diff --git a/ipc/chromium/src/third_party/libevent/kqueue.c b/ipc/chromium/src/third_party/libevent/kqueue.c
--- a/ipc/chromium/src/third_party/libevent/kqueue.c
+++ b/ipc/chromium/src/third_party/libevent/kqueue.c
@@ -158,26 +158,20 @@ kq_init(struct event_base *base)
 	base->evsigsel = &kqsigops;
 
 	return (kqueueop);
 err:
 	if (kqueueop)
 		kqop_free(kqueueop);
 
 	return (NULL);
 }
 
-static void
-kq_sighandler(int sig)
-{
-	/* Do nothing here */
-}
-
 #define ADD_UDATA 0x30303
 
 static void
 kq_setup_kevent(struct kevent *out, evutil_socket_t fd, int filter, short change)
 {
 	memset(out, 0, sizeof(struct kevent));
 	out->ident = fd;
 	out->filter = filter;
 
 	if (change & EV_CHANGE_ADD) {
@@ -431,24 +425,31 @@ kq_sig_add(struct event_base *base, int 
 	kev.ident = nsignal;
 	kev.filter = EVFILT_SIGNAL;
 	kev.flags = EV_ADD;
 
 	/* Be ready for the signal if it is sent any
 	 * time between now and the next call to
 	 * kq_dispatch. */
 	if (kevent(kqop->kq, &kev, 1, NULL, 0, &timeout) == -1)
 		return (-1);
 
-	/* XXXX The manpage suggest we could use SIG_IGN instead of a
-	 * do-nothing handler */
-	if (_evsig_set_handler(base, nsignal, kq_sighandler) == -1)
+	/* Backported from
+	 * https://github.com/nmathewson/Libevent/commit/148458e0a1fd25e167aa2ef229d1c9a70b27c3e9 */
+	/* We can set the handler for most signals to SIG_IGN and
+	 * still have them reported to us in the queue.  However,
+	 * if the handler for SIGCHLD is SIG_IGN, the system reaps
+	 * zombie processes for us, and we don't get any notification.
+	 * This appears to be the only signal with this quirk. */
+	if (_evsig_set_handler(base, nsignal,
+	                       nsignal == SIGCHLD ? SIG_DFL : SIG_IGN) == -1) {
 		return (-1);
+	}
 
 	return (0);
 }
 
 static int
 kq_sig_del(struct event_base *base, int nsignal, short old, short events, void *p)
 {
 	struct kqop *kqop = base->evbase;
 	struct kevent kev;