What's new in 0.50.0:

Compatibility changes:

 - The TCP/UDP network code now uses port 3493 by default since IANA
   has assigned that to the project.  If you have older installations
   present on port 3305, be sure to specify the port numbers explicitly
   until they are upgraded.

   Just tack on a :3305 on the new systems, and :3493 on the old ones.

   MONITOR ups@oldbox:3305 1 mypass slave
   MONITOR ups@newbox:3493 1 mypass slave

   If you have firewall rules protecting port 3305, remember to change
   them to protect 3493 instead.

 - All old style drivers have been removed from the tree.  The following
   drivers are no longer being distributed as a result:

   aeg, bestfort, bestuferrups, engetron, ipt-anzen, mgeups, multilink,
   mustekups, optiups, sms, toshiba1500, upseyeux

   For more information on why this happened and what you can do about
   it, see the FAQ.

Key features:

 - This release was focused on cleaning up the drivers.

   One benefit is that upsdrvctl will now control all of the drivers
   that are distributed with the tree.

Other new features:

 - upsstats and upsimage have been cleaned up with respect to
   host names and commands that are received from the user.

 - More man pages have been written.

 - upsc and upsct now allow you to poll a single variable by adding it
   as an argument after the host name.  This saves the trouble of 
   fetching the whole list and then throwing it away when grepping for
   the one you really want to see.	[Richard Muratti]

 - The statepath is now created as mode 0700 by default.  This will
   make upsd happy, as it will let you know when the path is world
   readable.  Your statepath should be locked down if your system doesn't
   protect named pipes, as the .sock files live in there.
   [Mark Bergman]

Driver news:

 - 0.45.5's newvictron replaced the old victronups.

 - 0.45.5's newpowercom replaced the old powercom.  ups-trust425+625
   was also removed as the same time, as this driver should support both.

 - snmp-ups is a new experimental driver for RFC 1628 SNMP UPS support.
   It requires UCD SNMP/Net-SNMP, so it's not built by default.
   [Arnaud Quette]

 - The fentonups and bestups drivers now send a 'cancel shutdown'
   command during a normal startup just to be safe.

 - The new powerware driver supports Powerware UPS 9120 'Binary Computer 
   Mode'.	[Martin Schroeder]

 - The old mgeups driver has been replaced by mge-utalk, which supports
   the MGE UPS Systems UTALK protocol.
   [Hans Ekkehard Plesser, Arnaud Quette, Nicholas Reilly, Dave Abbott
   and Marek Kralewski]

Bug fixes:

 - upsset's showsettings mode should work again.

 - Setting read-write variables wasn't working due to a subtle ) 
   placement error in upsd.  Fixed.

 - The config parser should no longer hang when compiled with gcc 2.95.4 
   due to improper variable types being used with fgetc.

 - bestups now delays slightly when sending queries.  This should make
   status polls more reliable by not rushing the UPS.
   [Frank T. Lofaro Jr.]

 - upsd no longer returns "ERR <varname> VAR-UNKNOWN" in some situations,
   which was confusing some of the clients.

 - A case in the config parser where it might call memset after getting
   an error from malloc was squashed.

 - bestups and fentonups now ignore BOOST/TRIM when on battery since
   it's meaningless while the inverter is active.

 - The drivers now unlink the .sock file in the statepath when starting.
   Previously they would fail if the driver or system crashed and they
   encountered those files upon startup.

 - mmap support is now explicitly disabled on HP-UX since it doesn't seem
   to get along with our use of it.	[Mark Bergman]

 - upsd no longer sends a double newline when reporting 
   'ERR VAR-NOT-SUPPORTED', which could confuse clients.

 - upssched actually routes log activity to the syslog now.  Previously
   it thought it had a tty and was using stderr.  This didn't work since
   it's usually called by upsmon which is in the background.
   [Pascal Pichon]

 - Instances of 'sigmask' were renamed throughout the code to avoid
   conflicts with a variable of the same name in libraries on AIX.

 - Similarly, sa_len was renamed to avoid conflicts on IRIX.
   [Aleksandar Topuzovic]

 - The cyberpower driver should now compile on IRIX.  It uses CRTSCTS
   and IRIX doesn't have that #define.  This may not actually work, 
   though.	[Aleksandar Topuzovic]

 - The Makefiles no longer have a rule for .c.o.  This was breaking
   compiles on OpenBSD 3.0, which complained about GNU-isms.

Out there:

 - Establish a channel from the drivers back to the clients for
   reporting on instant commands

 - Fix newapc for old hardware and retire apcsmart

 - Document everything in the tree

 - Slap a 1.0 tag on this thing