Repository navigation

Test

Set the coverage floor from what the runners measure #230

Workflow file for this run

	name: Test

	on:
	push:
	pull_request:
	workflow_call:

	concurrency:
	group: test-${{ github.ref }}
	cancel-in-progress: true

	# Verbose pytest defaults: long tracebacks, locals, and a short test summary
	# (-ra) so failed/skipped/xfailed tests are listed at the end of the log
	# instead of buried in the middle. Applied via env so every pytest invocation
	# in this workflow inherits the same diagnostics.
	env:
	PYTEST_ADDOPTS: "-vv --tb=long --showlocals -ra --color=yes"
	PIP_DISABLE_PIP_VERSION_CHECK: "1"
	PYTHONUNBUFFERED: "1"

	jobs:
	# Fast pytest selftests + rawterm/menuconfig smoke test on all platforms.
	# Independent of the kernel tree, so it returns feedback quickly and runs
	# in parallel with the slower conformance job below.
	selftest:
	name: Selftest (Python ${{ matrix.target.python }}, ${{ matrix.target.os }})
	runs-on: ${{ matrix.target.builder }}

	defaults:
	run:
	shell: bash

	strategy:
	fail-fast: false
	matrix:
	# NOTE: Windows runs only the headless-portable subset (no Unix shell).
	# Linux spans the supported Python floor (3.8) through current (3.13)
	# to keep the python_requires claim honest; macOS/Windows pin 3.12
	# since the version-sensitive surface is pure Python and OS-independent.
	target:
	- python: '3.8'
	os: Linux
	builder: ubuntu-24.04
	- python: '3.10'
	os: Linux
	builder: ubuntu-24.04
	- python: '3.12'
	os: Linux
	builder: ubuntu-24.04
	- python: '3.13'
	os: Linux
	builder: ubuntu-24.04
	- python: '3.12'
	os: macOS
	builder: macos-15
	- python: '3.12'
	os: Windows
	builder: windows-2022
	headless-only: true

	steps:
	- name: Set up Python
	uses: actions/setup-python@v6
	with:
	python-version: ${{ matrix.target.python }}

	- name: Install Python dependencies
	run: \|
	set -euo pipefail
	python -m pip install --user --upgrade pip
	python -m pip install --user setuptools wheel pytest pytest-cov

	- name: Check out Kconfiglib source code
	uses: actions/checkout@v6

	- name: Run pytest selftests
	run: \|
	set -euo pipefail
	if [ "${{ matrix.target.headless-only }}" == "true" ]; then
	# Windows: portable subset -- excludes tests that need gcc or
	# Unix shell commands (Kpreprocess uses $(shell,...) with
	# bash-specific syntax).
	python -m pytest tests/test_preprocess.py \
	-k "test_user_defined or test_success_failure or test_python_fn or test_kconfig_warn"
	else
	# Coverage on the jobs that run the whole suite. The floor is a
	# backstop against a large drop, not a ratchet; raise it as
	# coverage improves.
	#
	# 35 is set against what these runners report, which is 40. Do not
	# calibrate it from a local run on Python 3.14 or newer: coverage
	# switches to the sys.monitoring backend there and scores the same
	# suite at 51, and a floor taken from that number fails every job
	# here. COVERAGE_CORE=ctrace reproduces the runner figure locally.
	python -m pytest tests/ --ignore=tests/test_conformance.py \
	--cov=kconfiglib --cov=menuconfig --cov=guiconfig --cov=rawterm \
	--cov-report=term --cov-report=xml:coverage.xml \
	--cov-fail-under=35
	fi

	- name: Store coverage report
	# Only the jobs that ran the full suite produce one. !cancelled() rather
	# than always(), so that a coverage failure still uploads the report
	# explaining it but a cancelled run doesn't stop to upload anything.
	if: ${{ !cancelled() && matrix.target.headless-only != true }}
	uses: actions/upload-artifact@v6
	with:
	name: coverage-${{ matrix.target.os }}-py${{ matrix.target.python }}
	path: coverage.xml
	if-no-files-found: ignore

	- name: Validate rawterm and menuconfig (Unix)
	# Exercises rawterm Color/Style/Region compositing, terminal init/close
	# (termios on Unix), and menuconfig headless mode with style parsing.
	if: ${{ matrix.target.os != 'Windows' }}
	run: \|
	set -euo pipefail
	python .ci/validate-rawterm.py

	- name: Validate rawterm and menuconfig (Windows)
	# On Windows, use cmd so Python gets real console handles (bash
	# runs through MSYS2/mintty which lacks a native Windows console).
	if: ${{ matrix.target.os == 'Windows' }}
	shell: cmd
	run: \|
	python .ci/validate-rawterm.py
	if errorlevel 1 exit /b %errorlevel%

	- name: Diagnostic dump on failure
	if: failure()
	shell: bash
	run: \|
	# Best-effort diagnostics. Never let this step itself fail the job;
	# the real failure has already been recorded above.
	set +e
	echo "::group::Python environment"
	python --version
	python -c "import sys, platform; print('exec:', sys.executable); print('arch:', platform.architecture())"
	python -m pip list
	echo "::endgroup::"
	echo "::group::Relevant environment variables"
	env \| grep -E '^(PYTHON\|PIP\|PYTEST\|KCONFIG_)' \| sort
	echo "::endgroup::"

	# Slow kernel-conformance tests: checkout Linux v6.18, build the C kconfig
	# tools, then diff our output against them. Runs in parallel with selftest.
	conformance:
	name: Conformance (Python ${{ matrix.target.python }}, ${{ matrix.target.os }})
	runs-on: ${{ matrix.target.builder }}

	defaults:
	run:
	shell: bash

	strategy:
	fail-fast: false
	matrix:
	target:
	- python: '3.12'
	os: Linux
	builder: ubuntu-24.04
	- python: '3.12'
	os: macOS
	builder: macos-15

	steps:
	- name: Set up Python
	uses: actions/setup-python@v6
	with:
	python-version: ${{ matrix.target.python }}

	- name: Install Python dependencies
	run: \|
	set -euo pipefail
	python -m pip install --user --upgrade pip
	python -m pip install --user setuptools wheel pytest

	- name: Install GNU toolchain (macOS)
	# v6.18 kernel Kconfig probes require GNU Make >= 4.0 and a
	# supported linker (GNU ld or LLD). macOS ships Make 3.81
	# and Apple ld, neither of which is compatible.
	# We add ld.lld to PATH and set LD=ld.lld; reltest forwards LD
	# as a make command-line variable to override the kernel
	# Makefile's unconditional LD=$(CROSS_COMPILE)ld assignment.
	if: ${{ matrix.target.os == 'macOS' }}
	run: \|
	set -euo pipefail
	brew install make lld
	echo "$(brew --prefix make)/libexec/gnubin" >> "$GITHUB_PATH"
	echo "$(brew --prefix lld)/bin" >> "$GITHUB_PATH"
	echo "LD=ld.lld" >> "$GITHUB_ENV"

	- name: Check out Linux source code
	uses: actions/checkout@v6
	with:
	repository: torvalds/linux
	ref: v6.18

	- name: Check out Kconfiglib source code
	uses: actions/checkout@v6
	with:
	path: Kconfiglib

	- name: Apply Linux Kconfig Makefile patch
	run: \|
	set -euo pipefail
	git apply Kconfiglib/makefile.patch

	- name: Build kernel C kconfig tools
	run: \|
	set -euo pipefail
	${MAKE:-make} -j"$(getconf _NPROCESSORS_ONLN)" ARCH=x86 ${LD:+LD=$LD} allnoconfig
	test -x scripts/kconfig/conf
	rm -f .config .config.old

	- name: Run conformance tests and example scripts
	run: \|
	set -euo pipefail
	Kconfiglib/tests/reltest python

	- name: Diagnostic dump on failure
	if: failure()
	run: \|
	set +e
	echo "::group::Kernel kconfig build artifacts"
	ls -la scripts/kconfig/ 2>/dev/null
	echo "::endgroup::"
	echo "::group::Leftover .config (if any)"
	cat .config 2>/dev/null \| head -200
	echo "::endgroup::"
	echo "::group::Toolchain versions"
	${MAKE:-make} --version 2>/dev/null
	${LD:-ld} --version 2>/dev/null
	cc --version 2>/dev/null
	python --version
	echo "::endgroup::"
	echo "::group::Relevant environment variables"
	env \| grep -E '^(PYTHON\|PIP\|PYTEST\|CC\|HOSTCC\|HOSTCXX\|MAKE\|LD\|ARCH\|SRCARCH\|srctree\|KERNELVERSION\|KCONFIG_)' \| sort
	echo "::endgroup::"

Navigation Menu

Repository navigation

Set the coverage floor from what the runners measure #230

Workflow file

Set the coverage floor from what the runners measure #230

Uh oh!

Workflow file for this run