Re: [PATCH v6 1/2] t1092: add tests for `git diff-files`

From: Victoria Dye <vdye@github.com>
Date: Mar 21, 2023, 22:19 UTC
Message-ID: <3e2371e9-2d7d-27c4-58dd-296fcee49e88@github.com>
In-Reply-To: <xmqq1qlhhk4s.fsf@gitster.g>

Junio C Hamano wrote:

Show 10 quoted lines

> Victoria Dye <vdye@github.com> writes:
> 
>> 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.
> 
> Is that a bug, or just a common "ah, you forgot to refresh the index"?

Ah, you're right - I completely forgot about 'diff-files' not refreshing the index (since 'diff', by default, does). The ctime is (often, but not always) different on the copied file than what's in the index, so 'diff-files' shows the file as "modified" if the index isn't refreshed.

Going back to these tests, the goal is to make sure that 'diff-files' finds the correct index entry (possibly in a sparse directory) and compares that correctly to what's on disk. But since we want to ignore ctime differences, we could use '--stat' (or '-p', or '--num-stat':

	run_on_all mkdir -p folder1 &&
	run_on_all cp a folder1/a &&

	# file present on-disk without modifications
	# use `--stat` to ignore file creation time differences in
	# unrefreshed index
	test_all_match git diff-files --stat &&
	test_all_match git diff-files --stat folder1/a &&

I don't think that makes the test any less comprehensive (especially since later on in the same test, we modify the contents of 'folder1/a' and get the expected "modified" status in 'git diff-files' without '--stat'), but it avoids potential breakages related to inconsistency in file creation time.