C Testing

C Testing

Pragmatic workflow for reliable C tests and bug-finding.

When to activate

• Adding unit tests for C functions
• Building a small test harness around a module
• Debugging failing tests or memory corruption
• Enabling sanitizers or fuzzing to reproduce crashes

Core rules

• Keep unit tests deterministic and isolated.
• Avoid real network/time in unit tests.
• Run ASan/UBSan in CI for memory and UB signal.
• Reproduce failures in the smallest possible command first.
• Do not hide memory/UB bugs behind test-only flags, retries, or inflated timeouts.
• Pair with test-driven-development when implementing persistent code or bug fixes test-first.
• Pair with testing-reliability for fixture/timing anti-patterns and systematic-debugging when the root cause is unclear.

Recommended workflow

Phase 1 — Reproduce with smallest scope

• Run the single failing test first (ctest -R ... or direct binary invocation).
• Capture exact command, seed/input, and environment variables.
• Eliminate unrelated test noise before debugging.

Phase 2 — Strengthen harness determinism

• Keep each test independent (no shared mutable global state).
• Use explicit fixtures/setup-teardown for temp files and state reset.
• Prefer table-driven tests for parsers and boundary-heavy functions.

Phase 3 — Sanitizer-first diagnosis

• Run AddressSanitizer for memory corruption/UAF/OOB.
• Run UndefinedBehaviorSanitizer for integer/shift/null/alignment UB.
• Use stack-symbolized reports before stepping into debugger.

Phase 4 — CTest execution strategy

• Label tests (unit, integration, slow, flaky) and run by label.
• Use --output-on-failure and --rerun-failed in local loops.
• Add per-test TIMEOUT to avoid hanging pipelines.

Phase 5 — Fuzzing escalation

• Add libFuzzer target for parser/decoder/validator style APIs.
• Seed corpus with valid and invalid minimal inputs.
• Keep fuzz target stateless, deterministic, and side-effect free.

Fast triage checklist

• Single failing test reproduced in isolation.
• Failure reproduced under sanitizer build (if memory/UB suspected).
• Backtrace/symbolization points to root cause function.
• Regression test added for the fixed bug.
• Full suite rerun (or label-targeted suite if justified) with no new failures.

Resources

Load on demand:

• references/harness.md — harness structure, assertions, Unity/cmocka options
• references/cmake-ctest.md — CMake/CTest patterns, labels, sanitizer presets, MinGW cross-compile
• references/sanitizers-fuzzing.md — ASan/UBSan usage and a minimal fuzz target outline
• references/debugging.md — gdb/lldb, Valgrind, MinGW objdump/nm, MSVC dumpbin/WinDbg, ASan env vars