What's new in 0.44.2:

Key feature:

 - upsmon now has a mini-scheduler available.  It lives in a separate
   program called upssched, and is used to make things happen some number
   of seconds after the UPS changes state.

   If you only want to send a page about power status after the UPS has
   been running on battery for 30 seconds, this is the solution.  It
   can also be used right now to approximate an early shutdown if that's
   what you really want.  Better ways of doing this will be coming,
   though.

   See upssched.txt in the docs subdirectory for more information.

Binary change notice:

 - upsmon and upsd now live in the sbin directory along with upssched.
   If you're upgrading from an old version, this may affect you.  Be
   sure to update your scripts so that they don't start the old versions.

   The old versions will still work, but you'll be missing the goodies
   introduced in this release.

upsmon.conf reminder:

 - Your SHUTDOWNCMD probably needs quotes around it.

   SHUTDOWNCMD "/sbin/shutdown -h +0"

   Otherwise, it will just run the first word after the space!  This
   changed back in 0.44.1.

Other new features:

 - upsct2 can now handle non-interactive variable changing with
   -s <var>=<value> and -p <password> on the command line. 
   [Nigel Metheringham]

 - newapc: New experimental driver for APC smart protocol units.  This
   is intended to replace apcsmart at some future date, and aims to 
   better support the wide range of their hardware with a less hackish
   approach.  This driver also supports just about all of the possible
   variables within the UPS, whereas apcsmart is missing many.
   [Nigel Metheringham]

 - Lots of compatibility changes to make the source build on SunOS 4.
   [Mark Powell]

 - More compatibility changes to help building on Solaris 5.x.
   [Ben Collver]

 - upsmon will now use ERR severity when syslogging login failures
   so they are easier to spot.	[Nick Hibma]

 - RH script improvements: upsd, upsmon, upspowerdown 
   [Nigel Metheringham]

 - The apcsmart driver can now turn off the UPS after the configurable
   grace period with -s 3.  The meanings of these numbers are also
   now explained in the -h help text.	[Patrik Rak]

 - upsmon can now provide lots of debug information on demand without
   being recompiled.  Use -d to enable this mode.	[Oliver Jones]

 - upsmon will now notice the replace battery status flag (RB) and 
   provides a notify type to handle it - NOTIFY_REPLBATT.  By default
   it will only invoke it every 12 hours, but this can be changed in
   the config file.

 - The apcsmart driver now supports the BATTDATE variable for keeping
   track of the last battery change.

 - Failures in upsmon's system() call to the final shutdown command
   are now syslogged for diagnostic purposes.	[Bill Richter]

 - The genericups driver now allows overriding the settings for model
   and manufacturer with -m and -M, respectively.  Use this when your
   UPS matches a type already in the database but the identifying
   strings are different.

 - Various Win32 compatibility patches to make it compile there.
   [Ben Collver]

 - upsd's messages to the syslog about data staleness now include more
   information on the cause, so you can track it down.	[Kenneth Porter]

 - More 'compat' functions and other patches for HPUX compiles.
   [Ben Collver]

 - The man pages (both of them) will now go into a specified directory
   if you use --mandir=<path> during configure.  Man pages added and
   updated by Shaul Karl and Bill Richter.

 - The bestups driver now recognizes PR2 as the Patriot Pro II.
   ["Mark"]

 - The upscommon driver will now show the received buffer during debug
   runs to make serial diagnosis easier.	[Kenneth Porter]

 - New driver: toshiba1500 - supports Toshiba 1500 units.
   [Kenneth Porter]

 - New driver: bestuferrups - supports Best Power Micro-Ferrups units.
   [John Stone]

 - New shutdown mode for apcsmart: -s 2 will turn the UPS off entirely.
   [Igor Khristophorov]

 - upsd no longer requires math.h, and no longer links to libm, since
   the netmask calculations are now done with 1<<x rather than pow.

 - Security domains.  You can now run upsd as one user and the model
   driver(s) as another.  This way, compromising the upsd account won't
   yield direct control of the serial port.  See Changes and the FAQ
   for more information.	[Daniel Roesen]

Bug fixes:

 - upsd's complaints about "NULL password" now get syslogged. 
   [Nick Hibma]

 - Buffer lengths throughout the source have been increased significantly
   to handle drivers with large datasets like newapc.  
   [Nigel Metheringham]

 - apcsmart's shutdown type 1 now sends the right number of characters 
   to the serial port.	[Patrik Rak]

 - In upsmon, broken MONITOR lines that are missing arguments will no
   longer cause segfaults and will now generate a warning instead.

 - The fentonups driver no longer mixes up the output and utility
   voltages.	[Damion de Soto]

 - The genericups driver should no longer believe that every call to
   ioctl is broken.	[Eric Lawson]

 - upscommon should now compile on systems without <sys/shm.h>.

 - upsd can handle a /32 style ACL again without wrapping around and
   eventually arriving at 0.  Overflowing variables is bad.

 - NOTIFY_COMMBAD should actually work in upsmon now.	[Shaul Karl]

 - Calls to ioctl are actually checked for saneness now.	
   [Allen Mcintosh]

 - Systems that don't have cfsetispeed() available now generate an error
   at compile-time, since they will need to hack in some other way to
   set the speed.	[Kenneth Porter]

 - The models will actually be rebuilt when you edit one and run make.
   The catch is that _all_ of them will be rebuilt.  This is due to
   an evil hack to support BSD make.  See Changes for more, and send
   in a fix if you know of one.

 - SIGSYS errors due to ignoring the return value of shmget() in
   upscommon should be gone now.

 - fds 0-2 are now forced into /dev/null to keep other processes from
   stealing them when calls to system() are made.

 - The programs now call setsid() to avoid signals that would otherwise
   kill them at startup on some systems.	[Bob Pekarske]

 - Lots of pedantic cleanups.  More on the way.

What's next:

 - Callbacks from upssched to upsmon for coordinated early shutdowns.
   Right now you can have upssched invoke /sbin/shutdown directly, but
   that doesn't tell the slaves!  Expect pid files to arrive too.

Out there:

 - Finer grained user control as dictated by the server, and the silly
   hack that is upsset's password scheme can go away.