C++ Testing with GoogleTest and CTest

Flaky tests, silent memory corruption in test fixtures, and slow test discovery degrade confidence in test suites. Establishing a deterministic test harness with fast discovery and active sanitizers catches memory leaks and race conditions before code merges.

ctest --test-dir build --output-on-failure -j4

Detailed test fixture syntax, parameterized test suites, and CMake integration patterns are documented in references/gtest-patterns.md.

Core Concepts

1. Red-Green-Refactor loop: Write the failing assertion first, implement the minimum code to turn it green, then refactor with test verification.
2. Isolation over integration: Prefer fakes and explicit dependency injection rather than relying on shared global state.
3. Structured directory layout: Keep tests mirrored against source under tests/unit/, tests/integration/, and tests/fixtures/.
4. Dynamic discovery: Use gtest_discover_tests() in CMake so CTest registers tests automatically without manual listing.

Writing Deterministic Tests

Prefer EXPECT_* for non-fatal checks that allow remaining assertions in the test to execute, and ASSERT_* when failure invalidates subsequent logic (such as checking pointer nullity before dereferencing).

#include <gtest/gtest.h>

TEST(RingBufferTest, OverwriteAdvancesTail) {
    RingBuffer<int, 4> buffer;
    for (int i = 0; i < 5; ++i) {
        buffer.push(i);
    }
    EXPECT_EQ(buffer.size(), 4u);
    EXPECT_EQ(buffer.pop(), 1);
}

When verifying mocked interactions with GoogleMock:

class MockHardwareBus : public HardwareBus {
public:
    MOCK_METHOD(bool, WriteRegister, (uint16_t reg, uint8_t val), (override));
    MOCK_METHOD(uint8_t, ReadRegister, (uint16_t reg), (override));
};

Diagnosing Memory and Concurrency Bugs

Run tests under sanitizers by enabling compiler flags during configuration:

• AddressSanitizer (-fsanitize=address): Catches out-of-bounds accesses, use-after-free, and stack corruption.
• UndefinedBehaviorSanitizer (-fsanitize=undefined): Catches integer overflows, null pointer dereferences, and alignment violations.
• ThreadSanitizer (-fsanitize=thread): Detects data races in multithreaded test cases. Do not combine TSan with ASan in the same binary.

Honest limits

Unit tests demonstrate the presence of specific verified behaviors, but cannot prove the absence of subtle hardware interactions, cache coherency anomalies, or platform-specific driver timing errors. Sanitizers introduce CPU and memory overhead, meaning timing-sensitive bugs in real-time loops may not manifest during sanitized test runs.