Re: [PATCH] pathspec: die on empty strings as pathspec

[Junio C Hamano <gitster@pobox.com>]

Junio C Hamano <gitster@pobox.com> writes:

> I am not sure if we even want the dot there, but at least that is
> what the original author of the test intended to do when s/he
> decided to pass an empty string as the pathspec.
>
>  t/t0027-auto-crlf.sh | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

I'll queue the following _before_ your "final step", so that the
whole thing can be advanced to 'next' and hopefully to 'master' in
some future releases.  

Given that we ourselves did not even notice until now that one of
our scripts get broken by this "final step", even though we kept
issuing the warning message, we may want to re-think our overall
strategy for deprecation.  We've assumed that "keep behaving as
before but warn for a few years and then finally give a hard
failure" would be sufficient, but depending on how the script that
employ our programs hide the warnings from the end users, that may
not be a good enough transition strategy.

At the same time we re-think the deprecation strategy, we also need
to see if we can update our test framework to help us better.
Ideally, we could have caught this existing breakage of passing ""
as pathspec in the test soon after we went into "keep behaving as
before but warn" stage.  We didn't and found it out only when we
switched to "finally give a hard failure" stage.  That is very
unsatisfactory.

Needless to say, neither of the above two comes from _your_ change.
It's just something we need to improve in a larger scope of the
whole project I realized.

Thanks.

-- >8 --
Subject: [PATCH] t0027: do not use an empty string as a pathspec element

In an upcoming update, we will finally make an empty string illegal
as an element in a pathspec; it traditionally meant the same as ".",
i.e. include everything, so update this test that passes "" to pass
a dot instead.

At this point in the test sequence, there is no modified path that
need to be further added before committing; the working tree is
empty except for .gitattributes which was just added to the index.
So we could instead pass no pathspec, but this is a conversion more
faithful to the original.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 t/t0027-auto-crlf.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/t/t0027-auto-crlf.sh b/t/t0027-auto-crlf.sh
index 93725895a4..e41c9b3bb2 100755
--- a/t/t0027-auto-crlf.sh
+++ b/t/t0027-auto-crlf.sh
@@ -322,7 +322,7 @@ test_expect_success 'setup master' '
 	echo >.gitattributes &&
 	git checkout -b master &&
 	git add .gitattributes &&
-	git commit -m "add .gitattributes" "" &&
+	git commit -m "add .gitattributes" . &&
 	printf "\$Id: 0000000000000000000000000000000000000000 \$\nLINEONE\nLINETWO\nLINETHREE"     >LF &&
 	printf "\$Id: 0000000000000000000000000000000000000000 \$\r\nLINEONE\r\nLINETWO\r\nLINETHREE" >CRLF &&
 	printf "\$Id: 0000000000000000000000000000000000000000 \$\nLINEONE\r\nLINETWO\nLINETHREE"   >CRLF_mix_LF &&
-- 
2.13.1-678-g93553a431c