permissions so that clients can check if they have permission to
      change parameters (David Dawes).
 681. Fix read-only XFree86-VidModeExtension requests for remote connections
      (David Dawes, reported by Jamie Zawinski).
+ minor extension man page updates

      xkb action' in a format that xkbcomp cannot understand, and
      was causing problems when calling XkbGetKeyboard (Ivan Pascal).

  Message from Ivan Pascal describing the problem below:
--
     Hi,

>   But there is yet another (probably) simple problem that I have noticed,
> and just did patches to not allow xf86cfg to crash, when it starts, it
> tries to load the current xkb configuration, in previous versions of
> XFree86 the current code in
> xc/programs/Xserver/hw/xfree86/xf86cfg/keywboard-cfg.c:InitializeKeyboard()
> was working fine, but now it always timeouts (actually, the timeout
> check and several tries was added because an already resolved problem
> with signals and interrupted reads when forking the Xserver to call
> xkbcomp).

  Fortunatly it was a simple bug.
The thing is when you call XkbGetKeyboard the server converts a current
xkb description to the text form, calls the xkbcomp and feeds it with this
text, reads an XKM format file that the xkbcomp produces and finally sends
this data to application.  (I didn't know it is so complicated. I thought
the server simply sends its internal structures. :( )
  The problem is (was) there are some bugs in the procedures which write
the xkb format file.  And in some case they produce the text that the xkbcomp
can't understand. (I say 'bugs' becouse I suspect there are some other besides
one I caught.)
  In problem you describe the bug is a printing the name of 'the private xkb
action'.  As you know there are xkb actions recently added for server actions
such as a VT mode switching.  And in original files they looks like
  Private(type=0x86, data="...")
and it is the form that the xkbcomp understands.
  But the 'reverse procudures' print it as
  0x86(type=0x86, data[0]=...)
As the result the xkbcomp can't read the text description, doesn't produce
the XKM file and the XkbGetKeyboard procedure fails.

  I attached a patch.  As you can see it is simplest. :)
--
 Ivan U. Pascal         |   e-mail: pascal@tsu.ru
   Administrator of     |   Tomsk State University
     University Network |       Tomsk, Russia
--
  Thanks to Ivan Pascal for fixing this problem.

      Keith Packard).

      (using setxkbmap or xf86cfg) the LEDs become 'frozen' and can't
      be switched off. The patch fixes this bug. (#5544, Ivan Pascal).
 675. Fix some problems with the addition of multi-layout scheme to
      xkb: some keys that aren't 'alphabetic' should be treated as
      such to allow CapsLock+Shift working as expected; sometimes it
      is required to load an include file more than once due to
      multi-layout configuration, also fix some typos
      (#5545, Ivan Pascal).
 674. XKB programable 'special combinations' (such as Ctrl+Alt+<key>)
      also send usual key evants (press and release).  Although the
      keysym generated in this case is NoSymbol such events can confuse
      some applications (#5546, Ivan Pascal).
 673. Fix some remaining memory leaks in xkb initialization code
      (Paulo César Pereira de Andrade).

  Many thanks to Ivan Pascal for fixing these bugs and reviewing my patch
for fixing remaining memory leaks.

directly. Fixes some segfaults.

   not be traversed in every gc call, also, associate it with the atom
   symbol structure instead of storing documentation in a global list.
 o Add setf method for documentation, argument to the setf method must
   be either a string to replace/create the doc-string, or nil to
   explicitly release the memory used for documentation. Also,
   documentation memory is never automatically released, that is, it
   is possible to have a symbol with no function associated, but with
   a function doc-string, this should make it easier to, for example,
   store documentation in a file, and optionally load it, as most times
   documentation is not required.
 o Fix bug in indentation definition for the C language, preprocessor
   lines were sometimes interfering incorrectly with the indentation
   calculation, simple case is:
	#else
		<c-code>
   the indentation code was treating "#else" as a normal "else".
 o Correct pattern for identifier in progmodes/lisp.lsp, after previous
   patch was not highlighting package names.

      until a better fallback mechanism is implemented (Egbert Eich).
 671. Allow XKB-remappable hot keys to work with the "kbd" driver
      (Egbert Eich).
 670. Export xf86inSuspend as needed by the "kbd" driver (Egbert Eich).

       - support for SiS 330 Xabre (untested)
       - Bugfix for boxes with two VGA connectors (tested)
       - Autodetection of second monitor (tested)
       - TV detection improved (tested)
       - fixed error in 1280x1024 panel support (yet untested)
       - Fixed bug in TV output on SiS30xB/30xLV (PAL still only BW, not
         even the Windows driver can display PAL in color)
       - Fixed positioning of TV picture. Works now for all video bridges
         except Chrontel 7019 (still untested).
      (Thomas Winischhofer).

      Rene Cougnenc, Nicolas Mailhot).

      pc105 xkb maps (based on #A.1445, Göran Uddeborg).

      xkb maps (#A.1440, Robin Rosenberg).

      Séamus Ó Ciardhuáin).

      and laptops (#5542, Séamus Ó Ciardhuáin).
 663. Update the Irish xkb map, including adding support for laptops (#5541,
      Séamus Ó Ciardhuáin).

      Scott Penrose).

      Raphaël Poss).

      by people with low dexterity (#A.1224, Paul Elliott).

      is a VGA mode with large horizontal and/or vertical blanking pulses
      (Marc La France).