Re: [Outreachy][PATCH] abspath: reconcile `dir_exists()` and `is_directory()`

From: SZEDER Gábor <szeder.dev@gmail.com>
Date: Oct 24, 2019, 11:41 UTC
Message-ID: <20191024114148.GK4348@szeder.dev>
In-Reply-To: <20191024092745.97035-1-mirucam@gmail.com>

On Thu, Oct 24, 2019 at 11:27:45AM +0200, Miriam Rubio wrote:

Show 6 quoted lines

> The dir_exists() function in builtin/clone.c is marked as static, so
> nobody can use it outside builtin/clone.c.
> 
> There is also is_directory() which obviously tries to do the very same, but it uses a name that few developers will think of when they see file_exists() and look for the equivalent function to see whether a given directory exists.
> 
> Let's reconcile these functions by renaming is_directory() to dir_exists() and using it also in builtin/clone.c.

Please wrap the proposed log message at about 70 characters width; that way it will look much better in 'git log' in a standard 80 char wide terminal.

I think this is a cleanup worth doing, but...

Show 13 quoted lines

> diff --git a/abspath.c b/abspath.c
> index 9857985329..13bd92eca5 100644
> --- a/abspath.c
> +++ b/abspath.c
> @@ -5,7 +5,7 @@
>   * symlink to a directory, we do not want to say it is a directory when
>   * dealing with tracked content in the working tree.
>   */
> -int is_directory(const char *path)
> +int dir_exists(const char *path)
>  {
>  	struct stat st;
>  	return (!stat(path, &st) && S_ISDIR(st.st_mode));

Note the '&& S_ISDIR(st.st_mode)', making sure that the given path is in fact a directory. Good.

Show 12 quoted lines

> diff --git a/builtin/clone.c b/builtin/clone.c
> index c46ee29f0a..f89938bf94 100644
> --- a/builtin/clone.c
> +++ b/builtin/clone.c
> @@ -899,12 +899,6 @@ static void dissociate_from_references(void)
>  	free(alternates);
>  }
>  
> -static int dir_exists(const char *path)
> -{
> -	struct stat sb;
> -	return !stat(path, &sb);

But look at this, it only checks that the given path exists, but it could be a regular file or any other kind of path other than a directory as well!

So this function clearly doesn't do what it's name suggests. That's bad.

Unfortunately, it gets worse: some of its callsites in 'builtin/clone.c' do expect it to check the existence of _any_ path, not just a directory.

The first callsite is:

    dest_exists = dir_exists(dir);
    if (dest_exists && !is_empty_dir(dir))
            die(_("destination path '%s' already exists and is not "
                    "an empty directory."), dir);

I think this actually means path_exists(): if a file, or any other kind of path with the given name were to exist, then we should die() showing this error message, but after changing dir_exists() to make sure that the path is indeed a directory we won't:

  # create a 'git' _file_
  $ >git
  # current git master:
  $ git clone https://github.com/git/git
  fatal: destination path 'git' already exists and is not an empty directory.
  # with this patch:
  $ ~/src/git/git clone https://github.com/git/git
  fatal: could not create work tree dir 'git': File exists

So the command still fails, which is good, but with a different error message. The test suite doesn't catch this, because the test case looking at this scenario ('clone to an existing path' in './t5601-clone.sh') only checks that 'git clone' fails, but it doesn't check whether it failed with the right error message.

Now, that other error message comes after a failed mkdir() call later on which should have created the work tree. So it begs the question what would happen when a file is in the way of a bare clone:

  $ >git.git
  $ git clone --bare https://github.com/git/git
  fatal: destination path 'git.git' already exists and is not an empty directory.
  $ ~/src/git/git clone --bare https://github.com/git/git
  Cloning into bare repository 'git.git'...
  fatal: invalid gitfile format: /home/szeder/src/git/tmp/git.git

Then the next callsite looks like it meant path_exists() as well, but I didn't try to make it fail or show different behavior:

    work_tree = getenv("GIT_WORK_TREE");
    if (work_tree && dir_exists(work_tree))
            die(_("working tree '%s' already exists."), work_tree);

And there is a third callsite, but I'm not sure what it is about, and, consequently, what is really meant with dir_exists() here:

    if (real_git_dir) {
            if (dir_exists(real_git_dir))
                    junk_git_dir_flags |= REMOVE_DIR_KEEP_TOPLEVEL;