>From f79b643dd4d6858a3bca4f4b1e0e0e8953409022 Mon Sep 17 00:00:00 2001
From: Nils Goroll <[email protected]>
Date: Tue, 11 Oct 2011 21:14:28 +0200
Subject: [PATCH] synchronize ccf_listen_address() on hack_ready
Partly fixes #865
This fixes a race condition between the cli thread and the acceptor
thread where the cli thread would respond to a debug.listen_address
command before the acceptor thread got around to call listen() on all
accept sockets.
This was probably only relevant for VTC tests: False negatives which
would surface for any test which connected to the Varnish server with
a low probability. Seen on Solaris.
---
bin/varnishd/cache_acceptor.c | 8 ++++++++
1 files changed, 8 insertions(+), 0 deletions(-)
diff --git a/bin/varnishd/cache_acceptor.c b/bin/varnishd/cache_acceptor.c
index e6ad6e3..a3d7f73 100644
--- a/bin/varnishd/cache_acceptor.c
+++ b/bin/varnishd/cache_acceptor.c
@@ -376,6 +376,14 @@ ccf_listen_address(struct cli *cli, const char * const
*av, void *priv)
(void)cli;
(void)av;
(void)priv;
+
+ /*
+ * don't tell the vtc child where to connect to before we're
+ * ready - otherwise it may die with an ECONNREFUSED (seen on Solaris)
+ */
+ while(!hack_ready)
+ (void)usleep(100*1000);
+
VTAILQ_FOREACH(ls, &heritage.socks, list) {
if (ls->sock < 0)
continue;
--
1.5.6.5
_______________________________________________
varnish-dev mailing list
[email protected]
https://www.varnish-cache.org/lists/mailman/listinfo/varnish-dev
