Re: [PATCH v2 2/2] mv: check for missing destination directory before renaming

From: Junio C Hamano <gitster@pobox.com>
Date: Jul 26, 2026, 14:59 UTC
Message-ID: <xmqqik61yeyn.fsf@gitster.g>
In-Reply-To: <xmqqcxwdcmln.fsf@gitster.g>

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

Show 24 quoted lines

> Lucas Zamboni Orioli <lucaszam0@gmail.com> writes:
>
>>> lstat() can succeed and 'dir_st' may indicate something other than a
>>> directory (for example, a symbolic link or a regular file).
>>> Alternatively, it can fail with ENOTDIR when, for example, 'dst_dir'
>>> is 'a/b/c' and 'a/b' is a file rather than a directory.
>>>
>>> Both cases will cause 'git mv' into a path assumed to be a directory
>>> to fail.  Shouldn't we handle these conditions as well?
>>
>> Yes, agreed, both should be handled. For v3 I switched from lstat()
>> to stat() so that the check follows symlinks the same way rename()
>> does, and I handle the non-directory cases:
>
> Generally, a symbolic link in a Git-managed working tree should not
> be followed.  Following a symbolic link would mean that 'git mv x y'
> could move 'x' outside the working tree if 'y' is a tracked symbolic
> link pointing to a directory outside the working tree.  'git apply',
> for example, avoids being fooled by a symbolic link for the same
> reason.
>
> I doubt that using stat() instead of lstat() is the right approach.
> Doing so essentially amounts to ignoring the presence of symbolic
> links.

I actually think "outside the working tree" is an irrelevant red herring. What is relevant is the fact that Git tracks symbolic links.

If you have x (file) and y (another file), you would want to complain when the user says:

    $ git mv x y

because the location y is "taken" and the command line tells us only about what it wants to do to x, without saying anything about what you want to do to that existing y. If y were a symbolic link instead, you should behave exactly the same way.

It actually takes even more care, and I do not know if the implementation of git-mv is done carefully enough, but think about what should happen to:

    $ git mv x a/b/c

when 'a' is a tracked symbolic link, and it points at, say, '.'. Should it behave exactly the same as:

    $ git mv x b/c

or should it simply error out? I think the latter, "I see a symlink in the middle, so I refuse to follow," is the right behavior.

Think carefully about cases where 'a' is a directory and 'a/b' is a symlink, or where 'a' and 'a/b' are directories and 'a/b/c' is a symlink, and so on. We do not want to craft an arbitrary rule that says we allow or refuse to operate depending on the link target.