test_must_be_empty: make sure the file exists, not just empty

The helper function test_must_be_empty is meant to make sure the
given file is empty, but its implementation is:

	if test -s "$1"
	then
		... not empty, we detected a failure ...
	fi

Surely, the file having non-zero size is a sign that the condition
"the file must be empty" is violated, but it misses the case where
the file does not even exist.  It is an accident waiting to happen
with a buggy test like this:

	git frotz 2>error-message &&
	test_must_be_empty errro-message

that won't get caught until you deliberately break 'git frotz' and
notice why the test does not fail.

Signed-off-by: Junio C Hamano <gitster@pobox.com>

diff --git a/t/test-lib-functions.sh b/t/test-lib-functions.sh
index 1701fe2..d2eaf5a 100644 (file)
--- a/t/test-lib-functions.sh
+++ b/t/test-lib-functions.sh
@@ -718,7 +718,11 @@ verbose () {
 # otherwise.
 
 test_must_be_empty () {
-       if test -s "$1"
+       if ! test -f "$1"
+       then
+               echo "'$1' is missing"
+               return 1
+       elif test -s "$1"
        then
                echo "'$1' is not empty, it contains:"
                cat "$1"