A test and testsuite modifier to state dependencies. The test(s) will not run until all dependencies are resolved, i.e. the referred tests have succeeded.

Used in: The modifier list of a test definition or testsuite definition. See TEST(name, ...) and TESTSUITE(name, ...)

The parameter to DEPENDS_ON(...) is a comma separated list of dependencies. Each dependency can be either a the name of a test, or ALL_TESTS(testsuitename)

When DEPENDS_ON(...) is used for a testsuite, the dependencies applies for all tests in the testsuite. It is possible to constrain a single test further than the testsuite it resides in, by adding a DEPENDS_ON(...) modifier to the test.

It is not possible to state a dependency on a DISABLED_TEST(name, ...).

	Tip
	It is often a good idea to let testsuites depend on testsuites, and tests to depend on other tests inside the same testsuite.

	Note
	DEPENDS_ON(...) can only be included once in the modifier list for each TEST(name, ...) or TESTSUITE(name, ...).

Example: The test program

     
     #include <crpcut.hpp>
     
     struct A
     {
       A() {}
       ~A() { exit(0); }
       void do_something() const {}
     };
     
     TESTSUITE(basics)
     {
       TEST(construct)
       {
         (void)new A(); // intentionally leak
       }
     
       TEST(destroy, DEPENDS_ON(construct))
       {
         A *p = new A();
         delete p;
       }
     }
     
     TESTSUITE(toy, DEPENDS_ON(ALL_TESTS(basics)))
     {
       TEST(work)
       {
         A obj;
         obj.do_something();
       }
     }
     
     int main(int argc, char *argv[])
     {
       return crpcut::run(argc, argv);
     }

reports one failed test, and one that did not run:

     FAILED: basics::destroy
     phase="post_mortem"  ----------------------------------------------------------
     Exited with code 0
     Expected normal exit
     -------------------------------------------------------------------------------
     ===============================================================================
     The following tests were blocked from running:
       toy::work
     Total 3 test cases selected
     UNTESTED : 1
     PASSED   : 1
     FAILED   : 1

See also the -n command line flag.