@@ -29,8 +29,12 @@ set_and_save_state () {
# verify_state <path> <expected-worktree-content> <expected-index-content>
verify_state () {
- test "$(cat "$1")" = "$2" &&
- test "$(git show :"$1")" = "$3"
+ echo "$2" >expect &&
+ test_cmp expect "$1" &&
+
+ echo "$3" >expect &&
+ git show :"$1" >actual &&
+ test_cmp expect actual
}
# verify_saved_state <path>
@@ -46,5 +50,6 @@ save_head () {
}
verify_saved_head () {
- test "$(cat _head)" = "$(git rev-parse HEAD)"
+ git rev-parse HEAD >actual &&
+ test_cmp _head actual
}
Fix code added in b319ef70a94 (Add a small patch-mode testing library, 2009-08-13) to use &&-chaining. This avoids losing both the exit code of a "git" and the "cat" processes. This fixes cases where we'd have e.g. missed memory leaks under SANITIZE=leak, this code doesn't leak now as far as I can tell, but I discovered it while looking at leaks in related code. For "verify_saved_head()" we could make use of "test_cmp_rev" with some changes, but it uses "git rev-parse --verify", and this existing test does not. I think it could safely use it, but let's avoid the while-at-it change, and narrowly fix the exit code problem. Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> --- t/lib-patch-mode.sh | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)