Skip to content

Testing and CI

How the driver is checked: host tests against a model of the chip, a fuzzer that corrupts what the chip sends back, compiled documentation, an install test, ESP32 and Cortex-M builds, static analysis and measured quality gates, on every pull request.

Host tests

The tests build the real driver against a register model of the PCA9685 written from the data sheet, which changes state on its own as the chip does: a power-on reset or a general-call software reset at any moment, auto-increment only with MODE1.AI, PRE_SCALE that ignores writes while awake, a sticky EXTCLK, RESTART that sets when the chip sleeps with a live channel and restarts the channels only 500 µs after the wake, written bytes that reach the outputs at the STOP (or per ACK with OCH = 1), the outputs through full OFF, full ON, INVRT, OUTDRV and OE with OUTNE, and the LED All Call and Sub Call addresses. The mock bus injects every fault a real bus has: an address NACK, a data NACK part-way (the bytes before it stored), a lost ACK, an acknowledged write that is dropped or corrupted, SDA stuck low or reading all ones, a reset just before any transfer. They build with the firmware's flags, -fno-exceptions -fno-rtti, and warnings as errors, on GCC and Clang.

cmake -S . -B build -DPCA9685_BUILD_TESTS=ON
cmake --build build
ctest --test-dir build --output-on-failure
File Covers
tests/test_pca9685.cpp Deinitialize() with OE high before the bus and no I/O on a refused address; the start-up's exact frames and waits, from a fresh, running, held or OCH = 1 chip; every channel call, banks and masks; the frequency sequence; sleep, wake and restart under each reading; OE; diagnostics; NotInitialized before the start
tests/test_reset.cpp Power-on and software resets found by every path, with Config{} too (the SUBADR1 sentinel); a reset before every transfer of a session; one corrupted read never taken for a reset; switch-offs and Deinitialize() after a found reset; nothing written over a reset chip; a reset before a wake or a RESTART write caught by the wake check; pending checks; the group addresses
tests/test_failed_writes.cpp A NACK at every byte of every switch-off never makes a phase-shifted channel more active; NACK part-way, lost ACKs and address NACKs; unknown channels and prescaler; every switch-off sent under every doubt; bank calls that refuse switch-ons and still switch off; two frames for a state change
tests/test_bus_faults.cpp Stuck and floating lines at start-up and after; SDA stuck low from a channel read on never clears a channel's doubt; InvalidReply; NoDevice against Bus; no empty transfer
tests/test_errors.cpp The base Error values are 0 to 9; every chip-specific value (16 and up) is returned by some call; Validate() and the arithmetic
tests/test_model.cpp The model against the data sheet, and its register map against the driver's with static_assert
tests/fuzz/ The fuzz target; CTest replays its corpus (pca9685_fuzz_pca9685_corpus), see below
tests/test_version.cpp The version getters follow the generated header (also against a 9.8.7 probe)
tests/test_kconfig.cpp KconfigDefaults() with and without CONFIG_ values
tests/compile_fail/ Code that must not compile, each for its EXPECT: diagnostic: a port without DelayUs(), a void GpioSet(), a void GetLastFailure(), a Log() without noexcept, a DelayUs(uint16_t), a hook with a narrowed parameter or without noexcept, an invalid CheckedConfig, a class that is not a port
tests/mock_pca9685.hpp The register model, with its own register map; the bus and its faults are in mock_pca9685_bus.hpp
tests/check_version_copy.cmake _config/Doxyfile carries the CMake version
examples/host/ The host example, built with the test flags and run (pca9685_host_example)
tests/check_install.cmake Installs into a scratch prefix and builds tests/install_consumer/ with find_package(pca9685)
tests/check_doxygen.cmake Doxygen with _config/Doxyfile and warnings as errors, when doxygen is installed

The model stands alone

tests/mock_pca9685.hpp is written from the data sheet, not from the driver: it does not include pca9685_registers.hpp or pca9685_types.hpp. Its register map, bits, reset values and timing are its own (namespace pca9685_test::model_reg, each with the data sheet's table), so a wrong constant in the driver cannot hide behind the same wrong constant in the model (handbook KB-TST-030). tests/test_model.cpp compares the two at compile time:

// Register addresses (table 4).
static_assert(chip::kMode1 == reg::kMode1 && chip::kMode2 == reg::kMode2 &&
              chip::kSubAdr1 == reg::kSubAdr1 && chip::kAllCallAdr == reg::kAllCallAdr);
static_assert(chip::kLed0OnL == reg::kLed0OnL && chip::kLastLed == reg::kLed15OffH &&
              chip::kAllLedOnL == reg::kAllLedOnL && chip::kAllLedOffH == reg::kAllLedOffH &&
              chip::kPreScale == reg::kPreScale);
// MODE1 and MODE2 bits (tables 5, 6), the full bits (table 7).
static_assert(chip::kRestart == reg::kMode1Restart && chip::kExtClk == reg::kMode1ExtClk &&
              chip::kAi == reg::kMode1Ai && chip::kSleep == reg::kMode1Sleep &&
              chip::kSub1 == reg::kMode1Sub1 && chip::kAllCall == reg::kMode1AllCall);
static_assert(chip::kInvrt == reg::kMode2Invrt && chip::kOch == reg::kMode2Och &&
              chip::kOutDrv == reg::kMode2OutDrv && chip::kOutneMask == reg::Mode2Outne::kMask &&
              chip::kFull == reg::kLedFull);
// Reset values and timing (tables 5, 6, 8; table 5 note 2).
static_assert(chip::kMode1Reset == reg::kMode1Reset && chip::kMode2Reset == reg::kMode2Reset &&
              chip::kPreScaleReset == reg::kPreScaleReset);
static_assert(chip::kStartUs == pca9685::kOscillatorStartUs);
// The LED All Call address at power-up: ALLCALLADR E0h, 7-bit 0x70 (section 7.1.2).
static_assert((0xE0U >> 1U) == pca9685::kAllCallAddress);

When the build stops on one of these, one of the two maps is wrong: check it against the section the model names, and fix that one. A new register goes into both, with its check. The model has a virtual clock (Elapse(), driven by the mock port's DelayUs() and 10 µs per transfer), so the 500 µs oscillator start-up is real.

Both readings of every open question

Where the data sheet leaves a question open, the model's Readings offers both answers and the safety tests run under each:

Question Reading A (default) Reading B
Asleep, what does an output do? (table 5, note 4) Keeps the level it had Goes inactive
With OCH = 1, when does a channel change? (section 7.3.2, note 3) Once all four of its registers were written On every byte
After a wake without a RESTART write, do the channels run? (section 7.3.1.1) Yes They stay stopped until a RESTART write or a channel write

Fuzz tests

Everything the driver reads comes from the chip over the bus, so the data read is what a noisy cable, a wrong part or a broken chip controls. tests/fuzz/fuzz_pca9685.cpp lets libFuzzer choose a configuration and an address, then up to 64 driver calls, and on every bus transfer what the bus does: pass it to the model, NACK the address, NACK a data byte part-way, lose the ACK of a completed write, drop or corrupt an acknowledged write, flip or replace the bytes of a read, or reset the chip just before it. Between calls it can reset the chip (power-on or software reset), flip any of the three readings above, let another master write through the LED All Call address, or take the bus down. After every call it checks the driver's promises and aborts, naming the line, when one breaks:

  • a call returns an Error the enum defines; a failed value is T{};
  • before EnsureInitialized() succeeds no call but DisableOutputs() makes a transfer, and InvalidArgument means nothing was sent (except a bank call's switch-offs);
  • DeviceReset leaves the driver uninitialised;
  • a switch-off on a driver that has started once is always sent, one write per channel asked for, also after a reset was found;
  • a switch-off, failed or not, never leaves a channel more active than before (its live duty on the chip);
  • an Ok Deinitialize() leaves every channel inactive and OE high;
  • an Ok switch-off leaves the channel inactive on the chip (unless a write of the run was acknowledged and then dropped or corrupted, which I2C does not promise against);
  • an Ok DisableOutputs() leaves OE high;
  • after an Ok call that ran the wake check, a running chip holds the driver's MODE2 and prescaler (unless a read of the run was tampered with);
  • every log line is a non-empty plain-text message.

Three corpus entries pin the design rules these checks found, each also a model test: a RESTART write clears SLEEP too, so the wake check runs again after it (reset-before-restart); with OCH = 1 a switch-off could reach the output part-written when the chip counts registers across transfers, so the driver keeps OCH = 0 (och-ack-stale-bytes); and a single read never clears the doubt that an output is held (held-after-lost-wake).

# libFuzzer needs Clang (on macOS, Homebrew's LLVM: link its libc++, and add -Wl,-ld_classic if ld asks)
cmake -S . -B build-fuzz -DCMAKE_CXX_COMPILER=clang++ -DPCA9685_BUILD_TESTS=ON -DPCA9685_BUILD_FUZZERS=ON
cmake --build build-fuzz --target fuzz_pca9685
./build-fuzz/tests/fuzz/fuzz_pca9685 -max_total_time=300 -max_len=512 tests/fuzz/corpus/fuzz_pca9685

The same file also builds without libFuzzer as fuzz_replay_fuzz_pca9685, and CTest replays every input in tests/fuzz/corpus/fuzz_pca9685/ on GCC and Clang (pca9685_fuzz_pca9685_corpus): named seeds, the inputs that broke a promise, and a set that covers every edge the fuzzer reached, so a fixed bug stays fixed. CI fuzzes for 120 s on every pull request, from the seeds plus a corpus it keeps between runs (job Host quality / fuzz).

Quality gates

The Host quality job measures the library code, inc/ and src/ only (tests and examples are left out), and fails when a value is over the limits of this repository's assurance level, high (set in .kokeb/handbook.yml, defined by the handbook's ADR-0020):

Gate Tool and how it is measured Limit at high
Complexity lizard 1.24.1, modified cyclomatic complexity (-m: a switch counts once) per function CCN at most 15
Function length lizard, lines of code per function at most 60
Parameters lizard, per function at most 6
Line coverage GCC 14 --coverage -O0, every CTest, gcovr 8.7 --merge-lines (a template's lines count once across its instances), --exclude-unreachable-branches --exclude-throw-branches at least 85 %
Branch coverage same run at least 75 %
MC/DC GCC -fcondition-coverage, same run measured, not gated at high (100 % at safety)
Sanitizers Clang, -fsanitize=address,undefined -fno-sanitize-recover=all, every CTest no report
Fuzzing libFuzzer with ASan and UBSan, 120 s per target no broken promise

The reports are in the run's host-quality-reports artifact and its step summary. To run the gates locally:

# Coverage (GCC 14 or later), then the complexity limits
cmake -S . -B build-cov -DCMAKE_CXX_COMPILER=g++-14 -DPCA9685_BUILD_TESTS=ON \
  -DCMAKE_CXX_FLAGS="--coverage -O0 -fcondition-coverage"
cmake --build build-cov && ctest --test-dir build-cov
gcovr --root . --filter inc/ --filter src/ --merge-lines --exclude-unreachable-branches --txt build-cov
lizard -m -C 15 -L 60 -a 6 inc src

# Sanitizers (Clang)
cmake -S . -B build-san -DCMAKE_CXX_COMPILER=clang++ -DPCA9685_BUILD_TESTS=ON \
  -DCMAKE_CXX_FLAGS="-fsanitize=address,undefined -fno-sanitize-recover=all -O1 -g"
cmake --build build-san && ctest --test-dir build-san --output-on-failure

The fuzz gate runs locally as shown in Fuzz tests.

clang-tidy 20 (_config/.clang-tidy: bugprone, CERT, clang-analyzer, cppcoreguidelines, modernize, performance, portability and readability, each disabled check with its reason) and cppcheck (strict) run in the C++ lint and Repository checks jobs and fail on any finding.

The MISRA job (ci-misra.yml) runs the shared ru-misra workflow of ci-tools at level high: it traces a host build and the Cortex-M build (pca9685_optimised_build) of every translation unit, checks them against MISRA C++:2023 and the handbook's C++20 profile (the CodeQL coding-standards and profile packs, plus the validated clang-tidy and cppcheck checks), and reports each finding against the handbook's compliance matrix and deviation records. A finding is fixed in the code; a deliberate exception names its deviation on the line before it (// codeql::misra_deviation_next_line(MD-CPP-009) above the .ipp include of inc/pca9685.hpp). Until the repository is public the job runs with open_source_prerelease: true, the declaration that the code is released under the licence in LICENSE and is to be published.

Documentation snippets

Every C++ block on this site is included from tests/docs/, which the host build compiles (target pca9685_doc_snippets) with the driver instantiated over the documented port. A change to the API that breaks a page breaks the build first. The README's example is a generated copy of the same file; kokeb-docs check fails when it drifts.

CI workflows

Everything runs from ci.yml, which calls one workflow file per check and ends in the CI gate job, the single required check (CI gate / gate).

Job What it checks
Host tests GCC, then Clang (one job), build and run every test above, then build and run examples/host on its own
Host quality The gates above: complexity, coverage, sanitizers, fuzzing
Cross-compile arm-none-eabi-g++ for Cortex-M0+, M4F and M7F with the strict flags, warnings as errors (nothing linked)
C++ lint clang-format and clang-tidy over the sources
MISRA MISRA C++:2023 and the C++20 profile on traced host and Cortex-M builds, at level high (see Quality gates)
Docs Strict site build (_config/mkdocs.yml), figure and generated-region drift, Doxygen (_config/Doxyfile), offline link check, Lighthouse
Repository checks One job: DCO on pull requests, the org standard, Markdown and YAML lint, the Kokeb Labs handbook rules for .kokeb/handbook.yml (by rule ID), template drift (warns only) and cppcheck over the driver sources; a second job runs the handbook's source checks (include guards, bus requirements, raw access)
CI gate Passes only if every job above passed: the one required check

The ESP32 examples job builds every app in examples/esp32/app_config.yml, for each listed ESP-IDF version and build type, with the repository root as the driver's component, and lints the example sources. The gate needs it too.

The workflows call the shared, versioned workflows of kokeb-labs/ci-tools at @v1.

On hardware

Everything above runs against the model. These bench steps confirm what only the chip can, each not verified on hardware until someone runs it and records the board and the result in the pull request's bench-status section:

  1. Start-up from any state. Run the ESP32 example, then reset only the MCU while LED0 is lit: the next EnsureInitialized() switches it off before anything else. Repeat with the chip left asleep by Sleep(SleepMode::HoldOutputs).
  2. Reset detection. Pull VDD below 0.2 V for a moment while the outputs run: every output goes low (off for an active-high load), and the next CheckForReset() returns DeviceReset. Send the general call 0x00 0x06 from a second master: the same.
  3. Sleep and restart. With LED0 at 50 %, Sleep(SleepMode::HoldOutputs): record the level LED0 holds (the data sheet leaves it open; the driver assumes it may stay on). Wake(): LED0 runs again; capture SDA, SCL and LED0 and check the RESTART write comes at least 500 µs after the SLEEP = 0 write.
  4. Frequency and pulse width. Measure the period on LED0 at Config{} (197 Hz expected from 25 MHz) and at ServoPreset() (50 Hz); compute the oscillator from it and check SetPulseWidthUs(1, 1500) against a scope after setting Config::oscillator_hz to the measured value.
  5. A NACK part-way. With a logic analyser on SDA and SCL, interrupt SDA during a 4-byte channel write: the output after the STOP matches the bytes acknowledged, and the driver reports the channel unknown.
  6. OE. With OE on a GPIO, DisableOutputs() takes every output to the off level for each configuration of the OUTNE table on the Hardware page; measure the delay against 60 ns.