# [PATCH v2] diff: fix --relative without arguments

3 messages from 2017-12-07 to 2017-12-07. Participants: Jacob Keller, Junio C Hamano.
Thread: https://gitlist.dev/t/47392

## Jacob Keller, 2017-12-07 00:59

Subject: [PATCH v2] diff: fix --relative without arguments
Message-ID: <20171207005929.24109-1-jacob.e.keller@intel.com>
URL: https://gitlist.dev/e/20171207005929.24109-1-jacob.e.keller%40intel.com

```
From: Jacob Keller <jacob.keller@gmail.com>

Recently, commit f7923a5ece00 ("diff: use skip_to_optional_val()",
2017-12-04) changed how we parsed some diff options, preferring
skip_to_optional_val instead of a more complex skip_prefix() setup.

Unfortunately, this breaks --relative, because of the semantics of
skip_to_optional_val. If no optional_val is given, then the function
puts the empty string "" into the arg. Unfortunately, relative passes in
options->prefix as the location for the optional argument, and expects
this value to remain unchanged if no optional value was given.

This results in breaking --relative when no arguments are provided. This
cascades into many failures of code flow that rely on this.

Partially revert the commit, and restore --relative option parsing to
the previous working code.

Since no tests exist for --relative without options, add a new test
which will help ensure no future regressions.

Signed-off-by: Jacob Keller <jacob.keller@gmail.com>
---
Actually perform the revert of --relative handling.

 diff.c                   | 6 +++++-
 t/t4045-diff-relative.sh | 5 +++++
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/diff.c b/diff.c
index b62e4b61b564..1919afb4f104 100644
--- a/diff.c
+++ b/diff.c
@@ -4581,8 +4581,12 @@ int diff_opt_parse(struct diff_options *options,
 		options->flags.rename_empty = 1;
 	else if (!strcmp(arg, "--no-rename-empty"))
 		options->flags.rename_empty = 0;
-	else if (skip_to_optional_val(arg, "--relative", &options->prefix))
+	else if (!strcmp(arg, "--relative"))
 		options->flags.relative_name = 1;
+	else if (skip_prefix(arg, "--relative=", &arg)) {
+		options->flags.relative_name = 1;
+		options->prefix = arg;
+	}
 
 	/* xdiff options */
 	else if (!strcmp(arg, "--minimal"))
diff --git a/t/t4045-diff-relative.sh b/t/t4045-diff-relative.sh
index 3950f5034d31..41e4f59b2ffb 100755
--- a/t/t4045-diff-relative.sh
+++ b/t/t4045-diff-relative.sh
@@ -70,4 +70,9 @@ for type in diff numstat stat raw; do
 	check_$type dir/file2 --relative=sub
 done
 
+cd subdir
+for type in diff numstat stat raw; do
+	check_$type file2 --relative
+done
+
 test_done
-- 
2.15.1.477.g3ed0a2a61da8


```

## Junio C Hamano, 2017-12-07 15:26

Subject: Re: [PATCH v2] diff: fix --relative without arguments
Message-ID: <xmqq4lp2ftjt.fsf@gitster.mtv.corp.google.com>
URL: https://gitlist.dev/e/xmqq4lp2ftjt.fsf%40gitster.mtv.corp.google.com
In-Reply-To: <20171207005929.24109-1-jacob.e.keller@intel.com>

```
Jacob Keller <jacob.e.keller@intel.com> writes:

> From: Jacob Keller <jacob.keller@gmail.com>
>
> Recently, commit f7923a5ece00 ("diff: use skip_to_optional_val()",
> 2017-12-04) changed how we parsed some diff options, preferring
> skip_to_optional_val instead of a more complex skip_prefix() setup.

I'd expect a moral equivalent of this squashed in when Christian
rerolls the skip-to-optional-val series (with helped-by: attribution
to you, probably).  It does not make much sense to leave the initial
breakage like this one in the history.

Thanks.

```

## Jacob Keller, 2017-12-07 19:01

Subject: Re: [PATCH v2] diff: fix --relative without arguments
Message-ID: <CA+P7+xpAFJcvcQ9PN5ZsnQxf8GpSG2yYKSYEg5wHebyCLtvdsw@mail.gmail.com>
URL: https://gitlist.dev/e/CA%2BP7%2BxpAFJcvcQ9PN5ZsnQxf8GpSG2yYKSYEg5wHebyCLtvdsw%40mail.gmail.com
In-Reply-To: <xmqq4lp2ftjt.fsf@gitster.mtv.corp.google.com>

```
On Thu, Dec 7, 2017 at 7:26 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Jacob Keller <jacob.e.keller@intel.com> writes:
>
>> From: Jacob Keller <jacob.keller@gmail.com>
>>
>> Recently, commit f7923a5ece00 ("diff: use skip_to_optional_val()",
>> 2017-12-04) changed how we parsed some diff options, preferring
>> skip_to_optional_val instead of a more complex skip_prefix() setup.
>
> I'd expect a moral equivalent of this squashed in when Christian
> rerolls the skip-to-optional-val series (with helped-by: attribution
> to you, probably).  It does not make much sense to leave the initial
> breakage like this one in the history.
>
> Thanks.

Quite correct. I'll be sending a patch with just new tests shortly.

Thanks,
Jake

```
