Some servers use ports_manage_port_operations_one_thread to process
requests and terminate when it returns. Since many of them don't detach
before shutting down, a client may receive an error if its request
arrived while the server is shutting down. Prevent those spurious errors
by forcing ports_manage_port_operations_one_thread not to return.
This is the same change as 235491231bdd1fd93507c835767503f047e10b91
introduced for ports_manage_port_operations_multithread.
* libports/manage-one-thread.c
(ports_manage_port_operations_one_thread): Force timeout to 0.
---
libports/manage-one-thread.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/libports/manage-one-thread.c b/libports/manage-one-thread.c
index 4ea740b..cbd2df7 100644
--- a/libports/manage-one-thread.c
+++ b/libports/manage-one-thread.c
@@ -85,7 +85,14 @@ ports_manage_port_operations_one_thread (struct port_bucket
*bucket,
return status;
}
-
+
+ /* XXX It is currently unsafe for most servers to terminate based on
+ inactivity because a request may arrive after a server has
+ started shutting down, causing the client to receive an error.
+ Prevent the service loop from terminating by setting TIMEOUT to
+ zero. */
+ timeout = 0;
+
do
err = mach_msg_server_timeout (internal_demuxer, 0, bucket->portset,
timeout ? MACH_RCV_TIMEOUT : 0, timeout);
--
1.9.2
