reasons for this, quite a few of them technical, and not all of them
in response to Intel's broken ia32e crud.  The gcc toolchain stays at
x86_64 for now.

ok mcbride@

ok millert@

diffs from Paul de Weerd;
ok deraadt@

- refer to the section 4 manual page for every device target, whenever
possible
- repair a few minor quoting errors
- repair a few sorting errors
- fix a few device descriptions (ideally they would be synched with the
.Nd line of their manpages, but we are not there... yet)
- minor mdoc fixes
Thanks to jmc@ for review and a lot of constructive criticism.

reviewed by deraadt@, tedu@

designed to output mdoc rather than shell script, and use them to produce
much better MAKEDEV.8 manual pages.
Note that MAKEDEV.man relies upon BSD m4 behaviour, which all m4 binaries
might not follow.
mdoc help and comments jmc@, disabused ok deraadt@

- homogenize section order
- sort devices inside sections
- make sure all platforms provide a local target (sigh)
- remove a few dead devices not worth mentioning on a few arches while there.
disabused ok deraadt@

While there, allow the comment char and the divert section for md script
fragments to be overriden.
No functional change (yet).
disabused ok deraadt@

"looks ok" drahn@ deraadt@

AFS functionality (enough to to pkg_add's)

theo ok

suggested by deraadt@

ioctl(2), an unprivileged process may allocate a pty and have its owner
and mode set appropriately.  This means that programs such as xterm and
screen no longer need to be setuid.  Programs using the openpty()
function require zero changes and will "just work".
Designed by beck@ and deraadt@; changes by beck@ with cleanup (and
a rewrite of the vnode bits) by art@ and tweaks/bugfixes by me.
Tested by many.

The latest versions are at http://www.ata-atapi.com/hiw.htm.
Discussed with nick@ and jmc@.  ok deraadt@.

some of this ok espie@, all of it ok deraadt@