Make git-pull-script do the right thing for symlinked HEAD's.

Also exit gracefully if the HEAD pull failed, rather than use
a possibly stale MERGE_HEAD.

git-pull-script

@@ -5,11 +5,13 @@
 #
 merge_repo=$1
 
+rm -f .git/MERGE_HEAD
+
 echo "Getting object database"
 rsync -avz --ignore-existing $merge_repo/objects/. .git/objects/.
 
 echo "Getting remote head"
-rsync -avz $merge_repo/HEAD .git/MERGE_HEAD
+rsync -L $merge_repo/HEAD .git/MERGE_HEAD || exit 1
 
 head=$(cat .git/HEAD)
 merge_head=$(cat .git/MERGE_HEAD)