marks=[pytest.mark.esp32, pytest.mark.esp32s2]),
    pytest.param('coredump_uart_elf_crc', marks=[pytest.mark.esp32, pytest.mark.esp32s2]),
    pytest.param('gdbstub', marks=[pytest.mark.esp32, pytest.mark.esp32s2]),
    pytest.param('panic', marks=[pytest.mark.esp32, pytest.mark.esp32s2]),
]

@pytest.mark.parametrize('config', CONFIGS, indirect=True)
...

Use Custom Class

Usually, you can write a custom class in these conditions:

Add more reusable functions for a certain number of DUTs
Add custom setup and teardown functions in different phases described here

This code example is taken from :idf_file:`panic/conftest.py <tools/test_apps/system/panic/conftest.py>`

class PanicTestDut(IdfDut):
    ...

@pytest.fixture(scope='module')
def monkeypatch_module(request: FixtureRequest) -> MonkeyPatch:
    mp = MonkeyPatch()
    request.addfinalizer(mp.undo)
    return mp


@pytest.fixture(scope='module', autouse=True)
def replace_dut_class(monkeypatch_module: MonkeyPatch) -> None:
    monkeypatch_module.setattr('pytest_embedded_idf.dut.IdfDut', PanicTestDut)

monkeypatch_module provide a module-scoped monkeypatch fixture.

replace_dut_class is a module-scoped autouse fixture. This function replaces the IdfDut class with your custom class.

Mark Flaky Tests

Sometimes, our test is based on ethernet or wifi. The network may cause the test flaky. We could mark the single test case within the code repo.

This code example is taken from :idf_file:`pytest_esp_eth.py <components/esp_eth/test_apps/pytest_esp_eth.py>`

@pytest.mark.flaky(reruns=3, reruns_delay=5)
def test_esp_eth_ip101(dut: Dut) -> None:
    ...

This flaky marker means that if the test function failed, the test case would rerun for a maximum of 3 times with 5 seconds delay.

Mark Known Failure Cases

Sometimes a test couldn't pass for the following reasons:

Has a bug
The success ratio is too low because of environment issue, such as network issue. Retry couldn't help

Now you may mark this test case with marker xfail with a user-friendly readable reason.

This code example is taken from :idf_file:`pytest_panic.py <tools/test_apps/system/panic/pytest_panic.py>`

@pytest.mark.xfail('config.getvalue("target") == "esp32s2"', reason='raised IllegalInstruction instead')
def test_cache_error(dut: PanicTestDut, config: str, test_func_name: str) -> None:

This marker means that if the test would be a known failure one on esp32s2.

Mark Nightly Run Test Cases

Some tests cases are only triggered in nightly run pipelines due to a lack of runners.

@pytest.mark.nightly_run

This marker means that the test case would only be run with env var NIGHTLY_RUN or INCLUDE_NIGHTLY_RUN.

Run the Tests in CI

The workflow in CI is simple, build jobs -> target test jobs.

Build Jobs

Build Job Names

Component-based Unit Tests: build_pytest_components_<target>
Example Tests: build_pytest_examples_<target>
Custom Tests: build_pytest_test_apps_<target>

Build Job Command

The command used by CI to build all the relevant tests is: python $IDF_PATH/tools/ci/ci_build_apps.py <parent_dir> --target <target> -vv --pytest-apps

All apps which supported the specified target would be built with all supported sdkconfig files under build_<target>_<config>.

For example, If you run python $IDF_PATH/tools/ci/ci_build_apps.py $IDF_PATH/examples/system/console/basic --target esp32 --pytest-apps, the folder structure would be like this:

basic
├── build_esp32_history/
│   └── ...
├── build_esp32_nohistory/
│   └── ...
├── main/
├── CMakeLists.txt
├── pytest_console_basic.py
└── ...

All the binaries folders would be uploaded as artifacts under the same directories.

Target Test Jobs

Target Test Job Names

Component-based Unit Tests: component_ut_pytest_<target>_<test_env>
Example Tests: example_test_pytest_<target>_<test_env>
Custom Tests: test_app_test_pytest_<target>_<test_env>

Target Test Job Command

The command used by CI to run all the relevant tests is: pytest <parent_dir> --target <target> -m <test_env_marker>

All test cases with the specified target marker and the test env marker under the parent folder would be executed.

The binaries in the target test jobs are downloaded from build jobs, the artifacts would be placed under the same directories.

Run the Tests Locally

The local executing process is the same as the CI process.

For example, if you want to run all the esp32 tests under the $IDF_PATH/examples/system/console/basic folder, you may:

$ pip install pytest-embedded-serial-esp pytest-embedded-idf
$ cd $IDF_PATH
$ . ./export.sh
$ cd examples/system/console/basic
$ python $IDF_PATH/tools/ci/ci_build_apps.py . --target esp32 -vv --pytest-apps
$ pytest --target esp32

Tips and Tricks

Filter the Test Cases

filter by target with pytest --target <target>

pytest would run all the test cases that support specified target.
filter by sdkconfig file with pytest --sdkconfig <sdkconfig>

if <sdkconfig> is default, pytest would run all the test cases with the sdkconfig file sdkconfig.defaults.

In other cases, pytest would run all the test cases with sdkconfig file sdkconfig.ci.<sdkconfig>.

Add New Markers

We’re using two types of custom markers, target markers which indicate that the test cases should support this target, and env markers which indicate that the test case should be assigned to runners with these tags in CI.

You can add new markers by adding one line under the ${IDF_PATH}/pytest.ini markers = section. The grammar should be: <marker_name>: <marker_description>

Generate JUnit Report

You can call pytest with --junitxml <filepath> to generate the JUnit report. In ESP-IDF, the test case name would be unified as "<target>.<config>.<function_name>".

Skip Auto Flash Binary

Skipping auto-flash binary every time would be useful when you're debugging your test script.

You can call pytest with --skip-autoflash y to achieve it.

Record Statistics

Sometimes you may need to record some statistics while running the tests, like the performance test statistics.

You can use record_xml_attribute fixture in your test script, and the statistics would be recorded as attributes in the JUnit report.

Logging System

Sometimes you may need to add some extra logging lines while running the test cases.

You can use python logging module to achieve this.

Known Limitations and Workarounds

Avoid Using `Thread` for Performance Test

pytest-embedded is using some threads internally to help gather all stdout to the pexpect process. Due to the limitation of Global Interpreter Lock, if you're using threads to do performance tests, these threads would block each other and there would be great performance loss.

workaround

Use Process instead, the APIs should be almost the same as Thread.

Files

esp-idf-tests-with-pytest.rst

Latest commit

History