Skip to content

testing.md

Latest commit

 

History

History
17 lines (16 loc) · 1.72 KB

testing.md

File metadata and controls

17 lines (16 loc) · 1.72 KB
Raw

Tests & benchmarks

tests/ is a separate CMake project (the library itself stays header-only):

cmake -S tests -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release
ctest --test-dir build -C Release --output-on-failure
./build/bench_scan  [module file]   # single-pattern speed; with a file, 40 real signatures sampled from it
./build/bench_multi [module file]   # many patterns: find() per pattern vs find_all vs threaded
  • test_patterns: edge cases, fuzzing against a naive scanner, per-definition XOR keys. Also built with PATTERNS_NO_SIMD.
  • test_artifacts_*: every compile-time pattern type (macros, and C++20 literals) must leave no pattern text in its binary, and the XOR ones no plaintext bytes, words or masks. All patterns are scanned from one function, as in a project's signature file, and built at each optimization level (-O0..-O3, -Os; MSVC /Od /O1 /O2) plus LTO, with debug info, as C++17 and C++20: optimizers fold constants differently at each.
  • test_cpp17: the macro path.
  • test_arm: ArmPattern encodings checked against clang's assembler (concrete instructions must encode exactly, wildcard patterns must match / not match), line formats, RuntimeArmPattern parsing the same, and /d resolution. On Apple arm64 it also re-assembles its tables and resolves /d on linker-relocated ADRP/ADD/LDR/BL code.
  • example_usage: usage styles 3-5 from usage.md against a module that has to be decrypted first, logging each result as [count] name address (rebased address).
  • test_tls (Windows): style 6, find_all run from a TLS callback over the executable's own image.
  • multi_scan.hpp: other multi-pattern strategies (per-pattern threads, chunks over threads) for the benchmark.