Repair longstanding error in btree xlog replay: XLogReadBuffer should be

passed extend = true whenever we are reading a page we intend to reinitialize
completely, even if we think the page "should exist".  This is because it
might indeed not exist, if the relation got truncated sometime after the
current xlog record was made and before the crash we're trying to recover
from.  These two thinkos appear to explain both of the old bug reports
discussed here:
http://archives.postgresql.org/pgsql-hackers/2005-05/msg01369.php

diff --git a/src/backend/access/nbtree/nbtxlog.c b/src/backend/access/nbtree/nbtxlog.c
index 95df7213ee73a8af5c88338e487bcf8e7959ab89..4803efa9e6a2ce270e6ae309a31b4821b76a8232 100644 (file)
--- a/src/backend/access/nbtree/nbtxlog.c
+++ b/src/backend/access/nbtree/nbtxlog.c
@@ -258,7 +258,7 @@ btree_xlog_split(bool redo, bool onleft, bool isroot,
        rightsib = (onleft) ? xlrec->otherblk : targetblk;
 
        /* Left (original) sibling */
-       buffer = XLogReadBuffer(false, reln, leftsib);
+       buffer = XLogReadBuffer(true, reln, leftsib);
        if (!BufferIsValid(buffer))
                elog(PANIC, "btree_split_%s: lost left sibling", op);
 
@@ -551,7 +551,7 @@ btree_xlog_delete_page(bool redo, bool ismeta,
        }
 
        /* Rewrite target page as empty deleted page */
-       buffer = XLogReadBuffer(false, reln, target);
+       buffer = XLogReadBuffer(true, reln, target);
        if (!BufferIsValid(buffer))
                elog(PANIC, "btree_delete_page_%s: lost target page", op);
        page = (Page) BufferGetPage(buffer);