Date: 03 Aug 2003 14:10:44 +0200
Message-Id: <ur843djqz.fsf@elta.co.il>
From: Eli Zaretskii <eliz AT elta DOT co DOT il>
To: "Andrew Cottrell" <AndrewCottrell AT swiftdsl DOT com DOT au>
CC: djgpp-workers AT delorie DOT com
In-reply-to: <000001c3598f$d4f02850$0101a8c0@acp42g>
	(AndrewCottrell AT swiftdsl DOT com DOT au)
Subject: Re: mv bug (filutils? library?)
References:  <000001c3598f$d4f02850$0101a8c0 AT acp42g>
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From: "Andrew Cottrell" <AndrewCottrell AT swiftdsl DOT com DOT au>
> Date: Sun, 3 Aug 2003 17:21:11 +1000
> 
> The initial call to rename and _rename are okay, but after the return from
> _rename() the recusive issue strikes. After the return from _rename() the
> directory result is 100% perfect.

There shouldn't be any recursive calls; see below.

> I have further traced the problem to the following line in rename(). Before
> the call all is okay, but after the recusive junk1/junk1/junk1 exists....
>       /* Process all of its siblings.  */
>       retval = __file_tree_walk(source, mover);

This is a bug, I think: on Windows, there should be no need to do this
file-tree traversal, since Windows can move a directory with all its
files and subdirectories in one go.  `rename' has code to see if a
simple call to `_rename' succeeds, and only do the recursive thing if
it doesn't.  Only plain DOS needs the recursive renaming, IIRC.

Furthermore, there should be code _before_ the call to `_rename' that
detects the fact that you are trying to move a directory into its own
subdirectory, and refuse to do so.  Here's that code, I think:

  else if ((source_attr & 0x10) == 0x10)
    {
      /* Fail to rename OLD -> OLD/whatever.  */
      char new_true[FILENAME_MAX], old_true[FILENAME_MAX];

      errno = 0;
      if (is_parent(_truename(old, old_true), _truename(new, new_true)))
        {
          errno = EINVAL;
          return -1;
        }
      else if (errno)
        return -1;
    }

Now, this `is_parent' thing depends heavily on what `_truename'
returns for each of the two file names, so perhaps it's `_truename'
that messes up the code on XP.  Or maybe there's some simple bug in
the case that `whatever' is the same string as OLD.

Finally, I still don't understand how come `mv' did say "cannot move
`junk1' to a subdirectory of itself": that seems to suggest that it
correctly diagnosed the problem and refused to move the file, no?

Last, but not least, thanks for working on this.