Past list: (items tagged as bug fix [bug], or feature [ftr]. Big ones starred.)
---------------
6.04:
  bug		Fixed seg-fault in newer compilers for the UNC joystick box
			running under Linux.
6.03:
  ftr		Code to control the Zaber Serial Positioning Interface
			linear actuators.
  ftr		vrpn_Analog_Remotes can now send requests to the server to
			update the value.  Used to control the Zaber and
			other devices.
  ftr		New driver for Intersense trackers that uses their library
			directly, rather than talking to the raw devices.
			For IS600 and such, probably want to use the
			Fastrak server still.  You need to get their include
			files and libraries to link, then define the proper
			thing in the compiler VRPN_USE_ISENSE.
 *ftr		Java port of several of the VRPN remote objects.  Still in
			beta, but enables reading from trackers, buttons, and
			analogs from Java applications.  Also code to build
			VRPN as a DLL so that it can be loaded by Java.
  bug		Serial port code changed on Windows to raise the DTR line when
			the port is opened.  When this was not done explicitly
			on Windows XP, the TNG3 died because it uses this line
			for power.
 *bug		TNG3 driver fixed.  It used to hang forever on the read routine
			because I made it loop while getting reports and the read
			routine held out waiting for all of the characters.  It
			now polls.
 *ftr		Includes Phantom example programs (Tracker, Button, Forcedevice)
			in client_src/haptic that show how to use VRPN with openGL
			on SGIs.
 *ftr		Includes code from John Stone at Beckman Institute for driving
			various versions of the Spaceball 6DOF controller with
			buttons.
 *bug		3Space driver orientation was decoded incorrectly for some
			orientations.  There is still a bug in the driver that will
			make it only work on machines that have the same byte-order
			as PCs.
  bug		Some IS900s report 54-byte status messages, which does not match
			the 55-byte messages reported by Fastraks.  This modifies
			the Fastrak driver to accept messages of either length.
 *bug		Fixed a problem with the IS900 wand driver that prevented it from
			ever reporting analog values (the mainloop() function in the
			Analog_Server base class changed its behavior and the server
			was not modified to match).
 *ftr		Modified project files to enable the compilation of the Phantom
			server using a standard GHOST 3.1 installation, rather than
			requiring the "external" system that we have at UNC.
  bug		Re-ordered unpacking of parameters to control button toggle
			or momentary from the client.  This re-enables client-side
			control over whether a button is a toggle or momentary
 *ftr		Build files in vrpn_embedded to compile the client-side VRPN
			library in Windows CE development environment.  Currently
			doesn't build the vrpn_print_devices file because of a link
			problem, but the library does compile.
 *ftr		Program to display traces of tracker data is built into the
			Vlib/openGL (vogl) library.  Requires external UNC version
			of Glut library to run.
  ftr		National Instruments D/A board server that will try to connect
			to an analog server and will set the D/A channels to the
			voltages that correspond to the Analog outputs.
  ftr		q_to_euler() function for quatlib to go directly to Euler angles
			by Rich Holloway
  clean	Massive clean-up of Quatlib code by Rich Holloway.
6.02:
  *bug	Fixed a bug introduced by the "send all packets when a connection is
			closing" fix reported below.
  *bug	Compiles the Phantom server with the same settings for all parts,
			to avoid seg faults on print messages (the binary file was
			bad last relese).
  *bug	Fixes the project file for the Phantom server so user-built ones will
			also compile correctly.
   bug	Undosify run on the source files, so they will compile on Unix.
6.01:
   ftr	Upgraded to version 3.1 of GHOST library for Phantom server.
  *bug	Removed dynamic allocation of comm handles, which was causing seg
			faults.
   bug	Added check for negative sensor number in callback for handler (this
			happened when connected to a malfunctioning tracker server).
  *bug	Re-ordered the delivery of VRPN messages so that messages are packed
			to endpoints before being delivered locally.  This makes it be
			the case that the messages are delivered in the same order on
			the local and remote host if the callback handler for one message
			sends a message of its own.
   ftr	Making error messages from more VRPN devices use vrpn_TEXT rather than
			fprintf to stderr on the server machine when there is a
			connection.
   bug	Making many serial devices keep reading in case there are multiple
			reports in the serial buffer during one call to mainloop()
   clean	Refactoring serial trackers to put the main loop in the base class.
   bug	Making serial trackers read multiple reports if they are present during
			a single call to mainloop().
   bug	Connections that are deleted send all of their pending messages
			before shutting down.  This is supposed to help stop race
			conditions.
   clean	Removing un-called functions, adding comments, rearranging code
			to be cleaner.
6.00:
   bug	Not a fix, but at least the Flock code now prints an error if you
			try and open a Flock that has the transmitter set at other
			than the first unit.  A general solution will have to wait
			until someone who wants it badly enough upgrades the driver,
			or pays 3rdTech to.
  *bug	The timing code under Win32 tried to use the performance counter on
			the Pentium chip to get timing resolution better than that
			provided by _ftime() (which is about 6ms).  Unfortunately,
			the code was fragile and didn't work across all the flavors
			(98, NT, and 2000).  That code has been removed, so now the
			resolution is about 6ms on Win32 (boo!), but at least it works
			reliably.
		***	Take not those who relied on better timing information!
  *ftr	Better client application for printing the output from devices.
			vrpn_print_devices is given a list of devices and it will
			print all common message types coming from them (tracker,
			button, analog, dial, warning/error) or a subset of them.
			User can select the frequency of tracker reports for each
			sensor of each tracker -- reports from all sensors are
			printed.
  *ftr	Includes a driver for the buttons on a mouse that is plugged into
			a serial port (courtesy of Rob King at NRL)
  *ftr	Includes a driver for the buttons on a TNG3 (Totally Neat Gadget)
			(again, courtesy of Rob King at NRL)
  *bug	Modified the various serial-device servers so that they check for
			a valid record before checking for timeout.  This prevents
			infinite-reset loops, where one resets and causes the other
			to time out, and then it returns the favor.
  *ftr	Includes a driver for PC joysticks under Linux, courtesy of
			Harald Barth.
   ftr	Includes 3rdTech modifications to button code to enable use of
			more general parallel port setups and button devices besides
			the UNC Python parallel-port button.
  *ftr	Includes a driver for the 5DT glove (five fingers plus pitch and
			roll are reported as analog values).  The driver was written
			by David Philippe from Direction de la Recherche in Paris.
   ftr	The Timecode generator server code has been combined into the
			standard vrpn.C server code, guarded by INCLUDE_TIMECODE_SERVER
			so that changes to the server will track.  The project file
			has been modified to look for the Adrienne code in a directory
			at the same level as vrpn named Adrienne/AEX_DLL.
   bug	Fixed the driver for the UNC serial joysticks.  Also, renamed it
			vrpn_UNC_Joystick.[hC], rather than just vrpn_Joystick.[hC]
   ftr	Modified all of the message types to be legal "C" variable names.
			This will enable auto-generation by PERL of the code to
			handle them.
   ftr	Made it so that you can log both incoming and outgoing streams at
			the server, requested by the client.
  *ftr	Added the ability to connect using only an outgoing TCP connection
			to the server (to punch through a firewall or Network Address
			Translation).  When a connection is made this way, no data
			is sent via UDP over that connection between the client and
			server.
  *bug	Non-word-aligned code had sneaked back into the tracker drivers for
			some devices.  Now the vrpn_buffer() and vrpn_unbuffer()
			routines have been changed to work for unaligned buffers as
			well, so this should not recur.  This should be fixed for
			everything including the vrpn_float64 type now.
5.04:
   ftr	Adds a driver for the AEC PCI-LTC time-code reader board.  Emits an
			analog value for each connected board that has the current
			timecode in seconds.
5.03:
   ftr	Adds a driver for the Wanda device, written by Brown University
   ftr	Code to handle Sierra video switches
  *bug	It turns out that ^C handlers in Windows are multithreaded, causing
			printvals much confusion.  Handling this properly fixes the
			crash-at-the-end bug on 98/NT.  HOWEVER, this does not work
			properly when the program is run in at least some versions of
			the Cygwin shell when run on Windows 98.
  *bug	Removed buggy clock-check code introduced in version 5.01
			Now checks to see if it is Windows NT and disables if not
   bug	Cleaned up the gethostbyname() code on the NT version
   bug	Seg fault when using printvals to open file://stored.vrpn, where
			file:stored.vrpn works.
   bug	Fixed vrpn_drain_output_buffer() to call drain rather than
			flush.  Should make Flock driver work better.
   bug	Made -millisleep 0 the default on vrpn_server again.  Setting
			it to 1 broke both the Radamac SPI and the Fastrak
  *bug	Removes code that causes a seg fault when a VRPN application
			exits (due to multiple deletions of connections in the
			system textprinter)
5.02:
   ftr	Clean-up of vrpn_Tracker.[hC]
			Reduction of the number of register/unregister functions in vrpn_Tracker
			Switching to the use of vrpn_buffer() and vrpn_unbuffer()
   ftr	Phantom button changed to button_filter to allow app to make it a toggle
  *bug	Fixes bugs in analogfly that caused it to seg fault
  *ftr	Driver for the Radamec Serial Position Interface camera tracker
   bug	Fixes deletion bug in printcereal
  *ftr	vrpn_AnalogFly now has to flavors: absolute and relative.
			Also, added offset to the calculation chain
			*** See new vrpn.cfg.SAMPLE file if you are using AnalogFly ***
5.01:
   bug	Zero-length VRPN messages were not logged.
   bug	If you deleted a vrpn_Connection with active Endpoints, they didn't close their sockets
  *ftr	Fastrak driver updated
		*Support for InterSense 900 tracker (with Wand and/or Stylus) added to Fastrak driver
		No longer issues "dangerous" commands during reset (dangerous to Isense)
		Tries a fast reset first before doing full reset
		Uses VRPN warning/error/info for many messages, rather than printf
   bug	Fixed gettimeofday() implementation on Win98 when perfcounter isn't close
   bug	Fixed some "const" declarations keeping it from compiling under Solaris
   bug	Buffer-send code to clear up space for new messages had been broken.
		Fixing this allows sending gobs of data between calls to mainloop()
   ftr	Clipping/scaling analog server to map values to [-1..1]
  *bug	Fixes seg-fault on some architectures when reading from files
5.0:
   bug	Removed the mainloop() calls from within the ForceDevice methods that set
		parameters. This will allow them to be used from within callback
		handlers.
   bug	Modified "delete msgbuf" commands to become "delete [] msgbuf", which
		matches the fact that they were created with "new char [len]".
  *ftr	The timeout parameter has been removed from the mainloop() functions of
		the object classes (such as Analog) that had it.  Waiting is more properly
		done outside this function (using vrpn_SleepMsecs()).
  *ftr	Make it so that vrpn_Text messages don't get sent as the maximum-length text message
		no matter how short the actual message is.  THIS FEATURE MAKES THE CODE
		NON-INTEROPERABLE WITH VERSION 4 OF VRPN.  THIS FEATURE WILL BREAK ANY CODE
		THAT USED vrpn_Text TO PASS STRINGS THAT CONTAINED THE NULL CHARACTER.
  *bug	Fixed the packing/unpacking of vrpn_Bool so that it uses the proper-length
		byte ordering.  THIS FEATURE MAY MAKE THE CODE NON-INTEROPERABLE WITH
		VERSION 4 OF VRPN.
   bug	vrpn_Text send_message() should not call mainloop() inside it. Causes segfaults if
		messages are send inside a callback handler.
  *ftr	vrpn_BaseClass object created, from which all object types should derive.  Most
		system object types now derive from this (clock, analog, button, dial,
		forcedevice, sound, tracker, text).  This pulls sender_id registration
		and several other
		functions together so that they are only implemented once.  Also allows
		extension of function (needed for the following features) by extending
		only the vrpn_BaseClass object functions.
  *ftr	Text-handling code pulled into the vrpn_BaseClass object, so that all
		objects can send text messages -- allows moving towards this way
		of printing error messages from within objects.
  *ftr	vrpn_System_TextPrinter object created to print warnings and errors
		from any created object (by default).  User can change the ostream
		that it uses, as well as adjust its level of verbosity.  This should
		allow all objects to print errors/warnings to the user console by
		emitting VRPN text messages.  This addresses a long-standing problem
		with VRPN where you couldn't tell if there was a problem with the
		server because its error messages go to some console on another
		machine.  Now they will go to both places for objects that are
		derived from vrpn_BaseClass.
  *ftr	Handling the Ping/Pong messages that tell the client if the server is alive
		The client warns of no hearbeat in 3 seconds, error if no heartbeat
		in 10 second.  Warn/error 1/second.  Gets printed by vrpn_System_TextPrinter.
4.13:	(This version was never released publicly, changes are in 5.0, except the Logitech Magellan,
		look for it in 5.1 or email vrpn to get a beta driver).
  *ftr	Driver for the Logitech Magellan 6DOF motion controller.  It is a button and
		analog driver; you put the AnalogFly on top of its axes to make it
		act as a tracker.
  *ftr	Multiple connections to the same server has been implemented. This allows
		more than one client to connect to the same server at the same
		time.  Messages are unicast from the server to each client in this
		version.
   bug	More general type/sender ID setup -- the old version didn't look at the
		incoming index, but just assumed that they were sent in order.
   bug	Makes it so that logged messages from the client are not clock-adjusted
		(they should never have been).
   ftr	Removes the endpointPtr() method from vrpn_Connection, since there can now be
		more than one endpoint. This function provided a violation of encapsulation
		in any case. Use the newconnection/droppedconnection messages to track
		information on who is connected.
   ftr	Make it so that the server clears out all incoming requests when started
		(in case client has been trying to call for a while). Also, bug fixed where
		the client would pause for a second between a few calls, then request new
		connections as fast as it could.
4.12:	(This version was never released publicly, changes are in 5.0).
4.11:
	Adds server and remote versions of vrpn_float64 and vrpn_int32
		that allow easy sharing of information between networked
		applications. Implemented to allow shared state for multiple
		users of the nanoManipulator connected across a network.
	Make sure the millisleep() works without busywait on Unix and NT
	Shutdown/restart server w/o restarting app
		If the client starts before the server, this is okay -- it will retry
		If the server stops, the client will restore the connection later
		If you are logging on the client, it logs before and after reconnect
			The break is marked in the log file, in case we need to know
			this for some reason.
		Both fullsync and interval clock synchronization work
	Fixes the code to control a raw SGI dial/button box. The old version would
		write too fast and sometimes choke the box.
	Fixing the output on the Intersense
		It failed to send any line feed to the tracker
		It sends add a cr/lf to the screen
	Rearranged headers to get rid of vrpn_cygwin_hack.h.
	Problem with Cygwin hanging on a packet with 0-byte payload fixed
	Removing the quat directory from within server_src. Quat should now be
		checked out separately from CVS, at the same level as the VRPN
		main directory.
	Fixed problems with the output buffers in all of the classes that buffer messages. This causes
		seg faults on some architectures when it is not done. Switched all the classes
		to using the vrpn_buffer() calls, which handle unaligned buffers.
	Figured out what was making the unaligned unbuffer seg-fault happen on some machines when
		reading. PROBABLY FOUND THIS -- the TCP buffer was allocated on the fly
		based on the size of the message received. I've modified it to match the
		UDP way, which allocates a float64 buffer and points the character buffer there.
	The last two fixes should solve problems found on solaris, aix and older
		SGI machines that were causing segmentation violations.
4.10:
	Example classes (templates) for device drivers (server objects)
	Make sure you can create/delete/create remotes [destructor unregister handlers]
	test_vrpn code that verifies handler registration/unregistration
4.09:
	Fixes to allow compilation on sparc_solaris
	Client_and_server example program added
	Added sleep commands to the extended Fastrak reset commands
	Document the example vrpn.cfg on the web pages
	Document client_and_server.C in the web pages
4.08:
	Fastrak driver improved for clarity and simplicity, augmented to work with Intersense
	CerealBox driver for analog, digital and encoder inputs
	Example server configuration file (vrpn.cfg)
	Making client not complain when it cannot find vrpn_Tracker.cfg file
	AnalogFly device created to allow flying with CerealBox joystick