The plan for eliminating lp_posix_pathnames()

[Jeremy Allison]

On Sun, Feb 14, 2016 at 08:09:43AM +0200, Uri Simchoni wrote:
> Thanks for the writeup!
> 
> Another thing we can do to streamline/simplify code for VFS is to
> guarantee that this smb_fname has a valid struct. I see that lots of

That's a little more difficult and touches more code
paths. There are many cases where we create a filename
by getting the parent path from an existing smb_fname,
and then pass into a VFS call. Not all VFS calls need
a valid ex_st component, and those that do check by
using VALID_STAT and re-filling using stat/lstat
if needed.

Not sure if we should 100% guarentee that the
smb_fname is valid for VFS calls (and indeed for
stat/lstat calls we know it isn't).