Shuqi Liang wrote:

Before integrating the 'git diff-files' builtin
with the sparse index feature, add tests to
t1092-sparse-checkout-compatibility.sh to ensure it currently works
with sparse-checkout and will still work with sparse index
after that integration.

When adding tests against a sparse-checkout
definition, we test two modes: all changes are
within the sparse-checkout cone and some changes are outside
the sparse-checkout cone.

In order to have staged changes outside of
the sparse-checkout cone, make a directory called 'folder1' and
copy `a` into 'folder1/a'. 'folder1/a' is identical to `a` in the base
commit. These make 'folder1/a' in the index, while leaving it outside of
the sparse-checkout definition. Test 'folder1/a'being present on-disk
without modifications, then change content inside 'folder1/a' in order
to test 'folder1/a' being present on-disk with modifications.

The word wrapping on this message (and your other commits/cover letter) is a
bit strange. By convention, it should be consistently wrapped to 72 columns
per line. Most text editors have some way of configuring that so you don't
need to do it manually.

Makes sense. In a sparse-checkout, folder2/a isn't in the working tree, so 
'diff-files' (which compares working tree vs. index) doesn't really apply.

+
+	write_script edit-contents <<-\EOF &&
+	echo text >>"$1"
+	EOF
+
+	# Add file to the index but outside of cone for sparse-checkout cases.
+	# Add file to the index without sparse-checkout cases to ensure all have 
+	# same output.
+	run_on_all mkdir folder1 &&

The test is failing because of this line. It should be 'mkdir -p folder1';
as you have it now, the command fails because 'folder1' already exists in
'full-checkout'. Alternatively, you could use 'run_on_sparse mkdir folder1',
but using '-p' seems like the less fragile approach. 

For reference, the way I figured that out was to run 't1092' as follows:

	cd t && ./t1092-sparse-checkout-compatibility.sh -xvdi --run=1,82

What the options correspond to:
* -x: print the commands called in the test (equivalent of calling 'set -x'
      in the test script).
* -v: print stdout/stderr to the console.
* -d: do not remove the "trash directory" of the test.
* -i: stop test execution once it encounters a failure.
* --run: run only the specified tests (1 is the first test - 'setup' - and
         82 is 'diff-files with pathspec outside sparse definition').

+	run_on_all cp a folder1/a &&
+
+	# file present on-disk without modifications
+	test_all_match git diff-files &&
+	test_all_match git diff-files folder1/a &&

The strange thing is, once I fixed the 'mkdir' issue in my local copy of
these patches, these 'test_all_match diff-files' calls succeeded. It turns
out that 'git diff-files' in the 'full-checkout', like in 'sparse-checkout',
reports a difference in 'folder1/a' that doesn't actually exist. So the bug
isn't in sparse-checkout as I initially assumed [1], but rather in
diff-files itself.

At this point, I'd say the diff-files bug is out-of-scope of this sparse
index integration; in your implementation, sparse-checkout and sparse index
work the same as a full checkout, it's just that the "normal" full checkout
behavior is wrong. My recommendation would be that you keep this test as-is
and, to force the failure, add a check that 'full-checkout-out' is empty.
Then, in a "NEEDSWORK" comment on the test (like the one on 'diff with
renames and conflicts') that indicates that the 'diff-files' behavior is
wrong.

I'll try to make some time this week to look into the 'diff-files' bug.
Sorry for the back-and-forth and distraction from your sparse index
integration. Other than the 'mkdir' issue, these updated tests look great!

[1] https://lore.kernel.org/git/b537d855-edb7-4f67-de08-d651868247a5@github.com/ (local)

+
+	# file present on-disk with modifications
+	run_on_all ../edit-contents folder1/a &&
+	test_all_match git diff-files &&
+	test_all_match git diff-files folder1/a
+'
+
 test_done