capstone/tests
Rot127 5026c2c4e9
Some checks failed
Run Test / ${{ matrix.config.name }} (map[arch:x64 build-system:cmake diet-build:OFF enable-asan:OFF name:ubuntu-22.04 x64 cmake os:ubuntu-22.04]) (push) Waiting to run
Run Test / ${{ matrix.config.name }} (map[arch:x64 build-system:cmake diet-build:OFF enable-asan:ON name:ubuntu-24.04 x64 ASAN os:ubuntu-24.04]) (push) Waiting to run
Run Test / ${{ matrix.config.name }} (map[arch:x64 build-system:make diet-build:OFF enable-asan:OFF name:ubuntu-22.04 x64 make os:ubuntu-22.04]) (push) Waiting to run
Run Test / ${{ matrix.config.name }} (map[arch:x64 name:windows x64 MSVC 64bit os:windows-latest platform:windows python-arch:x64 python-version:3.9]) (push) Waiting to run
Run clang-tidy / clang-tidy (push) Waiting to run
RELEASE BUILD - PyPI 📦 Distribution / Build wheels on ${{ matrix.os }} (macos-latest) (push) Waiting to run
RELEASE BUILD - PyPI 📦 Distribution / Build wheels on ${{ matrix.os }} (ubuntu-latest) (push) Waiting to run
RELEASE BUILD - PyPI 📦 Distribution / Build wheels on ${{ matrix.os }} (windows-latest) (push) Waiting to run
RELEASE BUILD - PyPI 📦 Distribution / Make SDist (push) Waiting to run
RELEASE BUILD - PyPI 📦 Distribution / publish (push) Blocked by required conditions
Python Package CI / build (macOS-14, 3.12) (push) Waiting to run
Python Package CI / build (macOS-14, 3.8) (push) Waiting to run
Python Package CI / build (ubuntu-24.04, 3.12) (push) Waiting to run
Python Package CI / build (ubuntu-24.04, 3.8) (push) Waiting to run
Python Package CI / build (windows-2022, 3.12) (push) Waiting to run
Python Package CI / build (windows-2022, 3.8) (push) Waiting to run
Auto-Sync / check (push) Has been cancelled
Merge pull request #2507 from thestr4ng3r/no-varargs-aarch64
AArch64: Replace vararg add_cs_detail by multiple concrete functions
2024-10-22 13:11:33 +00:00
..
details Fix LDR not assigning immediate as memory offset. (#2487) 2024-09-30 17:31:34 +08:00
features Modern Testing (#2456) 2024-08-31 21:33:38 +08:00
integration Fixing UB santizer, LITBASE and assert errors. (#2499) 2024-10-06 08:45:13 +08:00
issues AArch64: Replace vararg add_cs_detail by multiple concrete functions 2024-10-21 22:03:00 +02:00
MC xtensa: Fix Branch Target (#2516) 2024-10-22 16:20:37 +08:00
unit Coverity defects (#2469) 2024-09-18 21:19:42 +08:00
README.md Modern Testing (#2456) 2024-08-31 21:33:38 +08:00

Testing in Capstone

Running tests

Types of test and their location

YAML test files

These test files are consumed by the various cstest tools. They contain all detail tests. As well as the LLVM regression tests (MC tests).

Directories group tests by the category they intent to test.

Legacy (integration)

Legacy tests which only printed to stdout. In practice they only test if the code segfaults. Checking the produced output was not implemented.

Testing tools and usage

cstest

cstest is the testing tool written in C. It is implemented in suite/cstest/ It consumes the yaml files and reports errors or mismatches for disassembled instructions and their details.

Building

Dependencies: cstest requires the libyaml library.

You build cstest by adding the -DCAPSTONE_BUILD_CSTEST=1 option during configuration of the Capstone build.

If you build and install Capstone cstest gets installed as well. Otherwise you find it in the build directory.

# Install libyaml
# sudo apt install libyaml-dev
# or
# sudo dnf install libyaml-devel
cd "<capstone-repo-root>"
# Optionally add the `-DENABLE_ASAN=1` flag.
cmake -B build -DCMAKE_BUILD_TYPE=Debug -DCAPSTONE_BUILD_CSTEST=ON
cmake --build build --config Debug
cmake --install build --prefix "<install-prefix>"

Run the integration tests for cstest itself

./suite/cstest/test/integration_tests.py cstest

Run the tests

# Check supported options
cstest -h
# Run all
cstest tests/

Alternatively, you can use the CMake test manager.

# List available tests
ctest --test-dir build -N
# Run a specific test
ctest --test-dir build -R "<name>"

cstest_py

cstest_py is the testing tool written in Python. It is implemented in bindings/python/cstest_py It consumes the yaml files and reports errors or mismatches for disassembled instructions and their details.

Installing

You need to install the Capstone Python bindings first and afterwards the cstest_py.

# Optionally, create a new virtual environment
python3 -m venv .venv
source .venv/bin/activate

cd bindings/python
pip install -e .
pip install -e cstest_py
cd ../..

Run the integration tests for cstest_py itself

./suite/cstest/test/integration_tests.py cstest_py

And run the tests

# Check supported options
cstest_py -h
# Run all
cstest_py tests/

Add new tests

Unit and integration tests

Add the source into test/integration or test/unit respectively and update the CMakeLists.txt file.

YAML

There are very few fields which are mandatory to fill. Check suite/cstest/test/min_valid_test_file.yaml to see which one.

  • In general it is useful to just copy a previous test file and rewrite it accordingly.
  • If you assign C enumeration identifiers to some fields (to check enumeration values), ensure they are added on the suite/cstest/include/test_mapping.h. Otherwise, cstest cannot map the strings to the values for comparison.
  • Rarely used, but useful fields are: name, skip, skip_reason.

MC regression tests

The MCUpdater translates most test files of the LLVM MC regression tests into our YAML files.

The LLVM regression tests, check the bytes and assembly for all instructions of an architecture. They do it by passing bytes or assembly to the llvm-mc and FileCheck tool and compare the output. We capture this output and process it into YAML. So you need to install llvm-mc and FileCheck for our updater to work.

To update the YAML MC regression tests, you need to install Auto-Sync and run the MCUpdater.

cd suite/auto-sync/
# Follow install instructions of Auto-Sync described in the README
# And run the updater:
./src/autosync/MCUpdater.py -a ARCH
ls build/mc_out/
# The produce yaml files. Copy them manually to tests/MC/ARCH

Please note:

Each of the LLVM test files can contain several llvm-mc commands to run on the same file. This is done to test the same file with different CPU features enabled. So it can test different assembly flavors etc.

In Capstone all modules enable always all CPU features (even if this is not possible in reality). Due to this, we always parse all llvm-mc commands but only write the last version of them to disk. So if the same test file is tested with three different features enables, once with FeatureA, FeatureB and FeatureC we only save the output with FeatureC enabled.

This might give you MC test files which fail due to valid but mismatching disassembly. You can set the skip field for those tests and add a skip_reason.

Once https://github.com/capstone-engine/capstone/issues/1992 is resolved, we can test all variants.