Broken

Lightweight Unit Testing for C++.

• Official Repository (kobalicek/broken)
• Public Domain

Introduction

Broken is a lightweight framework that can be used to unit-test your C++ code. It has been designed to be lightweight so it can be shipped with the source code it tests. A typical use-case of the framework is to compile an extra version of your library or executable that is statically compiled with the code it tests. This provides some advantages - internals of the library can be easily tested and all issues related to a static compilation of the code can be eliminated.

Broken test-set is a single executable that can run one or multiple tests, which can be specified as command line options. This can be handy when one test corrupts some data that automatically breaks another test. Running all tests together simplifies continuous integration, running tests individually provides more control and checks.

Design Goals

• Lightweight, to be distributed with the source code it tests.
• Do not use C++ STL library or any APIs that allocate memory.
• Do not depend on C++ exception handling or dynamic casts.
• Do tests your way, the framework just helps with the boilerplate.
• Minimum command line interface to help listing and running tests.

Documentation

Broken defines only 3 macros:

• UNIT(_Name_) - Use this to create a new unit test. _Name_ can only contain ASCII characters, numbers and underscore. It has the same rules as identifiers in C and C++. Dashes - are automatically converted to underscores _ when specifying a test from command line.

• INFO(...) - Print an informative message to stdout or other FILE* if configured.

• EXPECT(_Ext_, ...) - Expect _Ext_ to be truthy, fail otherwise. This is the main macro used during testing. If the expression _Exp_ converts to false your unit test failed and you will receive a report.

Installation

There is nothing to be installed to use Broken. Just copy the files broken.h, broken.cpp and optionally broken_main.cpp somewhere to your project and use it in your unit-test enabled build target. The preferred location of these files is in a source subdirectory called test, but it's not required, just a convention.

Beginners Guide

Let's start with a simple function that you plan to test and assume that it's stored in a .cpp file in our project.

static bool safeToMul(uint32_t a, uint32_t b) {
  uint64_t result = static_cast<uint64_t>(a) * b;
  return result < 0xFFFFFFFFU;
}

This function contains a bug of course, but we want to explore the bug by running the unit-test. Broken doesn't dictate the way how the tests should be written, but there are in general two possibilities - put your test into the .cpp implementation file or create a new file with _test.cpp suffix (you can use any other suffix, _test is just used in other projects). I would go with a test inside the implementation file as it allows to test also internals available only in the file itself.

As a convention there should be defined macro when building the unit-test itself. The name is by convention YOUR_PROJECT_NAME_TEST, but can be anything else. The example below is a complete example where BROKEN_EXAMPLE_TEST is used to indicate it's a unit-test build:

#include <stdint.h>
#include "./broken.h"

static bool safeToMul(uint32_t a, uint32_t b) {
  uint64_t result = static_cast<uint64_t>(a) * b;
  return result < 0xFFFFFFFFU;
}

#if defined(BROKEN_EXAMPLE_TEST)
UNIT(base_safe_to_mul) {
  EXPECT(safeToMul(0, 0), "Should allow 0 * 0");
  EXPECT(safeToMul(1, 1), "Should allow 1 * 1");

  EXPECT(safeToMul(0xFFFFFFFF, 0), "Doesn't overflow");
  EXPECT(safeToMul(0xFFFFFFFF, 1), "Doesn't overflow"); // This test fails!

  EXPECT(!safeToMul(0xFFFFFFFF, 2), "Unsafe as it overflows");
}
#endif // BROKEN_EXAMPLE_TEST

Well, to make the test executable working it's required to provide an entry point. If you copied also broken_main.cpp there is nothing to do, otherwise you have to create a file that will define an application entry point and call BrokenAPI::run():

#include "./broken.h"

int main(int argc, const char* argv[]) {
  return BrokenAPI::run(argc, argv);
}

That's it. If you compile the unit-test executable you should be able to run it. The output should look like:

Running base_safe_to_mul
  Failed! Doesn't overflow
  File: ..\broken_example.cpp (Line: 15)

Since there is a bug in the sample test-case it shows it, including the message associated with EXPECT and file/line information. If you fix the code (quiz for you) it should print something similar to:

Running base_safe_to_mul

Success:
  All tests passed!

Broken package contains a simple CMakeLists.txt that can be used by a CMake to generate project files to run the example shown above. Integration with your project and build system is purely on you.

Command Line

The unit-test executable provides several command line arguments:

• --help - Display all command line arguments supported.
• --list - List all tests.
• --run-some-test - Run some_test only, specifying more tests works as expected.
• --run-some-* - Run all tests matching some-*, for simplicity wildcard matching is only supported if it is specified at the end of the string.
• --run-all - Run all tests (enabled by default).