What's new in 0.45.0:

Key features:

 - Sanity in user handling.  upsset.cgi no longer does its own
   authentication, so you don't have to worry about how to secure the
   upsset.users file.  All user authentication now takes place inside
   upsd.  This also makes other things possible:

 - upsd can now restrict access to certain instant commands by username.
   You can now have a user called "admin" that gets to do all of them,
   and also create a user called "peon" that can only do battery tests.
   
   You can also control whether a user can do SET commands to change the
   value of variables inside the UPS hardware.  This is all configured
   in the upsd.users file.

 * Take note that your upsset.cgi will not function until you assure it
   of your CGI directory's security.  See the sample upsset.conf for
   more information about how to do this.  This is intended to keep you
   from opening holes in the system by doing 'make install-cgi'.
  
Other new features:

 - upscmd and upsct2 (the command line tools for administration) also
   honor the new "user" level access as you would expect.

 - statedebug has been reworked and now supports shared memory mode. 
   This is only of interest to developers working on new drivers and
   others who need to debug the interprocess communications.

 - A new common driver design is now in the works.  The fentonups and
   bestups drivers have already been converted.  This approach will
   unify the startup process for all drivers and allow some other helpful
   features in the future.

 - The bestups driver now reports battery voltage.

 - Liebert Multilink UPSes are now supported with the multilink driver.
   [Rick Lyons]

 - Config files are now installed as <name>.sample so you can compare them
   to your existing files to see what's new.	[Ben Collver]

 - upslog now writes a pid file for easy signal sending.
   [Dirk-Willem van Gulik]

 - Failures to open the UPS device will now yield a verbose message
   showing the current user ID, intended device, and the permissions of
   that device to help users set things right.	[Alan Kennington]

 - SEC protocol hardware is now supported with the sec driver.
   [John Marley]

 - Repoteck RPT-800A and RPT-162A models are now supported as genericups
   type 13.	[Theodor Milkov]

 - Note that QUICKSTART is now called INSTALL, and Changes is now CHANGES.
   [Michael Wayne]

Bug fixes:

 - The getupsmsg() stub in upscommon could have sucked 100% CPU if the
   system lacked SysV message queue support.

 - The normal getupsmsg() function was eating the CPU when message queues
   failed at runtime.  This will no longer happen.  If you find that your
   drivers can't create a queue, run ipcs and make sure that you don't
   have runaway processes consuming your IPC resources.

 - Using a hostname-only identifier for a UPS in upsmon will no longer
   try to build a string from a NULL pointer.  This was causing the
   forced shutdown routine to fail due to a mismatch on some systems, and
   crashes on other depending on the interaction of snprintf, %s, and NULL.
   [Matthew Dharm]

 - Calling upsmon with a bogus argument to -c no longer starts the
   program.	[Andrew Delpha]

 - upsmon: another NULL + %s format bug squashed.	[Matthew Dharm]

 - basename already exists in some C libraries, so work around it by
   calling our version xbasename.	[Alan Kennington]

 - The newapc driver no longer tries to open the serial port in the
   forceshutdown call, as it's already open.  This was causing the -k
   command to fail, disabling the forced power cycle of the load.
   [John Pochmara]

 - cfmakeraw compatibility routine added for systems which lack it,
   since the multilink driver uses it.  This should help Solaris people.
   I don't have one of those boxes, so let me know if it doesn't work.
   [Marty Lee]

 - genericups type 2 (APC 940-0020B) now uses RTS for cable power.  This
   may make things better for some people, although it apparently may
   never work on some systems.  See the discussion in the mailing list
   archives for the full story, including the pin routing and possibly
   invalid RS232 voltage levels.

 - upsmon will now generate a notify event if the first poll of a UPS
   is "on battery".

 - time.h and sys/time.h are now both included where appropriate.
   glibc 2.2 started throwing warnings, and this should fix it.

 - Lots of sanity checks added to the code to make sure we don't try
   to parse NULL values from config files or similar.

 - powercom: the TIOCMSET ioctl now uses the address of the argument
   rather than the argument itself, and checks the return codes to
   detect errors.	[Igor Podlesny]

What's next:

 - Unified driver configuration and startup routines.

 - USB HID UPS support under Linux at last.

Out there:

 - More drivers switch to the unified scheme.

 - The final march to 1.0?  The todo list is getting pretty small...