white-box testing

knowledge of code under test is used when designing a test

Test-Driven Development

small development cycles of write test, write minimal implementation, refactor

Cons

extensive usage of preprocessor macros

not available for Objectiv-C and iOS

..anything else?

Alternatives

Boost Test
^{http://www.boost.org/doc/libs/1_55_0/libs/test/doc/html}
CppUnit
^{http://cppunit.sourceforge.net/doc/lastest/cppunit_cookbook.html}

a whole jungle of small developments and domain-specific libraries
^{http://gamesfromwithin.com/exploring-the-c-unit-testing-framework-jungle}

Single test skeleton:

TEST(test_case_name, test_name)
{
 ... test body ...
}

• organize test code in test cases and single tests
• tests should be independent and repeatable
• automatic test discovery: no need to register tests manually

EXPECT_TRUE(condition);
EXPECT_FALSE(condition);
EXPECT_EQ(expected, actual);
EXPECT_NE(expected, actual);
EXPECT_LT(val1, val2);
EXPECT_LE(val1, val2);
EXPECT_GT(val1, val2);
EXPECT_GE(val1, val2);

ASSERT_TRUE(condition);
ASSERT_FALSE(condition);
ASSERT_EQ(expected, actual);
ASSERT_NE(expected, actual);
ASSERT_LT(val1, val2);
ASSERT_LE(val1, val2);
ASSERT_GT(val1, val2);
ASSERT_GE(val1, val2);

• use expect by default: test execution continues after failture
• use assert for hard pre-conditions on remaining test body:
  test execution does not continue after failture — this may also affect clean-up code!
• both offer well-structured error descriptions in system console and IDE out of the box

• each TEST_F gets a fresh instance of TestFixtureClass
• order of invokation:
    → TestFixtureClass()
    → SetUp() → test body → TearDown()
    → ~TestFixtureClass()
• prefer SetUp() / TearDown() over constructor / destructor for anything but resource allocation / deallocation
• if possible use the C++11 override keyword to avoid common mistakes like misspelling SetUp() as Setup()

The problem with subroutines..

Function under Test:

int fib(int n)
{
  return (n < 2) ? n : (fib(n - 1) + fib(n - 2));
}

Helper function:

bool isEven(int n) 
{
  return (n % 2) == 0;
}

Test code:

TEST(Fibonacci, eachThirdIsEven)
{
  EXPECT_TRUE(isEven(fib(0)));
  EXPECT_TRUE(isEven(fib(3)));
  EXPECT_TRUE(isEven(fib(4))); //< this fails
  EXPECT_TRUE(isEven(fib(6)));
}

..solved using SCOPED_TRACE

Helper macro:

#define M_CALL_SCOPE_TRACED(statement) { \
  SCOPED_TRACE("in " #statement); statement; }

Helper function:

void isEven(int n) {
  EXPECT_EQ(0, n % 2);
}

Modified test code:

TEST(Fibonacci, eachThirdIsEven) {
  M_CALL_SCOPE_TRACED(isEven(fib(0)));
  M_CALL_SCOPE_TRACED(isEven(fib(3)));
  M_CALL_SCOPE_TRACED(isEven(fib(4)));
  M_CALL_SCOPE_TRACED(isEven(fib(6)));
}

• affects helper function and invoking test code
• requires another preprocessor macro

..solved using AssertionResult

::testing::AssertionResult isEven(int n) 
{
  if ((n % 2) == 0)
    return ::testing::AssertionSuccess();
  else
    return ::testing::AssertionFailure() << n << " is odd";
}