From: Heikki Linnakangas <heikki.linnakangas@iki.fi>
Date: Sun, 15 Jan 2012 20:03:09 +0000 (+0200)
Subject: Fix poll() implementation of WaitLatchOrSocket to notice postmaster death.
X-Git-Tag: REL9_2_BETA1~581
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=b2b4af535eb733ba0c2ea6eeb2b14cac7f1ca4be;p=postgresql

Fix poll() implementation of WaitLatchOrSocket to notice postmaster death.

When the remote end of the pipe is closed, select() reports the fd as
readable, but poll() has a separate POLLHUP return code for that.

Spotted by Peter Geoghegan.
---

diff --git a/src/backend/port/unix_latch.c b/src/backend/port/unix_latch.c
index fc1a579ad2..10bf2dbec7 100644
--- a/src/backend/port/unix_latch.c
+++ b/src/backend/port/unix_latch.c
@@ -310,8 +310,13 @@ WaitLatchOrSocket(volatile Latch *latch, int wakeEvents, pgsocket sock,
 		{
 			result |= WL_SOCKET_WRITEABLE;
 		}
+		/*
+		 * We expect a POLLHUP when the remote end is closed, but because we
+		 * don't expect the pipe to become readable or to have any errors
+		 * either, treat those as postmaster death, too.
+		 */
 		if ((wakeEvents & WL_POSTMASTER_DEATH) &&
-			(pfds[nfds - 1].revents & POLLIN))
+			(pfds[nfds - 1].revents & (POLLHUP | POLLIN | POLLERR | POLLNVAL)))
 		{
 			result |= WL_POSTMASTER_DEATH;
 		}