# [PATCH 1/2] expand_user_path: do not look at NULL path

2 messages from 2014-01-28 to 2014-01-28. Participants: Jeff King.
Thread: https://gitlist.dev/t/35745

## Jeff King, 2014-01-28 01:36

Subject: [PATCH 1/2] expand_user_path: do not look at NULL path
Message-ID: <20140128013612.GA29548@sigill.intra.peff.net>
URL: https://gitlist.dev/e/20140128013612.GA29548%40sigill.intra.peff.net

```
We explicitly check for and handle the case that the
incoming "path" variable is NULL, but before doing so we
call strchrnul on it, leading to a potential segfault.

We can fix this simply by moving the strchrnul call down; as
a bonus, we can tighten the scope on the associated
variable.

Signed-off-by: Jeff King <peff@peff.net>
---
Most of the callers already check for NULL, so the only way to trigger
this is via an empty "include.path". That is addressed separately in the
next patch, so technically this is not needed if we apply that one. But
in that case, the "path == NULL" check here is useless, so I think this
makes things safer overall.

 path.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/path.c b/path.c
index 24594c4..f9c5062 100644
--- a/path.c
+++ b/path.c
@@ -265,12 +265,12 @@ static struct passwd *getpw_str(const char *username, size_t len)
 char *expand_user_path(const char *path)
 {
 	struct strbuf user_path = STRBUF_INIT;
-	const char *first_slash = strchrnul(path, '/');
 	const char *to_copy = path;
 
 	if (path == NULL)
 		goto return_null;
 	if (path[0] == '~') {
+		const char *first_slash = strchrnul(path, '/');
 		const char *username = path + 1;
 		size_t username_len = first_slash - username;
 		if (username_len == 0) {
-- 
1.8.5.2.500.g8060133

```

## Jeff King, 2014-01-28 01:37

Subject: [PATCH 2/2] handle_path_include: don't look at NULL value
Message-ID: <20140128013729.GA21120@sigill.intra.peff.net>
URL: https://gitlist.dev/e/20140128013729.GA21120%40sigill.intra.peff.net
In-Reply-To: <20140128013612.GA29548@sigill.intra.peff.net>

```
When we see config like:

  [include]
  path

the expand_user_path helper notices that the config value is
empty, but we then dereference NULL while printing the error
message (glibc will helpfully print "(null)" for us here,
but we cannot rely on that).

  $ git -c include.path rev-parse
  error: Could not expand include path '(null)'
  fatal: unable to parse command-line config

Instead of tweaking our message, let's actually use
config_error_nonbool to match other config variables that
expect a value:

  $ git -c include.path rev-parse
  error: Missing value for 'include.path'
  fatal: unable to parse command-line config

Signed-off-by: Jeff King <peff@peff.net>
---
 config.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/config.c b/config.c
index d969a5a..314d8ee 100644
--- a/config.c
+++ b/config.c
@@ -84,8 +84,12 @@ static int handle_path_include(const char *path, struct config_include_data *inc
 {
 	int ret = 0;
 	struct strbuf buf = STRBUF_INIT;
-	char *expanded = expand_user_path(path);
+	char *expanded;
 
+	if (!path)
+		return config_error_nonbool("include.path");
+
+	expanded = expand_user_path(path);
 	if (!expanded)
 		return error("Could not expand include path '%s'", path);
 	path = expanded;
-- 
1.8.5.2.500.g8060133

```