Add regression tests for bugs 220459 and 220398

[asomers]

Summary

Add regression tests for bugs 220459 and 220398

Bug 220398 - lio_listio(2) never sends asynchronous notification if nent==0
Bug 220459 - lio_listio(2) doesn't support SIGEV_THREAD

[cem]

Only some minor style nits below.

tests/sys/aio/lio_test.c

40–44 ↗

(On Diff #30382)

style(9) puts a few extra newlines in here; between types and function names, as well as before opening function braces.

96 ↗

(On Diff #30382)

I believe the space between ! and signaled should be removed, per style(9).

[asomers]

I'll fix the style. In the meantime, since you seem to be interested in aio(4), do you want to review D11468 ?

[jhb]

Another option instead of 'signaled' might be to use a POSIX semaphore as you can 'sem_post' from a signal handler and you could then just do 'sem_wait' in the test bodies.

tests/sys/aio/lio_test.c

77 ↗

(On Diff #30382)

Do you want to assert the fields in kq_returned? (EVFILT_LIO, &list and possibly set a value in sigev_value that will be in udata)

96 ↗

(On Diff #30382)

It might be cleaner to use sigsuspend() instead of a loop with sleep? That is, something like:

sigset_t oset, set;

sigemptyset(&set);
sigaddset(&set, SIGUSR1);
sigprocmask(SIG_BLOCK, &set, &oset);
ATF_REQUIRE(SIG_ERR != signal(....));
ATF_REQUIRE(0 == lio_listio(...));
ATF_REQUIRE(-1 == sigsuspend(&oset));
ATF_REQUIRE_MSG(signaled, ...);
sigprocmask(SIG_SETMASK, &oset, NULL);

OTOH, maybe that is more code overall?

[asomers]

• Use semaphore instead of polling
• Validate the returned kevent's fields