On 13/11/2024 01:07, Junio C Hamano wrote:

kristofferhaugsbakk@fastmail.com writes:

quoted

+test_expect_success 'git revert --reference with core.commentChar' '
+	test_when_finished "git reset --hard to-ident" &&
+	git checkout --detach to-ident &&
+	git -c core.commentChar=% revert \
+		--edit --reference HEAD &&
+	git log -1 --format=%B HEAD >actual &&
+	printf "This reverts commit $(git show -s \
+ 		--pretty=reference HEAD^).\n\n" \
+		>expect &&
+	test_cmp expect actual
+'

I guess this fails by leaving the "# *** SAY WHY" in the resulting
message, because the stripspace wants to see '%' to start commented
out lines to be stripped?  If we inspect with this test what the
temporary file we give to the editor looks like to make sure that
'%' is used for commenting, that would be a more direct test, but
without going that far, at least can we have a comment describing
how this is expected to fail without the fix?

For me something like

	GIT_EDITOR="cat >actual" git -c core.commentChar=% revert \
		--edit --reference HEAD &&
	test_grep "^% \*\*\* SAY WHY WE ARE REVERTING THE COMMIT \*\*\*" \
		  actual

Would be a more convincing test as it actually checks that the user sees
the line that we expect strbuf_stripspace() to remove from the final
message. If we want to check the commit message as well that's fine but
I'm not sure its necessary. (if we do we should use test_commit_message
like patch 3)

Best Wishes

Phillip