Skip to content
cccskills
FreePublished on GitHub

unity-test-framework-c

Author and run ThrowTheSwitch Unity (the C unit-testing library) for bare-metal and RTOS C code. Distinct from the Unity game-engine Test Framework at docs.unity3d.com: this is the ThrowTheSwitch C testing library at throwtheswitch.org/unity, a single C file plus headers that runs on 8-bit MCUs through 64-bit hosts. Anchored on the Unity assertion API and configuration macros regardless of execution environment: the TEST_ASSERT_EQUAL_* / _FLOAT / _DOUBLE / _STRING / _MEMORY / _BITS assertion families, setUp/tearDown/RUN_TEST/UNITY_BEGIN/UNITY_END semantics and the exit-code contract, the generate_test_runner.rb generator, build-time config defines (UNITY_INCLUDE_DOUBLE, UNITY_OUTPUT_CHAR, UNITY_EXCLUDE_SETJMP), and CI integration via Ceedling JUnit XML; applies to host builds, cross-builds, and QEMU-run targets alike. For QEMU machine flags, semihosting, and exit-code capture, see qemu-system-test-runner. Use when the unit-under-test is pure C and the target ranges from 8-bit AVR to Cortex-M0 to Linux ARM.

The English summary is being prepared. Showing the original description.

testland/qa13 stars(GitHub 13 + cccskills 0)
Works withGitHub

Run score Not tested

See how to install

Files included (4)

  • SKILL.md11.7 KB
  • .tessl-plugin/plugin.json1.1 KB
  • references/assertion-api.md2.2 KB
  • references/ci-integration.md1.5 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

unity-test-framework-c

Overview

This skill covers ThrowTheSwitch Unity - a C unit-testing library ("a single C file and a pair of headers") at throwtheswitch.org/unity and github.com/ThrowTheSwitch/Unity, running on 8-bit MCUs through 64-bit hosts with no preprocessor magic, no auto-generated main, and no C++ requirement.

It is distinct from the Unity game-engine Test Framework (com.unity.test-framework, a runner inside the Unity 3D editor at docs.unity3d.com, covered by unity-test-framework). The two share only a name.

Composes with:

  • ceedling-build-runner - the build orchestration that calls generate_test_runner.rb and stitches Unity + CMock + the test binary.
  • ceedling-build-runner's references/cmock.md - the CMock-generated mocks Unity asserts against.
  • qemu-system-test-runner - for running the cross-built binary on a virtual Cortex-M.
  • embedded-coverage-strategy-reference - for the gcov / llvm-cov instrumentation.

When to use

  • Unit-under-test is pure C (not C++). For C++ use googletest-embedded-arm.
  • Target may be tiny - Unity works on 8-bit AVR / PIC, 16-bit MSP430, and 32-bit Cortex-M0/M3/M4/M7/M33.
  • The team already has Ceedling, or wants the canonical ThrowTheSwitch-stack pairing of Unity + CMock + CException.
  • The MCU has limited RAM - Unity's footprint is dramatically smaller than GoogleTest's.

Authoring

Minimal test

#include "unity.h"
#include "ringbuffer.h"

void setUp(void) {
    /* runs before each test */
}

void tearDown(void) {
    /* runs after each test */
}

void test_ringbuffer_initially_empty(void) {
    ringbuffer_t rb;
    ringbuffer_init(&rb);
    TEST_ASSERT_TRUE(ringbuffer_is_empty(&rb));
    TEST_ASSERT_EQUAL_size_t(0, ringbuffer_size(&rb));
}

void test_ringbuffer_push_increments_size(void) {
    ringbuffer_t rb;
    ringbuffer_init(&rb);
    TEST_ASSERT_EQUAL_INT(0, ringbuffer_push(&rb, 42));
    TEST_ASSERT_EQUAL_size_t(1, ringbuffer_size(&rb));
}

int main(void) {
    UNITY_BEGIN();
    RUN_TEST(test_ringbuffer_initially_empty);
    RUN_TEST(test_ringbuffer_push_increments_size);
    return UNITY_END();
}

A Unity test is a C function taking no arguments, named with a test or spec prefix. UNITY_BEGIN() initialises counters; RUN_TEST(name) invokes setUp -> test fn -> tearDown and captures the result; UNITY_END() prints the summary and returns a non-zero exit code on failure.

Assertions

The common assertions - TEST_ASSERT_TRUE(c), TEST_ASSERT_EQUAL_INT(a,b) (with _INT8 / _INT16 / _INT32 / _INT64 and _UINT / _HEX widths), TEST_ASSERT_EQUAL_FLOAT(a,b) / TEST_ASSERT_FLOAT_WITHIN(delta,a,b), TEST_ASSERT_EQUAL_STRING(a,b), TEST_ASSERT_EQUAL_MEMORY(a,b,len), TEST_ASSERT_NULL(p). Append _ARRAY to compare arrays (third arg is element count) and _MESSAGE to attach a custom failure string.

The full assertion-family table and the build-time config defines are in references/assertion-api.md.

Skipping and failing explicitly

void test_only_when_calibrated(void) {
    if (!device_calibrated()) {
        TEST_IGNORE_MESSAGE("device not calibrated; skip");
    }
    /* normal asserts */
}

TEST_IGNORE() records an "ignored" result, distinct from pass/fail in the summary.

generate_test_runner.rb

The Ruby script auto/generate_test_runner.rb scans a test file, finds every test_* and spec_* function, and emits a <file>_Runner.c that wires them up:

ruby /path/to/Unity/auto/generate_test_runner.rb \
    test/test_ringbuffer.c test/test_ringbuffer_Runner.c

Then compile test_ringbuffer.c + test_ringbuffer_Runner.c + unity.c and link. Ceedling does this implicitly - see ceedling-build-runner.

Building

Host build (fast inner loop)

gcc -Wall -O0 -g -DUNITY_INCLUDE_DOUBLE \
    -I src -I ext/Unity/src \
    src/ringbuffer.c \
    ext/Unity/src/unity.c \
    test/test_ringbuffer.c test/test_ringbuffer_Runner.c \
    -o test_ringbuffer
./test_ringbuffer

-DUNITY_INCLUDE_DOUBLE enables the _DOUBLE macros (off by default to save flash on tiny MCUs - per the Unity config docs).

Cortex-M0 cross-build (under QEMU)

arm-none-eabi-gcc -mcpu=cortex-m0 -mthumb -O0 -g \
    -DUNITY_OUTPUT_COLOR \
    --specs=rdimon.specs \
    -I src -I ext/Unity/src \
    src/ringbuffer.c ext/Unity/src/unity.c \
    test/test_ringbuffer.c test/test_ringbuffer_Runner.c \
    -o test_ringbuffer.elf -lrdimon
qemu-system-arm -M mps2-an385 -cpu cortex-m0 \
    -nographic -semihosting-config enable=on,target=native \
    -kernel test_ringbuffer.elf

--specs=rdimon.specs provides the ARM semihosting library (developer.arm.com GNU Toolchain), so Unity's printf-based reporting reaches QEMU's stdio.

Build-time configuration

Key defines (UNITY_INCLUDE_DOUBLE, UNITY_OUTPUT_CHAR, UNITY_OUTPUT_COLOR, UNITY_FIXTURE_NO_EXTRAS, UNITY_EXCLUDE_SETJMP) are tabulated in references/assertion-api.md, per the Unity configuration guide.

Running

Console output

Unity prints one line per result + a summary:

test/test_ringbuffer.c:34:test_ringbuffer_initially_empty:PASS
test/test_ringbuffer.c:42:test_ringbuffer_push_increments_size:PASS

-----------------------
2 Tests 0 Failures 0 Ignored
OK

The format is <file>:<line>:<test_name>:<PASS|FAIL|IGNORE>. Failures include the assertion details:

test/test_ringbuffer.c:48:test_overflow_returns_minus_one:FAIL: Expected -1 Was 0

Exit code

UNITY_END() returns the failure count (int). Use it from main:

int main(void) {
    UNITY_BEGIN();
    RUN_TEST(test_x);
    return UNITY_END();   /* 0 on all-pass; non-zero on any failure */
}

CI tools that gate on exit code (CMake CTest, GitHub Actions run: step) work natively.

Parsing results

Console parsing

Unity's text output is intentionally simple - a grep -c ':FAIL:' is enough for a smoke gate:

./test_ringbuffer | tee results.txt
fails=$(grep -c ':FAIL:' results.txt || true)
[ "$fails" -eq 0 ] || exit 1

JUnit XML via Ceedling

Ceedling wraps Unity and emits a JUnit XML report at build/artifacts/test/report.xml - see ceedling-build-runner. The schema matches GoogleTest's, so the same JUnit pipeline works for both.

Custom output

Override UNITY_OUTPUT_CHAR(c) at build time:

/* unity_config.h */
#define UNITY_OUTPUT_CHAR(c)    serial_putc(c)
#define UNITY_OUTPUT_FLUSH()    serial_flush()

On bare-metal, send results over UART → host serial → CI log file.

CI integration

The standalone (no Ceedling) GitHub Actions pipeline - generate runners, build + run on host, then cross-build and run under QEMU - is in references/ci-integration.md. For Ceedling-driven projects, see ceedling-build-runner for the canonical ceedling test:all + JUnit XML flow.

Anti-patterns

Anti-patternWhy it failsFix
Hand-maintaining the test runnerDrift: a new test_* function is silently skippedRun generate_test_runner.rb in the build; or use Ceedling
TEST_ASSERT_EQUAL_INT on size_tWidth mismatch warns on 64-bit hosts, may overflow on 8-bitUse TEST_ASSERT_EQUAL_size_t or width-specific _UINT32
TEST_ASSERT_TRUE(strcmp(a,b) == 0)Failure message reports "Expected true, got false" - uselessUse TEST_ASSERT_EQUAL_STRING(a,b)
TEST_ASSERT_EQUAL_MEMORY with len=sizeof(*p) on a struct with paddingPadding bytes vary; intermittent failuresInitialise structs with memset(.., 0, sizeof) before fill, or compare fields individually
Calling RUN_TEST outside UNITY_BEGIN/UNITY_ENDAsserts work but the summary is wrongAlways bracket runs with UNITY_BEGIN / UNITY_END
Mixing the C library with the game-engine Test FrameworkBuild sees two unity.h headers; one wins randomlyDon't co-locate; keep unity-test-framework to game-engine projects
Float comparison with TEST_ASSERT_EQUAL_FLOAT and exact valuesFloating-point equality is fragileUse TEST_ASSERT_FLOAT_WITHIN(epsilon, a, b)
UNITY_EXCLUDE_SETJMP on a target that has setjmpLoses early-abort on fatal assert; tests run on after corruptionOnly exclude when the toolchain genuinely lacks setjmp

Limitations

  • No native parameterised tests. Unity has no TEST_P equivalent. Loop with a fixture struct and call TEST_ASSERT_* inside; report the iteration via _MESSAGE suffix.
  • No native test discovery. Without generate_test_runner.rb (or Ceedling), each test must be manually listed in main.
  • Single-threaded. Unity's failure-recovery uses setjmp/ longjmp; concurrent tests in the same process collide. RTOS tests should run tests serially on the test thread.
  • No GoogleMock-style matchers. Mocks live in CMock; matcher expressivity is per ceedling-build-runner's references/cmock.md.
  • TEST_ASSERT_EQUAL_FLOAT precision is configurable but global. UNITY_FLOAT_PRECISION applies to every float compare in the suite; per-test precision needs _WITHIN.
  • 8-bit targets benefit from UNITY_FIXTURE_NO_EXTRAS. Default builds include features (per-test color, fixture hooks) that bloat tiny MCUs.

References

Cited inline. Foundational documents:

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Builds a CI gate that fails the build on **new** WCAG / a11y violations introduced by a PR while grandfathering pre-existing violations on a per-rule / per-page baseline. Aggregates verdicts from axe-core / pa11y / Lighthouse a11y / WAVE / IBM Equal Access scans. Use when a project has accumulated a11y debt and a strict "zero violations" gate would block every PR - the ratchet pattern lets the team ship while preventing regressions.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

Workflow skill that builds an A/B-test validity checklist from an experiment proposal, walking the canonical design-correctness gates - pre-registered OEC/power/guardrails, randomization unit + SRM check, assignment integrity, telemetry, peeking discipline, novelty/primacy, post-experiment SRM re-check - into a per-experiment checklist + sign-off form. Use when launching, auditing, or governing an experiment. For pitfall mechanics (guardrails, peeking) see experiment-results-interpreter's references; to read an already-valid result use experiment-results-interpreter; for per-SDK harness tests use experiment-sdk-testing - this gates DESIGN, not SDK code.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

The single install-flow skill: the reference contract for the PWA install flow (installability gate fields, `beforeinstallprompt` handshake, per-platform paths - Android WebAPK / iOS Share menu / Firefox no-op - and the `display-mode` post-install signal, in references/install-flow-reference.md) plus the build-an-X workflow that emits the Add-to-Home-Screen suite. Walks the four-stage timeline, emitting one test per gate cell per [web.dev/articles/install-criteria][install-criteria], the deferred-prompt → `prompt()` → `userChoice` chain per [web.dev/articles/customize-install][customize-install], the iOS Safari manual-metadata branch (`apple-touch-icon`) per [web.dev/learn/pwa/installation][learn-pwa], and the post-install `(display-mode: standalone)` MQ assertion. Output: a Playwright spec with per-stage cells plus a coverage matrix. Use when a PWA's manifest, icons, or install handler change, when install conversion drops at an unknown stage, or when triaging a flaky install assertion.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

Build-an-X workflow that uses an LLM to map existing tests to spec sections - given a spec doc + the test suite, the LLM identifies which tests cover which sections, surfaces uncovered sections (gap), and recommends specific tests to add. Output is a coverage matrix per spec ID. Scope is mapping tests that already exist and naming the gaps, not authoring tests for new acceptance criteria. Use when a spec doc and a test suite both exist but nobody can say which requirements are actually covered - before a release sign-off, an audit, or a decision about where to spend the next round of test effort.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

Generates tests from natural-language specs (acceptance criteria, user stories, requirements) using an LLM, with confidence scoring per test case (LLM self-assessment plus heuristics: assertion quality, naming, completeness), batching uncertain cases for human review, and integration with the team's existing test framework. Use when the user asks to generate unit tests from acceptance criteria, convert user stories to test cases, automate test creation from requirements, or augment a spec-driven test suite with AI-generated stubs that are then curated before merge.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

Configures Allure Report (test-runner adapter install, `allure-results` directory wiring, `categories.json` for failure classification, `history-trend.json` retention via the copy-history-between-runs pattern), runs the Allure CLI to convert `allure-results` to a static HTML site, and uploads the report as a CI artifact. Use when the team needs richer test reporting than JUnit XML - step-level attachments, per-test history, retry tracking, and severity / epic / feature labeling across framework-agnostic adapters (pytest, Jest, JUnit, TestNG, NUnit, Mocha). As a rich static HTML report generator, it is the open-source alternative to the sunset ExtentReports (JVM/.NET per-test HTML narrative); for hosted cross-run flakiness analytics rather than a static per-run report use currents-integration.

The English summary is being prepared. Showing the original description.

testland/qa13Updated Oct 20, 2026

See all skills by testland

Report a problem with this skill