Added notifications about error-exit values:

- The receiver notifies the generator if it is exiting with an error,
  and then, if it is a server, waits around for the generator to die.
  This ensures that the client side has time to read the error.
- The generator or sender will notifiy the other side of the transfer of
  an error-exit value if protocol 31 is in effect.  This will get rid of
  some "connection unexpectedly closed" errors that are really expected
  events due to a fatal exit on the other side.

cleanup.c

@@ -24,9 +24,12 @@
 
 extern int am_server;
 extern int am_daemon;
+extern int am_sender;
+extern int am_generator;
 extern int io_error;
 extern int keep_partial;
 extern int got_xfer_error;
+extern int protocol_version;
 extern int output_needs_newline;
 extern char *partial_dir;
 extern char *logfile_name;
@@ -127,6 +130,17 @@ NORETURN void _exit_cleanup(int code, const char *file, int line)
 				code, file, line);
 		}
 
+		/* FALLTHROUGH */
+#include "case_N.h"
+
+		if (exit_code && exit_code != RERR_RCVR_ERROR
+		 && exit_code != RERR_STREAMIO && exit_code != RERR_SIGNAL1
+		 && (protocol_version >= 31 || (!am_sender && !am_generator))) {
+			send_msg_int(MSG_ERROR_EXIT, exit_code);
+			if (am_server && !am_sender && !am_generator)
+				noop_io_until_death();
+		}
+
 		/* FALLTHROUGH */
 #include "case_N.h"
 
@@ -160,7 +174,8 @@ NORETURN void _exit_cleanup(int code, const char *file, int line)
 		/* FALLTHROUGH */
 #include "case_N.h"
 
-		io_flush(FULL_FLUSH);
+		if (!code || am_server || (!am_sender && !am_generator))
+			io_flush(FULL_FLUSH);
 
 		/* FALLTHROUGH */
 #include "case_N.h"
@@ -184,7 +199,8 @@ NORETURN void _exit_cleanup(int code, const char *file, int line)
 				code = exit_code = RERR_PARTIAL;
 		}
 
-		if (code || am_daemon || (logfile_name && (am_server || !INFO_GTE(STATS, 1))))
+		if ((code && code != RERR_RCVR_ERROR)
+		 || am_daemon || (logfile_name && (am_server || !INFO_GTE(STATS, 1))))
 			log_exit(code, file, line);
 
 		/* FALLTHROUGH */

errcode.h

@@ -47,6 +47,8 @@
 #define RERR_TIMEOUT    30      /* timeout in data send/receive */
 #define RERR_CONTIMEOUT 35      /* timeout waiting for daemon connection */
 
+#define RERR_RCVR_ERROR 42      /* receiver is exiting with an error */
+
 /* Although it doesn't seem to be specified anywhere,
  * ssh and the shell seem to return these values:
  *

io.c

@@ -752,7 +752,7 @@ static char *perform_io(size_t needed, int flags)
 					/* Don't write errors on a dead socket. */
 					msgs2stderr = 1;
 					out->len = iobuf.raw_flushing_ends_before = out->pos = 0;
-					rsyserr(FERROR_SOCKET, errno, "write error");
+					rsyserr(FERROR_SOCKET, errno, "[%s] write error", who_am_i());
 					exit_cleanup(RERR_STREAMIO);
 				}
 			}
@@ -802,6 +802,16 @@ static char *perform_io(size_t needed, int flags)
 	return data;
 }
 
+void noop_io_until_death(void)
+{
+	char buf[1024];
+
+	kluge_around_eof = 1;
+
+	while (1)
+		read_buf(iobuf.in_fd, buf, sizeof buf);
+}
+
 /* Buffer a message for the multiplexed output stream.  Is never used for MSG_DATA. */
 int send_msg(enum msgcode code, const char *buf, size_t len, int convert)
 {
@@ -1426,6 +1436,16 @@ static void read_a_msg(void)
 			first_message = 0;
 		}
 		break;
+	case MSG_ERROR_EXIT:
+		if (msg_bytes != 4)
+			goto invalid_msg;
+		data = perform_io(4, PIO_INPUT_AND_CONSUME);
+		val = IVAL(data, 0);
+		if (am_generator && protocol_version >= 31)
+			send_msg_int(MSG_ERROR_EXIT, val);
+		if (am_generator)
+			val = RERR_RCVR_ERROR; /* avoids duplicate errors */
+		exit_cleanup(val);
 	default:
 		rprintf(FERROR, "unexpected tag %d [%s%s]\n",
 			tag, who_am_i(), inc_recurse ? "/inc" : "");

log.c

@@ -97,13 +97,14 @@ struct {
 	{ RERR_MALLOC     , "error allocating core memory buffers" },
 	{ RERR_PARTIAL    , "some files/attrs were not transferred (see previous errors)" },
 	{ RERR_VANISHED   , "some files vanished before they could be transferred" },
+	{ RERR_DEL_LIMIT  , "the --max-delete limit stopped deletions" },
 	{ RERR_TIMEOUT    , "timeout in data send/receive" },
 	{ RERR_CONTIMEOUT , "timeout waiting for daemon connection" },
+	{ RERR_RCVR_ERROR , "exiting due to receiver error" },
 	{ RERR_CMD_FAILED , "remote shell failed" },
 	{ RERR_CMD_KILLED , "remote shell killed" },
 	{ RERR_CMD_RUN    , "remote command could not be run" },
 	{ RERR_CMD_NOTFOUND,"remote command not found" },
-	{ RERR_DEL_LIMIT  , "the --max-delete limit stopped deletions" },
 	{ 0, NULL }
 };
 

rsync.h

@@ -98,7 +98,7 @@
 /* This is used when working on a new protocol version in CVS, and should
  * be a new non-zero value for each CVS change that affects the protocol.
  * It must ALWAYS be 0 when the protocol goes final (and NEVER before)! */
-#define SUBPROTOCOL_VERSION 10
+#define SUBPROTOCOL_VERSION 11
 
 /* We refuse to interoperate with versions that are not in this range.
  * Note that we assume we'll work with later versions: the onus is on
@@ -233,6 +233,7 @@ enum msgcode {
 	MSG_IO_ERROR=22,/* the sending side had an I/O error */
 	MSG_IO_TIMEOUT=33,/* tell client about a daemon's timeout value */
 	MSG_NOOP=42,	/* a do-nothing message */
+	MSG_ERROR_EXIT=86, /* used by siblings and by protocol-31 */
 	MSG_SUCCESS=100,/* successfully updated indicated flist index */
 	MSG_DELETED=101,/* successfully deleted a file on receiving side */
 	MSG_NO_SEND=102,/* sender failed to open a file we wanted */