The recent changes to extended character conversion support for nmbd can allow
all sorts of 'non unix' things into our unix strings. It can also allow
unconverted unix strings into unicode bufferes, specificly those used for
StrCaseCmp() and fnmatch().
At the very least, we should make StrCaseCmp() and fnmatch() check the errno,
but we shoul simply allow the conversion to fail, where we are in a position to
handle the erorr.
*** This bug has been marked as a duplicate of 830 ***
originally reported against one of the 3.0.0rc[1-4] releases.
Cleaning up non-production versions.