"-T" option help text

[James R. Van Zandt]

Philip Rowlands wrote:
> On Sun, 30 Aug 2009, James R. Van Zandt wrote:
> 
> 
> >  For the help text, here are some alternatives:
> >
> >
> >  if DEST is a directory, then delete it first 
> 
> 
> This isn't what -T does. If DEST is an empty directory then it's
> overwritten with the rename(2) system call. Otherwise mv will fail
> e.g. if SOURCE isn't a directory or DEST is non-empty.

Let's see...

  vanzandt:/tmp$ date >source
  vanzandt:/tmp$ mkdir dest
  vanzandt:/tmp$ mv -T source dest
  mv: cannot overwrite directory `dest' with non-directory
  vanzandt:/tmp$ rm -rf dest
  vanzandt:/tmp$ touch test
  vanzandt:/tmp$ mv -T source dest
  vanzandt:/tmp$ 
  
I agree my proposed help text is wrong.  However, I think the current
description in the info file is wrong too.  It says "Do not treat the
last operand specially when it is a directory or a symbolic link to a
directory."  Yet, this is an example where mv fails when DEST is a
directory, and succeeds when it is a normal file.  So a directory *is*
treated differently.

The extended discussion on the "Target directory" page says:

     For example, when the command `mv /tmp/source /tmp/dest'
     succeeds, there is no guarantee that `/tmp/source' was renamed to
     `/tmp/dest': it could have been renamed to `/tmp/dest/source'
     instead, if some other process created `/tmp/dest' as a
     directory.  However, if `mv -T /tmp/source /tmp/dest' succeeds,
     there is no question that `/tmp/source' was renamed to
     `/tmp/dest'.

I read the last sentence as saying /tmp/source would get renamed as
/tmp/dest whether or not some other process had previously created
/tmp/dest as a directory.  The fact that it fails instead certainly
should be stated.  (To me it seems a much less useful behavior.)

            - Jim Van Zandt