Delete file so db_load doesn't run incrementally.

It turns out that db_load incrementally updates the database (instead
of writing a new one).  Remove the old database to force db_load to
write a new one.

This also ensures that we handle write failure gracefully.

Cherry picked from commit 67493bfca44d436143d825b155bf2bc38990ea82
and backported to stable-1.10 branch.

diff --git a/images/Makefile.am b/images/Makefile.am
index d45e699..e4a910a 100644 (file)
--- a/images/Makefile.am
+++ b/images/Makefile.am
@@ -174,7 +174,9 @@ fedora.img: guest-aux/make-fedora-img.sh guest-aux/fedora-name.db
        bash $<
 
 guest-aux/fedora-name.db: guest-aux/fedora-name.db.txt
-       $(DB_LOAD) $@ < $<
+       rm -f $@ $@-t
+       $(DB_LOAD) $@-t < $<
+       mv $@-t $@
 
 # Make a (dummy) Debian image.
 debian.img: guest-aux/make-debian-img.sh