git.samba.org
/
ira
/
wip.git
/ commitdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
| commitdiff |
tree
raw
|
patch
| inline |
side by side
(parent:
09753bf
)
r7431: this should fix the bug that mkaplan and I noticed which is that
author
Andrew Tridgell
<tridge@samba.org>
Thu, 9 Jun 2005 08:58:12 +0000
(08:58 +0000)
committer
Gerald (Jerry) Carter
<jerry@samba.org>
Wed, 10 Oct 2007 18:17:52 +0000
(13:17 -0500)
BENCH-NBENCH sometimes incorrectly reports that it failed.
(This used to be commit
167c29bec6f28e58b5d36325b4285748a47adc58
)
source4/torture/nbench/nbench.c
patch
|
blob
|
history
diff --git
a/source4/torture/nbench/nbench.c
b/source4/torture/nbench/nbench.c
index 8e19461b9559417e94e7a1d2e49fbe2cbaed72c8..9b3c9cdb5ac4120e8bbbcf140d5ee77ce77cbd68 100644
(file)
--- a/
source4/torture/nbench/nbench.c
+++ b/
source4/torture/nbench/nbench.c
@@
-198,6
+198,14
@@
BOOL torture_nbench(void)
printf("Running for %d seconds with load '%s' and warmup %d secs\n",
timelimit, loadfile, warmup);
+ /* we need to reset SIGCHLD here as the name resolution
+ library may have changed it. We rely on correct signals
+ from childs in the main torture code which reaps
+ children. This is why smbtorture BENCH-NBENCH was sometimes
+ failing */
+ signal(SIGCHLD, SIG_DFL);
+
+
signal(SIGALRM, nb_alarm);
alarm(1);
torture_create_procs(run_netbench, &correct);