This test creates three threads (and main one) and in each does preapare_read and read_events. This is normal workflow, but since we have no data on socket, the read_events will return with EAGAIN. On this path we do not wake up threads and this test blocks until timeout.
Signed-off-by: Marek Chalupa <mchqwe...@gmail.com> --- tests/display-test.c | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/tests/display-test.c b/tests/display-test.c index c420cbe..0a84ef3 100644 --- a/tests/display-test.c +++ b/tests/display-test.c @@ -483,6 +483,48 @@ TEST(threading_cancel_read_tst) display_destroy(d); } +static void +threading_read_eagain(void) +{ + struct client *c = client_connect(); + pthread_t th1, th2, th3; + + register_reading(c->wl_display); + + th1 = create_thread(c); + th2 = create_thread(c); + th3 = create_thread(c); + + /* All the threads are sleeping, waiting until read or cancel + * is called. Since we have no data on socket waiting, + * the wl_connection_read should end up with error and set errno + * to EAGIN => wl_display_read_events() returns 0 and set errno + * to EAGAIN. Problem is that on this path we did not wake up + * threads until now. Test it*/ + assert(wl_display_read_events(c->wl_display) == 0); + assert(errno == EAGAIN); + + /* kill test in 3 seconds. This should be enough time for the + * thread to exit if it's not blocking. If everything is OK, than + * the thread was woken up and the test will end before the SIGALRM */ + alarm(3); + pthread_join(th1, NULL); + pthread_join(th2, NULL); + pthread_join(th3, NULL); + + client_disconnect(c); +} + +TEST(threading_read_eagin_tst) +{ + struct display *d = display_create(); + + client_create(d, threading_read_eagain); + display_run(d); + + display_destroy(d); +} + static void * thread_prepare_and_read2(void *data) { -- 1.9.3 _______________________________________________ wayland-devel mailing list wayland-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/wayland-devel