once it reaches a largish size (16 million files) and make it
actually match the comments so it may be used to initialize
flists of known size (create_flist_from_batch()).

	Hardlink handling is improved.

	String area code is gone for other reasons.

show_malloc_stats.

  does not have permissions to set one or more of the gids.
- If the add_[ug]id() routine couldn't find a name for an ID, we
  still note the ID in our list of IDs so that we don't try to
  look it up again (and again).  (The name-less IDs are stripped
  when sending the IDs from the sender to the receiver.)
- Created a better add_to_list() function that handles the linking
  of the new node into the list (rather than the caller doing it).
- Added new internal functions for some receiver-side work that
  needed to be done in multiple places (the recv_add_[ug]id()
  functions).
- We output debug info even for ID mappings that have no names.
  This make the list complete and also notes which gids get
  mapped to -1 (GID_NONE).
- Turned several while loops into for loops.

won't see it in the list of IDs to remap.  That means that we need to
call the is_in_group() function from inside match_gid() to know if we
can set the group or not (if we're not root).

value.

a higher debug level instead of the default of level-l verbosity.

chance to map the uid/gid info in the receiver.

gid (depending on applicability).