The riddle: how can a file both exist and not exist? The answer: symlinks!
When a symlink is checked into perforce running 'p4 print -o' will create a
symlink, not the file that it points to. This naturally makes postreview.py
choke...

~% post-review
Traceback (most recent call last):
  File "/share/rbtools/rbtools/postreview.py", line 4115, in <module>
    main()  
  File "/share/rbtools/rbtools/postreview.py", line 4061, in main 
    diff, parent_diff = tool.diff(args)
  File "/share/rbtools/rbtools/postreview.py", line 2002, in diff 
    return self._changenum_diff(changenum)
  File "/share/rbtools/rbtools/postreview.py", line 2299, in _changenum_diff
    self._write_file(old_depot_path, tmp_diff_from_filename)
  File "/share/rbtools/rbtools/postreview.py", line 2475, in _write_file
    os.chmod(tmpfile, stat.S_IREAD | stat.S_IWRITE)
OSError: [Errno 2] No such file or directory: '/tmp/tmpsCS-fw'

Resolving symlinks, especially those with a relative path would be tricky. 
Also, there's no certainty that the file it points to is even in version 
control. Hence I'm just dropping it from the review with a warning.

Tested this change both when reviewing a single symlink file and a batch of
files containing a symlink.

We've been using this change against RBTools 0.3. I've adapted the change for
the current master but this copy of it hasn't been exercised.