Re: [PATCH v4 10/10] selftests/harness: Fix TEST_F()'s exit codes

[Mickaël Salaün]

On Thu, May 02, 2024 at 03:45:22PM GMT, Sean Christopherson wrote:
> On Thu, May 02, 2024, Mickaël Salaün wrote:
> > @@ -462,8 +462,10 @@ static inline pid_t clone3_vfork(void)
> >  		munmap(teardown, sizeof(*teardown)); \
> >  		if (self && fixture_name##_teardown_parent) \
> >  			munmap(self, sizeof(*self)); \
> > -		if (!WIFEXITED(status) && WIFSIGNALED(status)) \
> > -			/* Forward signal to __wait_for_test(). */ \
> > +		/* Forward exit codes and signals to __wait_for_test(). */ \
> > +		if (WIFEXITED(status)) \
> > +			_exit(_metadata->exit_code); \
> 
> This needs to be:
> 
> 		if (WIFEXITED(status)) \
> 			_exit(WEXITSTATUS(status)); \
> 
> otherwise existing tests that communicate FAIL/SKIP via exit() continue to yield
> exit(0) and thus false passes.

Yes of course.

> 
> If that conflicts with tests that want to communicate via _metadata->exit_code,
> then maybe this?
> 
> 		if (WIFEXITED(status)) \
> 			_exit(WEXITSTATUS(status) ?: _metadata->exit_code); \

I prefer this approach handling failed expectations in the fixture
teardown too.

However, the direct call to _exit() doesn't handle failed asserts.  I'll
fix that.


> 
> Or I suppose _metadata->exit_code could have priority, but that seems weird to
> me, e.g. if a test sets exit_code and then explodes, it seems like the explosion
> should be reported.
> 
> > +		if (WIFSIGNALED(status)) \
> >  			kill(getpid(), WTERMSIG(status)); \
> >  		__test_check_assert(_metadata); \
> >  	} \
> > -- 
> > 2.45.0
> > 
>