Hi,

Enclosed is a patch, mostly containing schwarze@'s original man.conf 
work, that lets apropos(1) and mandocdb(8) (and man.cgi, but it's not in 
the patch) behave much better.  Specifically:

apropos(1) now does as it traditionally has:
   If [-M manpath] is used, it sets the base path.
   If $MANPATH is used, it sets the base path (overriding -M).
   If neither -M or $MANPATH, man.conf/manpath(1)** is used.
   If [-m manpath] is used, it augments the base paths.

mandocdb(8) changes:
   If [dirs...] is empty, use man.conf/manpath(1).

manpath.c (was man.conf.c) changes:
   De-dupe the path listings.
   Resolve paths using realpath.
   Use manpath(1)** if -DUSE_MANPATH is specified.

Food for thought: MANPATH.  Does anybody set this?  It makes sense that 
it would be used by PATH.  But it's never set [on OpenBSD, at least], so 
users can't nicely just say "MANPATH=$MANPATH:foo/bar".  These days 
man.conf does that.  Unfortunately, setting MANPATH wipes these out, so 
users have no easy way of having their own MANPATH additions.  I think 
the MANPATH behaviour should be changed to augment the directory path 
(i.e., -m) instead of replace it (-M).  Does this make any sense or am I 
just pissing in the wind?

**What's manpath(1), you're asking?  It's a utility common to man-db and 
the new/old-man package.  It provides similar functionality to scanning 
man.conf, but also scans local manpath configuration files.  FreeBSD 
link below.  I think it's way over-engineered, but as mentioned above, I 
think our method is a little confusing too.

http://www.freebsd.org/cgi/man.cgi?query=manpath&sektion=1&apropos=0&manpath=FreeBSD+8.2-RELEASE+and+Ports

Thoughts?

Kristaps