summaryrefslogtreecommitdiff
path: root/tests
diff options
context:
space:
mode:
authorBastien Nocera <hadess@hadess.net>2016-11-18 14:12:12 +0100
committerBastien Nocera <hadess@hadess.net>2016-11-18 14:17:15 +0100
commita176b9a03d04d3f1657ffeaff51bd72a2e6ae21e (patch)
tree99e0269d21b00a52e6f185511bf00604867a4e01 /tests
parent2802eaf147dd7c2cb6ce3a3d2a39b612b5cc35f6 (diff)
downloadgrilo-a176b9a03d04d3f1657ffeaff51bd72a2e6ae21e.tar.gz
tests: Fix common test failures in GrlNet tests
The small-delay test fails fairly often, about 20% of the time for a 100 runs on a non-loaded machine. We're lucky it doesn't happen more often, as g_timeout_add_seconds() as used by grl-net-wc.c will ultimately use g_timeout_set_expiration() which might round up timeouts to the above second: " /* We want the microseconds part of the timeout to land on the * 'timer_perturb' mark, but we need to make sure we don't try to * set the timeout in the past. We do this by ensuring that we * always only *increase* the expiration time by adding a full * second in the case that the microsecond portion decreases. */ " Bump the threshold to a full second, and verify by raising the delay in the grl_net_wc_set_throttling() call. https://bugzilla.gnome.org/show_bug.cgi?id=774578
Diffstat (limited to 'tests')
-rw-r--r--tests/lib-net.c2
1 files changed, 1 insertions, 1 deletions
diff --git a/tests/lib-net.c b/tests/lib-net.c
index 5830729..4fa16b7 100644
--- a/tests/lib-net.c
+++ b/tests/lib-net.c
@@ -44,7 +44,7 @@ typedef struct {
#define DELAY 2
#define BIG_DELAY 15 /* Big enough for a timeout */
-#define THRESHOLD (G_USEC_PER_SEC / 10) /* 0.1 second */
+#define THRESHOLD 1 /* 1 second, see https://bugzilla.gnome.org/show_bug.cgi?id=774578 */
#define NUM_STRESS_TEST 100