barbu110 / ctest Goto Github PK

Failed assertions should leave the test in a FAILED state and should not allow the execution to continue.

Required changes:

The test function should receive an assertion_runner_fn instead of an assertion_container. This will not only help in the above mentioned problem, but also eliminate the limitation created by MAX_ASSERTIONS_PER_TEST definition.

The assertion_runner_fn should be defined as follows:

typedef int (*assertion_runner_fn)(struct assertion *);

This function should return 0 when the assertion successes and 1 when the assertion fails (similar to Unix syscalls).

The library should provide a main function for the test executables.

The program should accept the following command line arguments:

• --all or -a to run all the registered tests
• --suite <suite_name> or -s <suite_name> to run all the tests registered within the given suite_name
• --failed to skip running the tests that failed on the last run

The following assertions should be implemented in assertions.h:

• equality of integer values (ASSERT_EQ)
• not equal for integer values (ASSERT_NE)
• less than for integer values (ASSERT_LT)
• greater than for integer values (ASSERT_GT)
• less than or equal for integer values (ASSERT_LTE)
• greater than or equal for integer values (ASSERT_GTE)
• is 0 or was successful (ASSERT_ZERO, ASSERT_SUCCESS); the successful status comes from the philosophy of Unix syscalls which return 0 on success
• string relations (ASSERT_STRCMP(a, b, expected_relation))
• equality for floating point values (ASSERT_FLOAT_EQ)