Peter Eisentraut wrote:

Andrew Dunstan writes:

So if the data directory previously existed and was empty, we don't
clean it out on error, even if we didn't use the noclean flag. Is this
intended behaviour or a bug? (If a bug it's trivially easy to fix.)

If the data directory already existed, we don't want to delete it. We should possibly delete the content, though.

yes, that's what I meant.

Here's a patch which should work, I think:

Index: src/bin/initdb/
RCS file: /projects/cvsroot/pgsql-server/src/bin/initdb/,v
retrieving revision 1.203
diff -c -w -r1.203
*** src/bin/initdb/ 27 Sep 2003 16:27:57 -0000 1.203
--- src/bin/initdb/ 1 Oct 2003 21:06:07 -0000
*** 44,49 ****
--- 44,53 ----
if [ "$made_new_pgdata" = yes ]; then
echo "$CMDNAME: removing data directory \"$PGDATA\"" 1>&2
rm -rf "$PGDATA" || echo "$CMDNAME: failed" 1>&2
+ else
+ echo "$CMDNAME: removing contents of data directory \"$PGDATA\"" 1>&2
+ rm -rf "$PGDATA"/* || echo "$CMDNAME: failed" 1>&2
echo "$CMDNAME: data directory \"$PGDATA\" not removed at user's request" 1>&2

---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
     joining column's datatypes do not match

Reply via email to