fix: clarify that fs::rename on unix accepts targets that don't exist

[sftse]

No description provided.

[rustbot]

r? @tgross35

rustbot has assigned @tgross35.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[hkBst]

@sftse do you have link to a good source that documents this?

[sftse]

This is the current behavior on Linux

    fn dir_to_nonexist_dir() {
        let dir = Path::new("__testdir");
        std::fs::create_dir(dir).unwrap();
        std::fs::create_dir(dir.join("foo")).unwrap();
        std::fs::rename(dir.join("foo"), dir.join("bar")).unwrap();
        assert!(dir.join("bar").is_dir());

        std::fs::remove_dir(dir.join("bar")).unwrap();
        std::fs::remove_dir(dir).unwrap();
    }

rename

int rename(const char *oldpath, const char *newpath);
...
oldpath can specify a directory. In this case, newpath must
either not exist, or it must specify an empty directory.

[tgross35]

The top of the paragraph already has the qualifier "when both from and to exist", so I think this is slightly redundant as written. This is a bit dense anyway so it wouldn't be bad to clean up the whole section. How about switching to a list format, something like:

This function currently corresponds to the `rename` function on Unix, and
`MoveFileExW` with a fallback to `SetFileInformationByHandle` on Windows.
The exact behavior differs:

- On Unix, when `from` is a directory and `to` exists, `to` must be an empty directory.
- On Unix, when `from` is not a directory and `to` exists, `to` may not be a directory.
- On Windows 10 version 1607 and above, the behavior is the same as Unix if the
  filesystem supports  `FileRenameInfoEx`.
- Otherwise on Windows, `from` can be anything but `to` must not be a directory.
- If `to` does not exist, `from` can be anything.

[sftse]

Sounds good, although grouping the list by first behavior that is common to (most) platforms and then mentioning the caveats and edge cases might be better?

I can't currently tell what all the behaviors are, so I'll either test them or have to take your word for it.

[tgross35]

I kind of mirrored the ordering that exists, which I think makes enough sense: you read from top to bottom and stop when you have a match. The final item is a fallback.

I can't currently tell what all the behaviors are, so I'll either test them or have to take your word for it.

I don't think I added anything that wasn't there. I did double check the Windows case, and it should be accurate given we use FILE_RENAME_POSIX_SEMANTICS. The posix/unix behavior is specified at https://pubs.opengroup.org/onlinepubs/9799919799/functions/rename.html and says:

If the old argument names a file that is not a directory and the new argument names a directory, or old names a directory and new names a file that is not a directory, or new names a directory that is not empty, rename() shall fail. Otherwise, if the directory entry named by new exists, it shall be removed and old renamed to new.

So I think what we have matches.

[tgross35]

Ah, you were probably talking about the "with a fallback" part. For context there see #131072 and #137528.

It might also be worth turning `rename` into a link to that posix page so anyone who needs more details can find it quick.

[tgross35]

This needs followup to the above,
@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[tgross35]

@sftse gentle ping, would you be able to do the update from #149267 (comment)?

[sftse]

Yes, thanks for the ping, I'll get to it.