rsync: race condition can cause loss of diagnostic output

[Dick Streefland]

On Wednesday 2002-05-22 19:00, Charles C. Fu ?????? wrote:
| [This is a copy of the contents of Debian bug report #147842.]

Sorry, but the patch in my earlier reply was incorrect. An essential
continue statement was missing. I also found out that it was not a
complete solution to the problem. Here is a revised patch:

--- rsync-2.5.5/io.c.orig	Fri Mar 22 06:14:44 2002
+++ rsync-2.5.5/io.c	Tue May 28 18:49:52 2002
@@ -214,6 +214,7 @@
 
 		if (io_error_fd != -1 && FD_ISSET(io_error_fd, &fds)) {
 			read_error_fd();
+			continue;
 		}
 
 		if (!FD_ISSET(fd, &fds)) continue;
@@ -435,6 +436,7 @@
 
 		if (io_error_fd != -1 && FD_ISSET(io_error_fd, &r_fds)) {
 			read_error_fd();
+			continue;
 		}
 
 		if (FD_ISSET(fd, &w_fds)) {
--- rsync-2.5.5/log.c.orig	Mon Feb 18 20:51:12 2002
+++ rsync-2.5.5/log.c	Tue May 28 18:51:01 2002
@@ -116,7 +116,14 @@
 		int n = write(log_error_fd, el->buf+el->written, el->len - el->written);
 		/* don't check for an error if the best way of handling the error is
 		   to ignore it */
-		if (n == -1) break;
+		if (n == -1) {
+			if (errno == EAGAIN)
+			{
+				msleep(20);
+				continue;
+			}
+			break;
+		}
 		if (n > 0) {
 			el->written += n;
 		}

-- 
Dick Streefland                      ////             Altium Software BV
dick.streefland at altium.nl           (@ @)          http://www.altium.com
--------------------------------oOO--(_)--OOo---------------------------