Tests in Python packages

Why is running tests important?

Since Python performs only minimal build-time (or more precisely, import-time) checking of correctness, it is important to run tests of Python packages in order to catch any problems early. This is especially important for permitting others to verify support for new Python implementations.

Using distutils_enable_tests

Basic use case

The simplest way of enabling tests is to call distutils_enable_tests in global scope, passing the test runner name as the first argument. This function takes care of declaring test phase, setting appropriate dependencies and test USE flag if necessary. If called after setting RDEPEND, it also copies it to test dependencies.

 # Copyright 1999-2020 Gentoo Authors
 # Distributed under the terms of the GNU General Public License v2

 EAPI=7

 PYTHON_COMPAT=( python2_7 python3_{6,7,8} pypy3 )
 inherit distutils-r1

 DESCRIPTION="An easy whitelist-based HTML-sanitizing tool"
 HOMEPAGE="https://github.com/mozilla/bleach https://pypi.org/project/bleach/"
 SRC_URI="mirror://pypi/${PN:0:1}/${PN}/${P}.tar.gz"

 LICENSE="Apache-2.0"
 SLOT="0"
 KEYWORDS="~alpha ~amd64 ~arm ~arm64 ~hppa ~ia64 ~mips ~ppc ~ppc64 ~s390 ~sparc ~x86"

 RDEPEND="
     dev-python/six[${PYTHON_USEDEP}]
     dev-python/webencodings[${PYTHON_USEDEP}]
 "

 distutils_enable_tests pytest

The valid values include:

• pytest for dev-python/pytest

• setup.py to call setup.py test (deprecated)

• unittest to use built-in unittest discovery

See choosing the correct test runner for more information.

Adding more test dependencies

Additional test dependencies can be specified in test? conditional. The flag normally does not need to be explicitly declared, as distutils_enable_tests does that in the majority of cases.

Please read the section on undesirable test dependencies too.

 # Copyright 1999-2020 Gentoo Authors
 # Distributed under the terms of the GNU General Public License v2

 EAPI=6

 PYTHON_COMPAT=( python2_7 python3_{6,7,8} pypy3 )
 inherit distutils-r1

 DESCRIPTION="Universal encoding detector"
 HOMEPAGE="https://github.com/chardet/chardet https://pypi.org/project/chardet/"
 SRC_URI="https://github.com/chardet/chardet/archive/${PV}.tar.gz -> ${P}.tar.gz"

 LICENSE="LGPL-2.1"
 SLOT="0"
 KEYWORDS="~alpha amd64 arm arm64 hppa ia64 ~m68k ~mips ppc ppc64 s390 ~sh sparc x86 ~x64-cygwin ~amd64-linux ~x86-linux ~x64-macos ~x86-macos ~x64-solaris"

 DEPEND="
     test? ( dev-python/hypothesis[${PYTHON_USEDEP}] )
 "

 distutils_enable_tests pytest

Note that distutils_enable_tests modifies variables directly in the ebuild environment. This means that if you wish to change their values, you need to append to them, i.e. the bottom part of that ebuild can be rewritten as:

 distutils_enable_tests pytest

 DEPEND+="
     test? ( dev-python/hypothesis[${PYTHON_USEDEP}] )
 "

Installing the package before running tests

In PEP 517 mode, the eclass automatically exposes a venv-style install tree to the test phase. No explicit action in necessary.

In the legacy mode, distutils_enable_tests has an optional --install option that can be used to force performing an install to a temporary directory. More information can be found in the legacy section.

Customizing the test phase

If additional pre-/post-test phase actions need to be performed, they can be easily injected via overriding src_test() and making it call distutils-r1_src_test:

 # Copyright 1999-2020 Gentoo Authors
 # Distributed under the terms of the GNU General Public License v2

 EAPI=7

 PYTHON_COMPAT=( python3_{6,7,8} )
 inherit distutils-r1

 DESCRIPTION="Extra features for standard library's cmd module"
 HOMEPAGE="https://github.com/python-cmd2/cmd2"
 SRC_URI="mirror://pypi/${PN:0:1}/${PN}/${P}.tar.gz"

 LICENSE="MIT"
 SLOT="0"
 KEYWORDS="~amd64 ~arm ~arm64 ~ppc64 ~x86 ~amd64-linux ~x86-linux"

 RDEPEND="
     dev-python/attrs[${PYTHON_USEDEP}]
     >=dev-python/colorama-0.3.7[${PYTHON_USEDEP}]
     >=dev-python/pyperclip-1.6[${PYTHON_USEDEP}]
     dev-python/six[${PYTHON_USEDEP}]
     dev-python/wcwidth[${PYTHON_USEDEP}]
 "
 BDEPEND="
     dev-python/setuptools_scm[${PYTHON_USEDEP}]
 "

 distutils_enable_tests pytest

 src_test() {
     # tests rely on very specific text wrapping...
     local -x COLUMNS=80
     distutils-r1_src_test
 }

If the actual test command needs to be customized, or a non-standard test tool needs to be used, you can define a python_test() sub-phase function. This function is called for every enabled Python target by the default src_test implementation. This can either be combined with distutils_enable_tests call, or used instead of it. In fact, the former function simply defines a python_test() function as part of its logic.

 # Copyright 1999-2020 Gentoo Authors
 # Distributed under the terms of the GNU General Public License v2

 EAPI=7

 PYTHON_COMPAT=( python{2_7,3_6,3_7,3_8} pypy3 )
 inherit distutils-r1

 DESCRIPTION="Bash tab completion for argparse"
 HOMEPAGE="https://pypi.org/project/argcomplete/"
 SRC_URI="mirror://pypi/${PN:0:1}/${PN}/${P}.tar.gz"

 LICENSE="Apache-2.0"
 SLOT="0"
 KEYWORDS="~amd64 ~arm ~arm64 ~hppa ~x86 ~amd64-linux ~x86-linux ~x64-macos"
 IUSE="test"
 RESTRICT="!test? ( test )"

 RDEPEND="
     $(python_gen_cond_dep '
         <dev-python/importlib_metadata-2[${PYTHON_USEDEP}]
     ' -2 python3_{5,6,7} pypy3)"
 # pip is called as an external tool
 BDEPEND="
     dev-python/setuptools[${PYTHON_USEDEP}]
     test? (
         app-shells/fish
         app-shells/tcsh
         dev-python/pexpect[${PYTHON_USEDEP}]
         dev-python/pip
     )"

 python_test() {
     "${EPYTHON}" test/test.py -v || die
 }

Note that python_test is called by distutils-r1_src_test, so you must make sure to call it if you override src_test.

Customizing the test phase for pytest

For the relatively frequent case of pytest-based packages needing additional customization, a epytest helper is provided. The helper runs pytest with a standard set of options and automatic handling of test failures.

For example, if upstream uses network marker to disable network-based tests, you can override the test phase in the following way:

distutils_enable_tests pytest

python_test() {
    epytest -m 'not network'
}

Running tests with virtualx

Test suites requiring a display to work correctly can often be appeased usng Xvfb. If the package in question does not start Xvfb directly, virtualx.eclass can be used to do that. Whenever possible, it is preferable to run a single server in src_test() for all passes of the test suite, e.g.:

distutils_enable_tests pytest

src_test() {
    virtx distutils-r1_src_test
}

Note that virtx implicitly enables nonfatal mode. This means that e.g. epytest will no longer terminate the ebuild on failure, and you need to use die explicitly for it:

src_test() {
    virtx distutils-r1_src_test
}

python_test() {
    epytest -m "not network" || die "Tests failed with ${EPYTHON}"
}

Warning

Explicit || die is only necessary when overriding python_test and running epytest inside a nonfatal. The virtx command runs its arguments via a nonfatal. The default python_test implementation created by distutils_enable_tests accounts for this. In other contexts, epytest will die on its own.

Choosing the correct test runner

There are a few modules used to run tests in Python packages. The most common include the built-in unittest module and pytest. There are also some rarely used test tools and domain-specific solutions, e.g. django has its own test runner. This section will help you determining which test runner to use and depend on.

Firstly, it is a good idea to look at test sources. Explicit imports clearly indicate that a particular test runner needs to be installed, and most likely used. For example, if at least one test file has import pytest, pytest is the obvious choice.

In some rare cases the tests may use multiple test packages simultaneously. In this case, you need to choose one of the test runners (see other suggestions) but depend on all of them.

Secondly, some test suites are relying on implicit features of a test runner. For example, pytest has less strict naming and structural requirements for test cases. In some cases, unittest runner will simply be unable to find all tests.

Thirdly, there are cases when a particular feature of a test runner is desired even if it is not strictly necessary to run tests. This is particularly the case with pytest’s output capture that can make test output much more readable with particularly verbose packages.

Upstream documentation, tox configuration, CI pipelines can provide tips on the test runner to be used. However, you should establish whether this information is wholly correct and up-to-date, and whether the particular test tool is really desirable.

If the test suite requires no particular runner (i.e. works with built-in unittest module), using it is preferable to avoid unnecessary dependencies. However, you need to make sure that it finds all tests correctly (i.e. runs no less tests than the alternative) and that it does not spew too much irrelevant output.

Undesirable test dependencies

There is a number of packages that are frequently listed as test dependencies upstream but have little to no value for Gentoo users. It is recommended to skip those test dependencies whenever possible. If tests fail to run without them, it is often preferable to strip the dependencies and/or configuration values enforcing them.

Coverage testing establishes how much of the package’s code is covered by the test suite. While this is useful statistic upstream, it has no value for Gentoo users who just want to install the package. This is often represented by dependencies on dev-python/coverage, dev-python/pytest-cov. In the latter case, you usually need to strip --cov parameter from pytest.ini.

PEP-8 testing enforces standard coding style across Python programs. Issues found by it are relevant to upstream but entirely irrelevant to Gentoo users. If the package uses dev-python/pep8, dev-python/pycodestyle, dev-python/flake8, strip that dependency.

dev-python/pytest-runner is a thin wrapper to run pytest from setup.py. Do not use it, just call pytest directly.

dev-python/tox is a convenient wrapper to run tests for multiple Python versions, in a virtualenv. The eclass already provides the logic for the former, and an environment close enough to the latter. Do not use tox in ebuilds.

Missing test files in PyPI packages

One of the more common test-related problems is that PyPI packages (generated via setup.py sdist) often miss some or all test files. The latter results in no tests being run, the former in test failures or errors.

The simplest solution is to use a VCS snapshot instead of the PyPI tarball:

# pypi tarballs are missing test data
#SRC_URI="mirror://pypi/${PN:0:1}/${PN}/${P}.tar.gz"
SRC_URI="https://github.com/${PN}/${PN}/archive/${PV}.tar.gz -> ${P}.gh.tar.gz"

ImportErrors for C extensions

Tests are often invoked in such a way that the Python packages and modules from the current directory take precedence over these found in the staging area or build directory. In fact, this is often necessary to prevent import collisions — e.g. when modules would be loaded first from the staging area due to explicit imports then again from the current directory due to test discovery.

Warning

Not all packages fail explicitly like that. In particular, if the C extensions are optional, the package may either skip tests requiring them or silently fall back to testing the pure Python variant. Special caution is advised when packaging software using C extensions with top-level source layout.

Unfortunately, this does not work correctly if C extensions are built as part of these packages. Since the package imported relatively does not include the necessary extensions, the imports fail, e.g.:

____________________ ERROR collecting systemd/test/test_login.py ____________________
ImportError while importing test module '/tmp/portage/dev-python/python-systemd-234-r
2/work/python-systemd-234/systemd/test/test_login.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
/usr/lib/python3.8/site-packages/_pytest/python.py:578: in _importtestmodule
    mod = import_path(self.fspath, mode=importmode)
/usr/lib/python3.8/site-packages/_pytest/pathlib.py:524: in import_path
    importlib.import_module(module_name)
/usr/lib/python3.8/importlib/__init__.py:127: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
<frozen importlib._bootstrap>:1014: in _gcd_import
    ???
<frozen importlib._bootstrap>:991: in _find_and_load
    ???
<frozen importlib._bootstrap>:975: in _find_and_load_unlocked
    ???
<frozen importlib._bootstrap>:671: in _load_unlocked
    ???
/usr/lib/python3.8/site-packages/_pytest/assertion/rewrite.py:170: in exec_module
    exec(co, module.__dict__)
systemd/test/test_login.py:6: in <module>
    from systemd import login
E   ImportError: cannot import name 'login' from 'systemd' (/tmp/portage/dev-python/python-systemd-234-r2/work/python-systemd-234/systemd/__init__.py)

Note

Historically, the primary recommendation was to change directory prior to running the test suite. However, this was change since the rm approach is easier in PEP 517 mode, and less likely to trigger other issues (e.g. pytest missing its configuration, tests relying on relative paths).

The modern preference is to remove the package directory prior to running the test suite. In PEP 517 mode, the build system is only run as part of src_compile, and therefore the original sources are not needed afterwards. For example, when the tests are in a separate directory:

python_test() {
    rm -rf frozendict || die
    epytest
}

When the tests are installed as a part of the installed package, --pyargs option can be used to find them via package lookup:

python_test() {
    rm -rf numpy || die
    epytest --pyargs numpy
}

The other possible solution is to change the working directory prior to running the test suite, either into an arbitrary directory that does not feature the collision, or into the install directory. The latter could possible be necessary if the tests are installed as part of the package, and assume paths relative to the source directory:

python_test() {
    cd "${BUILD_DIR}/install$(python_get_sitedir)" || die
    epytest
}

However, please note that changing the working directory leads to pytest missing its configuration (pytest.ini, setup.cfg, pyproject.toml) which in turn could lead to warnings about missing marks or misleading test suite problems.

Checklist for dealing with test failures

If you see some test failures but do not have a guess as to why they would be happening, try the following for a start:

1. Check upstream CI (if any). That’s a quick way of verifying that there is no known breakage at the relevant tag.

2. Try running tests as your regular user, the way upstream suggests (e.g. via tox). Try using a git checkout at the specific tag. This is the basic way of determining whether the package is actually broken or if it is something on our end.

3. If the tests fail at the specified tag, try upstream master branch. Maybe there was a fix committed.

If it seems that the issue is on our end, try the following and see if it causes the subset of failing tests to change:

1. Make sure that the test runner is started via ${EPYTHON} (the eclass-provided epytest and eunittest wrappers do that). Calling system executables directly (either Python via absolute path or system-installed tools that use absolute path in their shebangs) may cause just-built modules not to be imported correctly.

2. Try running the test suite from another directory. If you’re seeing failures to load compiled extensions, Python may be wrongly importing modules from the current directory instead of the build/install tree. Some test suite also depend on paths relative to where upstream run tests.

3. Either switch to PEP 517 mode (preferred), or add distutils_install_for_testing to the test sub-phase or --install to the distutils_enable_tests call. This resolves the majority of problems that arise from the test suite requiring the package to be installed prior to testing.

4. Actually install the package to the system (with tests disabled). This can confirm cases of package for whom the above function does not work. In the worst case, you can set a test self-dependency to force users to install the package before testing:

   test? ( ~dev-python/myself-${PV} )
   

5. Try testing a different Python implementation. If a subset of tests is failing with Python 3.6, see if it still happens with 3.7 or 3.8. If 3.8 is passing but 3.9 is not, it’s most likely some incompatibility upstream did not account for.

6. Run tests with FEATURES=-network-sandbox. Sometimes lack of Internet access causes non-obvious failures.

7. Try a different test runner. Sometimes the subtle differences in how tests are executed can lead to test failures. But beware: some test runners may not run the full set of tests, so verify that you have actually fixed them and not just caused them to be skipped.

Skipping problematic tests

While generally it is preferable to fix tests, sometimes you will face failures that cannot be easily resolved. This especially applies to tests that are broken themselves rather than indicating real problems with the software. However, in some cases you will even find yourself ignoring minor test failures.

Note

When possible, it is preferable to use pytest along with its convenient ignore/deselect options to skip problematic tests. Using pytest instead of unittest is usually possible.

Tests that are known to fail reliably can be marked as expected failures. This has the advantage that the test in question will continue being run and the test suite will report when it unexpectedly starts passing again.

Expected failures are not supported by the standard Python unittest module. It is supported e.g. by pytest.

sed -i -e \
    "/def test_start_params_bug():/i@pytest.mark.xfail(reason='Known to fail on Gentoo')" \
    statsmodels/tsa/tests/test_arima.py || die

Tests that cause inconsistent results, trigger errors, consume horrendous amounts of disk space or cause another kind of undesirable mayhem can be skipped instead. Skipping means that they will not be run at all.

There are multiple ways to skip a test. You can patch it to use a skip decorator, possibly with a condition:

# broken on py2.7, upstream knows
sed -i -e '5a\
import sys' \
    -e '/test_null_bytes/i\
@pytest.mark.skipif(sys.hexversion < 0x03000000, reason="broken on py2")' \
    test/server.py || die

The easy way to skip a test unconditioanlly is to prefix its name with an underscore:

# tests requiring specific locales
sed -i -e 's:test_babel_with_language_:_&:' \
    tests/test_build_latex.py || die
sed -i -e 's:test_polyglossia_with_language_:_&:' \
    tests/test_build_latex.py || die

Finally, if all tests in a particular file are problematic, you can simply remove that file. If all tests belonging to the package are broken, you can use RESTRICT=test to disable testing altogether.

Tests requiring Internet access

One of more common causes of test failures are attempts to use Internet. With Portage blocking network access by default, packages performing tests against remote servers often fail.

Ideally, packages would use mocking or replay tests rather than using real Internet services. Devmanual provides a detailed explanation why tests must not use Internet.

Some packages provide explicit methods of disabling network-based tests. For example, dev-python/tox provides a switch for that:

python_test() {
    distutils_install_for_testing
    epytest --no-network
}

There are packages that skip tests if they fail specifically due to connection errors, or detect whether Internet is accessible. Ideally, you should modify those packages to disable network tests unconditionally. For example, dev-python/pygit2 ebuild does this:

# unconditionally prevent it from using network
sed -i -e '/def no_network/a \
    return True' test/utils.py || die

In other cases, you will have to explicitly disable these tests. In some cases, it will be reasonable to remove whole test files or even restrict tests entirely.

If the package’s test suite relies on Internet access entirely and there is no point in running even a subset of tests, please implement running tests and combine test restriction with PROPERTIES=test_network to allow interested users to run tests when possible:

# users can use ALLOW_TEST=network to override this
PROPERTIES="test_network"
RESTRICT="test"

distutils_enable_tests pytest

Tests aborting (due to assertions)

There are cases of package’s tests terminating with an unclear error message and backtrace similar to the following:

============================= test session starts ==============================
platform linux -- Python 3.7.8, pytest-6.0.1, py-1.9.0, pluggy-0.13.1 -- /usr/bin/python3.7
cachedir: .pytest_cache
rootdir: /var/tmp/portage/dev-python/sabyenc-4.0.2/work/sabyenc-4.0.2, configfile: pytest.ini
collecting ... collected 24 items

[...]
tests/test_decoder.py::test_crc_pickles PASSED                           [ 54%]
tests/test_decoder.py::test_empty_size_pickles Fatal Python error: Aborted

Current thread 0x00007f748bc47740 (most recent call first):
  File "/var/tmp/portage/dev-python/sabyenc-4.0.2/work/sabyenc-4.0.2/tests/testsupport.py", line 74 in sabyenc3_wrapper
  File "/var/tmp/portage/dev-python/sabyenc-4.0.2/work/sabyenc-4.0.2/tests/test_decoder.py", line 119 in test_empty_size_pickles
  File "/usr/lib/python3.7/site-packages/_pytest/python.py", line 180 in pytest_pyfunc_call
  File "/usr/lib/python3.7/site-packages/pluggy/callers.py", line 187 in _multicall
  [...]
  File "/usr/lib/python-exec/python3.7/pytest", line 11 in <module>
/var/tmp/portage/dev-python/sabyenc-4.0.2/temp/environment: line 2934:    66 Aborted                 (core dumped) pytest -vv

This usually indicates that the C code of some Python extension failed an assertion. Since pytest does not print captured output when exiting due to a signal, you need to disable output capture (using -s) to get a more useful error, e.g.:

$ python3.7 -m pytest -s
=============================================================== test session starts ===============================================================
platform linux -- Python 3.7.8, pytest-6.0.1, py-1.9.0, pluggy-0.13.1
rootdir: /tmp/sabyenc, configfile: pytest.ini
plugins: asyncio-0.14.0, forked-1.3.0, xdist-1.34.0, hypothesis-5.23.9, mock-3.2.0, flaky-3.7.0, timeout-1.4.2, freezegun-0.4.2
collected 25 items

tests/test_decoder.py .............python3.7: src/sabyenc3.c:596: decode_usenet_chunks: Assertion `PyByteArray_Check(PyList_GetItem(Py_input_list, lp))' failed.
Fatal Python error: Aborted

Current thread 0x00007fb5db746740 (most recent call first):
  File "/tmp/sabyenc/tests/testsupport.py", line 73 in sabyenc3_wrapper
  File "/tmp/sabyenc/tests/test_decoder.py", line 117 in test_empty_size_pickles
  File "/usr/lib/python3.7/site-packages/_pytest/python.py", line 180 in pytest_pyfunc_call
  File "/usr/lib/python3.7/site-packages/pluggy/callers.py", line 187 in _multicall
  File "/usr/lib/python3.7/site-packages/pluggy/manager.py", line 87 in <lambda>
  [...]
  File "/usr/lib/python3.7/site-packages/pytest/__main__.py", line 7 in <module>
  File "/usr/lib/python3.7/runpy.py", line 85 in _run_code
  File "/usr/lib/python3.7/runpy.py", line 193 in _run_module_as_main
Aborted (core dumped)

Now the message clearly indicates the failed assertion.

It is also common that upstream is initially unable to reproduce the bug. This is because Ubuntu and many other common distributions build Python with -DNDEBUG and the flag leaks to extension builds. As a result, all assertions are stripped at build time. Upstream can work around that by explicitly setting CFLAGS for the build, e.g.:

$ CFLAGS='-O0 -g' python setup.py build build_ext -i
$ pytest -s

Installing extra dependencies in test environment (PEP 517 mode)

Rarely, the test suite expects some package being installed that does not fit being packaged and installed system-wide. For example, isort’s tests use a few example plugins that are not useful to end users, or pip’s test suite still requires old virtualenv that collides with the modern versions. These problems can be resolved by installing the packages locally within the ebuild.

The distutils-r1.eclass provides a distutils_pep517_install helper that can be used to install additional packages. Please note that this helper is intended for expert users only, and special care needs to be taken when using it. The function takes a single argument specifying the destination install root, and installs the package from the current directory.

For example, dev-python/isort uses the following test phase to duplicate the install tree and then install additional packages into it for the purpose of testing. Note that PATH is manipulated (rather than PYTHONPATH) to use virtualenv-style install root.

python_test() {
    cp -a "${BUILD_DIR}"/{install,test} || die
    local -x PATH=${BUILD_DIR}/test/usr/bin:${PATH}

    # Install necessary plugins
    local p
    for p in example*/; do
        pushd "${p}" >/dev/null || die
        distutils_pep517_install "${BUILD_DIR}"/test
        popd >/dev/null || die
    done

    epytest
}