Re: [PATCH v2] apply.c: fix -p argument parsing

From: Junio C Hamano <gitster@pobox.com>
Date: Mar 10, 2026, 03:31 UTC
Message-ID: <xmqqwlzkxsv5.fsf@gitster.g>
In-Reply-To: <20260310005408.2022216-1-mroik@delayed.space>

Mirko Faina <mroik@delayed.space> writes:

Show 13 quoted lines

> "git apply" has an option -p that takes an integer as its argument.
> Unfortunately the function apply_option_parse_p() in charge of parsing
> this argument uses atoi() to convert from string to integer, which
> allows a non-digit after the number (e.g. "1q") to be silently ignored.
> As a consequence, an argument that does not begin with a digit silently
> becomes a zero. Despite this command working fine when a non-positive
> argument is passed, it might be useful for the end user to know that
> their input contains non-digits that might've been unintended.
>
> Replace atoi() with strtol_i() to catch malformed inputs.
>
> Signed-off-by: Mirko Faina <mroik@delayed.space>
> ---

Show 5 quoted lines

>  apply.c                 |  3 ++-
>  t/t4103-apply-binary.sh | 19 +++++++++++++++++++
>  t/t4103/patch           | 16 ++++++++++++++++
>  3 files changed, 37 insertions(+), 1 deletion(-)
>  create mode 100644 t/t4103/patch

Curious. It is true that we need to parse the p_value correctly even when we are applying a binary patch, but the problem is not limited to binary patches, is it?

Show 14 quoted lines

> diff --git a/apply.c b/apply.c
> index b6dd1066a0..61df3bdcd0 100644
> --- a/apply.c
> +++ b/apply.c
> @@ -4981,7 +4981,8 @@ static int apply_option_parse_p(const struct option *opt,
>  
>  	BUG_ON_OPT_NEG(unset);
>  
> -	state->p_value = atoi(arg);
> +	if (strtol_i(arg, 10, &state->p_value) < 0 || state->p_value < 0)
> +		die("<num> has to be a non-negative integer");
>  	state->p_value_known = 1;
>  	return 0;
>  }

Sounds sensible.

I briefly wondered if it would have negative fallouts to change the type of .p_value member to "unsigned int" and use strtol_ui() to parse it, but the amount of work this part of the code needs to do does not change that much, so such a change is of dubious value. It looks like the above draws the line at the right place to stop.

Great execution.

Show 14 quoted lines

> diff --git a/t/t4103-apply-binary.sh b/t/t4103-apply-binary.sh
> index 8e302a5a57..d9dc884946 100755
> --- a/t/t4103-apply-binary.sh
> +++ b/t/t4103-apply-binary.sh
> @@ -53,6 +53,25 @@ test_expect_success 'setup' '
>  	)
>  '
>  
> +test_expect_success 'git apply -p 1 patch' '
> +	test_when_finished "rm -rf result t" &&
> +	git apply -p 1 $TEST_DIRECTORY/t4103/patch &&
> +	ls -l | sed -e "/[[:space:]]t$/!d" >result &&
> +	test_line_count = 1 result
> +'

Is this saying "in the directory there must be only a single file whose name is t?" Wouldn't it be more readable and direct to do something like

	test_path_is_dir t

or is there something more subtle going on here?

Show 7 quoted lines

> +test_expect_success 'git apply -p malformed patch' '
> +	test_must_fail git apply -p malformed $TEST_DIRECTORY/t4103/patch
> +'
>
> +test_expect_success 'git apply -p 2q patch' '
> +	test_must_fail git apply -p 2q $TEST_DIRECTORY/t4103/patch
> +'

If this did not fail and patch gets applied with some p_value that happens to be used when we fail to parse the number, then ...

> +test_expect_success 'git apply -p -1 patch' '
> +	test_must_fail git apply -p -1 $TEST_DIRECTORY/t4103/patch
> +'

... it would not be clear why this step fails. Perhaps with that same "unable to parse" p_value was used and this tried to create the same file as the previous step already created, or we detected parse failure. We cannot tell.

It probably is a good idea to prepare for the worst by doing something silly like

	test_when_finished "rm -f t/test/test test/test test" &&

at the beginning of each of these tests so that we would clean up whatever we could leave behind? I dunno.

Show 25 quoted lines

>  test_expect_success 'stat binary diff -- should not fail.' \
>  	'git checkout main &&
>  	 git apply --stat --summary B.diff'
> diff --git a/t/t4103/patch b/t/t4103/patch
> new file mode 100644
> index 0000000000..c4511bb708
> --- /dev/null
> +++ b/t/t4103/patch
> @@ -0,0 +1,16 @@
> +From 90ad11d5b2d437e82d4d992f72fb44c2227798b5 Mon Sep 17 00:00:00 2001
> +From: Mroik <mroik@delayed.space>
> +Date: Mon, 9 Mar 2026 23:25:00 +0100
> +Subject: [PATCH] Test
> +
> +---
> + t/test/test | 0
> + 1 file changed, 0 insertions(+), 0 deletions(-)
> + create mode 100644 t/test/test
> +
> +diff --git a/t/test/test b/t/test/test
> +new file mode 100644
> +index 0000000000..e69de29bb2
> +-- 
> +2.53.0.851.ga537e3e6e9
> +