Make archive recovery always start a new timeline, rather than only when a

recovery stop time was used.  This avoids a corner-case risk of trying to
overwrite an existing archived copy of the last WAL segment, and seems
simpler and cleaner all around than the original definition.  Per example
from Jon Colverson and subsequent analysis by Simon.

doc/src/sgml/backup.sgml

@@ -1,5 +1,5 @@
 <!--
-$PostgreSQL: pgsql/doc/src/sgml/backup.sgml,v 2.75.2.2 2006/11/04 18:20:40 tgl Exp $
+$PostgreSQL: pgsql/doc/src/sgml/backup.sgml,v 2.75.2.3 2007/09/29 01:36:29 tgl Exp $
 -->
 <chapter id="backup">
  <title>Backup and Restore</title>
@@ -1084,11 +1084,9 @@ restore_command = 'copy /mnt/server/archivedir/%f "%p"'  # Windows
 
    <para>
     To deal with these problems, <productname>PostgreSQL</> has a notion
-    of <firstterm>timelines</>.  Each time you recover to a point-in-time
-    earlier than the end of the WAL sequence, a new timeline is created
-    to identify the series of WAL records generated after that recovery.
-    (If recovery proceeds all the way to the end of WAL, however, we do not
-    start a new timeline: we just extend the existing one.)  The timeline
+    of <firstterm>timelines</>.  Whenever an archive recovery is completed,
+    a new timeline is created to identify the series of WAL records
+    generated after that recovery.  The timeline
     ID number is part of WAL segment file names, and so a new timeline does
     not overwrite the WAL data generated by previous timelines.  It is
     in fact possible to archive many different timelines.  While that might

src/backend/access/transam/xlog.c

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.222.2.5 2007/08/04 01:42:34 tgl Exp $
+ * $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.222.2.6 2007/09/29 01:36:29 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -4155,7 +4155,8 @@ exitArchiveRecovery(TimeLineID endTLI, uint32 endLogId, uint32 endLogSeg)
 	 *
 	 * Note that if we are establishing a new timeline, ThisTimeLineID is
 	 * already set to the new value, and so we will create a new file instead
-	 * of overwriting any existing file.
+	 * of overwriting any existing file.  (This is, in fact, always the case
+	 * at present.)
 	 */
 	snprintf(recoveryPath, MAXPGPATH, XLOGDIR "/RECOVERYXLOG");
 	XLogFilePath(xlogpath, ThisTimeLineID, endLogId, endLogSeg);
@@ -4330,7 +4331,7 @@ StartupXLOG(void)
 	XLogCtlInsert *Insert;
 	CheckPoint	checkPoint;
 	bool		wasShutdown;
-	bool		needNewTimeLine = false;
+	bool		reachedStopPoint = false;
 	XLogRecPtr	RecPtr,
 				LastRec,
 				checkPointLoc,
@@ -4598,7 +4599,7 @@ StartupXLOG(void)
 				 */
 				if (recoveryStopsHere(record, &recoveryApply))
 				{
-					needNewTimeLine = true;		/* see below */
+					reachedStopPoint = true;		/* see below */
 					recoveryContinue = false;
 					if (!recoveryApply)
 						break;
@@ -4653,11 +4654,10 @@ StartupXLOG(void)
 	 */
 	if (XLByteLT(EndOfLog, recoveryMinXlogOffset))
 	{
-		if (needNewTimeLine)	/* stopped because of stop request */
+		if (reachedStopPoint)	/* stopped because of stop request */
 			ereport(FATAL,
 					(errmsg("requested recovery stop point is before end time of backup dump")));
-		else
-			/* ran off end of WAL */
+		else					/* ran off end of WAL */
 			ereport(FATAL,
 					(errmsg("WAL ends before end time of backup dump")));
 	}
@@ -4665,12 +4665,18 @@ StartupXLOG(void)
 	/*
 	 * Consider whether we need to assign a new timeline ID.
 	 *
-	 * If we stopped short of the end of WAL during recovery, then we are
-	 * generating a new timeline and must assign it a unique new ID.
-	 * Otherwise, we can just extend the timeline we were in when we ran out
-	 * of WAL.
+	 * If we are doing an archive recovery, we always assign a new ID.  This
+	 * handles a couple of issues.  If we stopped short of the end of WAL
+	 * during recovery, then we are clearly generating a new timeline and must
+	 * assign it a unique new ID.  Even if we ran to the end, modifying the
+	 * current last segment is problematic because it may result in trying
+	 * to overwrite an already-archived copy of that segment, and we encourage
+	 * DBAs to make their archive_commands reject that.  We can dodge the
+	 * problem by making the new active segment have a new timeline ID.
+	 *
+	 * In a normal crash recovery, we can just extend the timeline we were in.
 	 */
-	if (needNewTimeLine)
+	if (InArchiveRecovery)
 	{
 		ThisTimeLineID = findNewestTimeLine(recoveryTargetTLI) + 1;
 		ereport(LOG,