Re: [PATCH 8/8] tree_entry_interesting(): support negative pathspec

Hi,

2010/9/8 Nguyễn Thái Ngọc Duy [off-list ref]:

+
+test_expect_success 'diff' '
+       cat >expected <<-\EOF &&
+:100644 100644 e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 d00491fd7e5bb6fa28c517a0bb32b8b506539d4d M     one/file
+:100644 100644 e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 d00491fd7e5bb6fa28c517a0bb32b8b506539d4d M     one/two/file
+:100644 100644 e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 d00491fd7e5bb6fa28c517a0bb32b8b506539d4d M     one/zoo
+EOF
+       git diff-tree -r HEAD^ HEAD -- one >result &&
+       test_cmp expected result
+'

Having the full mode & sha1sum seems to make the test harder to read
than necessary.  Perhaps you could use printf or git ls-files together
with --name-only to simplify and shorten these a bit?  Example
alternative versions for three of your testcases:

test_expect_success 'diff' '
	git ls-files one >expected &&
	git diff --name-only HEAD^ HEAD -- one >result &&
	test_cmp expected result
'

test_expect_success 'diff one ^one/two' '
	git ls-files one | grep -v one/two >expected &&
	git diff --name-only HEAD^ HEAD -- one ^one/two >result &&
	test_cmp expected result
'

test_expect_success 'diff ^one/two' '
	printf "file\none/file\none/zoo\n" >expected &&
	git diff --name-only HEAD^ HEAD -- ^one/two >result &&
	test_cmp expected result
'

Thoughts?