Simple testing frameworks are great, but instead of mixing presentation and testing, you could perhaps emit TAP https://lobste.rs/s/n6y1qt/test_anything_protocol and use existing test harnesses?

Warning don’t use this from two threads at once. I would have passed around a “test context” that can store the results such as int tt_passes, int tt_fails, etc. Prefixing the macros TT_RUN to avoid potential clashes would have been good too.

I like single-header unit test frameworks for C. If you’re interested in something with a few more features, I absolutely love Greatest: https://github.com/silentbicycle/greatest.

We used to use something like:

/* in testing.h or whatever */
typedef (int)(*testfunc_t);
typedef struct { const char* name; testfunc_t test } test_t;
typedef struct { const char* suiteName; test_t* tests; size_t numTests} test_suite_t;

/* in one or more test files */
test_t [] myFileTests = [
 { "Test foo", test_foo },
 { "Test foo doesn't assert", test_foo_no_assert},
 /* etc. */
];
size_t myFileTestsCount = sizeof(myFileTests)/(sizeof(test_t);

/* in main.c */
extern test_t[] myFileTests;
extern size_t myFileTestsCount;

test_suite suites[] = [
  { "myFile tests", myFileTests, myFileTestsCount }
  /* etc. */
];
size_t numSuites = sizeof(suites) / sizeof(test_suite_t);

/* in main function runner */
size_t failures = 0;
size_t testsRun = 0;
for (size_t i = 0; i < numSuites; i++) {
  printf("=== %s ===\n", suites[i].suiteName);
  for ( size_t j = 0; j < suites[i].numTests; j++ ) {
    testsRun++;
    printf("\t* %s \n", suites[i].tests[j].name);
    if (suites[i].tests[j].test() ) {
      failures++;
    }
  }
}

printf("Ran %d tests. %d failures.\n", testsRun, failures);
exit(failures > 0);

EDIT:

We really need to support code blocks in the markup…