On Wed, Dec 30, 2009 at 07:23:00PM +0000, konrad merz wrote:
> Hey,
>
> during writing on icb (client) I discoverd a bug by sending /topic w/o an
> argument to the daemon, which leads to a segfault.
i do not think we should wait till this gets "reimplemented"
and fix it first. then we can wait till it's "reimplemented"
i do not see anything wrong w/ the fix unless mikeb can
argue otherwise...
> The segfault of one process leads to a endless loop of the dns which allways
> returns to the 'dns read'.
yes the exit is better since any kind of failure indicates problem
at the parent process... (short read or failure or eof)
Index: usr.sbin/icbd/cmd.c
===================================================================
RCS file: /cvs/src/usr.sbin/icbd/cmd.c,v
retrieving revision 1.19
diff -u -p -r1.19 cmd.c
--- usr.sbin/icbd/cmd.c 23 Jun 2009 13:39:33 -0000 1.19
+++ usr.sbin/icbd/cmd.c 30 Dec 2009 19:06:18 -0000
@@ -227,9 +227,15 @@ icb_cmd_topic(struct icb_cmdarg *ca)
char buf[ICB_MAXGRPLEN + ICB_MAXTOPICLEN + 25];
struct icb_group *ig = ca->sess->group;
- if (strlen(ca->arg) == 0) {
- (void)snprintf(buf, sizeof buf, "Topic for %s is \"%s\"",
- ig->name, ig->topic);
+ if (ca->arg == NULL) {
+ if (strlen(ig->topic) > 0)
+ (void)snprintf(buf, sizeof buf,
+ "Topic for %s is \"%s\"",
+ ig->name, ig->topic);
+ else
+ (void)snprintf(buf, sizeof buf, "No topic set for %s",
+ ig->name);
+
icb_status(ca->sess, STATUS_TOPIC, buf);
return;
}
Index: usr.sbin/icbd/dns.c
===================================================================
RCS file: /cvs/src/usr.sbin/icbd/dns.c,v
retrieving revision 1.1
diff -u -p -r1.1 dns.c
--- usr.sbin/icbd/dns.c 23 Jun 2009 13:39:33 -0000 1.1
+++ usr.sbin/icbd/dns.c 30 Dec 2009 19:06:18 -0000
@@ -114,7 +114,7 @@ dns_dispatch(int fd, short event, void *
if (read(fd, &ss, ss_len) != ss_len) {
syslog(LOG_ERR, "dns read: %m");
- return;
+ exit(1);
}
if ((gerr = getnameinfo(sa, sa->sa_len,
