comparison etc/NEWS @ 90159:08185296b491

Revision: [e-mail]--gnu-2005/emacs--unicode--0--patch-44 Merge from emacs--cvs-trunk--0 Patches applied: * emacs--cvs-trunk--0 (patch 272-288) - src/xdisp.c (dump_glyph_row): Don't display overlay_arrow_p field. - Update from CVS - Merge from gnus--rel--5.10 * gnus--rel--5.10 (patch 67) - Update from CVS

90158:bf4846baba9a

eight-bit-control/eight-bit-graphic charsets aren't now in the range
128-255.

* Installation Changes in Emacs 22.1

** Emacs includes now support for loading image libraries on demand.
(Currently this feature is only used on MS Windows.)  You can configure
the supported image types and their associated dynamic libraries by
setting the variable `image-library-alist'.

---
** New translations of the Emacs Tutorial are available in the following
   languages: Brasilian, Bulgarian, Chinese (both with simplified and
   traditional characters), French, and Italian.  Type `C-u C-h t' to
   choose one of them in case your language setup doesn't automatically
   select the right one.

---
** You can build Emacs with Gtk+ widgets by specifying `--with-x-toolkit=gtk'
when you run configure.  This requires Gtk+ 2.0 or newer.  This port
provides a way to display multilingual text in menus (with some caveats).

---
** Emacs can now be built without sound support.

---
** The `emacsserver' program has been removed, replaced with elisp code.

---
** Emacs now supports new configure options `--program-prefix',
`--program-suffix' and `--program-transform-name' that affect the names of
installed programs.

---
** By default, Emacs now uses a setgid helper program to update game
scores.  The directory ${localstatedir}/games/emacs is the normal
place for game scores to be stored.  This may be controlled by the

Emacs with Leim.

+++
** The Emacs Lisp Reference Manual is now part of the distribution.

The ELisp reference manual in Info format is built as part of the
Emacs build procedure and installed together with the Emacs User
Manual.  A menu item was added to the menu bar that makes it easy
accessible (Help->More Manuals->Emacs Lisp Reference).

---

together with the Emacs User Manual, into the Info directory.  A menu
item was added to the menu bar that makes it easy accessible
(Help->More Manuals->Introduction to Emacs Lisp).

---
** Support for Cygwin was added.

---
** Support for FreeBSD/Alpha has been added.

** Mac OS 9 port now uses the Carbon API by default.  You can also
create non-Carbon build by specifying `NonCarbon' as a target.  See
the files mac/README and mac/INSTALL for build instructions.

---
** A French translation of the `Emacs Survival Guide' is available.

---
** Building with -DENABLE_CHECKING does not automatically build with union
types any more.  Add -DUSE_LISP_UNION_TYPE if you want union types.


* Changes in Emacs 22.1

** Improved Thai support.  A new minor mode `thai-word-mode' (which is
automatically activated if you select Thai as a language
environment) changes key bindings of most word-oriented commands to
versions which recognize Thai words.  Affected commands are
    M-f     (forward-word)

    M-d     (kill-word)
    M-DEL   (backward-kill-word)
    M-t     (transpose-words)
    M-q     (fill-paragraph)

** fast-lock.el and lazy-lock.el are obsolete.  Use jit-lock.el instead.

** iso-acc.el is now obsolete.  Use one of the latin input methods instead.

---
** Language environment and various default coding systems are setup
more correctly according to the current locale name.  If the locale
name doesn't specify a charset, the default is what glibc defines.
This change may result in using the different coding systems as
default in some locale (e.g. vi_VN).

+++
** The commands copy-file, rename-file, make-symbolic-link and
add-name-to-file, when given a directory as the "new name" argument,
convert it to a file name by merging in the within-directory part of
the existing file's name.  (This is the same convention that shell
commands cp, mv, and ln follow.)  Thus, M-x copy-file RET ~/foo RET
/tmp RET copies ~/foo to /tmp/foo.

+++
** M-o now is the prefix key for setting text properties;
M-o M-o requests refontification.

+++
** M-g is now a prefix key.

M-g g and M-g M-g run goto-line.
M-g n and M-g M-n run next-error (like C-x `).
M-g p and M-g M-p run previous-error.

+++
** font-lock-lines-before specifies a number of lines before the
current line that should be refontified when you change the buffer.
The default value is 1.

+++
** C-u M-x goto-line now switches to the most recent previous buffer,
and goes to the specified line in that buffer.

When goto-line starts to execute, if there's a number in the buffer at
point then it acts as the default argument for the minibuffer.

---
** Emacs now responds to mouse-clicks on the mode-line, header-line and
display margin, when run in an xterm.

+++
** M-SPC (just-one-space) when given a numeric argument N
converts whitespace around point to N spaces.

+++
** Control characters and escape glyphs are now shown in the new
escape-glyph face.

+++
** Non-breaking space and hyphens are now prefixed with an escape
character, unless the new user variable `show-nonbreak-escape' is set
to nil.

---
** The type-break package now allows `type-break-file-name' to be nil
and if so, doesn't store any data across sessions.  This is handy if
you don't want the .type-break file in your home directory or are
annoyed by the need for interaction when you kill Emacs.

---
** display-battery has been replaced by display-battery-mode.

---
** calculator.el now has radix grouping mode, which is available when
`calculator-output-radix' is non-nil.  In this mode a separator
character is used every few digits, making it easier to see byte
boundries etc.  For more info, see the documentation of the variable
`calculator-radix-grouping-mode'.

+++
** You can now follow links by clicking Mouse-1 on the link.

Traditionally, Emacs uses a Mouse-1 click to set point and a Mouse-2
click to follow a link, whereas most other applications use a Mouse-1
click for both purposes, depending on whether you click outside or
inside a link.  Now the behavior of a Mouse-1 click has been changed
to match this context-sentitive dual behavior.

Depending on the current mode, a Mouse-2 click in Emacs may do much
more than just follow a link, so the new Mouse-1 behavior is only
activated for modes which explicitly mark a clickable text as a "link"
(see the new function `mouse-on-link-p' for details).  The Lisp
packages that are included in release 22.1 have been adapted to do
this, but external packages may not yet support this.  However, there
is no risk in using such packages, as the worst thing that could
happen is that you get the original Mouse-1 behavior when you click
on a link, which typically means that you set point where you click.

If you want to get the original Mouse-1 action also inside a link, you
just need to press the Mouse-1 button a little longer than a normal
click (i.e. press and hold the Mouse-1 button for half a second before
you release it).

Dragging the Mouse-1 inside a link still performs the original
drag-mouse-1 action, typically copy the text.

You can customize the new Mouse-1 behavior via the new user options
`mouse-1-click-follows-link' and `mouse-1-click-in-non-selected-windows'.

+++
** require-final-newline now has two new possible values:

`visit' means add a newline (as an undoable change) if it's needed
when visiting the file.

`visit-save' means add a newline (as an undoable change) if it's
needed when visiting the file, and also add a newline if it's needed
when saving the file.

+++
** The new option mode-require-final-newline controls how certain
major modes enable require-final-newline.  Any major mode that's
designed for a kind of file that should normally end in a newline
sets require-final-newline based on mode-require-final-newline.
So you can customize mode-require-final-newline to control what these
modes do.

+++
** When the undo information of the current command gets really large
(beyond the value of `undo-outer-limit'), Emacs discards it and warns
you about it.

+++
** line-move-ignore-invisible now defaults to t.

+++
** In Outline mode, hide-body no longer hides lines at the top
of the file that precede the first header line.

+++
** In Enriched mode, `set-left-margin' and `set-right-margin' are now
by default bound to `C-c [' and `C-c ]' instead of the former `C-c C-l'
and `C-c C-r'.

+++
** In processing a local variables list, Emacs strips the prefix and
suffix are from every line before processing all the lines.

+++
** `apply-macro-to-region-lines' now operates on all lines that begin
in the region, rather than on all complete lines in the region.

---
** global-whitespace-mode is a new alias for whitespace-global-mode.

+++
** There are now two new regular expression operators, \_< and \_>,
for matching the beginning and end of a symbol.  A symbol is a
non-empty sequence of either word or symbol constituent characters, as
specified by the syntax table.
---
*** rx.el has new corresponding `symbol-end' and `symbol-start' elements.

+++
** Passing resources on the command line now works on MS Windows.
You can use --xrm to pass resource settings to Emacs, overriding any
existing values.  For example:

  emacs --xrm "Emacs.Background:red" --xrm "Emacs.Geometry:100x20"

will start up Emacs on an initial frame of 100x20 with red background,
irrespective of geometry or background setting on the Windows registry.

---
** The terminal emulation code in term.el has been improved, it can
run most curses applications now.

** New features in evaluation commands

+++
*** The function `eval-defun' (C-M-x) called on defface reinitializes
the face to the value specified in the defface expression.

+++
*** Typing C-x C-e twice prints the value of the integer result
in additional formats (octal, hexadecimal, character) specified
by the new function `eval-expression-print-format'.  The same
function also defines the result format for `eval-expression' (M-:),
`eval-print-last-sexp' (C-j) and some edebug evaluation functions.

---
** New input method chinese-sisheng for inputting Chinese Pinyin
characters.

+++
** New command quail-show-key shows what key (or key sequence) to type
in the current input method to input a character at point.

+++
** You can now switch buffers in a cyclic order with C-x C-left and
(prev-buffer) and C-x C-right (next-buffer).  C-x left and C-x right
can be used as well.

---
** Commands winner-redo and winner-undo, from winner.el, are now bound to
C-c <left> and C-c <right>, respectively.  This is an incompatible change.

---
** Help commands `describe-function' and `describe-key' now show function
arguments in lowercase italics on displays that support it.  To change the
default, customize face `help-argument-name' or redefine the function
`help-default-arg-highlight'.

---
** The comint prompt can now be made read-only, using the new user
option `comint-prompt-read-only'.  This is not enabled by default,
except in IELM buffers.  The read-only status of IELM prompts can be

not the case, then `comint-kill-region' behaves just like
`kill-region' if read-only are involved: it copies the text to the
kill-ring, but does not delete it.

+++
** You can now use next-error (C-x `) and previous-error to advance to
the next/previous matching line found by M-x occur.

+++
** Telnet now prompts you for a port number with C-u M-x telnet.

+++
** New command line option -Q or --quick.

This is like using -q --no-site-file, but in addition it also disables
the fancy startup screen.

+++
** New command line option -D or --basic-display.

Disables the menu-bar, the tool-bar, the scroll-bars, tool tips, and
the blinking cursor.

+++
** New command line option -nbc or --no-blinking-cursor disables
the blinking cursor on graphical terminals.

+++
** C-h v and C-h f commands now include a hyperlink to the C source for
variables and functions defined in C (if the C source is available).

---
** When used interactively, `format-write-file' now asks for confirmation
before overwriting an existing file, unless a prefix argument is
supplied.  This behavior is analogous to `write-file'.

+++
** You can now use Auto Revert mode to `tail' a file.
If point is at the end of a file buffer before reverting, Auto Revert
mode keeps it at the end after reverting.  Similarly if point is
displayed at the end of a file buffer in any window, it stays at
the end of the buffer in that window.  This allows to tail a file:
just put point at the end of the buffer and it stays there.  This

then you can tail the file more efficiently by using the new minor
mode Auto Revert Tail mode.  The function `auto-revert-tail-mode'
toggles this mode.

+++
** Auto Revert mode is now more careful to avoid excessive reverts and
other potential problems when deciding which non-file buffers to
revert.  This matters especially if Global Auto Revert mode is enabled
and `global-auto-revert-non-file-buffers' is non-nil.  Auto Revert
mode only reverts a non-file buffer if the buffer has a non-nil
`revert-buffer-function' and a non-nil `buffer-stale-function', which
decides whether the buffer should be reverted.  Currently, this means
that auto reverting works for Dired buffers (although this may not
work properly on all operating systems) and for the Buffer Menu.

+++
** If the new user option `auto-revert-check-vc-info' is non-nil, Auto
Revert mode reliably updates version control info (such as the version
control number in the mode line), in all version controlled buffers in
which it is active.  If the option is nil, the default, then this info
only gets updated whenever the buffer gets reverted.

+++
** New command `Buffer-menu-toggle-files-only' toggles display of file
buffers only in the Buffer Menu.  It is bound to `T' in Buffer Menu
mode.

---
** M-x compile has become more robust and reliable

Quite a few more kinds of messages are recognized.  Messages that are
recognized as warnings or informational come in orange or green, instead of
red.  Informational messages are by default skipped with `next-error'
(controlled by `compilation-skip-threshold').

Location data is collected on the fly as the *compilation* buffer changes.
This means you could modify messages to make them point to different files.
This also means you can not go to locations of messages you may have deleted.

The variable `compilation-error-regexp-alist' has now become customizable.  If
you had added your own regexps to this, you'll probably need to include a
leading `^', otherwise they'll match anywhere on a line.  There is now also a
`compilation-mode-font-lock-keywords' and it nicely handles all the checks
that configure outputs and -o options so you see at a glance where you are.

The new file etc/compilation.txt gives examples of each type of message.

** Compilation mode enhancements:

+++
*** New user option `compilation-environment'.
This option allows you to specify environment variables for inferior
compilation processes without affecting the environment that all
subprocesses inherit.

+++
** Grep has been decoupled from compilation mode setup.

---
*** There's a new separate package grep.el.

---
*** M-x grep has been adapted to new compile

Hits are fontified in green, and hits in binary files in orange.  Grep buffers
can be saved and automatically revisited with the new Grep mode.

---
*** Grep commands now have their own submenu and customization group.

+++
*** `grep-find' is now also available under the name `find-grep' where
people knowing `find-grep-dired' would probably expect it.

---
*** The new variables `grep-window-height', `grep-auto-highlight', and
`grep-scroll-output' can be used to override the corresponding
compilation mode settings for grep commands.

+++
*** New option `grep-highlight-matches' highlightes matches in *grep*
buffer.  It uses a special feature of some grep programs which accept
--color option to output markers around matches.  When going to the next
match with `next-error' the exact match is highlighted in the source
buffer.  Otherwise, if `grep-highlight-matches' is nil, the whole
source line is highlighted.

+++
*** New key bindings in grep output window:
SPC and DEL scrolls window up and down.  C-n and C-p moves to next and
previous match in the grep window.  RET jumps to the source line of
the current match.  `n' and `p' shows next and previous match in
other window, but does not switch buffer.  `{' and `}' jumps to the
previous or next file in the grep output.  TAB also jumps to the next
file.

+++
** New options `next-error-highlight' and `next-error-highlight-no-select'
specify the method of highlighting of the corresponding source line
in new face `next-error'.

+++
** A new minor mode `next-error-follow-minor-mode' can be used in
compilation-mode, grep-mode, occur-mode, and diff-mode (i.e. all the
modes that can use `next-error').  In this mode, cursor motion in the
buffer causes automatic display in another window of the corresponding
matches, compilation errors, etc.  This minor mode can be toggled with
C-c C-f.

+++
** M-x diff uses diff-mode instead of compilation-mode.

+++
** M-x compare-windows now can automatically skip non-matching text to
resync points in both windows.

---
** New command `strokes-global-set-stroke-string'.
This is like `strokes-global-set-stroke', but it allows you to bind
the stroke directly to a string to insert.  This is convenient for
using strokes as an input method.

** Gnus package

---
*** Gnus now includes Sieve and PGG
Sieve is a library for managing Sieve scripts.  PGG is a library to handle
PGP/MIME.

---
*** There are many news features, bug fixes and improvements.
See the file GNUS-NEWS or the node "Oort Gnus" in the Gnus manual for details.

+++
** Desktop package

+++

Customize the new option `save-place-forget-unreadable-files' to nil
to get the old behavior.  The new options `save-place-save-skipped'
and `save-place-skip-check-regexp' allow further fine-tuning of this
feature.

+++
** You can have several Emacs servers on the same machine.

	% emacs --eval '(setq server-name "foo")' -f server-start &
	% emacs --eval '(setq server-name "bar")' -f server-start &
	% emacsclient -s foo file1
	% emacsclient -s bar file2

+++
** On window systems, lines which are exactly as wide as the window
(not counting the final newline character) are no longer broken into
two lines on the display (with just the newline on the second line).
Instead, the newline now "overflows" into the right fringe, and the
cursor will be displayed in the fringe when positioned on that newline.

The new user option 'overflow-newline-into-fringe' may be set to nil to
revert to the old behavior of continuing such lines.

+++
** The buffer boundaries (i.e. first and last line in the buffer) may
now be marked with angle bitmaps in the fringes.  In addition, up and
down arrow bitmaps may be shown at the top and bottom of the left or
right fringe if the window can be scrolled in either direction.

This behavior is activated by setting the buffer-local variable
`indicate-buffer-boundaries' to a non-nil value.  The default value of
this variable is found in `default-indicate-buffer-boundaries'.

If value is `left' or `right', both angle and arrow bitmaps are
displayed in the left or right fringe, resp.

Value may also be an alist which specifies the presense and position
of each bitmap individually.

For example, ((top . left) (t .  right)) places the top angle bitmap
in left fringe, the bottom angle bitmap in right fringe, and both
arrow bitmaps in right fringe.  To show just the angle bitmaps in the
left fringe, but no arrow bitmaps, use ((top . left) (bottom . left)).

+++
** New command `display-local-help' displays any local help at point
in the echo area.  It is bound to `C-h .'.  It normally displays the
same string that would be displayed on mouse-over using the
`help-echo' property, but, in certain cases, it can display a more
keyboard oriented alternative.

+++
** New user option `help-at-pt-display-when-idle' allows to
automatically show the help provided by `display-local-help' on
point-over, after suitable idle time.  The amount of idle time is
determined by the user option `help-at-pt-timer-delay' and defaults
to one second.  This feature is turned off by default.

---
** New commands `scan-buf-next-region' and `scan-buf-previous-region'
move to the start of the next (previous, respectively) region with
non-nil help-echo property and display any help found there in the
echo area, using `display-local-help'.

+++
** Help mode now only makes hyperlinks for faces when the face name is
preceded or followed by the word `face'.  It no longer makes
hyperlinks for variables without variable documentation, unless
preceded by one of the words `variable' or `option'.  It now makes
hyperlinks to Info anchors (or nodes) if the anchor (or node) name is
enclosed in single quotes and preceded by `info anchor' or `Info
anchor' (in addition to earlier `info node' and `Info node').

+++
** The max size of buffers and integers has been doubled.
On 32bit machines, it is now 256M (i.e. 268435455).

+++
** The -f option, used from the command line to call a function,
now reads arguments for the function interactively if it is
an interactively callable function.

---
** sql changes.

*** The variable `sql-product' controls the highlightng of different

*** Added SQL->Start SQLi Session menu entry which calls the
appropriate sql-interactive-mode wrapper for the current setting of
`sql-product'.

---
** M-x view-file and commands that use it now avoid interfering
with special modes such as Tar mode.

** Enhancements to apropos commands:

+++
*** The apropos commands now accept a list of words to match.
When more than one word is specified, at least two of those words must
be present for an item to match.  Regular expression matching is still
available.

+++
*** The new option `apropos-sort-by-scores' causes the matching items
to be sorted according to their score.  The score for an item is a
number calculated to indicate how well the item matches the words or
regular expression that you entered to the apropos command.  The best
match is listed first, and the calculated score is shown for each
matching item.

+++
** The old bindings C-M-delete and C-M-backspace have been deleted,
since there are situations where one or the other will shut down
the operating system or your X server.

---
** New minor mode, Visible mode, toggles invisibility in the current buffer.
When enabled, it makes all invisible text visible.  When disabled, it
restores the previous value of `buffer-invisibility-spec'.

---
** New command `kill-whole-line' kills an entire line at once.
By default, it is bound to C-S-<backspace>.

---
** New commands to operate on pairs of open and close characters:
`insert-pair', `delete-pair', `raise-sexp'.

+++
** A prefix argument of C-M-q in Emacs Lisp mode pretty-printifies the
list starting after point.

** Dired mode:

---
*** New faces dired-header, dired-mark, dired-marked, dired-flagged,
dired-ignored, dired-directory, dired-symlink, dired-warning
introduced for Dired mode instead of font-lock faces.

*** New Dired command `dired-compare-directories' marks files
with different file attributes in two dired buffers.

+++
*** New Dired command `dired-do-touch' (bound to T) changes timestamps
of marked files with the value entered in the minibuffer.

+++
*** In Dired's ! command (dired-do-shell-command), `*' and `?' now
control substitution of the file names only when they are surrounded
by whitespace.  This means you can now use them as shell wildcards
too.  If you want to use just plain `*' as a wildcard, type `*""'; the
doublequotes make no difference in the shell, but they prevent
special treatment in `dired-do-shell-command'.

+++
*** In Dired, the w command now copies the current line's file name
into the kill ring.  With a zero prefix arg, copies absolute file names.

+++
** Dired-x:

+++
*** Omitting files is now a minor mode, dired-omit-mode. The mode toggling
command is bound to M-o. A new command dired-mark-omitted, bound to M-O,
marks omitted files. The variable dired-omit-files-p is obsoleted, use the
mode toggling function instead.

+++
** find-file-read-only visits multiple files in read-only mode,
when the file name contains wildcard characters.

+++
** find-alternate-file replaces the current file with multiple files,
when the file name contains wildcard characters.

** FFAP

+++
*** New ffap commands and keybindings: C-x C-r (`ffap-read-only'),
C-x C-v (`ffap-alternate-file'), C-x C-d (`ffap-list-directory'),
C-x 4 r (`ffap-read-only-other-window'), C-x 4 d (`ffap-dired-other-window'),

---
*** FFAP accepts wildcards in a file name by default.  C-x C-f passes
it to `find-file' with non-nil WILDCARDS argument, which visits
multiple files, and C-x d passes it to `dired'.

** Info mode:

+++
*** A numeric prefix argument of `info' selects an Info buffer
with the number appended to the *info* buffer name (e.g. "*info*<2>").

---
*** isearch in Info uses Info-search and searches through multiple nodes.
Before leaving the initial Info node isearch fails once with the error
message [initial node], and with subsequent C-s/C-r continues through
other nodes.  When isearch fails for the rest of the manual, it wraps
aroung the whole manual to the top/final node.  The user option
`Info-isearch-search' controls whether to use Info-search for isearch,
or the default isearch search function that wraps around the current
Info node.

*** New search commands: `Info-search-case-sensitively' (bound to S),
`Info-search-backward', and `Info-search-next' which repeats the last
search without prompting for a new search string.

*** New command `Info-history-forward' (bound to r and new toolbar icon)
moves forward in history to the node you returned from after using
`Info-history-back' (renamed from `Info-last').

*** New command `Info-history' (bound to L) displays a menu of visited nodes.

*** New command `Info-toc' (bound to T) creates a node with table of contents
from the tree structure of menus of the current Info file.

*** New command `info-apropos' searches the indices of the known
Info files on your system for a string, and builds a menu of the
possible matches.

*** New command `Info-copy-current-node-name' (bound to w) copies
the current Info node name into the kill ring.  With a zero prefix
arg, puts the node name inside the `info' function call.

---
*** New face `info-xref-visited' distinguishes visited nodes from unvisited
and a new option `Info-fontify-visited-nodes' to control this.

*** http and ftp links in Info are now operational: they look like cross
references and following them calls `browse-url'.

+++
*** Info now hides node names in menus and cross references by default.
If you prefer the old behavior, you can set the new user option
`Info-hide-note-references' to nil.

---
*** Images in Info pages are supported.
Info pages show embedded images, in Emacs frames with image support.
Info documentation that includes images, processed with makeinfo
version 4.7 or newer, compiles to Info pages with embedded images.

+++
*** The default value for `Info-scroll-prefer-subnodes' is now nil.

---
*** Info-index offers completion.

---
** Support for the SQLite interpreter has been added to sql.el by calling
'sql-sqlite'.

** BibTeX mode:
*** The new command bibtex-url browses a URL for the BibTeX entry at
point (bound to C-c C-l and mouse-2, RET on clickable fields).

*** The new command bibtex-entry-update (bound to C-c C-u) updates
an existing BibTeX entry.

*** New `bibtex-entry-format' option `required-fields', enabled by default.

*** bibtex-maintain-sorted-entries can take values `plain',
`crossref', and `entry-class' which control the sorting scheme used
for BibTeX entries.  `bibtex-sort-entry-class' controls the sorting
scheme `entry-class'.  TAB completion for reference keys and
automatic detection of duplicates does not require anymore that
bibtex-maintain-sorted-entries is non-nil.

*** If the new variable bibtex-parse-keys-fast is non-nil,
use fast but simplified algorithm for parsing BibTeX keys.

*** If the new variable bibtex-autoadd-commas is non-nil,
automatically add missing commas at end of BibTeX fields.

*** The new variable bibtex-autofill-types contains a list of entry
types for which fields are filled automatically (if possible).

*** The new command bibtex-complete completes word fragment before
point according to context (bound to M-tab).

*** The new commands bibtex-find-entry and bibtex-find-crossref
locate entries and crossref'd entries (bound to C-c C-s and C-c C-x).
Crossref fields are clickable (bound to mouse-2, RET).

*** In BibTeX mode the command fill-paragraph (bound to M-q) fills
individual fields of a BibTeX entry.

*** The new variables bibtex-files and bibtex-file-path define a set
of BibTeX files that are searched for entry keys.

*** The new command bibtex-validate-globally checks for duplicate keys
in multiple BibTeX files.

*** The new command bibtex-copy-summary-as-kill pushes summary
of BibTeX entry to kill ring (bound to C-c C-t).

+++
** When display margins are present in a window, the fringes are now
displayed between the margins and the buffer's text area, rather than
at the edges of the window.

+++
** A window may now have individual fringe and scroll-bar settings,
in addition to the individual display margin settings.

Such individual settings are now preserved when windows are split
horizontally or vertically, a saved window configuration is restored,
or when the frame is resized.

+++
** New functions frame-current-scroll-bars and window-current-scroll-bars.

These functions return the current locations of the vertical and
horizontal scroll bars in a frame or window.

+++
** Emacs now supports drag and drop for X.  Dropping a file on a window
   opens it, dropping text inserts the text.  Dropping a file on a dired
   buffer copies or moves the file to that directory.

+++
** Under X, mouse-wheel-mode is turned on by default.

+++
** The X resource useXIM can be used to turn off use of XIM, which may
speed up Emacs with slow networking to the X server.

If the configure option `--without-xim' was used to turn off use of
XIM by default, the X resource useXIM can be used to turn it on.

+++
** The X resource cursorBlink can be used to turn off cursor blinking.

+++
** `undo-only' does an undo which does not redo any previous undo.

---
** `uniquify-strip-common-suffix' tells uniquify to prefer
`file|dir1' and `file|dir2' to `file|dir1/subdir' and `file|dir2/subdir'.

+++
** If the user visits a file larger than `large-file-warning-threshold',
Emacs prompts her for confirmation.

---
** A UTF-7 coding system is available in the library `utf-7'.

---
** GUD mode has its own tool bar for controlling execution of the inferior
and other common debugger commands.

---
** recentf changes.

The recent file list is now automatically cleanup when recentf mode is
enabled.  The new option `recentf-auto-cleanup' controls when to do
automatic cleanup.

The `recentf-keep' option replaces `recentf-keep-non-readable-files-p'
and provides a more general mechanism to customize which file names to
keep in the recent list.

With the more advanced option: `recentf-filename-handler', you can
specify a function that transforms filenames handled by recentf.  For
example, if set to `file-truename', the same file will not be in the
recent list with different symbolic links.

To follow naming convention, `recentf-menu-append-commands-flag'
replaces the misnamed option `recentf-menu-append-commands-p'.  The
old name remains available as alias, but has been marked obsolete.

+++
** The default for the paper size (variable ps-paper-type) is taken
from the locale.

+++
** Init file changes

You can now put the init files .emacs and .emacs_SHELL under
~/.emacs.d or directly under ~.  Emacs will find them in either place.

---
** partial-completion-mode now does partial completion on directory names.

---
** skeleton.el now supports using - to mark the skeleton-point without
   interregion interaction.  @ has reverted to only setting
   skeleton-positions and no longer sets skeleton-point.  Skeletons
   which used @ to mark skeleton-point independent of _ should now use -
   instead.  The updated skeleton-insert docstring explains these new
   features along with other details of skeleton construction.

---
** MH-E changes.

Upgraded to MH-E version 7.82. There have been major changes since
version 5.0.2; see MH-E-NEWS for details.

+++
** The `emacsclient' command understands the options `--eval' and
`--display' which tell Emacs respectively to evaluate the given elisp
expression and to use the given display when visiting files.

+++
** User option `server-mode' can be used to start a server process.

+++
** The mode line position information now comes before the major mode.
When the file is maintained under version control, that information
appears between the position information and the major mode.

+++
** C-x s (save-some-buffers) now offers an option `d' to diff a buffer
against its file, so you can see what changes you would be saving.

+++
** You can now customize the use of window fringes.  To control this
for all frames, use M-x fringe-mode or the Show/Hide submenu of the
top-level Options menu, or customize the `fringe-mode' variable.  To
control this for a specific frame, use the command M-x
set-fringe-style.

+++
** There is a new user option `mail-default-directory' that allows you
to specify the value of `default-directory' for mail buffers.  This
directory is used for auto-save files of mail buffers.  It defaults to
"~/".

+++
** When you are root, and you visit a file whose modes specify
read-only, the Emacs buffer is now read-only too.  Type C-x C-q if you
want to make the buffer writable.  (As root, you can in fact alter the
file.)

+++
** The new command `revert-buffer-with-coding-system' (C-x RET r)
revisits the current file using a coding system that you specify.

+++
** The new command `recode-file-name' changes the encoding of the name
of a file.

---
** `ps-print' can now print characters from the mule-unicode charsets.

Printing text with characters from the mule-unicode-* sets works with
ps-print, provided that you have installed the appropriate BDF fonts.
See the file INSTALL for URLs where you can find these fonts.

---
** The new options `buffers-menu-show-directories' and
`buffers-menu-show-status' let you control how buffers are displayed
in the menu dropped down when you click "Buffers" from the menu bar.

`buffers-menu-show-directories' controls whether the menu displays
leading directories as part of the file name visited by the buffer.
If its value is `unless-uniquify', the default, directories are
shown unless uniquify-buffer-name-style' is non-nil.  The value of nil
and t turn the display of directories off and on, respectively.

`buffers-menu-show-status' controls whether the Buffers menu includes
the modified and read-only status of the buffers.  By default it is
t, and the status is shown.

Setting these variables directly does not take effect until next time
the Buffers menu is regenerated.

+++
** The commands M-x customize-face and M-x customize-face-other-window
now look at the character after point.  If a face or faces are
specified for that character, the commands by default customize those
faces.

---
** New language environments: French, Ukrainian, Tajik,
Bulgarian, Belarusian, Ukrainian, UTF-8, Windows-1255, Welsh, Latin-6,
Latin-7, Lithuanian, Latvian, Swedish, Slovenian, Croatian, Georgian,
Italian, Russian, Malayalam, Tamil, Russian, Chinese-EUC-TW.  (Set up
automatically according to the locale.)

---
** Indian support has been updated.
The in-is13194 coding system is now Unicode-based.  CDAC fonts are
assumed.  There is a framework for supporting various
Indian scripts, but currently only Devanagari, Malayalam and Tamil are
supported.

---
** New input methods: latin-alt-postfix, latin-postfix, latin-prefix,
ukrainian-computer, belarusian, bulgarian-bds, russian-computer,
vietnamese-telex, lithuanian-numeric, lithuanian-keyboard,
latvian-keyboard, welsh, georgian, rfc1345, ucs, sgml,
bulgarian-phonetic, dutch, slovenian, croatian, malayalam-inscript,
tamil-inscript.

---
** A new coding system `euc-tw' has been added for traditional Chinese
in CNS encoding; it accepts both Big 5 and CNS as input; on saving,
Big 5 is then converted to CNS.

---
** Many new coding systems are available by loading the `code-pages'
library.  These include complete versions of most of those in
codepage.el, based on Unicode mappings.  `codepage-setup' is now
obsolete and is used only in the MS-DOS port of Emacs.  windows-1252
and windows-1251 are preloaded since the former is so common and the
latter is used by GNU locales.

---
** The utf-8/16 coding systems have been enhanced.
By default, untranslatable utf-8 sequences are simply composed into
single quasi-characters.  User option `utf-translate-cjk-mode' (it is
turned on by default) arranges to translate many utf-8 CJK character
sequences into real Emacs characters in a similar way to the Mule-UCS
system.  As this loads a fairly big data on demand, people who are not
interested in CJK characters may want to customize it to nil.
You can augment/amend the CJK translation via hash tables
`ucs-mule-cjk-to-unicode' and `ucs-unicode-to-mule-cjk'.  The utf-8
coding system now also encodes characters from most of Emacs's
one-dimensional internal charsets, specifically the ISO-8859 ones.
The utf-16 coding system is affected similarly.

---
** New variable `utf-translate-cjk-unicode-range' controls which
Unicode characters to translate in `utf-translate-cjk-mode'.

---
** iso-10646-1 (`Unicode') fonts can be used to display any range of
characters encodable by the utf-8 coding system.  Just specify the
fontset appropriately.

---
** New command `ucs-insert' inserts a character specified by its
unicode.

+++
** Limited support for character `unification' has been added.
Emacs now knows how to translate between different representations of
the same characters in various Emacs charsets according to standard
Unicode mappings.  This applies mainly to characters in the ISO 8859
sets plus some other 8-bit sets, but can be extended.  For instance,
translation works amongst the Emacs ...-iso8859-... charsets and the
mule-unicode-... ones.

By default this translation happens automatically on encoding.
Self-inserting characters are translated to make the input conformant
with the encoding of the buffer in which it's being used, where
possible.

You can force a more complete unification with the user option
unify-8859-on-decoding-mode.  That maps all the Latin-N character sets
into Unicode characters (from the latin-iso8859-1 and
mule-unicode-0100-24ff charsets) on decoding.  Note that this mode
will often effectively clobber data with an iso-2022 encoding.

---
** There is support for decoding Greek and Cyrillic characters into
either Unicode (the mule-unicode charsets) or the iso-8859 charsets,
when possible.  The latter are more space-efficient.  This is
controlled by user option utf-fragment-on-decoding.

+++
** The new command `set-file-name-coding-system' (C-x RET F) sets
coding system for encoding and decoding file names.  A new menu item
(Options->Mule->Set Coding Systems->For File Name) invokes this
command.

---
** The scrollbar under LessTif or Motif has a smoother drag-scrolling.
On the other hand, the size of the thumb does not represent the actual
amount of text shown any more (only a crude approximation of it).

---
** The pop up menus for Lucid now stay up if you do a fast click and can
be navigated with the arrow keys (like Gtk+, Mac and W32).

+++
** The Lucid menus can display multilingual text in your locale.  You have
to explicitly specify a fontSet resource for this to work, for example
`-xrm "Emacs*fontSet:  -*-helvetica-medium-r-*--*-120-*-*-*-*-*-*,*"'.

---
** Dialogs for Lucid/Athena and Lesstif/Motif now pops down when pressing
ESC, like they do for Gtk+, Mac and W32.

---
**  Dialogs and menus pop down when pressing C-g.

---
** The menu item "Open File..." has been split into two items, "New File..."
and "Open File...".  "Open File..." now opens only existing files.  This is
to support existing GUI file selection dialogs better.

+++
** The file selection dialog for Gtk+, Mac, W32 and Motif/Lesstif can be
disabled by customizing the variable `use-file-dialog'.

+++
** For Gtk+ version 2.4, you can make Emacs use the old file dialog
by setting the variable `x-use-old-gtk-file-dialog' to t.  Default is to use
the new dialog.

+++
** Emacs can produce an underscore-like (horizontal bar) cursor.
The underscore cursor is set by putting `(cursor-type . hbar)' in
default-frame-alist.  It supports variable heights, like the `bar'
cursor does.

+++
** On X, MS Windows, and Mac OS, the blinking cursor's "off" state is
now controlled by the variable `blink-cursor-alist'.

** Filesets are collections of files.  You can define a fileset in
various ways, such as based on a directory tree or based on
program files that include other program files.

Once you have defined a fileset, you can perform various operations on
all the files in it, such as visiting them or searching and replacing
in them.

---
** PO translation files are decoded according to their MIME headers
when Emacs visits them.

---
** The game `mpuz' is enhanced.

`mpuz' now allows the 2nd factor not to have two identical digits.  By
default, all trivial operations involving whole lines are performed
automatically.  The game uses faces for better visual feedback.

---
** The new variable `x-select-request-type' controls how Emacs
requests X selection.  The default value is nil, which means that
Emacs requests X selection with types COMPOUND_TEXT and UTF8_STRING,
and use the more appropriately result.

+++
** The parameters of automatic hscrolling can now be customized.
The variable `hscroll-margin' determines how many columns away from
the window edge point is allowed to get before automatic hscrolling
will horizontally scroll the window.  The default value is 5.

The variable `hscroll-step' determines how many columns automatic
hscrolling scrolls the window when point gets too close to the
window edge.  If its value is zero, the default, Emacs scrolls the
window so as to center point.  If its value is an integer, it says how
many columns to scroll.  If the value is a floating-point number, it
gives the fraction of the window's width to scroll the window.

The variable `automatic-hscrolling' was renamed to
`auto-hscroll-mode'.  The old name is still available as an alias.

** TeX modes:

*** C-c C-c prompts for a command to run, and tries to offer a good default.

+++
*** The user option `tex-start-options-string' has been replaced
by two new user options: `tex-start-options', which should hold
command-line options to feed to TeX, and `tex-start-commands' which should hold
TeX commands to use at startup.

---
*** verbatim environments are now highlighted in courier by font-lock
and super/sub-scripts are made into super/sub-scripts.

+++
*** New major mode doctex-mode for *.dtx files.

+++
** New display feature: focus follows the mouse from one Emacs window
to another, even within a frame.  If you set the variable
mouse-autoselect-window to non-nil value, moving the mouse to a
different Emacs window will select that window (minibuffer window can
be selected only when it is active).  The default is nil, so that this
feature is not enabled.

+++
** On X, when the window manager requires that you click on a frame to
select it (give it focus), the selected window and cursor position
normally changes according to the mouse click position.  If you set
the variable x-mouse-click-focus-ignore-position to t, the selected
window and cursor position do not change when you click on a frame
to give it focus.

+++
** The new command `describe-char' (C-u C-x =) pops up a buffer with
description various information about a character, including its
encodings and syntax, its text properties, how to input, overlays, and
widgets at point.  You can get more information about some of them, by
clicking on mouse-sensitive areas or moving there and pressing RET.

+++
** The new command `multi-occur' is just like `occur', except it can
search multiple buffers.  There is also a new command
`multi-occur-by-filename-regexp' which allows you to specify the
buffers to search by their filename.  Internally, Occur mode has been
rewritten, and now uses font-lock, among other changes.

+++
** The default values of paragraph-start and indent-line-function have
been changed to reflect those used in Text mode rather than those used
in Indented-Text mode.

---
** New user option `query-replace-skip-read-only': when non-nil,
`query-replace' and related functions simply ignore
a match if part of it has a read-only property.

+++
** When used interactively, the commands `query-replace-regexp' and
`replace-regexp' allow \,expr to be used in a replacement string,
where expr is an arbitrary Lisp expression evaluated at replacement
time.  In many cases, this will be more convenient than using
`query-replace-regexp-eval'.  `\#' in a replacement string now refers
to the count of replacements already made by the replacement command.
All regular expression replacement commands now allow `\?' in the
replacement string to specify a position where the replacement string
can be edited for each replacement.

+++
** query-replace uses isearch lazy highlighting when the new user option
`query-replace-lazy-highlight' is non-nil.

---
** The current match in query-replace is highlighted in new face
`query-replace' which by default inherits from isearch face.

+++
** Emacs normally highlights mouse sensitive text whenever the mouse
is over the text.  By setting the new variable `mouse-highlight', you
can optionally enable mouse highlighting only after you move the
mouse, so that highlighting disappears when you press a key.  You can
also disable mouse highlighting.

+++
** You can now customize if selecting a region by dragging the mouse
shall not copy the selected text to the kill-ring by setting the new
variable mouse-drag-copy-region to nil.

+++
** font-lock: in modes like C and Lisp where the fontification assumes that
an open-paren in column 0 is always outside of any string or comment,
font-lock now highlights any such open-paren-in-column-zero in bold-red
if it is inside a string or a comment, to indicate that it can cause
trouble with fontification and/or indentation.

+++
** There's a new face `minibuffer-prompt'.
Emacs adds this face to the list of text properties stored in the
variable `minibuffer-prompt-properties', which is used to display the
prompt string.

+++
** The new face `mode-line-inactive' is used to display the mode line
of non-selected windows.  The `mode-line' face is now used to display
the mode line of the currently selected window.

The new variable `mode-line-in-non-selected-windows' controls whether
the `mode-line-inactive' face is used.

---
** A menu item "Show/Hide" was added to the top-level menu "Options".
This menu allows you to turn various display features on and off (such
as the fringes, the tool bar, the speedbar, and the menu bar itself).
You can also move the vertical scroll bar to either side here or turn
it off completely.  There is also a menu-item to toggle displaying of
current date and time, current line and column number in the
mode-line.

---
** Speedbar has moved from the "Tools" top level menu to "Show/Hide".

+++
** Emacs can now indicate in the mode-line the presence of new e-mail
in a directory or in a file.  See the documentation of the user option
`display-time-mail-directory'.

---
** LDAP support now defaults to ldapsearch from OpenLDAP version 2.

+++
** You can now disable pc-selection-mode after enabling it.
M-x pc-selection-mode behaves like a proper minor mode, and with no
argument it toggles the mode.

Turning off PC-Selection mode restores the global key bindings
that were replaced by turning on the mode.

+++
** Emacs now displays a splash screen by default even if command-line
arguments were given.  The new command-line option --no-splash
disables the splash screen; see also the variable
`inhibit-startup-message' (which is also aliased as
`inhibit-splash-screen').

** Changes in support of colors on character terminals

+++
*** The new command-line option --color=MODE lets you specify a standard
mode for a tty color support.  It is meant to be used on character
terminals whose capabilities are not set correctly in the terminal
database, or with terminal emulators which support colors, but don't
set the TERM environment variable to a name of a color-capable
terminal.  "emacs --color" uses the same color commands as GNU `ls'
when invoked with "ls --color", so if your terminal can support colors
in "ls --color", it will support "emacs --color" as well.  See the
user manual for the possible values of the MODE parameter.

---
*** Emacs now supports several character terminals which provide more
than 8 colors.  For example, for `xterm', 16-color, 88-color, and
256-color modes are supported.  Emacs automatically notes at startup
the extended number of colors, and defines the appropriate entries for
all of these colors.

+++
*** Emacs now uses the full range of available colors for the default
faces when running on a color terminal, including 16-, 88-, and
256-color xterms.  This means that when you run "emacs -nw" on an
88-color or 256-color xterm, you will see essentially the same face
colors as on X.

---
*** There's a new support for colors on `rxvt' terminal emulator.

+++
** Emacs can now be invoked in full-screen mode on a windowed display.

When Emacs is invoked on a window system, the new command-line options
`--fullwidth', `--fullheight', and `--fullscreen' produce a frame
whose width, height, or both width and height take up the entire
screen size.  (For now, this does not work with some window managers.)

---
** Emacs now tries to set up buffer coding systems for HTML/XML files
automatically.

+++
** The new command `comint-insert-previous-argument' in comint-derived
modes (shell-mode etc) inserts arguments from previous command lines,
like bash's `ESC .' binding.  It is bound by default to `C-c .', but
otherwise behaves quite similarly to the bash version.

+++
** Changes in C-h bindings:

C-h e displays the *Messages* buffer.

C-h followed by a control character is used for displaying files
    that do not change:

C-h C-f displays the FAQ.
C-h C-e displays the PROBLEMS file.

The info-search bindings on C-h C-f, C-h C-k and C-h C-i
have been moved to C-h F, C-h K and C-h S.

C-h c, C-h k, C-h w, and C-h f now handle remapped interactive commands.

- C-h c and C-h k report the actual command (after possible remapping)
  run by the key sequence.

- C-h w and C-h f on a command which has been remapped now report the
  command it is remapped to, and the keys which can be used to run
  that command.

For example, if C-k is bound to kill-line, and kill-line is remapped
to new-kill-line, these commands now report:

- C-h c and C-h k C-k reports:
  C-k runs the command new-kill-line

- C-h w and C-h f kill-line reports:
  kill-line is remapped to new-kill-line which is on C-k, <deleteline>

- C-h w and C-h f new-kill-line reports:
  new-kill-line is on C-k

+++
** Vertical scrolling is now possible within incremental search.
To enable this feature, customize the new user option
`isearch-allow-scroll'.  User written commands which satisfy stringent
constraints can be marked as "scrolling commands".  See the Emacs manual
for details.

+++
** C-w in incremental search now grabs either a character or a word,
making the decision in a heuristic way.  This new job is done by the
command `isearch-yank-word-or-char'.  To restore the old behavior,
bind C-w to `isearch-yank-word' in `isearch-mode-map'.

+++
** C-y in incremental search now grabs the next line if point is already
at the end of a line.

+++
** C-M-w deletes and C-M-y grabs a character in isearch mode.
Another method to grab a character is to enter the minibuffer by `M-e'
and to type `C-f' at the end of the search string in the minibuffer.

+++
** M-% typed in isearch mode invokes `query-replace' or
`query-replace-regexp' (depending on search mode) with the current
search string used as the string to replace.

+++
** Isearch no longer adds `isearch-resume' commands to the command
history by default.  To enable this feature, customize the new
user option `isearch-resume-in-command-history'.

+++
** New user option `history-delete-duplicates'.
If set to t when adding a new history element, all previous identical
elements are deleted.

+++
** Yanking text now discards certain text properties that can
be inconvenient when you did not expect them.  The variable
`yank-excluded-properties' specifies which ones.  Insertion
of register contents and rectangles also discards these properties.

+++
** Occur, Info, and comint-derived modes now support using
M-x font-lock-mode to toggle fontification.  The variable
`Info-fontify' is no longer applicable; to disable fontification,
remove `turn-on-font-lock' from `Info-mode-hook'.

+++
** M-x grep now tries to avoid appending `/dev/null' to the command line
by using GNU grep `-H' option instead.  M-x grep automatically
detects whether this is possible or not the first time it is invoked.
When `-H' is used, the grep command line supplied by the user is passed
unchanged to the system to execute, which allows more complicated
command lines to be used than was possible before.

---
** The face-customization widget has been reworked to be less confusing.
In particular, when you enable a face attribute using the corresponding
check-box, there's no longer a redundant `*' option in value selection
for that attribute; the values you can choose are only those which make
sense for the attribute.  When an attribute is de-selected by unchecking
its check-box, then the (now ignored, but still present temporarily in
case you re-select the attribute) value is hidden.

+++
** When you set or reset a variable's value in a Customize buffer,
the previous value becomes the "backup value" of the variable.
You can go back to that backup value by selecting "Use Backup Value"
under the "[State]" button.

---
** The new customization type `float' specifies numbers with floating
point (no integers are allowed).

+++
** In GUD mode, when talking to GDB, C-x C-a C-j "jumps" the program
counter to the specified source line (the one where point is).

---
** GUD mode improvements for jdb:

*** Search for source files using jdb classpath and class
    information. Fast startup since there is no need to scan all
    source files up front. There is also no need to create and maintain
    lists of source directories to scan. Look at `gud-jdb-use-classpath'
    and `gud-jdb-classpath' customization variables documentation.

*** Supports the standard breakpoint (gud-break, gud-clear)
    set/clear operations from java source files under the classpath, stack
    traversal (gud-up, gud-down), and run until current stack finish
    (gud-finish).

*** Supports new jdb (Java 1.2 and later) in addition to oldjdb
    (Java 1.1 jdb).

*** The previous method of searching for source files has been
    preserved in case someone still wants/needs to use it.
    Set gud-jdb-use-classpath to nil.

  Added Customization Variables

*** gud-jdb-command-name.  What command line to use to invoke jdb.

*** gud-jdb-use-classpath. Allows selection of java source file searching
    method: set to t for new method, nil to scan gud-jdb-directories for
    java sources (previous method).

*** gud-jdb-directories. List of directories to scan and search for java
    classes using the original gud-jdb method (if gud-jdb-use-classpath
    is nil).

  Minor Improvements

*** The STARTTLS elisp wrapper (starttls.el) can now use GNUTLS
instead of the OpenSSL based "starttls" tool.  For backwards
compatibility, it prefers "starttls", but you can toggle
`starttls-use-gnutls' to switch to GNUTLS (or simply remove the
"starttls" tool).

*** Do not allow debugger output history variable to grow without bounds.

+++
** hide-ifdef-mode now uses overlays rather than selective-display
to hide its text.  This should be mostly transparent but slightly
changes the behavior of motion commands like C-e and C-p.

+++
** Unquoted `$' in file names do not signal an error any more when
the corresponding environment variable does not exist.
Instead, the `$ENVVAR' text is left as is, so that `$$' quoting
is only rarely needed.

---
** JIT-lock changes
*** jit-lock can now be delayed with `jit-lock-defer-time'.

If this variable is non-nil, its value should be the amount of Emacs
idle time in seconds to wait before starting fontification.  For
example, if you set `jit-lock-defer-time' to 0.25, fontification will
only happen after 0.25s of idle time.

*** contextual refontification is now separate from stealth fontification.

jit-lock-defer-contextually is renamed jit-lock-contextually and
jit-lock-context-time determines the delay after which contextual
refontification takes place.

+++
** Marking commands extend the region when invoked multiple times.  If
you hit M-C-SPC (mark-sexp), M-@ (mark-word), M-h (mark-paragraph), or
C-M-h (mark-defun) repeatedly, the marked region extends each time, so
you can mark the next two sexps with M-C-SPC M-C-SPC, for example.
This feature also works for mark-end-of-sentence, if you bind that to
a key.  It also extends the region when the mark is active in Transient
Mark mode, regardless of the last command.  To start a new region with
one of marking commands in Transient Mark mode, you can deactivate the
active region with C-g, or set the new mark with C-SPC.

+++
** Some commands do something special in Transient Mark mode when the
mark is active--for instance, they limit their operation to the
region.  Even if you don't normally use Transient Mark mode, you might
want to get this behavior from a particular command.  There are two
ways you can enable Transient Mark mode and activate the mark, for one
command only.

One method is to type C-SPC C-SPC; this enables Transient Mark mode
and sets the mark at point.  The other method is to type C-u C-x C-x.
This enables Transient Mark mode temporarily but does not alter the
mark or the region.

After these commands, Transient Mark mode remains enabled until you
deactivate the mark.  That typically happens when you type a command
that alters the buffer, but you can also deactivate the mark by typing
C-g.

+++
** A prefix argument is no longer required to repeat a jump to a
previous mark, i.e. C-u C-SPC C-SPC C-SPC ... cycles through the
mark ring.  Use C-u C-u C-SPC to set the mark immediately after a jump.

+++
** Movement commands `beginning-of-buffer', `end-of-buffer',
`beginning-of-defun', `end-of-defun' do not set the mark if the mark
is already active in Transient Mark mode.

+++
** In the *Occur* buffer, `o' switches to it in another window, and
C-o displays the current line's occurrence in another window without
switching to it.

+++
** When you specify a frame size with --geometry, the size applies to
all frames you create.  A position specified with --geometry only
affects the initial frame.

+++
** M-h (mark-paragraph) now accepts a prefix arg.
With positive arg, M-h marks the current and the following paragraphs;
if the arg is negative, it marks the current and the preceding
paragraphs.

+++
** The variables dired-free-space-program and dired-free-space-args
have been renamed to directory-free-space-program and
directory-free-space-args, and they now apply whenever Emacs puts a
directory listing into a buffer.

---
** mouse-wheels can now scroll a specific fraction of the window
(rather than a fixed number of lines) and the scrolling is `progressive'.

---
** Unexpected yanking of text due to accidental clicking on the mouse
wheel button (typically mouse-2) during wheel scrolling is now avoided.
This behavior can be customized via the mouse-wheel-click-event and
mouse-wheel-inhibit-click-time variables.

+++
** The keyboard-coding-system is now automatically set based on your
current locale settings if you are not using a window system.  This
may mean that the META key doesn't work but generates non-ASCII
characters instead, depending on how the terminal (or terminal
emulator) works.  Use `set-keyboard-coding-system' (or customize
keyboard-coding-system) if you prefer META to work (the old default)
or if the locale doesn't describe the character set actually generated
by the keyboard.  See Info node `Single-Byte Character Support'.

+++
** Emacs now reads the standard abbrevs file ~/.abbrev_defs
automatically at startup, if it exists.  When Emacs offers to save
modified buffers, it saves the abbrevs too if they have changed.  It
can do this either silently or asking for confirmation first,
according to the value of `save-abbrevs'.

+++
** Display of hollow cursors now obeys the buffer-local value (if any)
of `cursor-in-non-selected-windows' in the buffer that the cursor
appears in.

+++
** The variable `cursor-in-non-selected-windows' can now be set to any
of the recognized cursor types.

---
** The variable `auto-save-file-name-transforms' now has a third element that
controls whether or not the function `make-auto-save-file-name' will
attempt to construct a unique auto-save name (e.g. for remote files).

+++
** There is a new calendar package, icalendar.el, that can be used to
convert Emacs diary entries to/from the iCalendar format.

+++
** Diary sexp entries can have custom marking in the calendar.
Diary sexp functions which only apply to certain days (such as
`diary-block' or `diary-cyclic') now take an optional parameter MARK,
which is the name of a face or a single-character string indicating
how to highlight the day in the calendar display.  Specifying a
single-character string as @var{mark} places the character next to the
day in the calendar.  Specifying a face highlights the day with that
face.  This lets you have different colors or markings for vacations,
appointments, paydays or anything else using a sexp.

+++
** The new function `calendar-goto-day-of-year' (g D) prompts for a
year and day number, and moves to that date. Negative day numbers
count backward from the end of the year.

+++
** The new Calendar function `calendar-goto-iso-week' (g w)
prompts for a year and a week number, and moves to the first
day of that ISO week.

---
** The new variable `calendar-minimum-window-height' affects the
window generated by the function `generate-calendar-window'.

---
** The functions `holiday-easter-etc' and `holiday-advent' now take
optional arguments, in order to only report on the specified holiday
rather than all. This makes customization of variables such as
`christian-holidays' simpler.

---
** The function `simple-diary-display' now by default sets a header line.
This can be controlled through the variables `diary-header-line-flag'
and `diary-header-line-format'.

+++
** The procedure for activating appointment reminders has changed: use
the new function `appt-activate'. The new variable
`appt-display-format' controls how reminders are displayed, replacing
appt-issue-message, appt-visible, and appt-msg-window.

+++
** The new functions `diary-from-outlook', `diary-from-outlook-gnus',
and `diary-from-outlook-rmail' can be used to import diary entries
from Outlook-format appointments in mail messages.  The variable
`diary-outlook-formats' can be customized to recognize additional
formats.


** VC Changes

+++
*** The key C-x C-q no longer checks files in or out, it only changes
the read-only state of the buffer (toggle-read-only).  We made this
change because we held a poll and found that many users were unhappy
with the previous behavior.  If you do prefer this behavior, you
can bind `vc-toggle-read-only' to C-x C-q in your .emacs:

    (global-set-key "\C-x\C-q" 'vc-toggle-read-only)

The function `vc-toggle-read-only' will continue to exist.

+++
*** There is a new user option `vc-cvs-global-switches' that allows
you to specify switches that are passed to any CVS command invoked
by VC.  These switches are used as "global options" for CVS, which
means they are inserted before the command name.  For example, this
allows you to specify a compression level using the "-z#" option for
CVS.

+++
*** New backends for Subversion and Meta-CVS.

** EDiff changes.

+++
***  When comparing directories.
Typing D brings up a buffer that lists the differences between the contents of
directories. Now it is possible to use this buffer to copy the missing files
from one directory to another.

+++
*** When comparing files or buffers.
Typing the = key now offers to perform the word-by-word comparison of the
currently highlighted regions in an inferior Ediff session. If you answer 'n'
then it reverts to the old behavior and asks the user to select regions for
comparison.

*** The new command `ediff-backup' compares a file with its most recent
backup using `ediff'.  If you specify the name of a backup file,
`ediff-backup' compares it with the file of which it is a backup.

+++
** Etags changes.

*** New regular expressions features

**** New syntax for regular expressions, multi-line regular expressions.
The syntax --ignore-case-regexp=/regex/ is now undocumented and retained
only for backward compatibility.  The new equivalent syntax is
--regex=/regex/i.  More generally, it is --regex=/TAGREGEX/TAGNAME/MODS,
where `/TAGNAME' is optional, as usual, and MODS is a string of 0 or
more characters among `i' (ignore case), `m' (multi-line) and `s'
(single-line).  The `m' and `s' modifiers behave as in Perl regular
expressions: `m' allows regexps to match more than one line, while `s'
(which implies `m') means that `.' matches newlines.  The ability to
span newlines allows writing of much more powerful regular expressions
and rapid prototyping for tagging new languages.

**** Regular expressions can use char escape sequences as in Gcc.
The escaped character sequence \a, \b, \d, \e, \f, \n, \r, \t, \v,
respectively, stand for the ASCII characters BEL, BS, DEL, ESC, FF, NL,
CR, TAB, VT,

**** Regular expressions can be bound to a given language.
The syntax --regex={LANGUAGE}REGEX means that REGEX is used to make tags
only for files of language LANGUAGE, and ignored otherwise.  This is
particularly useful when storing regexps in a file.

**** Regular expressions can be read from a file.
The --regex=@regexfile option means read the regexps from a file, one
per line.  Lines beginning with space or tab are ignored.

*** New language parsing features

**** The `::' qualifier triggers C++ parsing in C file.
Previously, only the `template' and `class' keywords had this effect.

**** The gnucc __attribute__ keyword is now recognised and ignored.

**** New language HTML.
Title and h1, h2, h3 are tagged.  Also, tags are generated when name= is
used inside an anchor and whenever id= is used.

**** In Makefiles, constants are tagged.
If you want the old behavior instead, thus avoiding to increase the
size of the tags file, use the --no-globals option.

**** New language Lua.
All functions are tagged.

**** In Perl, packages are tags.
Subroutine tags are named from their package.  You can jump to sub tags
as you did before, by the sub name, or additionally by looking for
package::sub.

**** In Prolog, etags creates tags for rules in addition to predicates.

**** New language PHP.
Tags are functions, classes and defines.
If the --members option is specified to etags, tags are variables also.

**** New default keywords for TeX.
The new keywords are def, newcommand, renewcommand, newenvironment and
renewenvironment.

*** Honour #line directives.
When Etags parses an input file that contains C preprocessor's #line
directives, it creates tags using the file name and line number
specified in those directives.  This is useful when dealing with code
created from Cweb source files.  When Etags tags the generated file, it
writes tags pointing to the source file.

*** New option --parse-stdin=FILE.
This option is mostly useful when calling etags from programs.  It can
be used (only once) in place of a file name on the command line.  Etags
reads from standard input and marks the produced tags as belonging to
the file FILE.

+++
** CC Mode changes.

*** Font lock support.
CC Mode now provides font lock support for all its languages.  This
supersedes the font lock patterns that have been in the core font lock
package for C, C++, Java and Objective-C.  Like indentation, font
locking is done in a uniform way across all languages (except the new
AWK mode - see below).  That means that the new font locking will be
different from the old patterns in various details for most languages.

The main goal of the font locking in CC Mode is accuracy, to provide a
dependable aid in recognizing the various constructs.  Some, like
strings and comments, are easy to recognize while others like
declarations and types can be very tricky.  CC Mode can go to great
lengths to recognize declarations and casts correctly, especially when
the types aren't recognized by standard patterns.  This is a fairly
demanding analysis which can be slow on older hardware, and it can
therefore be disabled by choosing a lower decoration level with the
variable font-lock-maximum-decoration.

Note that the most demanding font lock level has been tuned with lazy
fontification in mind, i.e.  there should be a support mode that waits
with the fontification until the text is actually shown
(e.g. Just-in-time Lock mode, which is the default, or Lazy Lock
mode).  Fontifying a file with several thousand lines in one go can
take the better part of a minute.

**** The (c|c++|objc|java|idl|pike)-font-lock-extra-types variables
are now used by CC Mode to recognize identifiers that are certain to
be types.  (They are also used in cases that aren't related to font
locking.)  At the maximum decoration level, types are often recognized
properly anyway, so these variables should be fairly restrictive and
not contain patterns for uncertain types.

**** Support for documentation comments.
There is a "plugin" system to fontify documentation comments like
Javadoc and the markup within them.  It's independent of the host
language, so it's possible to e.g. turn on Javadoc font locking in C
buffers.  See the variable c-doc-comment-style for details.

Currently two kinds of doc comment styles are recognized: Suns Javadoc
and Autodoc which is used in Pike.  This is by no means a complete
list of the most common tools; if your doc comment extractor of choice
is missing then please drop a note to [e-mail].

**** Better handling of C++ templates.
As a side effect of the more accurate font locking, C++ templates are
now handled much better.  The angle brackets that delimit them are
given parenthesis syntax so that they can be navigated like other
parens.

This also improves indentation of templates, although there still is
work to be done in that area.  E.g. it's required that multiline
template clauses are written in full and then refontified to be
recognized, and the indentation of nested templates is a bit odd and
not as configurable as it ought to be.

**** Improved handling of Objective-C and CORBA IDL.
Especially the support for Objective-C and IDL has gotten an overhaul.
The special "@" declarations in Objective-C are handled correctly.
All the keywords used in CORBA IDL, PSDL, and CIDL are recognized and
handled correctly, also wrt indentation.

*** Support for the AWK language.
Support for the AWK language has been introduced.  The implementation is
based around GNU AWK version 3.1, but it should work pretty well with
any AWK.  As yet, not all features of CC Mode have been adapted for AWK.
Here is a summary:

**** Indentation Engine
The CC Mode indentation engine fully supports AWK mode.

AWK mode handles code formatted in the conventional AWK fashion: `{'s
which start actions, user-defined functions, or compound statements are
placed on the same line as the associated construct; the matching `}'s
are normally placed under the start of the respective pattern, function
definition, or structured statement.

The predefined indentation functions haven't yet been adapted for AWK
mode, though some of them may work serendipitously.  There shouldn't be
any problems writing custom indentation functions for AWK mode.

The command C-c C-q (c-indent-defun) hasn't yet been adapted for AWK,
though in practice it works properly nearly all the time.  Should it
fail, explicitly set the region around the function (using C-u C-SPC:
C-M-h probably won't work either) then do C-M-\ (indent-region).

**** Font Locking
There is a single level of font locking in AWK mode, rather than the
three distinct levels the other modes have.  There are several
idiosyncrasies in AWK mode's font-locking due to the peculiarities of
the AWK language itself.

**** Comment Commands
M-; (indent-for-comment) works fine.  None of the other CC Mode
comment formatting commands have yet been adapted for AWK mode.

**** Movement Commands
Most of the movement commands work in AWK mode.  The most important
exceptions are M-a (c-beginning-of-statement) and M-e
(c-end-of-statement) which haven't yet been adapted.

The notion of "defun" has been augmented to include AWK pattern-action
pairs.  C-M-a (c-awk-beginning-of-defun) and C-M-e (c-awk-end-of-defun)
recognise these pattern-action pairs, as well as user defined
functions.

**** Auto-newline Insertion and Clean-ups
Auto-newline insertion hasn't yet been adapted for AWK.  Some of
the clean-ups can actually convert good AWK code into syntactically
invalid code.  These features are best disabled in AWK buffers.

*** New syntactic symbols in IDL mode.
The top level constructs "module" and "composition" (from CIDL) are
now handled like "namespace" in C++: They are given syntactic symbols
module-open, module-close, inmodule, composition-open,
composition-close, and incomposition.

*** New functions to do hungry delete without enabling hungry delete mode.
The functions c-hungry-backspace and c-hungry-delete-forward can be
bound to keys to get this feature without toggling a mode.
Contributed by Kevin Ryde.

*** Better control over require-final-newline.  The variable that
controls how to handle a final newline when the buffer is saved,
require-final-newline, is now customizable on a per-mode basis through
c-require-final-newline.  That is a list of modes, and only those
modes set require-final-newline.  By default that's C, C++ and
Objective-C.

The specified modes set require-final-newline based on
mode-require-final-newline, as usual.

*** Format change for syntactic context elements.
The elements in the syntactic context returned by c-guess-basic-syntax
and stored in c-syntactic-context has been changed somewhat to allow
attaching more information.  They are now lists instead of single cons
cells.  E.g. a line that previously had the syntactic analysis

((inclass . 11) (topmost-intro . 13))

is now analysed as

((inclass 11) (topmost-intro 13))

In some cases there are more than one position given for a syntactic
symbol.

This change might affect code that call c-guess-basic-syntax directly,
and custom lineup functions if they use c-syntactic-context.  However,
the argument given to lineup functions is still a single cons cell
with nil or an integer in the cdr.

*** API changes for derived modes.
There have been extensive changes "under the hood" which can affect
derived mode writers.  Some of these changes are likely to cause
incompatibilities with existing derived modes, but on the other hand
care has now been taken to make it possible to extend and modify CC
Mode with less risk of such problems in the future.

**** New language variable system.
See the comment blurb near the top of cc-langs.el.

**** New initialization functions.
The initialization procedure has been split up into more functions to
give better control: c-basic-common-init, c-font-lock-init, and
c-init-language-vars.

*** Changes in analysis of nested syntactic constructs.
The syntactic analysis engine has better handling of cases where
several syntactic constructs appear nested on the same line.  They are
now handled as if each construct started on a line of its own.

This means that CC Mode now indents some cases differently, and
although it's more consistent there might be cases where the old way
gave results that's more to one's liking.  So if you find a situation
where you think that the indentation has become worse, please report
it to [e-mail].

**** New syntactic symbol substatement-label.
This symbol is used when a label is inserted between a statement and
its substatement.  E.g:

    if (x)
      x_is_true:
        do_stuff();

*** Better handling of multiline macros.

**** Syntactic indentation inside macros.
The contents of multiline #define's are now analyzed and indented
syntactically just like other code.  This can be disabled by the new
variable c-syntactic-indentation-in-macros.  A new syntactic symbol
cpp-define-intro has been added to control the initial indentation
inside #define's.

**** New lineup function c-lineup-cpp-define.
Now used by default to line up macro continuation lines.  The behavior
of this function closely mimics the indentation one gets if the macro
is indented while the line continuation backslashes are temporarily
removed.  If syntactic indentation in macros is turned off, it works
much line c-lineup-dont-change, which was used earlier, but handles
empty lines within the macro better.

**** Automatically inserted newlines continues the macro if used within one.
This applies to the newlines inserted by the auto-newline mode, and to
c-context-line-break and c-context-open-line.

**** Better alignment of line continuation backslashes.
c-backslash-region tries to adapt to surrounding backslashes.  New
variable c-backslash-max-column which put a limit on how far out
backslashes can be moved.

**** Automatic alignment of line continuation backslashes.
This is controlled by the new variable c-auto-align-backslashes.  It
affects c-context-line-break, c-context-open-line and newlines
inserted in auto-newline mode.

**** Line indentation works better inside macros.
Regardless whether syntactic indentation and syntactic indentation
inside macros are enabled or not, line indentation now ignores the
line continuation backslashes.  This is most noticeable when syntactic
indentation is turned off and there are empty lines (save for the
backslash) in the macro.

*** indent-for-comment is more customizable.
The behavior of M-; (indent-for-comment) is now configurable through
the variable c-indent-comment-alist.  The indentation behavior based
on the preceding code on the line, e.g. to get two spaces after #else
and #endif but indentation to comment-column in most other cases
(something which was hardcoded earlier).

*** New function c-context-open-line.
It's the open-line equivalent of c-context-line-break.

*** New lineup functions

**** c-lineup-string-cont
This lineup function lines up a continued string under the one it
continues.  E.g:

result = prefix + "A message "
                  "string.";      <- c-lineup-string-cont

**** c-lineup-cascaded-calls
Lines up series of calls separated by "->" or ".".

**** c-lineup-knr-region-comment
Gives (what most people think is) better indentation of comments in
the "K&R region" between the function header and its body.

**** c-lineup-gcc-asm-reg
Provides better indentation inside asm blocks.  Contributed by Kevin
Ryde.

**** c-lineup-argcont
Lines up continued function arguments after the preceding comma.
Contributed by Kevin Ryde.

*** Better caching of the syntactic context.
CC Mode caches the positions of the opening parentheses (of any kind)
of the lists surrounding the point.  Those positions are used in many
places as anchor points for various searches.  The cache is now
improved so that it can be reused to a large extent when the point is
moved.  The less it moves, the less needs to be recalculated.

The effect is that CC Mode should be fast most of the time even when
opening parens are hung (i.e. aren't in column zero).  It's typically
only the first time after the point is moved far down in a complex
file that it'll take noticeable time to find out the syntactic
context.

*** Statements are recognized in a more robust way.
Statements are recognized most of the time even when they occur in an
"invalid" context, e.g. in a function argument.  In practice that can
happen when macros are involved.

*** Improved the way c-indent-exp chooses the block to indent.
It now indents the block for the closest sexp following the point
whose closing paren ends on a different line.  This means that the
point doesn't have to be immediately before the block to indent.
Also, only the block and the closing line is indented; the current
line is left untouched.

*** Added toggle for syntactic indentation.
The function c-toggle-syntactic-indentation can be used to toggle
syntactic indentation.

+++
** The command line option --no-windows has been changed to
--no-window-system.  The old one still works, but is deprecated.

+++
** The command `list-text-properties-at' has been deleted because
C-u C-x = gives the same information and more.

+++
** `buffer-menu' and `list-buffers' now list buffers whose names begin
with a space, when those buffers are visiting files.  Normally buffers
whose names begin with space are omitted.

+++
** You can now customize fill-nobreak-predicate to control where
filling can break lines.  The value is now normally a list of
functions, but it can also be a single function, for compatibility.

We provide two sample predicates, fill-single-word-nobreak-p and
fill-french-nobreak-p, for use in the value of fill-nobreak-predicate.

+++
** New user option `add-log-always-start-new-record'.
When this option is enabled, M-x add-change-log-entry always
starts a new record regardless of when the last record is.

+++
** SGML mode has indentation and supports XML syntax.
The new variable `sgml-xml-mode' tells SGML mode to use XML syntax.
When this option is enabled, SGML tags are inserted in XML style,
i.e., there is always a closing tag.
By default, its setting is inferred on a buffer-by-buffer basis
from the file name or buffer contents.

+++
** `xml-mode' is now an alias for `sgml-mode', which has XML support.

---
** Lisp mode now uses font-lock-doc-face for the docstrings.

---
** Perl mode has a new variable `perl-indent-continued-arguments'.

---
** Fortran mode does more font-locking by default. Use level 3
highlighting for the old default.

+++
** Fortran mode has a new variable `fortran-directive-re'.
Adapt this to match the format of any compiler directives you use.
Lines that match are never indented, and are given distinctive font-locking.

+++
** F90 mode and Fortran mode have new navigation commands
`f90-end-of-block', `f90-beginning-of-block', `f90-next-block',
`f90-previous-block', `fortran-end-of-block',
`fortran-beginning-of-block'.

---
** F90 mode and Fortran mode have support for hs-minor-mode (hideshow).
It cannot deal with every code format, but ought to handle a sizeable
majority.

---
** The new function `f90-backslash-not-special' can be used to change
the syntax of backslashes in F90 buffers.

---
** Prolog mode has a new variable `prolog-font-lock-keywords'
to support use of font-lock.

+++
** `special-display-buffer-names' and `special-display-regexps' now
understand two new boolean pseudo-frame-parameters `same-frame' and
`same-window'.

+++
** M-x setenv now expands environment variables of the form `$foo' and
`${foo}' in the specified new value of the environment variable.  To
include a `$' in the value, use `$$'.

+++
** File-name completion can now ignore directories.
If an element of the list in `completion-ignored-extensions' ends in a
slash `/', it indicates a subdirectory that should be ignored when
completing file names.  Elements of `completion-ignored-extensions'
which do not end in a slash are never considered when a completion
candidate is a directory.

+++
** The completion commands TAB, SPC and ? in the minibuffer apply only
to the text before point.  If there is text in the buffer after point,
it remains unchanged.

---
** Enhanced visual feedback in *Completions* buffer.

Completions lists use faces to highlight what all completions
have in common and where they begin to differ.

The common prefix shared by all possible completions uses the face
`completions-common-part', while the first character that isn't the
same uses the face `completions-first-difference'.  By default,
`completions-common-part' inherits from `default', and
`completions-first-difference' inherits from `bold'.  The idea of
`completions-common-part' is that you can use it to make the common
parts less visible than normal, so that the rest of the differing
parts is, by contrast, slightly highlighted.

+++
** New user option `inhibit-startup-buffer-menu'.
When loading many files, for instance with `emacs *', Emacs normally
displays a buffer menu.  This option turns the buffer menu off.

---
** Rmail now displays 5-digit message ids in its summary buffer.

+++
** Support for `movemail' from GNU mailutils was added to Rmail.
This version of `movemail' allows to read mail from a wide range of
mailbox formats, including remote POP3 and IMAP4 mailboxes with or
without TLS encryption.  If GNU mailutils is installed on the system
and its version of `movemail' can be found in exec-path, it will be
used instead of the native one.

---
** On MS Windows, the "system caret" now follows the cursor.
This enables Emacs to work better with programs that need to track
the cursor, for example screen magnifiers and text to speech programs.

multilingual text with other applications. On other versions of
MS Windows, Emacs now uses the appropriate locale coding-system, so
the clipboard should work correctly for your local language without
any customizations.

+++
** Under X11, it is possible to swap Alt and Meta (and Super and Hyper).
The new variables `x-alt-keysym', `x-hyper-keysym', `x-meta-keysym',
and `x-super-keysym' can be used to choose which keysyms Emacs should
use for the modifiers.  For example, the following two lines swap
Meta and Alt:
    (setq x-alt-keysym 'meta)
    (setq x-meta-keysym 'alt)

+++
** vc-annotate-mode enhancements

In vc-annotate mode, you can now use the following key bindings for
enhanced functionality to browse the annotations of past revisions, or
to view diffs or log entries directly from vc-annotate-mode:

    P:  annotates the previous revision
    N:  annotates the next revision
    J:  annotates the revision at line
    A:  annotates the revision previous to line
    D:  shows the diff of the revision at line with its previous revision
    L:  shows the log of the revision at line
    W:  annotates the workfile (most up to date) version

+++
** In pcl-cvs mode, there is a new `d y' command to view the diffs
between the local version of the file and yesterday's head revision
in the repository.

+++
** In pcl-cvs mode, there is a new `d r' command to view the changes
anyone has committed to the repository since you last executed
"checkout", "update" or "commit".  That means using cvs diff options
-rBASE -rHEAD.

---
** New variable `hs-set-up-overlay' allows customization of the overlay
used to effect hiding for hideshow minor mode.  Integration with isearch
handles the overlay property `display' specially, preserving it during
temporary overlay showing in the course of an isearch operation.

+++
** New command `recode-region' decodes the region again by a specified
coding system.

** On Mac OS, the value of the variable `keyboard-coding-system' is
now dynamically changed according to the current keyboard script.  The
variable `mac-keyboard-text-encoding' and the constants
`kTextEncodingMacRoman', `kTextEncodingISOLatin1', and
`kTextEncodingISOLatin2' are obsolete.

* New modes and packages in Emacs 22.1

+++
** The new package longlines.el provides a minor mode for editing text
files composed of long lines, based on the `use-hard-newlines'
mechanism.  The long lines are broken up by inserting soft newlines,
which are automatically removed when saving the file to disk or
copying into the kill ring, clipboard, etc.  By default, Longlines
mode inserts soft newlines automatically during editing, a behavior
referred to as "soft word wrap" in other text editors.  This is
similar to Refill mode, but more reliable.  To turn the word wrap
feature off, set `longlines-auto-wrap' to nil.

+++
** The new package conf-mode.el handles thousands of configuration files, with
varying syntaxes for comments (;, #, //, /* */ or !), assignment (var = value,
var : value, var value or keyword var value) and sections ([section] or
section { }).  Many files under /etc/, or with suffixes like .cf through
.config, .properties (Java), .desktop (KDE/Gnome), .ini and many others are
recognized.

+++
** The new package dns-mode.el add syntax highlight of DNS master files.
The key binding C-c C-s (`dns-mode-soa-increment-serial') can be used
to increment the SOA serial.

+++
** The new package flymake.el does on-the-fly syntax checking of program
source files.  See the Flymake's Info manual for more details.

** The library tree-widget.el provides a new widget to display a set
of hierarchical data as an outline.  For example, the tree-widget is
well suited to display a hierarchy of directories and files.

+++
** The wdired.el package allows you to use normal editing commands on Dired
buffers to change filenames, permissions, etc...

+++
** The thumbs.el package allows you to preview image files as thumbnails
and can be invoked from a Dired buffer.

+++
** The new python.el package is used to edit Python and Jython programs.

** The URL package (which had been part of W3) is now part of Emacs.

+++
** The new global minor mode `size-indication-mode' (off by default)
shows the size of accessible part of the buffer on the mode line.

---
** GDB-Script-mode is used for files like .gdbinit.

---
** Ido mode is now part of the Emacs distribution.

The ido (interactively do) package is an extension of the iswitchb
package to do interactive opening of files and directories in addition
to interactive buffer switching.  Ido is a superset of iswitchb (with
a few exceptions), so don't enable both packages.

---
** CUA mode is now part of the Emacs distribution.

The new cua package provides CUA-like keybindings using C-x for
cut (kill), C-c for copy, C-v for paste (yank), and C-z for undo.
With cua, the region can be set and extended using shifted movement
keys (like pc-selection-mode) and typed text replaces the active
region (like delete-selection-mode).  Do not enable these modes with
cua-mode.  Customize the variable `cua-mode' to enable cua.

In addition, cua provides unified rectangle support with visible
rectangle highlighting: Use S-return to start a rectangle, extend it
using the movement commands (or mouse-3), and cut or copy it using C-x
or C-c (using C-w and M-w also works).

Use M-o and M-c to `open' or `close' the rectangle, use M-b or M-f, to
fill it with blanks or another character, use M-u or M-l to upcase or
downcase the rectangle, use M-i to increment the numbers in the
rectangle, use M-n to fill the rectangle with a numeric sequence (such
as 10 20 30...), use M-r to replace a regexp in the rectangle, and use
M-' or M-/ to restrict command on the rectangle to a subset of the
rows.  See the commentary in cua-base.el for more rectangle commands.

Cua also provides unified support for registers:  Use a numeric
prefix argument between 0 and 9, i.e. M-0 .. M-9, for C-x, C-c, and
C-v to cut or copy into register 0-9, or paste from register 0-9.

The last text deleted (not killed) is automatically stored in
register 0.  This includes text deleted by typing text.

Finally, cua provides a global mark which is set using S-C-space.
When the global mark is active, any text which is cut or copied is
automatically inserted at the global mark position.  See the
commentary in cua-base.el for more global mark related commands.

The features of cua also works with the standard emacs bindings for
kill, copy, yank, and undo.  If you want to use cua mode, but don't
want the C-x, C-c, C-v, and C-z bindings, you may customize the
`cua-enable-cua-keys' variable.

Note: This version of cua mode is not backwards compatible with older
versions of cua.el and cua-mode.el.  To ensure proper operation, you
must remove older versions of cua.el or cua-mode.el as well as the
loading and customization of those packages from the .emacs file.

+++
** The new keypad setup package provides several common bindings for
the numeric keypad which is available on most keyboards.  The numeric
keypad typically has the digits 0 to 9, a decimal point, keys marked
+, -, /, and *, an Enter key, and a NumLock toggle key.  The keypad
package only controls the use of the digit and decimal keys.

By customizing the variables `keypad-setup', `keypad-shifted-setup',
`keypad-numlock-setup', and `keypad-numlock-shifted-setup', or by
using the function `keypad-setup', you can rebind all digit keys and
the decimal key of the keypad in one step for each of the four
possible combinations of the Shift key state (not pressed/pressed) and
the NumLock toggle state (off/on).

The choices for the keypad keys in each of the above states are:
`Plain numeric keypad' where the keys generates plain digits,
`Numeric keypad with decimal key' where the character produced by the
decimal key can be customized individually (for internationalization),
`Numeric Prefix Arg' where the keypad keys produce numeric prefix args
for emacs editing commands, `Cursor keys' and `Shifted Cursor keys'
where the keys work like (shifted) arrow keys, home/end, etc., and
`Unspecified/User-defined' where the keypad keys (kp-0, kp-1, etc.)
are left unspecified and can be bound individually through the global
or local keymaps.

+++
** The new kmacro package provides a simpler user interface to
emacs' keyboard macro facilities.

Basically, it uses two function keys (default F3 and F4) like this:
F3 starts a macro, F4 ends the macro, and pressing F4 again executes
the last macro.  While defining the macro, F3 inserts a counter value
which automatically increments every time the macro is executed.

There is now a keyboard macro ring which stores the most recently
defined macros.

The C-x C-k sequence is now a prefix for the kmacro keymap which
defines bindings for moving through the keyboard macro ring,
C-x C-k C-p and C-x C-k C-n, editing the last macro C-x C-k C-e,
manipulating the macro counter and format via C-x C-k C-c,
C-x C-k C-a, and C-x C-k C-f.  See the commentary in kmacro.el
for more commands.

The normal macro bindings C-x (, C-x ), and C-x e now interfaces to
the keyboard macro ring.

The C-x e command now automatically terminates the current macro
before calling it, if used while defining a macro.

In addition, when ending or calling a macro with C-x e, the macro can
be repeated immediately by typing just the `e'.  You can customize
this behavior via the variable kmacro-call-repeat-key and
kmacro-call-repeat-with-arg.

Keyboard macros can now be debugged and edited interactively.
C-x C-k SPC steps through the last keyboard macro one key sequence
at a time, prompting for the actions to take.

---
** The old Octave mode bindings C-c f and C-c i have been changed
to C-c C-f and C-c C-i.  The C-c C-i subcommands now have duplicate
bindings on control characters--thus, C-c C-i C-b is the same as
C-c C-i b, and so on.

** The printing package is now part of the Emacs distribution.

If you enable the printing package by including (require 'printing) in
the .emacs file, the normal Print item on the File menu is replaced
with a Print sub-menu which allows you to preview output through
ghostview, use ghostscript to print (if you don't have a PostScript
printer) or send directly to printer a PostScript code generated by
`ps-print' package.  Use M-x pr-help for more information.

+++
** Calc is now part of the Emacs distribution.

Calc is an advanced desk calculator and mathematical tool written in
Emacs Lisp.  Its documentation is in a separate manual; within Emacs,
type "C-h i m calc RET" to read that manual.  A reference card is
available in `etc/calccard.tex' and `etc/calccard.ps'.

+++
** Tramp is now part of the distribution.

This package is similar to Ange-FTP: it allows you to edit remote
files.  But whereas Ange-FTP uses FTP to access the remote host,
Tramp uses a shell connection.  The shell connection is always used
for filename completion and directory listings and suchlike, but for
the actual file transfer, you can choose between the so-called
`inline' methods (which transfer the files through the shell
connection using base64 or uu encoding) and the `out-of-band' methods
(which invoke an external copying program such as `rcp' or `scp' or
`rsync' to do the copying).

Shell connections can be acquired via `rsh', `ssh', `telnet' and also
`su' and `sudo'.  Ange-FTP is still supported via the `ftp' method.

If you want to disable Tramp you should set

  (setq tramp-default-method "ftp")

---
** The new global minor mode `file-name-shadow-mode' modifies the way
filenames being entered by the user in the minibuffer are displayed, so
that it's clear when part of the entered filename will be ignored due to
emacs' filename parsing rules.  The ignored portion can be made dim,
invisible, or otherwise less visually noticable.  The display method may
be displayed by customizing the variable `file-name-shadow-properties'.

---
** The ruler-mode.el library provides a minor mode for displaying an
"active" ruler in the header line.  You can use the mouse to visually
change the `fill-column', `window-margins' and `tab-stop-list'
settings.

---
** The minor mode Reveal mode makes text visible on the fly as you
move your cursor into hidden regions of the buffer.
It should work with any package that uses overlays to hide parts
of a buffer, such as outline-minor-mode, hs-minor-mode, hide-ifdef-mode, ...

There is also Global Reveal mode which affects all buffers.

---
** The new package ibuffer provides a powerful, completely
customizable replacement for buff-menu.el.

+++
** The new package table.el implements editable, WYSIWYG, embedded
`text tables' in Emacs buffers.  It simulates the effect of putting
these tables in a special major mode.  The package emulates WYSIWYG
table editing available in modern word processors.  The package also
can generate a table source in typesetting and markup languages such
as latex and html from the visually laid out text table.

+++
** SES mode (ses-mode) is a new major mode for creating and editing
spreadsheet files.  Besides the usual Emacs features (intuitive command
letters, undo, cell formulas in Lisp, plaintext files, etc.) it also offers
viral immunity and import/export of tab-separated values.

---
** Support for `magic cookie' standout modes has been removed.
Emacs will still work on terminals that require magic cookies in order
to use standout mode, however they will not be able to display
mode-lines in inverse-video.

---
** cplus-md.el has been removed to avoid problems with Custom.

+++
** New package benchmark.el contains simple support for convenient
timing measurements of code (including the garbage collection component).

---
** The new Lisp library fringe.el controls the appearance of fringes.

---
** `cfengine-mode' is a major mode for editing GNU Cfengine
configuration files.

* Incompatible Lisp Changes in Emacs 22.1

+++
** The new interactive-specification `G' reads a file name
much like `F', but if the input is a directory name (even defaulted),

:propertize and :eval forms in the value of a variable whose
`risky-local-variable' property is nil.

* Lisp Changes in Emacs 22.1

** New functions, macros, and commands

+++
*** The new function `filter-buffer-substring' extracts a buffer
substring, passes it through a set of filter functions, and returns
the filtered substring.  It is used instead of `buffer-substring' or
`delete-and-extract-region' when copying text into a user-accessible
data structure, like the kill-ring, X clipboard, or a register.  The
list of filter function is specified by the new variable
`buffer-substring-filters'.  For example, Longlines mode uses
`buffer-substring-filters' to remove soft newlines from the copied
text.

+++
*** (while-no-input BODY...) runs BODY, but only so long as no input
arrives.  If the user types or clicks anything, BODY stops as if a
quit had occurred.  while-no-input returns the value of BODY, if BODY
finishes.  It returns nil if BODY was aborted.

+++
*** New function `looking-back' checks whether a regular expression matches
the text before point.  Specifying the LIMIT argument bounds how far
back the match can start; this is a way to keep it from taking too long.

+++
*** New functions `make-progress-reporter', `progress-reporter-update',
`progress-reporter-force-update', `progress-reporter-done', and
`dotimes-with-progress-reporter' provide a simple and efficient way for
a command to present progress messages for the user.

+++
*** The new primitive `get-internal-run-time' returns the processor
run time used by Emacs since start-up.

+++
*** The new function `called-interactively-p' does what many people
have mistakenly believed `interactive-p' did: it returns t if the
calling function was called through `call-interactively'.  This should
only be used when you cannot add a new "interactive" argument to the
command.

+++
*** The new function `assoc-string' replaces `assoc-ignore-case' and
`assoc-ignore-representation', which are still available, but have
been declared obsolete.

---
*** New function quail-find-key returns a list of keys to type in the
current input method to input a character.

+++
*** New functions posn-at-point and posn-at-x-y return
click-event-style position information for a given visible buffer
position or for a given window pixel coordinate.

+++
*** The new primitive `set-file-times' sets a file's access and
modification times.  Magic file name handlers can handle this
operation.

+++
*** The new function syntax-after returns the syntax code
of the character after a specified buffer position, taking account
of text properties as well as the character code.

*** `syntax-class' extracts the class of a syntax code (as returned
by syntax-after).

+++
*** New function `line-number-at-pos' returns line number of current
line in current buffer, or if optional buffer position is given, line
number of corresponding line in current buffer.

+++
*** New function `macroexpand-all' expands all macros in a form.
It is similar to the Common-Lisp function of the same name.
One difference is that it guarantees to return the original argument
if no expansion is done, which may be tested using `eq'.

+++
*** New macro with-local-quit temporarily sets inhibit-quit to nil for use
around potentially blocking or long-running code in timers
and post-command-hooks.

---
** easy-mmode-define-global-mode has been renamed to
define-global-minor-mode.  The old name remains as an alias.

+++
** An element of buffer-undo-list can now have the form (apply FUNNAME
. ARGS), where FUNNAME is a symbol other than t or nil.  That stands
for a high-level change that should be undone by evaluating (apply
FUNNAME ARGS).

These entries can also have the form (apply DELTA BEG END FUNNAME . ARGS)
which indicates that the change which took place was limited to the
range BEG...END and increased the buffer size by DELTA.

+++
** The line-move, scroll-up, and scroll-down functions will now
modify the window vscroll to scroll through display rows that are
taller that the height of the window, for example in the presense of
large images.  To disable this feature, Lisp code may bind the new
variable `auto-window-vscroll' to nil.

+++
** If a buffer sets buffer-save-without-query to non-nil,
save-some-buffers will always save that buffer without asking
(if it's modified).

+++
** The function symbol-file tells you which file defined
a certain function or variable.

+++
** Lisp code can now test if a given buffer position is inside a
clickable link with the new function `mouse-on-link-p'.  This is the
function used by the new `mouse-1-click-follows-link' functionality.

+++
** `set-auto-mode' now gives the interpreter magic line (if present)
precedence over the file name.  Likewise an <?xml or <!DOCTYPE declaration
will give the buffer XML or SGML mode, based on the new var
`magic-mode-alist'.

---
** To manipulate the File menu using easy-menu, you must specify the
proper name "file".  In previous Emacs versions, you had to specify
"files", even though the menu item itself was changed to say "File"
several versions ago.

+++
** read-from-minibuffer now accepts an additional argument KEEP-ALL
saying to put all inputs in the history list, even empty ones.

+++
** The new variable search-spaces-regexp controls how to search
for spaces in a regular expression.  If it is non-nil, it should be a
regular expression, and any series of spaces stands for that regular
expression.  If it is nil, spaces stand for themselves.

Spaces inside of constructs such as [..] and *, +, ? are never
replaced with search-spaces-regexp.

---
** list-buffers-noselect now takes an additional argument, BUFFER-LIST.
If it is non-nil, it specifies which buffers to list.

---
** set-buffer-file-coding-system now takes an additional argument,
NOMODIFY.  If it is non-nil, it means don't mark the buffer modified.

+++
** An interactive specification may now use the code letter 'U' to get
the up-event that was discarded in case the last key sequence read for a
previous 'k' or 'K' argument was a down-event; otherwise nil is used.

+++
** Function `translate-region' accepts also a char-table as TABLE
argument.

+++
** Major mode functions now run the new normal hook
`after-change-major-mode-hook', at their very end, after the mode hooks.

+++
** `auto-save-file-format' has been renamed to
`buffer-auto-save-file-format' and made into a permanent local.

+++
** Both the variable and the function `disabled-command-hook' have
been renamed to `disabled-command-function'.  The variable
`disabled-command-hook' has been kept as an obsolete alias.

+++
** Function `compute-motion' now calculates the usable window
width if the WIDTH argument is nil.  If the TOPOS argument is nil,
the usable window height and width is used.

+++
** `visited-file-modtime' and `calendar-time-from-absolute' now return
a list of two integers, instead of a cons.

+++
** If a command sets transient-mark-mode to `only', that
enables Transient Mark mode for the following command only.
During that following command, the value of transient-mark-mode
is `identity'.  If it is still `identity' at the end of the command,
it changes to nil.

+++
** Cleaner way to enter key sequences.

You can enter a constant key sequence in a more natural format, the
same one used for saving keyboard macros, using the macro `kbd'.  For
example,

(kbd "C-x C-f") => "\^x\^f"

+++
** The sentinel is now called when a network process is deleted with
delete-process.  The status message passed to the sentinel for a
deleted network process is "deleted".  The message passed to the
sentinel when the connection is closed by the remote peer has been
changed to "connection broken by remote peer".

+++
** If the buffer's undo list for the current command gets longer than
undo-outer-limit, garbage collection empties it.  This is to prevent
it from using up the available memory and choking Emacs.

+++
** skip-chars-forward and skip-chars-backward now handle
character classes such as [:alpha:], along with individual characters
and ranges.

+++
** Function pos-visible-in-window-p now returns the pixel coordinates
and partial visiblity state of the corresponding row, if the PARTIALLY
arg is non-nil.

+++
** The function `eql' is now available without requiring the CL package.

+++
** The display space :width and :align-to text properties are now
supported on text terminals.

+++
** Support for displaying image slices

*** New display property (slice X Y WIDTH HEIGHT) may be used with
an image property to display only a specific slice of the image.

*** Function insert-image has new optional fourth arg to
specify image slice (X Y WIDTH HEIGHT).

*** New function insert-sliced-image inserts a given image as a
specified number of evenly sized slices (rows x columns).

+++
** New line-height and line-spacing properties for newline characters

A newline may now have line-height and line-spacing text or overlay
properties that control the height of the corresponding display row.

If the line-height property value is t, the newline does not
contribute to the height of the display row; instead the height of the
newline glyph is reduced.  Also, a line-spacing property on this
newline is ignored.  This can be used to tile small images or image
slices without adding blank areas between the images.

If the line-height property value is a positive integer, the value
specifies the minimum line height in pixels.  If necessary, the line
height it increased by increasing the line's ascent.

If the line-height property value is a float, the minimum line height
is calculated by multiplying the default frame line height by the
given value.

If the line-height property value is a cons (FACE . RATIO), the
minimum line height is calculated as RATIO * height of named FACE.
RATIO is int or float.  If FACE is t, it specifies the current face.

If the line-height property value is a cons (nil . RATIO), the line
height is calculated as RATIO * actual height of the line's contents.

If the line-height value is a cons (HEIGHT . TOTAL), HEIGHT specifies
the line height as described above, while TOTAL is any of the forms
described above and specifies the total height of the line, causing a
varying number of pixels to be inserted after the line to make it line
exactly that many pixels high.

If the line-spacing property value is an positive integer, the value
is used as additional pixels to insert after the display line; this
overrides the default frame line-spacing and any buffer local value of
the line-spacing variable.

If the line-spacing property may be a float or cons, the line spacing
is calculated as specified above for the line-height property.

+++
** The buffer local line-spacing variable may now have a float value,
which is used as a height relative to the default frame line height.

+++
** Enhancements to stretch display properties

The display property stretch specification form `(space PROPS)', where
PROPS is a property list now allows pixel based width and height
specifications, as well as enhanced horizontal text alignment.

The value of these properties can now be a (primitive) expression
which is evaluated during redisplay.  The following expressions
are supported:

EXPR ::= NUM | (NUM) | UNIT | ELEM | POS | IMAGE | FORM
NUM  ::= INTEGER | FLOAT | SYMBOL
UNIT ::= in | mm | cm | width | height
ELEM ::= left-fringe | right-fringe | left-margin | right-margin
      |  scroll-bar | text
POS  ::= left | center | right
FORM ::= (NUM . EXPR) | (OP EXPR ...)
OP   ::= + | -

The form `NUM' specifies a fractional width or height of the default
frame font size.  The form `(NUM)' specifies an absolute number of
pixels.  If a symbol is specified, its buffer-local variable binding
is used.  The `in', `mm', and `cm' units specifies the number of
pixels per inch, milli-meter, and centi-meter, resp.  The `width' and
`height' units correspond to the width and height of the current face
font.  An image specification corresponds to the width or height of
the image.

The `left-fringe', `right-fringe', `left-margin', `right-margin',
`scroll-bar', and `text' elements specify to the width of the
corresponding area of the window.

The `left', `center', and `right' positions can be used with :align-to
to specify a position relative to the left edge, center, or right edge
of the text area.  One of the above window elements (except `text')
can also be used with :align-to to specify that the position is
relative to the left edge of the given area.  Once the base offset for
a relative position has been set (by the first occurrence of one of
these symbols), further occurences of these symbols are interpreted as
the width of the area.

For example, to align to the center of the left-margin, use
    :align-to (+ left-margin (0.5 . left-margin))

If no specific base offset is set for alignment, it is always relative
to the left edge of the text area.  For example, :align-to 0 in a
header-line aligns with the first text column in the text area.

The value of the form `(NUM . EXPR)' is the value of NUM multiplied by
the value of the expression EXPR.  For example, (2 . in) specifies a
width of 2 inches, while (0.5 . IMAGE) specifies half the width (or
height) of the specified image.

The form `(+ EXPR ...)' adds up the value of the expressions.
The form `(- EXPR ...)' negates or subtracts the value of the expressions.

+++
** Normally, the cursor is displayed at the end of any overlay and
text property string that may be present at the current window
position.  The cursor may now be placed on any character of such
strings by giving that character a non-nil `cursor' text property.

** The first face specification element in a defface can specify
`default' instead of frame classification.  Then its attributes act as
defaults that apply to all the subsequent cases (and may be overridden
by them).

+++
** New face attribute `min-colors' can be used to tailor the face color
to the number of colors supported by a display, and define the
foreground and background colors accordingly so that they look best on
a terminal that supports at least this many colors.  This is now the
preferred method for defining default faces in a way that makes a good
use of the capabilities of the display.

+++
** Customizable fringe bitmaps

*** New function 'define-fringe-bitmap' can now be used to create new
fringe bitmaps, as well as change the built-in fringe bitmaps.

To change a built-in bitmap, do (require 'fringe) and use the symbol
identifing the bitmap such as `left-truncation or `continued-line'.

*** New function 'destroy-fringe-bitmap' may be used to destroy a
previously created bitmap, or restore a built-in bitmap.

*** New function 'set-fringe-bitmap-face' can now be used to set a
specific face to be used for a specific fringe bitmap.  The face is
automatically merged with the `fringe' face, so normally, the face
should only specify the foreground color of the bitmap.

*** There are new display properties, left-fringe and right-fringe,
that can be used to show a specific bitmap in the left or right fringe
bitmap of the display line.

Format is 'display '(left-fringe BITMAP [FACE]), where BITMAP is a
symbol identifying a fringe bitmap, either built-in or defined with
`define-fringe-bitmap', and FACE is an optional face name to be used
for displaying the bitmap instead of the default `fringe' face.
When specified, FACE is automatically merged with the `fringe' face.

*** New function `fringe-bitmaps-at-pos' returns the current fringe
bitmaps in the display line at a given buffer position.

+++
** Multiple overlay arrows can now be defined and managed via the new
variable `overlay-arrow-variable-list'.  It contains a list of
varibles which contain overlay arrow position markers, including
the original `overlay-arrow-position' variable.

Each variable on this list may have individual `overlay-arrow-string'
and `overlay-arrow-bitmap' properties that specify an overlay arrow
string (for non-window terminals) or fringe bitmap (for window
systems) to display at the corresponding overlay arrow position.
If either property is not set, the default `overlay-arrow-string' or
'overlay-arrow-fringe-bitmap' will be used.

+++
** The default value of `sentence-end' is now defined using the new
variable `sentence-end-without-space' which contains such characters
that end a sentence without following spaces.

+++
** The function `sentence-end' should be used to obtain the value of
the variable `sentence-end'.  If the variable `sentence-end' is nil,
then this function returns the regexp constructed from the variables
`sentence-end-without-period', `sentence-end-double-space' and
`sentence-end-without-space'.

+++
** The flags, width, and precision options for %-specifications in function
`format' are now documented.  Some flags that were accepted but not
implemented (such as "*") are no longer accepted.

+++
** New function `delete-dups' destructively removes `equal' duplicates
from a list.  Of several `equal' occurrences of an element in the list,
the first one is kept.

+++
** `declare' is now a macro.  This change was made mostly for
documentation purposes and should have no real effect on Lisp code.

+++
** The new hook `before-save-hook' is invoked by `basic-save-buffer'
before saving buffers.  This allows packages to perform various final
tasks, for example; it can be used by the copyright package to make
sure saved files have the current year in any copyright headers.

+++
** The function `insert-for-yank' now supports strings where the
`yank-handler' property does not span the first character of the
string.  The old behavior is available if you call
`insert-for-yank-1' instead.

+++
** New function `get-char-property-and-overlay' accepts the same
arguments as `get-char-property' and returns a cons whose car is the
return value of `get-char-property' called with those arguments and
whose cdr is the overlay in which the property was found, or nil if
it was found as a text property or not found at all.

+++ (lispref)
??? (man)
** The mouse pointer shape in void text areas (i.e. after the end of a
line or below the last line in the buffer) of the text window is now
controlled by the new variable `void-text-area-pointer'.  The default
is to use the `arrow' (non-text) pointer.  Other choices are `text'
(or nil), `hand', `vdrag', `hdrag', `modeline', and `hourglass'.

+++
** The mouse pointer shape over an image can now be controlled by the
:pointer image property.

+++
** The mouse pointer shape over ordinary text or images may now be
controlled/overriden via the `pointer' text property.

+++
** Images may now have an associated image map via the :map property.

An image map is an alist where each element has the format (AREA ID PLIST).
An AREA is specified as either a rectangle, a circle, or a polygon:
A rectangle is a cons (rect . ((x0 . y0) . (x1 . y1))) specifying the
pixel coordinates of the upper left and bottom right corners.
A circle is a cons (circle . ((x0 . y0) . r)) specifying the center
and the radius of the circle; r may be a float or integer.
A polygon is a cons (poly . [x0 y0 x1 y1 ...]) where each pair in the
vector describes one corner in the polygon.

When the mouse pointer is above a hot-spot area of an image, the
PLIST of that hot-spot is consulted; if it contains a `help-echo'
property it defines a tool-tip for the hot-spot, and if it contains
a `pointer' property, it defines the shape of the mouse cursor when
it is over the hot-spot.  See the variable 'void-area-text-pointer'
for possible pointer shapes.

When you click the mouse when the mouse pointer is over a hot-spot,
an event is composed by combining the ID of the hot-spot with the
mouse event, e.g. [area4 mouse-1] if the hot-spot's ID is `area4'.

** Mouse event enhancements:

+++
*** Mouse clicks on fringes now generates left-fringe or right-fringes
events, rather than a text area click event.

+++
*** Mouse clicks in the left and right marginal areas now includes a
sensible buffer position corresponding to the first character in the
corresponding text row.

+++
*** Function `mouse-set-point' now works for events outside text area.

+++
*** Mouse events now includes buffer position for all event types.

+++
*** `posn-point' now returns buffer position for non-text area events.

+++
*** New function `posn-area' returns window area clicked on (nil means
text area).

+++
*** Mouse events include actual glyph column and row for all event types.

+++
*** New function `posn-actual-col-row' returns actual glyph coordinates.

+++
*** Mouse events may now include image object in addition to string object.

+++
*** Mouse events include relative x and y pixel coordinates relative to
the top left corner of the object (image or character) clicked on.

+++
*** Mouse events include the pixel width and height of the object
(image or character) clicked on.

+++
*** New functions 'posn-object', 'posn-object-x-y', and
'posn-object-width-height' return the image or string object of a mouse
click, the x and y pixel coordinates relative to the top left corner
of that object, and the total width and height of that object.

+++
** New function `force-window-update' can initiate a full redisplay of
one or all windows.  Normally, this is not needed as changes in window
contents are detected automatically.  However, certain implicit
changes to mode lines, header lines, or display properties may require
forcing an explicit window update.

---
** New function `redirect-debugging-output' can be used to redirect
debugging output on the stderr file handle to a file.

+++
** `split-string' now includes null substrings in the returned list if
the optional argument SEPARATORS is non-nil and there are matches for
SEPARATORS at the beginning or end of the string.  If SEPARATORS is
nil, or if the new optional third argument OMIT-NULLS is non-nil, all
empty matches are omitted from the returned list.

+++
** `makehash' is now obsolete.  Use `make-hash-table' instead.

+++
** If optional third argument APPEND to `add-to-list' is non-nil, a
new element gets added at the end of the list instead of at the
beginning.  This change actually occurred in Emacs-21.1, but was not
documented.

** Major modes can define `eldoc-print-current-symbol-info-function'
locally to provide Eldoc functionality by some method appropriate to
the language.

---
** New coding system property `mime-text-unsuitable' indicates that
the coding system's `mime-charset' is not suitable for MIME text
parts, e.g. utf-16.

+++
** The argument to forward-word, backward-word, forward-to-indentation
and backward-to-indentation is now optional, and defaults to 1.

+++
** (char-displayable-p CHAR) returns non-nil if Emacs ought to be able
to display CHAR.  More precisely, if the selected frame's fontset has
a font to display the character set that CHAR belongs to.

Fontsets can specify a font on a per-character basis; when the fontset
does that, this value may not be accurate.

+++
** The new function `window-inside-edges' returns the edges of the
actual text portion of the window, not including the scroll bar or
divider line, the fringes, the display margins, the header line and
the mode line.

+++
** The new functions `window-pixel-edges' and `window-inside-pixel-edges'
return window edges in units of pixels, rather than columns and lines.

+++
** The kill-buffer-hook is now permanent-local.

+++
** `select-window' takes an optional second argument `norecord', like
`switch-to-buffer'.

+++
** The new macro `with-selected-window' temporarily switches the
selected window without impacting the order of buffer-list.

+++
** The `keymap' property now also works at the ends of overlays and
text-properties, according to their stickiness.  This also means that it
works with empty overlays.  The same hold for the `local-map' property.

+++
** (map-keymap FUNCTION KEYMAP) applies the function to each binding
in the keymap.

---
** VC changes for backends:
*** (vc-switches BACKEND OPERATION) is a new function for use by backends.
*** The new `find-version' backend function replaces the `destfile'
parameter of the `checkout' backend function.
Old code still works thanks to a default `find-version' behavior that
uses the old `destfile' parameter.

+++
** The new macro dynamic-completion-table supports using functions
as a dynamic completion table.

  (dynamic-completion-table FUN)

FUN is called with one argument, the string for which completion is required,
and it should return an alist containing all the intended possible
completions. This alist may be a full list of possible completions so that FUN
can ignore the value of its argument. If completion is performed in the
minibuffer, FUN will be called in the buffer from which the minibuffer was
entered. dynamic-completion-table then computes the completion.

+++
** The new macro lazy-completion-table initializes a variable
as a lazy completion table.

  (lazy-completion-table VAR FUN &rest ARGS)

If the completion table VAR is used for the first time (e.g., by passing VAR
as an argument to `try-completion'), the function FUN is called with arguments
ARGS. FUN must return the completion table that will be stored in VAR. If
completion is requested in the minibuffer, FUN will be called in the buffer
from which the minibuffer was entered. The return value of
`lazy-completion-table' must be used to initialize the value of VAR.

+++
** `minor-mode-list' now holds a list of minor mode commands.

+++
** The new function `modify-all-frames-parameters' modifies parameters
for all (existing and future) frames.

+++
** `sit-for' can now be called with args (SECONDS &optional NODISP).

+++
** New standard font-lock face `font-lock-preprocessor-face'.

+++
** The macro `with-syntax-table' does not copy the table any more.

+++
** The variable `face-font-rescale-alist' specifies how much larger
(or smaller) font we should use.  For instance, if the value is
'((SOME-FONTNAME-PATTERN . 1.3)) and a face requests a font of 10
point, we actually use a font of 13 point if the font matches
SOME-FONTNAME-PATTERN.

+++
** The function `number-sequence' returns a list of equally-separated
numbers.  For instance, (number-sequence 4 9) returns (4 5 6 7 8 9).
By default, the separation is 1, but you can specify a different separation
as the third argument.  (number-sequence 1.5 6 2) returns (1.5 3.5 5.5).

+++
** `file-chase-links' now takes an optional second argument LIMIT which
specifies the maximum number of links to chase through.  If after that
many iterations the file name obtained is still a symbolic link,
`file-chase-links' returns it anyway.

---
** `set-fontset-font', `fontset-info', `fontset-font' now operate on
the default fontset if the argument NAME is nil..

+++
** The escape sequence \s is now interpreted as a SPACE character,
unless it is followed by a `-' in a character constant (e.g. ?\s-A),
in which case it is still interpreted as the super modifier.
In strings, \s is always interpreted as a space.

+++
** New function `set-process-filter-multibyte' sets the multibyteness
of a string given to a process's filter.

+++
** New function `process-filter-multibyte-p' returns t if
a string given to a process's filter is multibyte.

+++
** A filter function of a process is called with a multibyte string if
the filter's multibyteness is t.  That multibyteness is decided by the
value of `default-enable-multibyte-characters' when the process is
created and can be changed later by `set-process-filter-multibyte'.

+++
** If a process's coding system is raw-text or no-conversion and its
buffer is multibyte, the output of the process is at first converted
to multibyte by `string-to-multibyte' then inserted in the buffer.
Previously, it was converted to multibyte by `string-as-multibyte',
which was not compatible with the behavior of file reading.

+++
** New function `string-to-multibyte' converts a unibyte string to a
multibyte string with the same individual character codes.

+++
** New variables `gc-elapsed' and `gcs-done' provide extra information
on garbage collection.

+++
** New function `decode-coding-inserted-region' decodes a region as if
it is read from a file without decoding.

+++
** New function `locale-info' accesses locale information.

+++
** `save-selected-window' now saves and restores the selected window
of every frame.  This way, it restores everything that can be changed
by calling `select-window'.

---
** `easy-menu-define' now allows you to use nil for the symbol name
if you don't need to give the menu a name.  If you install the menu
into other keymaps right away (MAPS is non-nil), it usually doesn't
need to have a name.

** Byte compiler changes:

---
*** `(featurep 'xemacs)' is treated by the compiler as nil.  This
helps to avoid noisy compiler warnings in code meant to run under both
Emacs and XEmacs and may sometimes make the result significantly more
efficient.  Since byte code from recent versions of XEmacs won't
generally run in Emacs and vice versa, this optimization doesn't lose
you anything.

+++
*** You can avoid warnings for possibly-undefined symbols with a
simple convention that the compiler understands.  (This is mostly
useful in code meant to be portable to different Emacs versions.)
Write forms like the following, or code that macroexpands into such
forms:

  (if (fboundp 'foo) <then> <else>)
  (if (boundp 'foo) <then> <else)

In the first case, using `foo' as a function inside the <then> form
won't produce a warning if it's not defined as a function, and in the
second case, using `foo' as a variable won't produce a warning if it's
unbound.  The test must be in exactly one of the above forms (after
macro expansion), but such tests may be nested.  Note that `when' and
`unless' expand to `if', but `cond' doesn't.

+++
*** The new macro `with-no-warnings' suppresses all compiler warnings
inside its body.  In terms of execution, it is equivalent to `progn'.

+++
** The new translation table `translation-table-for-input'
is used for customizing self-insertion.  The character to
be inserted is translated through it.

+++
** `load-history' can now have elements of the form (t . FUNNAME),
which means FUNNAME was previously defined as an autoload (before the
current file redefined it).

+++
** `load-history' now records (defun . FUNNAME) when a function is
defined.  For a variable, it records just the variable name.

+++
** New Lisp library testcover.el works with edebug to help you determine
whether you've tested all your Lisp code.  Function testcover-start
instruments all functions in a given file.  Then test your code.  Function
testcover-mark-all adds overlay "splotches" to the Lisp file's buffer to
show where coverage is lacking.  Command testcover-next-mark (bind it to
a key!) will move point forward to the next spot that has a splotch.

*** Normally, a red splotch indicates the form was never completely evaluated;
a brown splotch means it always evaluated to the same value.  The red
splotches are skipped for forms that can't possibly complete their evaluation,
such as `error'.  The brown splotches are skipped for forms that are expected
to always evaluate to the same value, such as (setq x 14).

*** For difficult cases, you can add do-nothing macros to your code to help
out the test coverage tool.  The macro `noreturn' suppresses a red splotch.
It is an error if the argument to `noreturn' does return.  The macro 1value
suppresses a brown splotch for its argument.  This macro is a no-op except
during test-coverage -- then it signals an error if the argument actually
returns differing values.

+++
** New function unsafep returns nil if the given Lisp form can't possibly
do anything dangerous; otherwise it returns a reason why the form might be
unsafe (calls dangerous function, alters global variable, etc).

+++
** The new variable `print-continuous-numbering', when non-nil, says
that successive calls to print functions should use the same
numberings for circular structure references.  This is only relevant
when `print-circle' is non-nil.

When you bind `print-continuous-numbering' to t, you should
also bind `print-number-table' to nil.

+++
** When using non-toolkit scroll bars with the default width,
the scroll-bar-width frame parameter value is nil.

+++
** The new function copy-abbrev-table returns a new abbrev table that
is a copy of a given abbrev table.

+++
** The option --script FILE runs Emacs in batch mode and loads FILE.
It is useful for writing Emacs Lisp shell script files, because they
can start with this line:

   #!/usr/bin/emacs --script

+++
** The option --directory DIR now modifies `load-path' immediately.
Directories are added to the front of `load-path' in the order they
appear on the command line.  For example, with this command line:

  emacs -batch -L .. -L /tmp --eval "(require 'foo)"

Emacs looks for library `foo' in the parent directory, then in /tmp, then
in the other directories in `load-path'.  (-L is short for --directory.)

+++
** A function's docstring can now hold the function's usage info on
its last line.  It should match the regexp "\n\n(fn.*)\\'".

---
** New CCL functions `lookup-character' and `lookup-integer' access
hash tables defined by the Lisp function `define-translation-hash-table'.

+++
** The new function `minibufferp' returns non-nil if its optional buffer
argument is a minibuffer.  If the argument is omitted it defaults to
the current buffer.

+++
** There is a new Warnings facility; see the functions `warn'
and `display-warning'.

+++
** The functions all-completions and try-completion now accept lists
of strings as well as hash-tables additionally to alists, obarrays
and functions.  Furthermore, the function `test-completion' is now
exported to Lisp.  The keys in alists and hash tables may be either
strings or symbols, which are automatically converted with to strings.

---
** When pure storage overflows while dumping, Emacs now prints how
much pure storage it will approximately need.

+++
** The new variable `auto-coding-functions' lets you specify functions
to examine a file being visited and deduce the proper coding system
for it.  (If the coding system is detected incorrectly for a specific
file, you can put a `coding:' tags to override it.)

---
** The new function `merge-coding-systems' fills in unspecified aspects
of one coding system from another coding system.

+++
** The variable `safe-local-eval-forms' specifies a list of forms that
are ok to evaluate when they appear in an `eval' local variables
specification.  Normally Emacs asks for confirmation before evaluating
such a form, but if the form appears in this list, no confirmation is
needed.

---
** If a function has a non-nil `safe-local-eval-function' property,
that means it is ok to evaluate some calls to that function when it
appears in an `eval' local variables specification.  If the property
is t, then any form calling that function with constant arguments is
ok.  If the property is a function or list of functions, they are called
with the form as argument, and if any returns t, the form is ok to call.

If the form is not "ok to call", that means Emacs asks for
confirmation as before.

+++
** Controlling the default left and right fringe widths.

The default left and right fringe widths for all windows of a frame
can now be controlled by setting the `left-fringe' and `right-fringe'
frame parameters to an integer value specifying the width in pixels.
Setting the width to 0 effectively removes the corresponding fringe.

The actual default fringe widths for the frame may deviate from the
specified widths, since the combined fringe widths must match an
integral number of columns.  The extra width is distributed evenly
between the left and right fringe.  For force a specific fringe width,
specify the width as a negative integer (if both widths are negative,
only the left fringe gets the specified width).

Setting the width to nil (the default), restores the default fringe
width which is the minimum number of pixels necessary to display any
of the currently defined fringe bitmaps.  The width of the built-in
fringe bitmaps is 8 pixels.

+++
** Per-window fringes settings

Windows can now have their own individual fringe widths and position
settings.

To control the fringe widths of a window, either set the buffer-local
variables `left-fringe-width', `right-fringe-width', or call
`set-window-fringes'.

To control the fringe position in a window, that is, whether fringes
are positioned between the display margins and the window's text area,
or at the edges of the window, either set the buffer-local variable
`fringes-outside-margins' or call `set-window-fringes'.

The function `window-fringes' can be used to obtain the current
settings.  To make `left-fringe-width', `right-fringe-width', and
`fringes-outside-margins' take effect, you must set them before
displaying the buffer in a window, or use `set-window-buffer' to force
an update of the display margins.

+++
** Per-window vertical scroll-bar settings

Windows can now have their own individual scroll-bar settings
controlling the width and position of scroll-bars.

To control the scroll-bar of a window, either set the buffer-local
variables `scroll-bar-mode' and `scroll-bar-width', or call
`set-window-scroll-bars'.  The function `window-scroll-bars' can be
used to obtain the current settings.  To make `scroll-bar-mode' and
`scroll-bar-width' take effect, you must set them before displaying
the buffer in a window, or use `set-window-buffer' to force an update
of the display margins.

+++
** The function `set-window-buffer' now has an optional third argument
KEEP-MARGINS which will preserve the window's current margin, fringe,
and scroll-bar settings if non-nil.

+++
** Renamed hooks to better follow the naming convention:
find-file-hooks to find-file-hook,
find-file-not-found-hooks to find-file-not-found-functions,
write-file-hooks to write-file-functions,
write-contents-hooks to write-contents-functions,
x-lost-selection-hooks to x-lost-selection-functions,
x-sent-selection-hooks to x-sent-selection-functions.
Marked local-write-file-hooks as obsolete (use the LOCAL arg of `add-hook').

+++
** The new variable `delete-frame-functions' replaces `delete-frame-hook'.
It was renamed to follow the naming conventions for abnormal hooks.  The old
name remains available as an alias, but has been marked obsolete.

+++
** The `read-file-name' function now takes an additional argument which
specifies a predicate which the file name read must satify.  The
new variable `read-file-name-predicate' contains the predicate argument

** The new function `read-directory-name' can be used instead of
`read-file-name' to read a directory name; when used, completion
will only show directories.

+++
** The new function `file-remote-p' tests a file name and returns
non-nil if it specifies a remote file (one that Emacs accesses using
its own special methods and not directly through the file system).
The value in that case is an identifier for the remote file system.

---
** When a Lisp file uses CL functions at run-time, compiling the file
now issues warnings about these calls, unless the file performs
(require 'cl) when loaded.

+++
** The `defmacro' form may contain declarations specifying how to
indent the macro in Lisp mode and how to debug it with Edebug.  The
syntax of defmacro has been extended to

   (defmacro NAME LAMBDA-LIST [DOC-STRING] [DECLARATION ...] ...)

DECLARATION is a list `(declare DECLARATION-SPECIFIER ...)'.  The
declaration specifiers supported are:

(indent INDENT)
	Set NAME's `lisp-indent-function' property to INDENT.

(edebug DEBUG)
	Set NAME's `edebug-form-spec' property to DEBUG.  (This is
	equivalent to writing a `def-edebug-spec' for the macro.

+++
** Interactive commands can be remapped through keymaps.

This is an alternative to using defadvice or substitute-key-definition
to modify the behavior of a key binding using the normal keymap
binding and lookup functionality.

- The new variable `this-original-command' contains the original
  command before remapping.  It is equal to `this-command' when the
  command was not remapped.

+++
** New variable emulation-mode-map-alists.

Lisp packages using many minor mode keymaps can now maintain their own
keymap alist separate from minor-mode-map-alist by adding their keymap
alist to this list.

will lead to undesirable results, so don't let it happen; the first
change group you start for any given buffer should be the last one
finished.

+++
** New variable char-property-alias-alist.

This variable allows you to create alternative names for text
properties.  It works at the same level as `default-text-properties',
although it applies to overlays as well.  This variable was introduced
to implement the `font-lock-face' property.

+++
** New special text property `font-lock-face'.

This property acts like the `face' property, but it is controlled by
M-x font-lock-mode.  It is not, strictly speaking, a builtin text
property.  Instead, it is implemented inside font-core.el, using the
new variable `char-property-alias-alist'.

+++
** New function remove-list-of-text-properties.

The new function `remove-list-of-text-properties' is almost the same
as `remove-text-properties'.  The only difference is that it takes
a list of property names as argument rather than a property list.

+++
** New function insert-for-yank.

This function normally works like `insert' but removes the text
properties in the `yank-excluded-properties' list.  However, if the
inserted text has a `yank-handler' text property on the first
character of the string, the insertion of the text may be modified in
a number of ways.  See the description of `yank-handler' below.

+++
** New function insert-buffer-substring-as-yank.

This function works like `insert-buffer-substring', but removes the
text properties in the `yank-excluded-properties' list.

+++
** New function insert-buffer-substring-no-properties.

This function is like insert-buffer-substring, but removes all
text properties from the inserted substring.

+++
** New `yank-handler' text property may be used to control how
previously killed text on the kill-ring is reinserted.

  If UNDO is present and non-nil, it is a function that will be called
by `yank-pop' to undo the insertion of the current object.  It is
called with two arguments, the start and end of the current region.
FUNCTION may set `yank-undo-function' to override the UNDO value.

+++
*** The functions kill-new, kill-append, and kill-region now have an
optional argument to specify the yank-handler text property to put on
the killed text.

+++
*** The function yank-pop will now use a non-nil value of the variable
`yank-undo-function' (instead of delete-region) to undo the previous
yank or yank-pop command (or a call to insert-for-yank).  The function
insert-for-yank automatically sets that variable according to the UNDO
element of the string argument's yank-handler text property if present.

+++
** New function display-supports-face-attributes-p may be used to test
whether a given set of face attributes is actually displayable.

A new predicate `supports' has also been added to the `defface' face
specification language, which can be used to do this test for faces
defined with defface.

---
** The function face-differs-from-default-p now truly checks whether the
given face displays differently from the default face or not (previously
it did only a very cursory check).

+++
** face-attribute, face-foreground, face-background, and face-stipple now
accept a new optional argument, INHERIT, which controls how face
inheritance is used when determining the value of a face attribute.

+++
** New functions face-attribute-relative-p and merge-face-attribute
help with handling relative face attributes.

+++
** The priority of faces in an :inherit attribute face-list is reversed.
If a face contains an :inherit attribute with a list of faces, earlier
faces in the list override later faces in the list; in previous releases
of Emacs, the order was the opposite.  This change was made so that
:inherit face-lists operate identically to face-lists in text `face'
properties.

+++
** Enhancements to process support

*** Function list-processes now has an optional argument; if non-nil,

is handled, suspending output from other processes.  If value is an
integer, also inhibit running timers.  This feature is generally not
recommended, but may be necessary for specific applications, such as
speech synthesis.

---
*** Adaptive read buffering of subprocess output.

On some systems, when emacs reads the output from a subprocess, the
output data is read in very small blocks, potentially resulting in
very poor performance.  This behavior can be remedied to some extent
by setting the new variable process-adaptive-read-buffering to a
non-nil value (the default), as it will automatically delay reading
from such processes, to allowing them to produce more output before
emacs tries to read it.

+++
** Enhanced networking support.

*** There is a new `make-network-process' function which supports

*** New function network-interface-info.

This function returns the network address, hardware address, current
status, and other information about a specific network interface.

+++
** New function copy-tree.

+++
** New function substring-no-properties.

+++
** New function minibuffer-selected-window.

+++
** New function `call-process-shell-command'.

+++
** New function `process-file'.

This is similar to `call-process', but obeys file handlers.  The file
handler is chosen based on default-directory.

---
** The dummy function keys made by easymenu
are now always lower case.  If you specify the
menu item name "Ada", for instance, it uses `ada'
as the "key" bound by that key binding.

This is relevant only if Lisp code looks for
the bindings that were made with easymenu.

+++
** The function `commandp' takes an additional optional
argument.  If it is non-nil, then `commandp' checks
for a function that could be called with `call-interactively',
and does not return t for keyboard macros.

---
** master-mode.el implements a minor mode for scrolling a slave
buffer without leaving your current buffer, the master buffer.

It can be used by sql.el, for example: the SQL buffer is the master
and its SQLi buffer is the slave.  This allows you to scroll the SQLi
buffer containing the output from the SQL buffer containing the

(add-hook 'sql-set-sqli-hook
   (function (lambda ()
	       (master-set-slave sql-buffer))))

+++
** File local variables.

A file local variables list cannot specify a string with text
properties--any specified text properties are discarded.

+++
** New function window-body-height.

This is like window-height but does not count the mode line
or the header line.

+++
** New function format-mode-line.

This returns the mode-line or header-line of the selected (or a
specified) window as a string with or without text properties.

+++
** New function `safe-get'.

This function is like `get', but never signals an error for
a malformed symbol property list.

+++
** New function `safe-plist-get'.

This function is like `plist-get', but never signals an error for
a malformed property list.

+++
** New functions `lax-plist-get' and `lax-plist-put'.

These functions are like `plist-get' and `plist-put' except that they
compare the property name using `equal' rather than `eq'.

+++
** New function `tool-bar-local-item-from-menu'

The `tool-bar-add-item-from-menu' must not be used (as previously
recommended) for making entries in the tool bar for local keymaps.
Instead, use the function `tool-bar-local-item-from-menu', which lets
you specify the map to use as an argument.

+++
** The function `atan' now accepts an optional second argument.

When called with 2 arguments, as in `(atan Y X)', `atan' returns the
angle in radians between the vector [X, Y] and the X axis.  (This is
equivalent to the standard C library function `atan2'.)

+++
** You can now make a window as short as one line.

A window that is just one line tall does not display either a mode
line or a header line, even if the variables `mode-line-format' and
`header-line-format' call for them.  A window that is two lines tall
cannot display both a mode line and a header line at once; if the
variables call for both, only the mode line actually appears.

+++
** The new frame parameter `tty-color-mode' specifies the mode to use
for color support on character terminal frames.  Its value can be a
number of colors to support, or a symbol.  See the Emacs Lisp
Reference manual for more detailed documentation.

+++
** The new mode-line construct `(:propertize ELT PROPS...)' can be
used to add text properties to mode-line elements.

+++
** The new `%i' and `%I' constructs for `mode-line-format' can be used
to display the size of the accessible part of the buffer on the mode
line.

---
** Indentation of simple and extended loop forms has been added to the
cl-indent package.  The new user options
`lisp-loop-keyword-indentation', `lisp-loop-forms-indentation', and
`lisp-simple-loop-indentation' can be used to customize the
indentation of keywords and forms in loop forms.

---
** Indentation of backquoted forms has been made customizable in the
cl-indent package.  See the new user option `lisp-backquote-indentation'.

+++
** Already true in Emacs 21.1, but not emphasized clearly enough:

Multibyte buffers can now faithfully record all 256 character codes
from 0 to 255.  As a result, most of the past reasons to use unibyte
buffers no longer exist.  We only know of three reasons to use them
now:

1. If you prefer to use unibyte text all of the time.

2. For reading files into temporary buffers, when you want to avoid
the time it takes to convert the format.

3. For binary files where format conversion would be pointless and
wasteful.

+++
** If text has a `keymap' property, that keymap takes precedence
over minor mode keymaps.

+++
** A hex escape in a string forces the string to be multibyte.
An octal escape makes it unibyte.

+++
** At the end of a command, point moves out from within invisible
text, in the same way it moves out from within text covered by an
image or composition property.

This makes it generally unnecessary to mark invisible text as intangible.
This is particularly good because the intangible property often has
unexpected side-effects since the property applies to everything
(including `goto-char', ...) whereas this new code is only run after
post-command-hook and thus does not care about intermediate states.

+++
** field-beginning and field-end now accept an additional optional
argument, LIMIT.

+++
** define-abbrev now accepts an optional argument SYSTEM-FLAG.  If
non-nil, this marks the abbrev as a "system" abbrev, which means that
it won't be stored in the user's abbrevs file if he saves the abbrevs.
Major modes that predefine some abbrevs should always specify this
flag.

---
** Support for Mocklisp has been removed.

---
** The function insert-string is now obsolete.

---
** The precedence of file-name-handlers has been changed.
Instead of blindly choosing the first handler that matches,
find-file-name-handler now gives precedence to a file-name handler
that matches near the end of the file name.  More specifically, the
handler whose (match-beginning 0) is the largest is chosen.
In case of ties, the old "first matched" rule applies.

---
** Dense keymaps now handle inheritance correctly.
Previously a dense keymap would hide all of the simple-char key
bindings of the parent keymap.

---
** jit-lock obeys a new text-property `jit-lock-defer-multiline'.
If a piece of text with that property gets contextually refontified
(see jit-lock-defer-contextually), then all of that text will
be refontified.  This is useful when the syntax of a textual element
depends on text several lines further down (and when font-lock-multiline
is not appropriate to solve that problem).  For example in Perl:

	s{
		foo
	}{
		bar
	}e

Adding/removing the last `e' changes the `bar' from being a piece of
text to being a piece of code, so you'd put a jit-lock-defer-multiline
property over the second half of the command to force (deferred)
refontification of `bar' whenever the `e' is added/removed.

---
** describe-vector now takes a second argument `describer' which is
called to print the entries' values.  It defaults to `princ'.

+++
** defcustom and other custom declarations now use a default group
(the last prior group defined in the same file) when no :group was given.

+++
** emacsserver now runs pre-command-hook and post-command-hook when
it receives a request from emacsclient.

---
** The variable `recursive-load-depth-limit' has been deleted.
Emacs now signals an error if the same file is loaded with more
than 3 levels of nesting.

---
** If a major mode function has a non-nil `no-clone-indirect'
property, `clone-indirect-buffer' signals an error if you use
it in that buffer.

---
** In `replace-match', the replacement text no longer inherits
properties from surrounding text.

+++
** The list returned by `(match-data t)' now has the buffer as a final
element, if the last match was on a buffer.  `set-match-data'
accepts such a list for restoring the match state.

+++
** New function `buffer-local-value'.

This function returns the buffer-local binding of VARIABLE (a symbol)
in buffer BUFFER.  If VARIABLE does not have a buffer-local binding in
buffer BUFFER, it returns the default value of VARIABLE instead.

---
** New function `text-clone-create'.  Text clones are chunks of text
that are kept identical by transparently propagating changes from one
clone to the other.

+++
** font-lock can manage arbitrary text-properties beside `face'.
*** the FACENAME returned in font-lock-keywords can be a list
of the form (face FACE PROP1 VAL1 PROP2 VAL2 ...) so you can set
other properties than `face'.
*** font-lock-extra-managed-props can be set to make sure those extra
properties are automatically cleaned up by font-lock.

---
** The special treatment of faces whose names are of the form `fg:COLOR'
or `bg:COLOR' has been removed.  Lisp programs should use the
`defface' facility for defining faces with specific colors, or use
the feature of specifying the face attributes :foreground and :background
directly in the `face' property instead of using a named face.

+++
** The new function `run-mode-hooks' and the new macro `delay-mode-hooks'
are used by define-derived-mode to make sure the mode hook for the
parent mode is run at the end of the child mode.

+++
** define-minor-mode now accepts arbitrary additional keyword arguments
and simply passes them to defcustom, if applicable.

+++
** define-derived-mode by default creates a new empty abbrev table.
It does not copy abbrevs from the parent mode's abbrev table.

+++
** `provide' and `featurep' now accept an optional second argument
to test/provide subfeatures.  Also `provide' now checks `after-load-alist'
and runs any code associated with the provided feature.

+++
** Functions `file-name-sans-extension' and `file-name-extension' now
ignore the leading dots in file names, so that file names such as
`.emacs' are treated as extensionless.

+++
** Functions `user-uid' and `user-real-uid' now return floats if the
user UID doesn't fit in a Lisp integer.  Function `user-full-name'
accepts a float as UID parameter.

---
** `define-key-after' now accepts keys longer than 1.

+++
** The local variable `no-byte-compile' in elisp files is now obeyed.

+++
** The Emacs Lisp byte-compiler now displays the actual line and
character position of errors, where possible.  Additionally, the form
of its warning and error messages have been brought more in line with
the output of other GNU tools.

+++
** New functions `keymap-prompt' and `current-active-maps'.

---
** New function `describe-buffer-bindings'.

+++
** New vars `exec-suffixes' and `load-suffixes' used when
searching for an executable resp. an elisp file.

+++
** Variable aliases have been implemented:

*** defvaralias ALIAS-VAR BASE-VAR [DOCSTRING]

This function defines the symbol ALIAS-VAR as a variable alias for
symbol BASE-VAR.  This means that retrieving the value of ALIAS-VAR
returns the value of BASE-VAR, and changing the value of ALIAS-VAR
changes the value of BASE-VAR.

DOCSTRING, if present, is the documentation for ALIAS-VAR; else it has
the same documentation as BASE-VAR.

*** indirect-variable VARIABLE

This function returns the variable at the end of the chain of aliases
of VARIABLE.  If VARIABLE is not a symbol, or if VARIABLE is not
defined as an alias, the function returns VARIABLE.

It might be noteworthy that variables aliases work for all kinds of
variables, including buffer-local and frame-local variables.

+++
** Functions from `post-gc-hook' are run at the end of garbage
collection.  The hook is run with GC inhibited, so use it with care.

+++
** If the second argument to `copy-file' is the name of a directory,
the file is copied to that directory instead of signaling an error.

+++
** The variables most-positive-fixnum and most-negative-fixnum
hold the largest and smallest possible integer values.

---
** On MS Windows, locale-coding-system is used to interact with the OS.
The Windows specific variable w32-system-coding-system, which was
formerly used for that purpose is now an alias for locale-coding-system.

---
** Functions y-or-n-p, read-char, read-key-sequence and the like, that
display a prompt but don't use the minibuffer, now display the prompt
using the text properties (esp. the face) of the prompt string.

---
** New function x-send-client-message sends a client message when
running under X.

+++
** Arguments for remove-overlays are now optional, so that you can remove
all overlays in the buffer by just calling (remove-overlay).

** New packages:

+++
*** The new package gdb-ui.el provides an enhanced graphical interface to
GDB. You can interact with GDB through the GUD buffer in the usual way, but
there are also further buffers which control the execution and describe the
state of your program.  It separates the input/output of your program from
that of GDB and watches expressions in the speedbar.  It also uses features of
Emacs 21 such as the display margin for breakpoints, and the toolbar.

Use M-x gdba to start GDB-UI.

*** The new package syntax.el provides an efficient way to find the
current syntactic context (as returned by parse-partial-sexp).

*** The new package bindat.el provides functions to unpack and pack
binary data structures, such as network packets, to and from Lisp
data structures.

---
*** The TCL package tcl-mode.el was replaced by tcl.el.
This was actually done in Emacs-21.1, and was not documented.

+++
*** The new package button.el implements simple and fast `clickable buttons'
in emacs buffers.  `buttons' are much lighter-weight than the `widgets'
implemented by widget.el, and can be used by lisp code that doesn't
require the full power of widgets.  Emacs uses buttons for such things
as help and apropos buffers.


* Installation changes in Emacs 21.3

** Support for GNU/Linux on little-endian MIPS and on IBM S390 has
been added.

90159:08185296b491

eight-bit-control/eight-bit-graphic charsets aren't now in the range
128-255.

* Installation Changes in Emacs 22.1

---
** Emacs now supports new configure options `--program-prefix',
`--program-suffix' and `--program-transform-name' that affect the names of
installed programs.

---
** Emacs can now be built without sound support.

---
** You can build Emacs with Gtk+ widgets by specifying `--with-x-toolkit=gtk'
when you run configure.  This requires Gtk+ 2.0 or newer.  This port
provides a way to display multilingual text in menus (with some caveats).

---
** The `emacsserver' program has been removed, replaced with elisp code.

---
** By default, Emacs now uses a setgid helper program to update game
scores.  The directory ${localstatedir}/games/emacs is the normal
place for game scores to be stored.  This may be controlled by the

Emacs with Leim.

+++
** The Emacs Lisp Reference Manual is now part of the distribution.

The Emacs Lisp Reference Manual in Info format is built as part of the
Emacs build procedure and installed together with the Emacs User
Manual.  A menu item was added to the menu bar that makes it easy
accessible (Help->More Manuals->Emacs Lisp Reference).

---

together with the Emacs User Manual, into the Info directory.  A menu
item was added to the menu bar that makes it easy accessible
(Help->More Manuals->Introduction to Emacs Lisp).

---
** New translations of the Emacs Tutorial are available in the
following languages: Brasilian Portuguese, Bulgarian, Chinese (both
with simplified and traditional characters), French, and Italian.
Type `C-u C-h t' to choose one of them in case your language setup
doesn't automatically select the right one.

---
** A French translation of the `Emacs Survival Guide' is available.

---
** Emacs now includes support for loading image libraries on demand.
(Currently this feature is only used on MS Windows.)  You can configure
the supported image types and their associated dynamic libraries by
setting the variable `image-library-alist'.

---
** Support for Cygwin was added.

---
** Support for FreeBSD/Alpha has been added.

** Mac OS 9 port now uses the Carbon API by default.  You can also
create non-Carbon build by specifying `NonCarbon' as a target.  See
the files mac/README and mac/INSTALL for build instructions.

---
** Building with -DENABLE_CHECKING does not automatically build with union
types any more.  Add -DUSE_LISP_UNION_TYPE if you want union types.

* Startup Changes in Emacs 22.1

** New command line option -Q or --quick.
This is like using -q --no-site-file, but in addition it also disables
the fancy startup screen.

+++
** New command line option -D or --basic-display.
Disables the menu-bar, the tool-bar, the scroll-bars, tool tips, and
the blinking cursor.

+++
** New command line option -nbc or --no-blinking-cursor disables
the blinking cursor on graphical terminals.

+++
** The command line option --no-windows has been changed to
--no-window-system.  The old one still works, but is deprecated.

+++
** The -f option, used from the command line to call a function,
now reads arguments for the function interactively if it is
an interactively callable function.

+++
** Emacs can now be invoked in full-screen mode on a windowed display.
When Emacs is invoked on a window system, the new command-line options
`--fullwidth', `--fullheight', and `--fullscreen' produce a frame
whose width, height, or both width and height take up the entire
screen size.  (For now, this does not work with some window managers.)

+++
** Emacs now displays a splash screen by default even if command-line
arguments were given.  The new command-line option --no-splash
disables the splash screen; see also the variable
`inhibit-startup-message' (which is also aliased as
`inhibit-splash-screen').

+++
** New user option `inhibit-startup-buffer-menu'.
When loading many files, for instance with `emacs *', Emacs normally
displays a buffer menu.  This option turns the buffer menu off.

+++
** Init file changes
You can now put the init files .emacs and .emacs_SHELL under
~/.emacs.d or directly under ~.  Emacs will find them in either place.

+++
** Emacs now reads the standard abbrevs file ~/.abbrev_defs
automatically at startup, if it exists.  When Emacs offers to save
modified buffers, it saves the abbrevs too if they have changed.  It
can do this either silently or asking for confirmation first,
according to the value of `save-abbrevs'.


* Editing Changes in Emacs 22.1

+++
** The mode line position information now comes before the major mode.
When the file is maintained under version control, that information
appears between the position information and the major mode.

+++
** M-g is now a prefix key.
M-g g and M-g M-g run goto-line.
M-g n and M-g M-n run next-error (like C-x `).
M-g p and M-g M-p run previous-error.

+++
** M-o now is the prefix key for setting text properties;
M-o M-o requests refontification.

+++
** C-u M-x goto-line now switches to the most recent previous buffer,
and goes to the specified line in that buffer.

When goto-line starts to execute, if there's a number in the buffer at
point then it acts as the default argument for the minibuffer.

+++
** You can now switch buffers in a cyclic order with C-x C-left and
(prev-buffer) and C-x C-right (next-buffer).  C-x left and C-x right
can be used as well.

+++
** The old bindings C-M-delete and C-M-backspace have been deleted,
since there are situations where one or the other will shut down
the operating system or your X server.

+++
** `undo-only' does an undo which does not redo any previous undo.

+++
** When the undo information of the current command gets really large
(beyond the value of `undo-outer-limit'), Emacs discards it and warns
you about it.

+++
** M-SPC (just-one-space) when given a numeric argument N
converts whitespace around point to N spaces.

+++
** line-move-ignore-invisible now defaults to t.

---
** New commands to operate on pairs of open and close characters:
`insert-pair', `delete-pair', `raise-sexp'.

---
** New command `kill-whole-line' kills an entire line at once.
By default, it is bound to C-S-<backspace>.

+++
** Yanking text now discards certain text properties that can
be inconvenient when you did not expect them.  The variable
`yank-excluded-properties' specifies which ones.  Insertion
of register contents and rectangles also discards these properties.

+++
** The default values of paragraph-start and indent-line-function have
been changed to reflect those used in Text mode rather than those used
in Indented-Text mode.

+++
** Movement commands `beginning-of-buffer', `end-of-buffer',
`beginning-of-defun', `end-of-defun' do not set the mark if the mark
is already active in Transient Mark mode.

** Mark Changes:

+++
*** A prefix argument is no longer required to repeat a jump to a
previous mark, i.e. C-u C-SPC C-SPC C-SPC ... cycles through the
mark ring.  Use C-u C-u C-SPC to set the mark immediately after a jump.

+++
*** Marking commands extend the region when invoked multiple times.  If
you hit M-C-SPC (mark-sexp), M-@ (mark-word), M-h (mark-paragraph), or
C-M-h (mark-defun) repeatedly, the marked region extends each time, so
you can mark the next two sexps with M-C-SPC M-C-SPC, for example.
This feature also works for mark-end-of-sentence, if you bind that to
a key.  It also extends the region when the mark is active in Transient
Mark mode, regardless of the last command.  To start a new region with
one of marking commands in Transient Mark mode, you can deactivate the
active region with C-g, or set the new mark with C-SPC.

+++
*** M-h (mark-paragraph) now accepts a prefix arg.
With positive arg, M-h marks the current and the following paragraphs;
if the arg is negative, it marks the current and the preceding
paragraphs.

+++
*** Some commands do something special in Transient Mark mode when the
mark is active--for instance, they limit their operation to the
region.  Even if you don't normally use Transient Mark mode, you might
want to get this behavior from a particular command.  There are two
ways you can enable Transient Mark mode and activate the mark, for one
command only.

One method is to type C-SPC C-SPC; this enables Transient Mark mode
and sets the mark at point.  The other method is to type C-u C-x C-x.
This enables Transient Mark mode temporarily but does not alter the
mark or the region.

After these commands, Transient Mark mode remains enabled until you
deactivate the mark.  That typically happens when you type a command
that alters the buffer, but you can also deactivate the mark by typing
C-g.

** Help command changes:

+++
*** Changes in C-h bindings:

C-h e displays the *Messages* buffer.

C-h followed by a control character is used for displaying files
    that do not change:

C-h C-f displays the FAQ.
C-h C-e displays the PROBLEMS file.

The info-search bindings on C-h C-f, C-h C-k and C-h C-i
have been moved to C-h F, C-h K and C-h S.

C-h c, C-h k, C-h w, and C-h f now handle remapped interactive commands.

- C-h c and C-h k report the actual command (after possible remapping)
  run by the key sequence.

- C-h w and C-h f on a command which has been remapped now report the
  command it is remapped to, and the keys which can be used to run
  that command.

For example, if C-k is bound to kill-line, and kill-line is remapped
to new-kill-line, these commands now report:

- C-h c and C-h k C-k reports:
  C-k runs the command new-kill-line

- C-h w and C-h f kill-line reports:
  kill-line is remapped to new-kill-line which is on C-k, <deleteline>

- C-h w and C-h f new-kill-line reports:
  new-kill-line is on C-k

---
*** Help commands `describe-function' and `describe-key' now show function
arguments in lowercase italics on displays that support it.  To change the
default, customize face `help-argument-name' or redefine the function
`help-default-arg-highlight'.

+++
*** C-h v and C-h f commands now include a hyperlink to the C source for
variables and functions defined in C (if the C source is available).

+++
*** Help mode now only makes hyperlinks for faces when the face name is
preceded or followed by the word `face'.  It no longer makes
hyperlinks for variables without variable documentation, unless
preceded by one of the words `variable' or `option'.  It now makes
hyperlinks to Info anchors (or nodes) if the anchor (or node) name is
enclosed in single quotes and preceded by `info anchor' or `Info
anchor' (in addition to earlier `info node' and `Info node').

+++
*** The new command `describe-char' (C-u C-x =) pops up a buffer with
description various information about a character, including its
encodings and syntax, its text properties, how to input, overlays, and
widgets at point.  You can get more information about some of them, by
clicking on mouse-sensitive areas or moving there and pressing RET.

+++
*** New command `display-local-help' displays any local help at point
in the echo area.  It is bound to `C-h .'.  It normally displays the
same string that would be displayed on mouse-over using the
`help-echo' property, but, in certain cases, it can display a more
keyboard oriented alternative.

+++
*** New user option `help-at-pt-display-when-idle' allows to
automatically show the help provided by `display-local-help' on
point-over, after suitable idle time.  The amount of idle time is
determined by the user option `help-at-pt-timer-delay' and defaults
to one second.  This feature is turned off by default.

** Buffer Menu changes

+++
*** New command `Buffer-menu-toggle-files-only' toggles display of file
buffers only in the Buffer Menu.  It is bound to `T' in Buffer Menu
mode.

+++
*** `buffer-menu' and `list-buffers' now list buffers whose names begin
with a space, when those buffers are visiting files.  Normally buffers
whose names begin with space are omitted.

---
*** The new options `buffers-menu-show-directories' and
`buffers-menu-show-status' let you control how buffers are displayed
in the menu dropped down when you click "Buffers" from the menu bar.

`buffers-menu-show-directories' controls whether the menu displays
leading directories as part of the file name visited by the buffer.
If its value is `unless-uniquify', the default, directories are
shown unless uniquify-buffer-name-style' is non-nil.  The value of nil
and t turn the display of directories off and on, respectively.

`buffers-menu-show-status' controls whether the Buffers menu includes
the modified and read-only status of the buffers.  By default it is
t, and the status is shown.

Setting these variables directly does not take effect until next time
the Buffers menu is regenerated.

** File Operation Changes:

+++
*** find-file-read-only visits multiple files in read-only mode,
when the file name contains wildcard characters.

+++
*** find-alternate-file replaces the current file with multiple files,
when the file name contains wildcard characters.

+++
*** Auto Compression mode is now enabled by default.

---
*** C-x C-f RET, typing nothing in the minibuffer, is no longer a special case.

Since the default input is the current directory, this has the effect
of specifying the current directory.  Normally that means to visit the
directory with Dired.

+++
*** When you are root, and you visit a file whose modes specify
read-only, the Emacs buffer is now read-only too.  Type C-x C-q if you
want to make the buffer writable.  (As root, you can in fact alter the
file.)

+++
*** C-x s (save-some-buffers) now offers an option `d' to diff a buffer
against its file, so you can see what changes you would be saving.

+++
*** The commands copy-file, rename-file, make-symbolic-link and
add-name-to-file, when given a directory as the "new name" argument,
convert it to a file name by merging in the within-directory part of
the existing file's name.  (This is the same convention that shell
commands cp, mv, and ln follow.)  Thus, M-x copy-file RET ~/foo RET
/tmp RET copies ~/foo to /tmp/foo.

---
*** When used interactively, `format-write-file' now asks for confirmation
before overwriting an existing file, unless a prefix argument is
supplied.  This behavior is analogous to `write-file'.

---
*** The variable `auto-save-file-name-transforms' now has a third element that
controls whether or not the function `make-auto-save-file-name' will
attempt to construct a unique auto-save name (e.g. for remote files).

+++
*** If the user visits a file larger than `large-file-warning-threshold',
Emacs prompts her for confirmation.

+++
*** require-final-newline now has two new possible values:

`visit' means add a newline (as an undoable change) if it's needed
when visiting the file.

`visit-save' means add a newline (as an undoable change) if it's
needed when visiting the file, and also add a newline if it's needed
when saving the file.

+++
*** The new option mode-require-final-newline controls how certain
major modes enable require-final-newline.  Any major mode that's
designed for a kind of file that should normally end in a newline
sets require-final-newline based on mode-require-final-newline.
So you can customize mode-require-final-newline to control what these
modes do.

+++
** The max size of buffers and integers has been doubled.
On 32bit machines, it is now 256M (i.e. 268435455).

** Minibuffer changes:

+++
*** There's a new face `minibuffer-prompt'.
Emacs adds this face to the list of text properties stored in the
variable `minibuffer-prompt-properties', which is used to display the
prompt string.

---
*** Enhanced visual feedback in *Completions* buffer.

Completions lists use faces to highlight what all completions
have in common and where they begin to differ.

The common prefix shared by all possible completions uses the face
`completions-common-part', while the first character that isn't the
same uses the face `completions-first-difference'.  By default,
`completions-common-part' inherits from `default', and
`completions-first-difference' inherits from `bold'.  The idea of
`completions-common-part' is that you can use it to make the common
parts less visible than normal, so that the rest of the differing
parts is, by contrast, slightly highlighted.

+++
*** File-name completion can now ignore directories.
If an element of the list in `completion-ignored-extensions' ends in a
slash `/', it indicates a subdirectory that should be ignored when
completing file names.  Elements of `completion-ignored-extensions'
which do not end in a slash are never considered when a completion
candidate is a directory.

+++
*** The completion commands TAB, SPC and ? in the minibuffer apply only
to the text before point.  If there is text in the buffer after point,
it remains unchanged.

+++
*** New user option `history-delete-duplicates'.
If set to t when adding a new history element, all previous identical
elements are deleted.

** Redisplay Changes

+++
*** Control characters and escape glyphs are now shown in the new
escape-glyph face.

+++
*** Non-breaking space and hyphens are now prefixed with an escape
character, unless the new user variable `show-nonbreak-escape' is set
to nil.

+++
*** The parameters of automatic hscrolling can now be customized.
The variable `hscroll-margin' determines how many columns away from
the window edge point is allowed to get before automatic hscrolling
will horizontally scroll the window.  The default value is 5.

The variable `hscroll-step' determines how many columns automatic
hscrolling scrolls the window when point gets too close to the
window edge.  If its value is zero, the default, Emacs scrolls the
window so as to center point.  If its value is an integer, it says how
many columns to scroll.  If the value is a floating-point number, it
gives the fraction of the window's width to scroll the window.

The variable `automatic-hscrolling' was renamed to
`auto-hscroll-mode'.  The old name is still available as an alias.

*** Moving or scrolling through images (and other lines) taller that
the window now works sensible, by automatically adjusting the window's
vscroll property.

+++
*** In graphical mode, with a C program, GUD Tooltips have been extended to
display the #define directive associated with an identifier when program is
not executing.

+++
*** The new face `mode-line-inactive' is used to display the mode line
of non-selected windows.  The `mode-line' face is now used to display
the mode line of the currently selected window.

The new variable `mode-line-in-non-selected-windows' controls whether
the `mode-line-inactive' face is used.

+++
*** You can now customize the use of window fringes.  To control this
for all frames, use M-x fringe-mode or the Show/Hide submenu of the
top-level Options menu, or customize the `fringe-mode' variable.  To
control this for a specific frame, use the command M-x
set-fringe-style.

+++
*** The buffer boundaries (i.e. first and last line in the buffer) may
now be marked with angle bitmaps in the fringes.  In addition, up and
down arrow bitmaps may be shown at the top and bottom of the left or
right fringe if the window can be scrolled in either direction.

This behavior is activated by setting the buffer-local variable
`indicate-buffer-boundaries' to a non-nil value.  The default value of
this variable is found in `default-indicate-buffer-boundaries'.

If value is `left' or `right', both angle and arrow bitmaps are
displayed in the left or right fringe, resp.

Value may also be an alist which specifies the presense and position
of each bitmap individually.

For example, ((top . left) (t .  right)) places the top angle bitmap
in left fringe, the bottom angle bitmap in right fringe, and both
arrow bitmaps in right fringe.  To show just the angle bitmaps in the
left fringe, but no arrow bitmaps, use ((top . left) (bottom . left)).

+++
*** On window systems, lines which are exactly as wide as the window
(not counting the final newline character) are no longer broken into
two lines on the display (with just the newline on the second line).
Instead, the newline now "overflows" into the right fringe, and the
cursor will be displayed in the fringe when positioned on that newline.

The new user option 'overflow-newline-into-fringe' may be set to nil to
revert to the old behavior of continuing such lines.

+++
*** When display margins are present in a window, the fringes are now
displayed between the margins and the buffer's text area, rather than
at the edges of the window.

+++
*** A window may now have individual fringe and scroll-bar settings,
in addition to the individual display margin settings.

Such individual settings are now preserved when windows are split
horizontally or vertically, a saved window configuration is restored,
or when the frame is resized.

** Cursor Display Changes

+++
*** On X, MS Windows, and Mac OS, the blinking cursor's "off" state is
now controlled by the variable `blink-cursor-alist'.

+++
*** The X resource cursorBlink can be used to turn off cursor blinking.

+++
*** Emacs can produce an underscore-like (horizontal bar) cursor.
The underscore cursor is set by putting `(cursor-type . hbar)' in
default-frame-alist.  It supports variable heights, like the `bar'
cursor does.

+++
*** Display of hollow cursors now obeys the buffer-local value (if any)
of `cursor-in-non-selected-windows' in the buffer that the cursor
appears in.

+++
*** The variable `cursor-in-non-selected-windows' can now be set to any
of the recognized cursor types.

+++
** font-lock-lines-before specifies a number of lines before the
current line that should be refontified when you change the buffer.
The default value is 1.

---
** JIT-lock changes

*** The default settings for JIT stealth lock parameters are changed.
The default value for the user option jit-lock-stealth-time is now 16
instead of 3, and the default value of jit-lock-stealth-nice is now
0.5 instead of 0.125.  The new defaults should lower the CPU usage
when Emacs is fontifying in the background.


*** jit-lock can now be delayed with `jit-lock-defer-time'.

If this variable is non-nil, its value should be the amount of Emacs
idle time in seconds to wait before starting fontification.  For
example, if you set `jit-lock-defer-time' to 0.25, fontification will
only happen after 0.25s of idle time.

*** contextual refontification is now separate from stealth fontification.

jit-lock-defer-contextually is renamed jit-lock-contextually and
jit-lock-context-time determines the delay after which contextual
refontification takes place.

** Menu Bar changes

---
*** A menu item "Show/Hide" was added to the top-level menu "Options".
This menu allows you to turn various display features on and off (such
as the fringes, the tool bar, the speedbar, and the menu bar itself).
You can also move the vertical scroll bar to either side here or turn
it off completely.  There is also a menu-item to toggle displaying of
current date and time, current line and column number in the
mode-line.

---
*** Speedbar has moved from the "Tools" top level menu to "Show/Hide".

+++
** You can now customize fill-nobreak-predicate to control where
filling can break lines.  The value is now normally a list of
functions, but it can also be a single function, for compatibility.

We provide two sample predicates, fill-single-word-nobreak-p and
fill-french-nobreak-p, for use in the value of fill-nobreak-predicate.

+++
** New display feature: focus follows the mouse from one Emacs window
to another, even within a frame.  If you set the variable
mouse-autoselect-window to non-nil value, moving the mouse to a
different Emacs window will select that window (minibuffer window can
be selected only when it is active).  The default is nil, so that this
feature is not enabled.

+++
** On X, when the window manager requires that you click on a frame to
select it (give it focus), the selected window and cursor position
normally changes according to the mouse click position.  If you set
the variable x-mouse-click-focus-ignore-position to t, the selected
window and cursor position do not change when you click on a frame
to give it focus.

+++
** When you specify a frame size with --geometry, the size applies to
all frames you create.  A position specified with --geometry only
affects the initial frame.

+++
** `special-display-buffer-names' and `special-display-regexps' now
understand two new boolean pseudo-frame-parameters `same-frame' and
`same-window'.

---
** New commands `scan-buf-next-region' and `scan-buf-previous-region'
move to the start of the next (previous, respectively) region with
non-nil help-echo property and display any help found there in the
echo area, using `display-local-help'.

+++
** In processing a local variables list, Emacs strips the prefix and
suffix are from every line before processing all the lines.

+++
** `apply-macro-to-region-lines' now operates on all lines that begin
in the region, rather than on all complete lines in the region.

+++
** You can now follow links by clicking Mouse-1 on the link.

Traditionally, Emacs uses a Mouse-1 click to set point and a Mouse-2
click to follow a link, whereas most other applications use a Mouse-1
click for both purposes, depending on whether you click outside or
inside a link.  Now the behavior of a Mouse-1 click has been changed
to match this context-sentitive dual behavior.

Depending on the current mode, a Mouse-2 click in Emacs may do much
more than just follow a link, so the new Mouse-1 behavior is only
activated for modes which explicitly mark a clickable text as a "link"
(see the new function `mouse-on-link-p' for details).  The Lisp
packages that are included in release 22.1 have been adapted to do
this, but external packages may not yet support this.  However, there
is no risk in using such packages, as the worst thing that could
happen is that you get the original Mouse-1 behavior when you click
on a link, which typically means that you set point where you click.

If you want to get the original Mouse-1 action also inside a link, you
just need to press the Mouse-1 button a little longer than a normal
click (i.e. press and hold the Mouse-1 button for half a second before
you release it).

Dragging the Mouse-1 inside a link still performs the original
drag-mouse-1 action, typically copy the text.

You can customize the new Mouse-1 behavior via the new user options
`mouse-1-click-follows-link' and `mouse-1-click-in-non-selected-windows'.

+++
** Emacs normally highlights mouse sensitive text whenever the mouse
is over the text.  By setting the new variable `mouse-highlight', you
can optionally enable mouse highlighting only after you move the
mouse, so that highlighting disappears when you press a key.  You can
also disable mouse highlighting.

+++
** You can now customize if selecting a region by dragging the mouse
shall not copy the selected text to the kill-ring by setting the new
variable mouse-drag-copy-region to nil.

---
** mouse-wheels can now scroll a specific fraction of the window
(rather than a fixed number of lines) and the scrolling is `progressive'.

---
** Unexpected yanking of text due to accidental clicking on the mouse
wheel button (typically mouse-2) during wheel scrolling is now avoided.
This behavior can be customized via the mouse-wheel-click-event and
mouse-wheel-inhibit-click-time variables.

+++
** Under X, mouse-wheel-mode is turned on by default.

+++
** M-x setenv now expands environment variables of the form `$foo' and
`${foo}' in the specified new value of the environment variable.  To
include a `$' in the value, use `$$'.

+++
** Unquoted `$' in file names do not signal an error any more when
the corresponding environment variable does not exist.
Instead, the `$ENVVAR' text is left as is, so that `$$' quoting
is only rarely needed.

---
** Language environment and various default coding systems are setup
more correctly according to the current locale name.  If the locale
name doesn't specify a charset, the default is what glibc defines.
This change may result in using the different coding systems as
default in some locale (e.g. vi_VN).

+++
** The default for the paper size (variable ps-paper-type) is taken
from the locale.

+++
** The keyboard-coding-system is now automatically set based on your
current locale settings if you are not using a window system.  This
may mean that the META key doesn't work but generates non-ASCII
characters instead, depending on how the terminal (or terminal
emulator) works.  Use `set-keyboard-coding-system' (or customize
keyboard-coding-system) if you prefer META to work (the old default)
or if the locale doesn't describe the character set actually generated
by the keyboard.  See Info node `Single-Byte Character Support'.

+++
** The new command `revert-buffer-with-coding-system' (C-x RET r)
revisits the current file using a coding system that you specify.

+++
** New command `recode-region' decodes the region again by a specified
coding system.

+++
** The new command `recode-file-name' changes the encoding of the name
of a file.

---
** New command `ucs-insert' inserts a character specified by its
unicode.

+++
** The new command `set-file-name-coding-system' (C-x RET F) sets
coding system for encoding and decoding file names.  A new menu item
(Options->Mule->Set Coding Systems->For File Name) invokes this
command.

+++
** New command quail-show-key shows what key (or key sequence) to type
in the current input method to input a character at point.

+++
** Limited support for character `unification' has been added.
Emacs now knows how to translate between different representations of
the same characters in various Emacs charsets according to standard
Unicode mappings.  This applies mainly to characters in the ISO 8859
sets plus some other 8-bit sets, but can be extended.  For instance,
translation works amongst the Emacs ...-iso8859-... charsets and the
mule-unicode-... ones.

By default this translation happens automatically on encoding.
Self-inserting characters are translated to make the input conformant
with the encoding of the buffer in which it's being used, where
possible.

You can force a more complete unification with the user option
unify-8859-on-decoding-mode.  That maps all the Latin-N character sets
into Unicode characters (from the latin-iso8859-1 and
mule-unicode-0100-24ff charsets) on decoding.  Note that this mode
will often effectively clobber data with an iso-2022 encoding.

---
** There is support for decoding Greek and Cyrillic characters into
either Unicode (the mule-unicode charsets) or the iso-8859 charsets,
when possible.  The latter are more space-efficient.  This is
controlled by user option utf-fragment-on-decoding.

---
** New language environments: French, Ukrainian, Tajik,
Bulgarian, Belarusian, Ukrainian, UTF-8, Windows-1255, Welsh, Latin-6,
Latin-7, Lithuanian, Latvian, Swedish, Slovenian, Croatian, Georgian,
Italian, Russian, Malayalam, Tamil, Russian, Chinese-EUC-TW.  (Set up
automatically according to the locale.)

---
** New input methods: latin-alt-postfix, latin-postfix, latin-prefix,
ukrainian-computer, belarusian, bulgarian-bds, russian-computer,
vietnamese-telex, lithuanian-numeric, lithuanian-keyboard,
latvian-keyboard, welsh, georgian, rfc1345, ucs, sgml,
bulgarian-phonetic, dutch, slovenian, croatian, malayalam-inscript,
tamil-inscript.

---
** New input method chinese-sisheng for inputting Chinese Pinyin
characters.

---
** Improved Thai support.  A new minor mode `thai-word-mode' (which is
automatically activated if you select Thai as a language
environment) changes key bindings of most word-oriented commands to
versions which recognize Thai words.  Affected commands are
    M-f     (forward-word)

    M-d     (kill-word)
    M-DEL   (backward-kill-word)
    M-t     (transpose-words)
    M-q     (fill-paragraph)

---
** Indian support has been updated.
The in-is13194 coding system is now Unicode-based.  CDAC fonts are
assumed.  There is a framework for supporting various
Indian scripts, but currently only Devanagari, Malayalam and Tamil are
supported.

---
** A UTF-7 coding system is available in the library `utf-7'.

---
** The utf-8/16 coding systems have been enhanced.
By default, untranslatable utf-8 sequences are simply composed into
single quasi-characters.  User option `utf-translate-cjk-mode' (it is
turned on by default) arranges to translate many utf-8 CJK character
sequences into real Emacs characters in a similar way to the Mule-UCS
system.  As this loads a fairly big data on demand, people who are not
interested in CJK characters may want to customize it to nil.
You can augment/amend the CJK translation via hash tables
`ucs-mule-cjk-to-unicode' and `ucs-unicode-to-mule-cjk'.  The utf-8
coding system now also encodes characters from most of Emacs's
one-dimensional internal charsets, specifically the ISO-8859 ones.
The utf-16 coding system is affected similarly.

---
** A new coding system `euc-tw' has been added for traditional Chinese
in CNS encoding; it accepts both Big 5 and CNS as input; on saving,
Big 5 is then converted to CNS.

---
** Many new coding systems are available by loading the `code-pages'
library.  These include complete versions of most of those in
codepage.el, based on Unicode mappings.  `codepage-setup' is now
obsolete and is used only in the MS-DOS port of Emacs.  windows-1252
and windows-1251 are preloaded since the former is so common and the
latter is used by GNU locales.

---
** New variable `utf-translate-cjk-unicode-range' controls which
Unicode characters to translate in `utf-translate-cjk-mode'.

---
** iso-10646-1 (`Unicode') fonts can be used to display any range of
characters encodable by the utf-8 coding system.  Just specify the
fontset appropriately.

+++
** Vertical scrolling is now possible within incremental search.
To enable this feature, customize the new user option
`isearch-allow-scroll'.  User written commands which satisfy stringent
constraints can be marked as "scrolling commands".  See the Emacs manual
for details.

+++
** C-w in incremental search now grabs either a character or a word,
making the decision in a heuristic way.  This new job is done by the
command `isearch-yank-word-or-char'.  To restore the old behavior,
bind C-w to `isearch-yank-word' in `isearch-mode-map'.

+++
** C-y in incremental search now grabs the next line if point is already
at the end of a line.

+++
** C-M-w deletes and C-M-y grabs a character in isearch mode.
Another method to grab a character is to enter the minibuffer by `M-e'
and to type `C-f' at the end of the search string in the minibuffer.

+++
** M-% typed in isearch mode invokes `query-replace' or
`query-replace-regexp' (depending on search mode) with the current
search string used as the string to replace.

+++
** Isearch no longer adds `isearch-resume' commands to the command
history by default.  To enable this feature, customize the new
user option `isearch-resume-in-command-history'.

---
** New user option `query-replace-skip-read-only': when non-nil,
`query-replace' and related functions simply ignore
a match if part of it has a read-only property.

+++
** When used interactively, the commands `query-replace-regexp' and
`replace-regexp' allow \,expr to be used in a replacement string,
where expr is an arbitrary Lisp expression evaluated at replacement
time.  In many cases, this will be more convenient than using
`query-replace-regexp-eval'.  `\#' in a replacement string now refers
to the count of replacements already made by the replacement command.
All regular expression replacement commands now allow `\?' in the
replacement string to specify a position where the replacement string
can be edited for each replacement.

+++
** query-replace uses isearch lazy highlighting when the new user option
`query-replace-lazy-highlight' is non-nil.

---
** The current match in query-replace is highlighted in new face
`query-replace' which by default inherits from isearch face.

+++
** M-x compare-windows now can automatically skip non-matching text to
resync points in both windows.

+++
** The commands M-x customize-face and M-x customize-face-other-window
now look at the character after point.  If a face or faces are
specified for that character, the commands by default customize those
faces.

---
** The face-customization widget has been reworked to be less confusing.
In particular, when you enable a face attribute using the corresponding
check-box, there's no longer a redundant `*' option in value selection
for that attribute; the values you can choose are only those which make
sense for the attribute.  When an attribute is de-selected by unchecking
its check-box, then the (now ignored, but still present temporarily in
case you re-select the attribute) value is hidden.

+++
** When you set or reset a variable's value in a Customize buffer,
the previous value becomes the "backup value" of the variable.
You can go back to that backup value by selecting "Use Backup Value"
under the "[State]" button.

** Dired mode:

---
*** New faces dired-header, dired-mark, dired-marked, dired-flagged,
dired-ignored, dired-directory, dired-symlink, dired-warning
introduced for Dired mode instead of font-lock faces.

+++
*** New Dired command `dired-compare-directories' marks files
with different file attributes in two dired buffers.

+++
*** New Dired command `dired-do-touch' (bound to T) changes timestamps
of marked files with the value entered in the minibuffer.

+++
*** In Dired's ! command (dired-do-shell-command), `*' and `?' now
control substitution of the file names only when they are surrounded
by whitespace.  This means you can now use them as shell wildcards
too.  If you want to use just plain `*' as a wildcard, type `*""'; the
doublequotes make no difference in the shell, but they prevent
special treatment in `dired-do-shell-command'.

+++
*** In Dired, the w command now copies the current line's file name
into the kill ring.  With a zero prefix arg, copies absolute file names.

+++
** The variables dired-free-space-program and dired-free-space-args
have been renamed to directory-free-space-program and
directory-free-space-args, and they now apply whenever Emacs puts a
directory listing into a buffer.

+++
** Dired-x:

+++
*** Omitting files is now a minor mode, dired-omit-mode.  The mode toggling
command is bound to M-o.  A new command dired-mark-omitted, bound to * O,
marks omitted files.  The variable dired-omit-files-p is obsoleted, use the
mode toggling function instead.

+++
** In Outline mode, hide-body no longer hides lines at the top
of the file that precede the first header line.

+++
** Occur, Info, and comint-derived modes now support using
M-x font-lock-mode to toggle fontification.  The variable
`Info-fontify' is no longer applicable; to disable fontification,
remove `turn-on-font-lock' from `Info-mode-hook'.

---
** The terminal emulation code in term.el has been improved, it can
run most curses applications now.

---
** The comint prompt can now be made read-only, using the new user
option `comint-prompt-read-only'.  This is not enabled by default,
except in IELM buffers.  The read-only status of IELM prompts can be

not the case, then `comint-kill-region' behaves just like
`kill-region' if read-only are involved: it copies the text to the
kill-ring, but does not delete it.

+++
** The new command `comint-insert-previous-argument' in comint-derived
modes (shell-mode etc) inserts arguments from previous command lines,
like bash's `ESC .' binding.  It is bound by default to `C-c .', but
otherwise behaves quite similarly to the bash version.

** `comint-use-prompt-regexp-instead-of-fields' has been renamed
`comint-use-prompt-regexp'.  The old name has been kept as an alias,
but declared obsolete.

+++
** Telnet now prompts you for a port number with C-u M-x telnet.

---
** M-x compile has become more robust and reliable

Quite a few more kinds of messages are recognized.  Messages that are
recognized as warnings or informational come in orange or green, instead of
red.  Informational messages are by default skipped with `next-error'
(controlled by `compilation-skip-threshold').

Location data is collected on the fly as the *compilation* buffer changes.
This means you could modify messages to make them point to different files.
This also means you can not go to locations of messages you may have deleted.

The variable `compilation-error-regexp-alist' has now become customizable.  If
you had added your own regexps to this, you'll probably need to include a
leading `^', otherwise they'll match anywhere on a line.  There is now also a
`compilation-mode-font-lock-keywords' and it nicely handles all the checks
that configure outputs and -o options so you see at a glance where you are.

The new file etc/compilation.txt gives examples of each type of message.

** Compilation mode enhancements:

+++
*** New user option `compilation-environment'.
This option allows you to specify environment variables for inferior
compilation processes without affecting the environment that all
subprocesses inherit.

+++
** Grep has been decoupled from compilation mode setup.

---
*** There's a new separate package grep.el.

---
*** M-x grep has been adapted to new compile

Hits are fontified in green, and hits in binary files in orange.  Grep buffers
can be saved and automatically revisited with the new Grep mode.

---
*** Grep commands now have their own submenu and customization group.

+++
*** `grep-find' is now also available under the name `find-grep' where
people knowing `find-grep-dired' would probably expect it.

---
*** The new variables `grep-window-height', `grep-auto-highlight', and
`grep-scroll-output' can be used to override the corresponding
compilation mode settings for grep commands.

+++
*** New option `grep-highlight-matches' highlightes matches in *grep*
buffer.  It uses a special feature of some grep programs which accept
--color option to output markers around matches.  When going to the next
match with `next-error' the exact match is highlighted in the source
buffer.  Otherwise, if `grep-highlight-matches' is nil, the whole
source line is highlighted.

+++
*** New key bindings in grep output window:
SPC and DEL scrolls window up and down.  C-n and C-p moves to next and
previous match in the grep window.  RET jumps to the source line of
the current match.  `n' and `p' shows next and previous match in
other window, but does not switch buffer.  `{' and `}' jumps to the
previous or next file in the grep output.  TAB also jumps to the next
file.

+++
** M-x grep now tries to avoid appending `/dev/null' to the command line
by using GNU grep `-H' option instead.  M-x grep automatically
detects whether this is possible or not the first time it is invoked.
When `-H' is used, the grep command line supplied by the user is passed
unchanged to the system to execute, which allows more complicated
command lines to be used than was possible before.

+++
** New options `next-error-highlight' and `next-error-highlight-no-select'
specify the method of highlighting of the corresponding source line
in new face `next-error'.

+++
** A new minor mode `next-error-follow-minor-mode' can be used in
compilation-mode, grep-mode, occur-mode, and diff-mode (i.e. all the
modes that can use `next-error').  In this mode, cursor motion in the
buffer causes automatic display in another window of the corresponding
matches, compilation errors, etc.  This minor mode can be toggled with
C-c C-f.

+++
** M-x diff uses diff-mode instead of compilation-mode.

+++
** In the *Occur* buffer, `o' switches to it in another window, and
C-o displays the current line's occurrence in another window without
switching to it.

+++
** You can now use next-error (C-x `) and previous-error to advance to
the next/previous matching line found by M-x occur.

+++
** The new command `multi-occur' is just like `occur', except it can
search multiple buffers.  There is also a new command
`multi-occur-by-filename-regexp' which allows you to specify the
buffers to search by their filename.  Internally, Occur mode has been
rewritten, and now uses font-lock, among other changes.

+++
** font-lock: in modes like C and Lisp where the fontification assumes that
an open-paren in column 0 is always outside of any string or comment,
font-lock now highlights any such open-paren-in-column-zero in bold-red
if it is inside a string or a comment, to indicate that it can cause
trouble with fontification and/or indentation.

** Enhancements to apropos commands:

+++
*** The apropos commands now accept a list of words to match.
When more than one word is specified, at least two of those words must
be present for an item to match.  Regular expression matching is still
available.

+++
*** The new option `apropos-sort-by-scores' causes the matching items
to be sorted according to their score.  The score for an item is a
number calculated to indicate how well the item matches the words or
regular expression that you entered to the apropos command.  The best
match is listed first, and the calculated score is shown for each
matching item.

+++
** You can have several Emacs servers on the same machine.

	% emacs --eval '(setq server-name "foo")' -f server-start &
	% emacs --eval '(setq server-name "bar")' -f server-start &
	% emacsclient -s foo file1
	% emacsclient -s bar file2

+++
** The `emacsclient' command understands the options `--eval' and
`--display' which tell Emacs respectively to evaluate the given elisp
expression and to use the given display when visiting files.

+++
** User option `server-mode' can be used to start a server process.

+++
** New user option `add-log-always-start-new-record'.
When this option is enabled, M-x add-change-log-entry always
starts a new record regardless of when the last record is.

** Menu support:

---
*** Dialogs and menus pop down if you type C-g.

---
*** The menu item "Open File..." has been split into two items, "New File..."
and "Open File...".  "Open File..." now opens only existing files.  This is
to support existing GUI file selection dialogs better.

+++
*** The file selection dialog for Gtk+, Mac, W32 and Motif/Lesstif can be
disabled by customizing the variable `use-file-dialog'.

** X Windows Support:

+++
*** Emacs now supports drag and drop for X.  Dropping a file on a window
   opens it, dropping text inserts the text.  Dropping a file on a dired
   buffer copies or moves the file to that directory.

+++
*** Under X11, it is possible to swap Alt and Meta (and Super and Hyper).
The new variables `x-alt-keysym', `x-hyper-keysym', `x-meta-keysym',
and `x-super-keysym' can be used to choose which keysyms Emacs should
use for the modifiers.  For example, the following two lines swap
Meta and Alt:
    (setq x-alt-keysym 'meta)
    (setq x-meta-keysym 'alt)

+++
*** The X resource useXIM can be used to turn off use of XIM, which may
speed up Emacs with slow networking to the X server.

If the configure option `--without-xim' was used to turn off use of
XIM by default, the X resource useXIM can be used to turn it on.

---
*** The new variable `x-select-request-type' controls how Emacs
requests X selection.  The default value is nil, which means that
Emacs requests X selection with types COMPOUND_TEXT and UTF8_STRING,
and use the more appropriately result.

---
*** The scrollbar under LessTif or Motif has a smoother drag-scrolling.
On the other hand, the size of the thumb does not represent the actual
amount of text shown any more (only a crude approximation of it).

---
*** The pop up menus for Lucid now stay up if you do a fast click and can
be navigated with the arrow keys (like Gtk+, Mac and W32).

+++
*** The Lucid menus can display multilingual text in your locale.  You have
to explicitly specify a fontSet resource for this to work, for example
`-xrm "Emacs*fontSet:  -*-helvetica-medium-r-*--*-120-*-*-*-*-*-*,*"'.

---
*** Dialogs for Lucid/Athena and Lesstif/Motif now pops down when pressing
ESC, like they do for Gtk+, Mac and W32.

+++
** For Gtk+ version 2.4, you can make Emacs use the old file dialog
by setting the variable `x-use-old-gtk-file-dialog' to t.  Default is to use
the new dialog.

** Xterm support:

---
*** Emacs now responds to mouse-clicks on the mode-line, header-line and
display margin, when run in an xterm.

---
*** Improved key bindings support when running in an xterm.
When emacs is running in an xterm more key bindings are available. The
following should work:
{C,S,C-S,A}-{right,left,up,down,prior,next,delete,insert,F1-12}.
These key bindings work on xterm from X.org 6.8, they might not work on
some older versions of xterm, or on some proprietary versions.

** Changes in support of colors on character terminals

+++
*** The new command-line option --color=MODE lets you specify a standard
mode for a tty color support.  It is meant to be used on character
terminals whose capabilities are not set correctly in the terminal
database, or with terminal emulators which support colors, but don't
set the TERM environment variable to a name of a color-capable
terminal.  "emacs --color" uses the same color commands as GNU `ls'
when invoked with "ls --color", so if your terminal can support colors
in "ls --color", it will support "emacs --color" as well.  See the
user manual for the possible values of the MODE parameter.

---
*** Emacs now supports several character terminals which provide more
than 8 colors.  For example, for `xterm', 16-color, 88-color, and
256-color modes are supported.  Emacs automatically notes at startup
the extended number of colors, and defines the appropriate entries for
all of these colors.

+++
*** Emacs now uses the full range of available colors for the default
faces when running on a color terminal, including 16-, 88-, and
256-color xterms.  This means that when you run "emacs -nw" on an
88-color or 256-color xterm, you will see essentially the same face
colors as on X.

---
*** There's a new support for colors on `rxvt' terminal emulator.

---
** M-x view-file and commands that use it now avoid interfering
with special modes such as Tar mode.

+++
** Filesets are collections of files.  You can define a fileset in
various ways, such as based on a directory tree or based on
program files that include other program files.

Once you have defined a fileset, you can perform various operations on
all the files in it, such as visiting them or searching and replacing
in them.

---
** Commands winner-redo and winner-undo, from winner.el, are now bound to
C-c <left> and C-c <right>, respectively.  This is an incompatible change.

---
** global-whitespace-mode is a new alias for whitespace-global-mode.

+++
** The command `list-text-properties-at' has been deleted because
C-u C-x = gives the same information and more.

* New modes and packages in Emacs 22.1

+++
** New package benchmark.el contains simple support for convenient
timing measurements of code (including the garbage collection component).

+++
** Calc is now part of the Emacs distribution.

Calc is an advanced desk calculator and mathematical tool written in
Emacs Lisp.  Its documentation is in a separate manual; within Emacs,
type "C-h i m calc RET" to read that manual.  A reference card is
available in `etc/calccard.tex' and `etc/calccard.ps'.

---
** `cfengine-mode' is a major mode for editing GNU Cfengine
configuration files.

+++
** The new package conf-mode.el handles thousands of configuration files, with
varying syntaxes for comments (;, #, //, /* */ or !), assignment (var = value,
var : value, var value or keyword var value) and sections ([section] or
section { }).  Many files under /etc/, or with suffixes like .cf through
.config, .properties (Java), .desktop (KDE/Gnome), .ini and many others are
recognized.

---
** CUA mode is now part of the Emacs distribution.

The new cua package provides CUA-like keybindings using C-x for
cut (kill), C-c for copy, C-v for paste (yank), and C-z for undo.
With cua, the region can be set and extended using shifted movement
keys (like pc-selection-mode) and typed text replaces the active
region (like delete-selection-mode).  Do not enable these modes with
cua-mode.  Customize the variable `cua-mode' to enable cua.

In addition, cua provides unified rectangle support with visible
rectangle highlighting: Use S-return to start a rectangle, extend it
using the movement commands (or mouse-3), and cut or copy it using C-x
or C-c (using C-w and M-w also works).

Use M-o and M-c to `open' or `close' the rectangle, use M-b or M-f, to
fill it with blanks or another character, use M-u or M-l to upcase or
downcase the rectangle, use M-i to increment the numbers in the
rectangle, use M-n to fill the rectangle with a numeric sequence (such
as 10 20 30...), use M-r to replace a regexp in the rectangle, and use
M-' or M-/ to restrict command on the rectangle to a subset of the
rows.  See the commentary in cua-base.el for more rectangle commands.

Cua also provides unified support for registers:  Use a numeric
prefix argument between 0 and 9, i.e. M-0 .. M-9, for C-x, C-c, and
C-v to cut or copy into register 0-9, or paste from register 0-9.

The last text deleted (not killed) is automatically stored in
register 0.  This includes text deleted by typing text.

Finally, cua provides a global mark which is set using S-C-space.
When the global mark is active, any text which is cut or copied is
automatically inserted at the global mark position.  See the
commentary in cua-base.el for more global mark related commands.

The features of cua also works with the standard emacs bindings for
kill, copy, yank, and undo.  If you want to use cua mode, but don't
want the C-x, C-c, C-v, and C-z bindings, you may customize the
`cua-enable-cua-keys' variable.

Note: This version of cua mode is not backwards compatible with older
versions of cua.el and cua-mode.el.  To ensure proper operation, you
must remove older versions of cua.el or cua-mode.el as well as the
loading and customization of those packages from the .emacs file.

+++
** The new package dns-mode.el add syntax highlight of DNS master files.
The key binding C-c C-s (`dns-mode-soa-increment-serial') can be used
to increment the SOA serial.

---
** The new global minor mode `file-name-shadow-mode' modifies the way
filenames being entered by the user in the minibuffer are displayed, so
that it's clear when part of the entered filename will be ignored due to
emacs' filename parsing rules.  The ignored portion can be made dim,
invisible, or otherwise less visually noticable.  The display method may
be displayed by customizing the variable `file-name-shadow-properties'.

+++
** The new package flymake.el does on-the-fly syntax checking of program
source files.  See the Flymake's Info manual for more details.

---
** The new Lisp library fringe.el controls the appearance of fringes.

---
** GDB-Script-mode is used for files like .gdbinit.

+++
** The new package gdb-ui.el provides an enhanced graphical interface to
GDB. You can interact with GDB through the GUD buffer in the usual way, but
there are also further buffers which control the execution and describe the
state of your program.  It separates the input/output of your program from
that of GDB and watches expressions in the speedbar.  It also uses features of
Emacs 21 such as the display margin for breakpoints, and the toolbar.

Use M-x gdba to start GDB-UI.

---
** The new package ibuffer provides a powerful, completely
customizable replacement for buff-menu.el.

---
** Ido mode is now part of the Emacs distribution.

The ido (interactively do) package is an extension of the iswitchb
package to do interactive opening of files and directories in addition
to interactive buffer switching.  Ido is a superset of iswitchb (with
a few exceptions), so don't enable both packages.

+++
** Image files are normally visited in Image mode, which lets you toggle
between viewing the image and viewing the text using C-c C-c.

+++
** The new keypad setup package provides several common bindings for
the numeric keypad which is available on most keyboards.  The numeric
keypad typically has the digits 0 to 9, a decimal point, keys marked
+, -, /, and *, an Enter key, and a NumLock toggle key.  The keypad
package only controls the use of the digit and decimal keys.

By customizing the variables `keypad-setup', `keypad-shifted-setup',
`keypad-numlock-setup', and `keypad-numlock-shifted-setup', or by
using the function `keypad-setup', you can rebind all digit keys and
the decimal key of the keypad in one step for each of the four
possible combinations of the Shift key state (not pressed/pressed) and
the NumLock toggle state (off/on).

The choices for the keypad keys in each of the above states are:
`Plain numeric keypad' where the keys generates plain digits,
`Numeric keypad with decimal key' where the character produced by the
decimal key can be customized individually (for internationalization),
`Numeric Prefix Arg' where the keypad keys produce numeric prefix args
for emacs editing commands, `Cursor keys' and `Shifted Cursor keys'
where the keys work like (shifted) arrow keys, home/end, etc., and
`Unspecified/User-defined' where the keypad keys (kp-0, kp-1, etc.)
are left unspecified and can be bound individually through the global
or local keymaps.

+++
** The new kmacro package provides a simpler user interface to
emacs' keyboard macro facilities.

Basically, it uses two function keys (default F3 and F4) like this:
F3 starts a macro, F4 ends the macro, and pressing F4 again executes
the last macro.  While defining the macro, F3 inserts a counter value
which automatically increments every time the macro is executed.

There is now a keyboard macro ring which stores the most recently
defined macros.

The C-x C-k sequence is now a prefix for the kmacro keymap which
defines bindings for moving through the keyboard macro ring,
C-x C-k C-p and C-x C-k C-n, editing the last macro C-x C-k C-e,
manipulating the macro counter and format via C-x C-k C-c,
C-x C-k C-a, and C-x C-k C-f.  See the commentary in kmacro.el
for more commands.

The normal macro bindings C-x (, C-x ), and C-x e now interfaces to
the keyboard macro ring.

The C-x e command now automatically terminates the current macro
before calling it, if used while defining a macro.

In addition, when ending or calling a macro with C-x e, the macro can
be repeated immediately by typing just the `e'.  You can customize
this behavior via the variable kmacro-call-repeat-key and
kmacro-call-repeat-with-arg.

Keyboard macros can now be debugged and edited interactively.
C-x C-k SPC steps through the last keyboard macro one key sequence
at a time, prompting for the actions to take.

+++
** The new package longlines.el provides a minor mode for editing text
files composed of long lines, based on the `use-hard-newlines'
mechanism.  The long lines are broken up by inserting soft newlines,
which are automatically removed when saving the file to disk or
copying into the kill ring, clipboard, etc.  By default, Longlines
mode inserts soft newlines automatically during editing, a behavior
referred to as "soft word wrap" in other text editors.  This is
similar to Refill mode, but more reliable.  To turn the word wrap
feature off, set `longlines-auto-wrap' to nil.

---
** The old Octave mode bindings C-c f and C-c i have been changed
to C-c C-f and C-c C-i.  The C-c C-i subcommands now have duplicate
bindings on control characters--thus, C-c C-i C-b is the same as
C-c C-i b, and so on.

** The printing package is now part of the Emacs distribution.

If you enable the printing package by including (require 'printing) in
the .emacs file, the normal Print item on the File menu is replaced
with a Print sub-menu which allows you to preview output through
ghostview, use ghostscript to print (if you don't have a PostScript
printer) or send directly to printer a PostScript code generated by
`ps-print' package.  Use M-x pr-help for more information.

+++
** The new python.el package is used to edit Python and Jython programs.

---
** The minor mode Reveal mode makes text visible on the fly as you
move your cursor into hidden regions of the buffer.
It should work with any package that uses overlays to hide parts
of a buffer, such as outline-minor-mode, hs-minor-mode, hide-ifdef-mode, ...

There is also Global Reveal mode which affects all buffers.

---
** The ruler-mode.el library provides a minor mode for displaying an
"active" ruler in the header line.  You can use the mouse to visually
change the `fill-column', `window-margins' and `tab-stop-list'
settings.

+++
** SES mode (ses-mode) is a new major mode for creating and editing
spreadsheet files.  Besides the usual Emacs features (intuitive command
letters, undo, cell formulas in Lisp, plaintext files, etc.) it also offers
viral immunity and import/export of tab-separated values.

+++
** The new global minor mode `size-indication-mode' (off by default)
shows the size of accessible part of the buffer on the mode line.

+++
** The new package table.el implements editable, WYSIWYG, embedded
`text tables' in Emacs buffers.  It simulates the effect of putting
these tables in a special major mode.  The package emulates WYSIWYG
table editing available in modern word processors.  The package also
can generate a table source in typesetting and markup languages such
as latex and html from the visually laid out text table.

+++
** The thumbs.el package allows you to preview image files as thumbnails
and can be invoked from a Dired buffer.

+++
** Tramp is now part of the distribution.

This package is similar to Ange-FTP: it allows you to edit remote
files.  But whereas Ange-FTP uses FTP to access the remote host,
Tramp uses a shell connection.  The shell connection is always used
for filename completion and directory listings and suchlike, but for
the actual file transfer, you can choose between the so-called
`inline' methods (which transfer the files through the shell
connection using base64 or uu encoding) and the `out-of-band' methods
(which invoke an external copying program such as `rcp' or `scp' or
`rsync' to do the copying).

Shell connections can be acquired via `rsh', `ssh', `telnet' and also
`su' and `sudo'.  Ange-FTP is still supported via the `ftp' method.

If you want to disable Tramp you should set

  (setq tramp-default-method "ftp")

---
** The library tree-widget.el provides a new widget to display a set
of hierarchical data as an outline.  For example, the tree-widget is
well suited to display a hierarchy of directories and files.

---
** The URL package (which had been part of W3) is now part of Emacs.

---
** New minor mode, Visible mode, toggles invisibility in the current buffer.
When enabled, it makes all invisible text visible.  When disabled, it
restores the previous value of `buffer-invisibility-spec'.

+++
** The wdired.el package allows you to use normal editing commands on Dired
buffers to change filenames, permissions, etc...

* Changes in specialized modes and packages:

+++
** There is a new user option `mail-default-directory' that allows you
to specify the value of `default-directory' for mail buffers.  This
directory is used for auto-save files of mail buffers.  It defaults to
"~/".

+++
** Emacs can now indicate in the mode-line the presence of new e-mail
in a directory or in a file.  See the documentation of the user option
`display-time-mail-directory'.

---
** PO translation files are decoded according to their MIME headers
when Emacs visits them.

** Info mode:

+++
*** A numeric prefix argument of `info' selects an Info buffer
with the number appended to the *info* buffer name (e.g. "*info*<2>").

---
*** isearch in Info uses Info-search and searches through multiple nodes.
Before leaving the initial Info node isearch fails once with the error
message [initial node], and with subsequent C-s/C-r continues through
other nodes.  When isearch fails for the rest of the manual, it wraps
aroung the whole manual to the top/final node.  The user option
`Info-isearch-search' controls whether to use Info-search for isearch,
or the default isearch search function that wraps around the current
Info node.

*** New search commands: `Info-search-case-sensitively' (bound to S),
`Info-search-backward', and `Info-search-next' which repeats the last
search without prompting for a new search string.

*** New command `Info-history-forward' (bound to r and new toolbar icon)
moves forward in history to the node you returned from after using
`Info-history-back' (renamed from `Info-last').

*** New command `Info-history' (bound to L) displays a menu of visited nodes.

*** New command `Info-toc' (bound to T) creates a node with table of contents
from the tree structure of menus of the current Info file.

*** New command `info-apropos' searches the indices of the known
Info files on your system for a string, and builds a menu of the
possible matches.

*** New command `Info-copy-current-node-name' (bound to w) copies
the current Info node name into the kill ring.  With a zero prefix
arg, puts the node name inside the `info' function call.

---
*** New face `info-xref-visited' distinguishes visited nodes from unvisited
and a new option `Info-fontify-visited-nodes' to control this.

*** http and ftp links in Info are now operational: they look like cross
references and following them calls `browse-url'.

+++
*** Info now hides node names in menus and cross references by default.
If you prefer the old behavior, you can set the new user option
`Info-hide-note-references' to nil.

---
*** Images in Info pages are supported.
Info pages show embedded images, in Emacs frames with image support.
Info documentation that includes images, processed with makeinfo
version 4.7 or newer, compiles to Info pages with embedded images.

+++
*** The default value for `Info-scroll-prefer-subnodes' is now nil.

---
*** Info-index offers completion.

** Lisp mode changes:

---
*** Lisp mode now uses font-lock-doc-face for the docstrings.

+++
*** A prefix argument of C-M-q in Emacs Lisp mode pretty-printifies the
list starting after point.

*** New features in evaluation commands

+++
*** The function `eval-defun' (C-M-x) called on defface reinitializes
the face to the value specified in the defface expression.

+++
*** Typing C-x C-e twice prints the value of the integer result
in additional formats (octal, hexadecimal, character) specified
by the new function `eval-expression-print-format'.  The same
function also defines the result format for `eval-expression' (M-:),
`eval-print-last-sexp' (C-j) and some edebug evaluation functions.

+++
** CC Mode changes.

*** Font lock support.
CC Mode now provides font lock support for all its languages.  This
supersedes the font lock patterns that have been in the core font lock
package for C, C++, Java and Objective-C.  Like indentation, font
locking is done in a uniform way across all languages (except the new
AWK mode - see below).  That means that the new font locking will be
different from the old patterns in various details for most languages.

The main goal of the font locking in CC Mode is accuracy, to provide a
dependable aid in recognizing the various constructs.  Some, like
strings and comments, are easy to recognize while others like
declarations and types can be very tricky.  CC Mode can go to great
lengths to recognize declarations and casts correctly, especially when
the types aren't recognized by standard patterns.  This is a fairly
demanding analysis which can be slow on older hardware, and it can
therefore be disabled by choosing a lower decoration level with the
variable font-lock-maximum-decoration.

Note that the most demanding font lock level has been tuned with lazy
fontification in mind, i.e.  there should be a support mode that waits
with the fontification until the text is actually shown
(e.g. Just-in-time Lock mode, which is the default, or Lazy Lock
mode).  Fontifying a file with several thousand lines in one go can
take the better part of a minute.

**** The (c|c++|objc|java|idl|pike)-font-lock-extra-types variables
are now used by CC Mode to recognize identifiers that are certain to
be types.  (They are also used in cases that aren't related to font
locking.)  At the maximum decoration level, types are often recognized
properly anyway, so these variables should be fairly restrictive and
not contain patterns for uncertain types.

**** Support for documentation comments.
There is a "plugin" system to fontify documentation comments like
Javadoc and the markup within them.  It's independent of the host
language, so it's possible to e.g. turn on Javadoc font locking in C
buffers.  See the variable c-doc-comment-style for details.

Currently two kinds of doc comment styles are recognized: Suns Javadoc
and Autodoc which is used in Pike.  This is by no means a complete
list of the most common tools; if your doc comment extractor of choice
is missing then please drop a note to [e-mail].

**** Better handling of C++ templates.
As a side effect of the more accurate font locking, C++ templates are
now handled much better.  The angle brackets that delimit them are
given parenthesis syntax so that they can be navigated like other
parens.

This also improves indentation of templates, although there still is
work to be done in that area.  E.g. it's required that multiline
template clauses are written in full and then refontified to be
recognized, and the indentation of nested templates is a bit odd and
not as configurable as it ought to be.

**** Improved handling of Objective-C and CORBA IDL.
Especially the support for Objective-C and IDL has gotten an overhaul.
The special "@" declarations in Objective-C are handled correctly.
All the keywords used in CORBA IDL, PSDL, and CIDL are recognized and
handled correctly, also wrt indentation.

*** Support for the AWK language.
Support for the AWK language has been introduced.  The implementation is
based around GNU AWK version 3.1, but it should work pretty well with
any AWK.  As yet, not all features of CC Mode have been adapted for AWK.
Here is a summary:

**** Indentation Engine
The CC Mode indentation engine fully supports AWK mode.

AWK mode handles code formatted in the conventional AWK fashion: `{'s
which start actions, user-defined functions, or compound statements are
placed on the same line as the associated construct; the matching `}'s
are normally placed under the start of the respective pattern, function
definition, or structured statement.

The predefined indentation functions haven't yet been adapted for AWK
mode, though some of them may work serendipitously.  There shouldn't be
any problems writing custom indentation functions for AWK mode.

The command C-c C-q (c-indent-defun) hasn't yet been adapted for AWK,
though in practice it works properly nearly all the time.  Should it
fail, explicitly set the region around the function (using C-u C-SPC:
C-M-h probably won't work either) then do C-M-\ (indent-region).

**** Font Locking
There is a single level of font locking in AWK mode, rather than the
three distinct levels the other modes have.  There are several
idiosyncrasies in AWK mode's font-locking due to the peculiarities of
the AWK language itself.

**** Comment Commands
M-; (indent-for-comment) works fine.  None of the other CC Mode
comment formatting commands have yet been adapted for AWK mode.

**** Movement Commands
Most of the movement commands work in AWK mode.  The most important
exceptions are M-a (c-beginning-of-statement) and M-e
(c-end-of-statement) which haven't yet been adapted.

The notion of "defun" has been augmented to include AWK pattern-action
pairs.  C-M-a (c-awk-beginning-of-defun) and C-M-e (c-awk-end-of-defun)
recognise these pattern-action pairs, as well as user defined
functions.

**** Auto-newline Insertion and Clean-ups
Auto-newline insertion hasn't yet been adapted for AWK.  Some of
the clean-ups can actually convert good AWK code into syntactically
invalid code.  These features are best disabled in AWK buffers.

*** New syntactic symbols in IDL mode.
The top level constructs "module" and "composition" (from CIDL) are
now handled like "namespace" in C++: They are given syntactic symbols
module-open, module-close, inmodule, composition-open,
composition-close, and incomposition.

*** New functions to do hungry delete without enabling hungry delete mode.
The functions c-hungry-backspace and c-hungry-delete-forward can be
bound to keys to get this feature without toggling a mode.
Contributed by Kevin Ryde.

*** Better control over require-final-newline.  The variable that
controls how to handle a final newline when the buffer is saved,
require-final-newline, is now customizable on a per-mode basis through
c-require-final-newline.  That is a list of modes, and only those
modes set require-final-newline.  By default that's C, C++ and
Objective-C.

The specified modes set require-final-newline based on
mode-require-final-newline, as usual.

*** Format change for syntactic context elements.
The elements in the syntactic context returned by c-guess-basic-syntax
and stored in c-syntactic-context has been changed somewhat to allow
attaching more information.  They are now lists instead of single cons
cells.  E.g. a line that previously had the syntactic analysis

((inclass . 11) (topmost-intro . 13))

is now analysed as

((inclass 11) (topmost-intro 13))

In some cases there are more than one position given for a syntactic
symbol.

This change might affect code that call c-guess-basic-syntax directly,
and custom lineup functions if they use c-syntactic-context.  However,
the argument given to lineup functions is still a single cons cell
with nil or an integer in the cdr.

*** API changes for derived modes.
There have been extensive changes "under the hood" which can affect
derived mode writers.  Some of these changes are likely to cause
incompatibilities with existing derived modes, but on the other hand
care has now been taken to make it possible to extend and modify CC
Mode with less risk of such problems in the future.

**** New language variable system.
See the comment blurb near the top of cc-langs.el.

**** New initialization functions.
The initialization procedure has been split up into more functions to
give better control: c-basic-common-init, c-font-lock-init, and
c-init-language-vars.

*** Changes in analysis of nested syntactic constructs.
The syntactic analysis engine has better handling of cases where
several syntactic constructs appear nested on the same line.  They are
now handled as if each construct started on a line of its own.

This means that CC Mode now indents some cases differently, and
although it's more consistent there might be cases where the old way
gave results that's more to one's liking.  So if you find a situation
where you think that the indentation has become worse, please report
it to [e-mail].

**** New syntactic symbol substatement-label.
This symbol is used when a label is inserted between a statement and
its substatement.  E.g:

    if (x)
      x_is_true:
        do_stuff();

*** Better handling of multiline macros.

**** Syntactic indentation inside macros.
The contents of multiline #define's are now analyzed and indented
syntactically just like other code.  This can be disabled by the new
variable c-syntactic-indentation-in-macros.  A new syntactic symbol
cpp-define-intro has been added to control the initial indentation
inside #define's.

**** New lineup function c-lineup-cpp-define.
Now used by default to line up macro continuation lines.  The behavior
of this function closely mimics the indentation one gets if the macro
is indented while the line continuation backslashes are temporarily
removed.  If syntactic indentation in macros is turned off, it works
much line c-lineup-dont-change, which was used earlier, but handles
empty lines within the macro better.

**** Automatically inserted newlines continues the macro if used within one.
This applies to the newlines inserted by the auto-newline mode, and to
c-context-line-break and c-context-open-line.

**** Better alignment of line continuation backslashes.
c-backslash-region tries to adapt to surrounding backslashes.  New
variable c-backslash-max-column which put a limit on how far out
backslashes can be moved.

**** Automatic alignment of line continuation backslashes.
This is controlled by the new variable c-auto-align-backslashes.  It
affects c-context-line-break, c-context-open-line and newlines
inserted in auto-newline mode.

**** Line indentation works better inside macros.
Regardless whether syntactic indentation and syntactic indentation
inside macros are enabled or not, line indentation now ignores the
line continuation backslashes.  This is most noticeable when syntactic
indentation is turned off and there are empty lines (save for the
backslash) in the macro.

*** indent-for-comment is more customizable.
The behavior of M-; (indent-for-comment) is now configurable through
the variable c-indent-comment-alist.  The indentation behavior based
on the preceding code on the line, e.g. to get two spaces after #else
and #endif but indentation to comment-column in most other cases
(something which was hardcoded earlier).

*** New function c-context-open-line.
It's the open-line equivalent of c-context-line-break.

*** New lineup functions

**** c-lineup-string-cont
This lineup function lines up a continued string under the one it
continues.  E.g:

result = prefix + "A message "
                  "string.";      <- c-lineup-string-cont

**** c-lineup-cascaded-calls
Lines up series of calls separated by "->" or ".".

**** c-lineup-knr-region-comment
Gives (what most people think is) better indentation of comments in
the "K&R region" between the function header and its body.

**** c-lineup-gcc-asm-reg
Provides better indentation inside asm blocks.  Contributed by Kevin
Ryde.

**** c-lineup-argcont
Lines up continued function arguments after the preceding comma.
Contributed by Kevin Ryde.

*** Better caching of the syntactic context.
CC Mode caches the positions of the opening parentheses (of any kind)
of the lists surrounding the point.  Those positions are used in many
places as anchor points for various searches.  The cache is now
improved so that it can be reused to a large extent when the point is
moved.  The less it moves, the less needs to be recalculated.

The effect is that CC Mode should be fast most of the time even when
opening parens are hung (i.e. aren't in column zero).  It's typically
only the first time after the point is moved far down in a complex
file that it'll take noticeable time to find out the syntactic
context.

*** Statements are recognized in a more robust way.
Statements are recognized most of the time even when they occur in an
"invalid" context, e.g. in a function argument.  In practice that can
happen when macros are involved.

*** Improved the way c-indent-exp chooses the block to indent.
It now indents the block for the closest sexp following the point
whose closing paren ends on a different line.  This means that the
point doesn't have to be immediately before the block to indent.
Also, only the block and the closing line is indented; the current
line is left untouched.

*** Added toggle for syntactic indentation.
The function c-toggle-syntactic-indentation can be used to toggle
syntactic indentation.

---
** Perl mode has a new variable `perl-indent-continued-arguments'.

** Fortran mode changes:

---
*** Fortran mode does more font-locking by default. Use level 3
highlighting for the old default.

+++
*** Fortran mode has a new variable `fortran-directive-re'.
Adapt this to match the format of any compiler directives you use.
Lines that match are never indented, and are given distinctive font-locking.

+++
*** F90 mode and Fortran mode have new navigation commands
`f90-end-of-block', `f90-beginning-of-block', `f90-next-block',
`f90-previous-block', `fortran-end-of-block',
`fortran-beginning-of-block'.

---
*** F90 mode and Fortran mode have support for hs-minor-mode (hideshow).
It cannot deal with every code format, but ought to handle a sizeable
majority.

---
*** The new function `f90-backslash-not-special' can be used to change
the syntax of backslashes in F90 buffers.

---
** Prolog mode has a new variable `prolog-font-lock-keywords'
to support use of font-lock.

** HTML/SGML changes:

---
*** Emacs now tries to set up buffer coding systems for HTML/XML files
automatically.

+++
*** SGML mode has indentation and supports XML syntax.
The new variable `sgml-xml-mode' tells SGML mode to use XML syntax.
When this option is enabled, SGML tags are inserted in XML style,
i.e., there is always a closing tag.
By default, its setting is inferred on a buffer-by-buffer basis
from the file name or buffer contents.

+++
*** `xml-mode' is now an alias for `sgml-mode', which has XML support.

** TeX modes:

+++
*** C-c C-c prompts for a command to run, and tries to offer a good default.

+++
*** The user option `tex-start-options-string' has been replaced
by two new user options: `tex-start-options', which should hold
command-line options to feed to TeX, and `tex-start-commands' which should hold
TeX commands to use at startup.

---
*** verbatim environments are now highlighted in courier by font-lock
and super/sub-scripts are made into super/sub-scripts.

+++
*** New major mode doctex-mode for *.dtx files.

** BibTeX mode:
*** The new command bibtex-url browses a URL for the BibTeX entry at
point (bound to C-c C-l and mouse-2, RET on clickable fields).

*** The new command bibtex-entry-update (bound to C-c C-u) updates
an existing BibTeX entry.

*** New `bibtex-entry-format' option `required-fields', enabled by default.

*** bibtex-maintain-sorted-entries can take values `plain',
`crossref', and `entry-class' which control the sorting scheme used
for BibTeX entries.  `bibtex-sort-entry-class' controls the sorting
scheme `entry-class'.  TAB completion for reference keys and
automatic detection of duplicates does not require anymore that
bibtex-maintain-sorted-entries is non-nil.

*** If the new variable bibtex-parse-keys-fast is non-nil,
use fast but simplified algorithm for parsing BibTeX keys.

*** If the new variable bibtex-autoadd-commas is non-nil,
automatically add missing commas at end of BibTeX fields.

*** The new variable bibtex-autofill-types contains a list of entry
types for which fields are filled automatically (if possible).

*** The new command bibtex-complete completes word fragment before
point according to context (bound to M-tab).

*** The new commands bibtex-find-entry and bibtex-find-crossref
locate entries and crossref'd entries (bound to C-c C-s and C-c C-x).
Crossref fields are clickable (bound to mouse-2, RET).

*** In BibTeX mode the command fill-paragraph (bound to M-q) fills
individual fields of a BibTeX entry.

*** The new variables bibtex-files and bibtex-file-path define a set
of BibTeX files that are searched for entry keys.

*** The new command bibtex-validate-globally checks for duplicate keys
in multiple BibTeX files.

*** The new command bibtex-copy-summary-as-kill pushes summary
of BibTeX entry to kill ring (bound to C-c C-t).

+++
** In Enriched mode, `set-left-margin' and `set-right-margin' are now
by default bound to `C-c [' and `C-c ]' instead of the former `C-c C-l'
and `C-c C-r'.

** GUD changes:

+++
*** In GUD mode, when talking to GDB, C-x C-a C-j "jumps" the program
counter to the specified source line (the one where point is).

---
*** GUD mode has its own tool bar for controlling execution of the inferior
and other common debugger commands.

---
** GUD mode improvements for jdb:

*** Search for source files using jdb classpath and class
    information. Fast startup since there is no need to scan all
    source files up front. There is also no need to create and maintain
    lists of source directories to scan. Look at `gud-jdb-use-classpath'
    and `gud-jdb-classpath' customization variables documentation.

*** Supports the standard breakpoint (gud-break, gud-clear)
    set/clear operations from java source files under the classpath, stack
    traversal (gud-up, gud-down), and run until current stack finish
    (gud-finish).

*** Supports new jdb (Java 1.2 and later) in addition to oldjdb
    (Java 1.1 jdb).

*** The previous method of searching for source files has been
    preserved in case someone still wants/needs to use it.
    Set gud-jdb-use-classpath to nil.

  Added Customization Variables

*** gud-jdb-command-name.  What command line to use to invoke jdb.

*** gud-jdb-use-classpath. Allows selection of java source file searching
    method: set to t for new method, nil to scan gud-jdb-directories for
    java sources (previous method).

*** gud-jdb-directories. List of directories to scan and search for java
    classes using the original gud-jdb method (if gud-jdb-use-classpath
    is nil).

  Minor Improvements

*** The STARTTLS elisp wrapper (starttls.el) can now use GNUTLS
instead of the OpenSSL based "starttls" tool.  For backwards
compatibility, it prefers "starttls", but you can toggle
`starttls-use-gnutls' to switch to GNUTLS (or simply remove the
"starttls" tool).

*** Do not allow debugger output history variable to grow without bounds.

** Auto-Revert changes:

+++
*** You can now use Auto Revert mode to `tail' a file.
If point is at the end of a file buffer before reverting, Auto Revert
mode keeps it at the end after reverting.  Similarly if point is
displayed at the end of a file buffer in any window, it stays at
the end of the buffer in that window.  This allows to tail a file:
just put point at the end of the buffer and it stays there.  This

then you can tail the file more efficiently by using the new minor
mode Auto Revert Tail mode.  The function `auto-revert-tail-mode'
toggles this mode.

+++
*** Auto Revert mode is now more careful to avoid excessive reverts and
other potential problems when deciding which non-file buffers to
revert.  This matters especially if Global Auto Revert mode is enabled
and `global-auto-revert-non-file-buffers' is non-nil.  Auto Revert
mode only reverts a non-file buffer if the buffer has a non-nil
`revert-buffer-function' and a non-nil `buffer-stale-function', which
decides whether the buffer should be reverted.  Currently, this means
that auto reverting works for Dired buffers (although this may not
work properly on all operating systems) and for the Buffer Menu.

+++
*** If the new user option `auto-revert-check-vc-info' is non-nil, Auto
Revert mode reliably updates version control info (such as the version
control number in the mode line), in all version controlled buffers in
which it is active.  If the option is nil, the default, then this info
only gets updated whenever the buffer gets reverted.

---
** recentf changes.

The recent file list is now automatically cleanup when recentf mode is
enabled.  The new option `recentf-auto-cleanup' controls when to do
automatic cleanup.

The `recentf-keep' option replaces `recentf-keep-non-readable-files-p'
and provides a more general mechanism to customize which file names to
keep in the recent list.

With the more advanced option: `recentf-filename-handler', you can
specify a function that transforms filenames handled by recentf.  For
example, if set to `file-truename', the same file will not be in the
recent list with different symbolic links.

To follow naming convention, `recentf-menu-append-commands-flag'
replaces the misnamed option `recentf-menu-append-commands-p'.  The
old name remains available as alias, but has been marked obsolete.

+++
** Desktop package

+++

Customize the new option `save-place-forget-unreadable-files' to nil
to get the old behavior.  The new options `save-place-save-skipped'
and `save-place-skip-check-regexp' allow further fine-tuning of this
feature.

** EDiff changes.

+++
***  When comparing directories.
Typing D brings up a buffer that lists the differences between the contents of
directories. Now it is possible to use this buffer to copy the missing files
from one directory to another.

+++
*** When comparing files or buffers.
Typing the = key now offers to perform the word-by-word comparison of the
currently highlighted regions in an inferior Ediff session. If you answer 'n'
then it reverts to the old behavior and asks the user to select regions for
comparison.

*** The new command `ediff-backup' compares a file with its most recent
backup using `ediff'.  If you specify the name of a backup file,
`ediff-backup' compares it with the file of which it is a backup.

+++
** Etags changes.

*** New regular expressions features

**** New syntax for regular expressions, multi-line regular expressions.
The syntax --ignore-case-regexp=/regex/ is now undocumented and retained
only for backward compatibility.  The new equivalent syntax is
--regex=/regex/i.  More generally, it is --regex=/TAGREGEX/TAGNAME/MODS,
where `/TAGNAME' is optional, as usual, and MODS is a string of 0 or
more characters among `i' (ignore case), `m' (multi-line) and `s'
(single-line).  The `m' and `s' modifiers behave as in Perl regular
expressions: `m' allows regexps to match more than one line, while `s'
(which implies `m') means that `.' matches newlines.  The ability to
span newlines allows writing of much more powerful regular expressions
and rapid prototyping for tagging new languages.

**** Regular expressions can use char escape sequences as in Gcc.
The escaped character sequence \a, \b, \d, \e, \f, \n, \r, \t, \v,
respectively, stand for the ASCII characters BEL, BS, DEL, ESC, FF, NL,
CR, TAB, VT,

**** Regular expressions can be bound to a given language.
The syntax --regex={LANGUAGE}REGEX means that REGEX is used to make tags
only for files of language LANGUAGE, and ignored otherwise.  This is
particularly useful when storing regexps in a file.

**** Regular expressions can be read from a file.
The --regex=@regexfile option means read the regexps from a file, one
per line.  Lines beginning with space or tab are ignored.

*** New language parsing features

**** The `::' qualifier triggers C++ parsing in C file.
Previously, only the `template' and `class' keywords had this effect.

**** The gnucc __attribute__ keyword is now recognised and ignored.

**** New language HTML.
Title and h1, h2, h3 are tagged.  Also, tags are generated when name= is
used inside an anchor and whenever id= is used.

**** In Makefiles, constants are tagged.
If you want the old behavior instead, thus avoiding to increase the
size of the tags file, use the --no-globals option.

**** New language Lua.
All functions are tagged.

**** In Perl, packages are tags.
Subroutine tags are named from their package.  You can jump to sub tags
as you did before, by the sub name, or additionally by looking for
package::sub.

**** In Prolog, etags creates tags for rules in addition to predicates.

**** New language PHP.
Tags are functions, classes and defines.
If the --members option is specified to etags, tags are variables also.

**** New default keywords for TeX.
The new keywords are def, newcommand, renewcommand, newenvironment and
renewenvironment.

*** Honour #line directives.
When Etags parses an input file that contains C preprocessor's #line
directives, it creates tags using the file name and line number
specified in those directives.  This is useful when dealing with code
created from Cweb source files.  When Etags tags the generated file, it
writes tags pointing to the source file.

*** New option --parse-stdin=FILE.
This option is mostly useful when calling etags from programs.  It can
be used (only once) in place of a file name on the command line.  Etags
reads from standard input and marks the produced tags as belonging to
the file FILE.

** VC Changes

+++
*** The key C-x C-q no longer checks files in or out, it only changes
the read-only state of the buffer (toggle-read-only).  We made this
change because we held a poll and found that many users were unhappy
with the previous behavior.  If you do prefer this behavior, you
can bind `vc-toggle-read-only' to C-x C-q in your .emacs:

    (global-set-key "\C-x\C-q" 'vc-toggle-read-only)

The function `vc-toggle-read-only' will continue to exist.

+++
*** There is a new user option `vc-cvs-global-switches' that allows
you to specify switches that are passed to any CVS command invoked
by VC.  These switches are used as "global options" for CVS, which
means they are inserted before the command name.  For example, this
allows you to specify a compression level using the "-z#" option for
CVS.

+++
*** New backends for Subversion and Meta-CVS.

+++
*** vc-annotate-mode enhancements

In vc-annotate mode, you can now use the following key bindings for
enhanced functionality to browse the annotations of past revisions, or
to view diffs or log entries directly from vc-annotate-mode:

    P:  annotates the previous revision
    N:  annotates the next revision
    J:  annotates the revision at line
    A:  annotates the revision previous to line
    D:  shows the diff of the revision at line with its previous revision
    L:  shows the log of the revision at line
    W:  annotates the workfile (most up to date) version

** pcl-cvs changes:

+++
*** In pcl-cvs mode, there is a new `d y' command to view the diffs
between the local version of the file and yesterday's head revision
in the repository.

+++
*** In pcl-cvs mode, there is a new `d r' command to view the changes
anyone has committed to the repository since you last executed
"checkout", "update" or "commit".  That means using cvs diff options
-rBASE -rHEAD.

** Gnus package

---
*** Gnus now includes Sieve and PGG
Sieve is a library for managing Sieve scripts.  PGG is a library to handle
PGP/MIME.

---
*** There are many news features, bug fixes and improvements.
See the file GNUS-NEWS or the node "Oort Gnus" in the Gnus manual for details.

---
** Rmail now displays 5-digit message ids in its summary buffer.

+++
** Support for `movemail' from GNU mailutils was added to Rmail.
This version of `movemail' allows to read mail from a wide range of
mailbox formats, including remote POP3 and IMAP4 mailboxes with or
without TLS encryption.  If GNU mailutils is installed on the system
and its version of `movemail' can be found in exec-path, it will be
used instead of the native one.

---
** MH-E changes.

Upgraded to MH-E version 7.82. There have been major changes since
version 5.0.2; see MH-E-NEWS for details.

** Calendar changes:

+++
*** There is a new calendar package, icalendar.el, that can be used to
convert Emacs diary entries to/from the iCalendar format.

+++
*** Diary sexp entries can have custom marking in the calendar.
Diary sexp functions which only apply to certain days (such as
`diary-block' or `diary-cyclic') now take an optional parameter MARK,
which is the name of a face or a single-character string indicating
how to highlight the day in the calendar display.  Specifying a
single-character string as @var{mark} places the character next to the
day in the calendar.  Specifying a face highlights the day with that
face.  This lets you have different colors or markings for vacations,
appointments, paydays or anything else using a sexp.

+++
*** The new function `calendar-goto-day-of-year' (g D) prompts for a
year and day number, and moves to that date. Negative day numbers
count backward from the end of the year.

+++
*** The new Calendar function `calendar-goto-iso-week' (g w)
prompts for a year and a week number, and moves to the first
day of that ISO week.

---
*** The new variable `calendar-minimum-window-height' affects the
window generated by the function `generate-calendar-window'.

---
*** The functions `holiday-easter-etc' and `holiday-advent' now take
optional arguments, in order to only report on the specified holiday
rather than all. This makes customization of variables such as
`christian-holidays' simpler.

---
*** The function `simple-diary-display' now by default sets a header line.
This can be controlled through the variables `diary-header-line-flag'
and `diary-header-line-format'.

+++
*** The procedure for activating appointment reminders has changed: use
the new function `appt-activate'. The new variable
`appt-display-format' controls how reminders are displayed, replacing
appt-issue-message, appt-visible, and appt-msg-window.

+++
*** The new functions `diary-from-outlook', `diary-from-outlook-gnus',
and `diary-from-outlook-rmail' can be used to import diary entries
from Outlook-format appointments in mail messages.  The variable
`diary-outlook-formats' can be customized to recognize additional
formats.

---
** sql changes.

*** The variable `sql-product' controls the highlightng of different

*** Added SQL->Start SQLi Session menu entry which calls the
appropriate sql-interactive-mode wrapper for the current setting of
`sql-product'.

---
*** Support for the SQLite interpreter has been added to sql.el by calling
'sql-sqlite'.

** FFAP changes:

+++
*** New ffap commands and keybindings: C-x C-r (`ffap-read-only'),
C-x C-v (`ffap-alternate-file'), C-x C-d (`ffap-list-directory'),
C-x 4 r (`ffap-read-only-other-window'), C-x 4 d (`ffap-dired-other-window'),

---
*** FFAP accepts wildcards in a file name by default.  C-x C-f passes
it to `find-file' with non-nil WILDCARDS argument, which visits
multiple files, and C-x d passes it to `dired'.

---
** skeleton.el now supports using - to mark the skeleton-point without
   interregion interaction.  @ has reverted to only setting
   skeleton-positions and no longer sets skeleton-point.  Skeletons
   which used @ to mark skeleton-point independent of _ should now use -
   instead.  The updated skeleton-insert docstring explains these new
   features along with other details of skeleton construction.

---
** New variable `hs-set-up-overlay' allows customization of the overlay
used to effect hiding for hideshow minor mode.  Integration with isearch
handles the overlay property `display' specially, preserving it during
temporary overlay showing in the course of an isearch operation.

+++
** hide-ifdef-mode now uses overlays rather than selective-display
to hide its text.  This should be mostly transparent but slightly
changes the behavior of motion commands like C-e and C-p.

---
** partial-completion-mode now does partial completion on directory names.

---
** The type-break package now allows `type-break-file-name' to be nil
and if so, doesn't store any data across sessions.  This is handy if
you don't want the .type-break file in your home directory or are
annoyed by the need for interaction when you kill Emacs.

---
** `ps-print' can now print characters from the mule-unicode charsets.

Printing text with characters from the mule-unicode-* sets works with
ps-print, provided that you have installed the appropriate BDF fonts.
See the file INSTALL for URLs where you can find these fonts.

---
** New command `strokes-global-set-stroke-string'.
This is like `strokes-global-set-stroke', but it allows you to bind
the stroke directly to a string to insert.  This is convenient for
using strokes as an input method.

---
** LDAP support now defaults to ldapsearch from OpenLDAP version 2.

+++
** You can now disable pc-selection-mode after enabling it.
M-x pc-selection-mode behaves like a proper minor mode, and with no
argument it toggles the mode.

Turning off PC-Selection mode restores the global key bindings
that were replaced by turning on the mode.

---
** `uniquify-strip-common-suffix' tells uniquify to prefer
`file|dir1' and `file|dir2' to `file|dir1/subdir' and `file|dir2/subdir'.

---
** Support for `magic cookie' standout modes has been removed.
Emacs will still work on terminals that require magic cookies in order
to use standout mode, however they will not be able to display
mode-lines in inverse-video.

---
** The game `mpuz' is enhanced.

`mpuz' now allows the 2nd factor not to have two identical digits.  By
default, all trivial operations involving whole lines are performed
automatically.  The game uses faces for better visual feedback.

---
** display-battery has been replaced by display-battery-mode.

---
** calculator.el now has radix grouping mode, which is available when
`calculator-output-radix' is non-nil.  In this mode a separator
character is used every few digits, making it easier to see byte
boundries etc.  For more info, see the documentation of the variable
`calculator-radix-grouping-mode'.

---
** fast-lock.el and lazy-lock.el are obsolete.  Use jit-lock.el instead.

---
** iso-acc.el is now obsolete.  Use one of the latin input methods instead.

---
** cplus-md.el has been removed to avoid problems with Custom.

* Changes for non-free operating systems

+++
** Passing resources on the command line now works on MS Windows.
You can use --xrm to pass resource settings to Emacs, overriding any
existing values.  For example:

  emacs --xrm "Emacs.Background:red" --xrm "Emacs.Geometry:100x20"

will start up Emacs on an initial frame of 100x20 with red background,
irrespective of geometry or background setting on the Windows registry.

---
** On MS Windows, the "system caret" now follows the cursor.
This enables Emacs to work better with programs that need to track
the cursor, for example screen magnifiers and text to speech programs.

multilingual text with other applications. On other versions of
MS Windows, Emacs now uses the appropriate locale coding-system, so
the clipboard should work correctly for your local language without
any customizations.

---
** On Mac OS, the value of the variable `keyboard-coding-system' is
now dynamically changed according to the current keyboard script.  The
variable `mac-keyboard-text-encoding' and the constants
`kTextEncodingMacRoman', `kTextEncodingISOLatin1', and
`kTextEncodingISOLatin2' are obsolete.

* Incompatible Lisp Changes in Emacs 22.1

+++
** The new interactive-specification `G' reads a file name
much like `F', but if the input is a directory name (even defaulted),

:propertize and :eval forms in the value of a variable whose
`risky-local-variable' property is nil.

* Lisp Changes in Emacs 22.1

+++
** The option --script FILE runs Emacs in batch mode and loads FILE.
It is useful for writing Emacs Lisp shell script files, because they
can start with this line:

   #!/usr/bin/emacs --script

+++
** The option --directory DIR now modifies `load-path' immediately.
Directories are added to the front of `load-path' in the order they
appear on the command line.  For example, with this command line:

  emacs -batch -L .. -L /tmp --eval "(require 'foo)"

Emacs looks for library `foo' in the parent directory, then in /tmp, then
in the other directories in `load-path'.  (-L is short for --directory.)

+++
** The default value of `sentence-end' is now defined using the new
variable `sentence-end-without-space', which contains such characters
that end a sentence without following spaces.

The function `sentence-end' should be used to obtain the value of the
variable `sentence-end'.  If the variable `sentence-end' is nil, then
this function returns the regexp constructed from the variables
`sentence-end-without-period', `sentence-end-double-space' and
`sentence-end-without-space'.

+++
** The argument to forward-word, backward-word, forward-to-indentation
and backward-to-indentation is now optional, and defaults to 1.

+++
** If a command sets transient-mark-mode to `only', that
enables Transient Mark mode for the following command only.
During that following command, the value of transient-mark-mode
is `identity'.  If it is still `identity' at the end of the command,
it changes to nil.

+++
** The new hook `before-save-hook' is invoked by `basic-save-buffer'
before saving buffers.  This allows packages to perform various final
tasks, for example; it can be used by the copyright package to make
sure saved files have the current year in any copyright headers.

+++
** If a buffer sets buffer-save-without-query to non-nil,
save-some-buffers will always save that buffer without asking
(if it's modified).

---
** list-buffers-noselect now takes an additional argument, BUFFER-LIST.
If it is non-nil, it specifies which buffers to list.

+++
** The kill-buffer-hook is now permanent-local.

+++
** `auto-save-file-format' has been renamed to
`buffer-auto-save-file-format' and made into a permanent local.

+++
** Functions `file-name-sans-extension' and `file-name-extension' now
ignore the leading dots in file names, so that file names such as
`.emacs' are treated as extensionless.

+++
** copy-file now takes an additional option arg MUSTBENEW.

This argument works like the MUSTBENEW argument of write-file.

+++
** If the second argument to `copy-file' is the name of a directory,
the file is copied to that directory instead of signaling an error.

+++
** `visited-file-modtime' and `calendar-time-from-absolute' now return
a list of two integers, instead of a cons.

+++
** `file-chase-links' now takes an optional second argument LIMIT which
specifies the maximum number of links to chase through.  If after that
many iterations the file name obtained is still a symbolic link,
`file-chase-links' returns it anyway.

+++
** The function `commandp' takes an additional optional
argument.  If it is non-nil, then `commandp' checks
for a function that could be called with `call-interactively',
and does not return t for keyboard macros.

+++
** An interactive specification may now use the code letter 'U' to get
the up-event that was discarded in case the last key sequence read for a
previous 'k' or 'K' argument was a down-event; otherwise nil is used.

---
** Functions y-or-n-p, read-char, read-key-sequence and the like, that
display a prompt but don't use the minibuffer, now display the prompt
using the text properties (esp. the face) of the prompt string.

+++
** read-from-minibuffer now accepts an additional argument KEEP-ALL
saying to put all inputs in the history list, even empty ones.

+++
** The `read-file-name' function now takes an additional argument which
specifies a predicate which the file name read must satify.  The
new variable `read-file-name-predicate' contains the predicate argument

** The new function `read-directory-name' can be used instead of
`read-file-name' to read a directory name; when used, completion
will only show directories.

+++
** The new variable search-spaces-regexp controls how to search
for spaces in a regular expression.  If it is non-nil, it should be a
regular expression, and any series of spaces stands for that regular
expression.  If it is nil, spaces stand for themselves.

Spaces inside of constructs such as [..] and *, +, ? are never
replaced with search-spaces-regexp.

+++
** There are now two new regular expression operators, \_< and \_>,
for matching the beginning and end of a symbol.  A symbol is a
non-empty sequence of either word or symbol constituent characters, as
specified by the syntax table.

+++
** skip-chars-forward and skip-chars-backward now handle
character classes such as [:alpha:], along with individual characters
and ranges.

---
** In `replace-match', the replacement text no longer inherits
properties from surrounding text.

+++
** The list returned by `(match-data t)' now has the buffer as a final
element, if the last match was on a buffer.  `set-match-data'
accepts such a list for restoring the match state.

+++
** Variable aliases have been implemented:

*** defvaralias ALIAS-VAR BASE-VAR [DOCSTRING]

This function defines the symbol ALIAS-VAR as a variable alias for
symbol BASE-VAR.  This means that retrieving the value of ALIAS-VAR
returns the value of BASE-VAR, and changing the value of ALIAS-VAR
changes the value of BASE-VAR.

DOCSTRING, if present, is the documentation for ALIAS-VAR; else it has
the same documentation as BASE-VAR.

*** indirect-variable VARIABLE

This function returns the variable at the end of the chain of aliases
of VARIABLE.  If VARIABLE is not a symbol, or if VARIABLE is not
defined as an alias, the function returns VARIABLE.

It might be noteworthy that variables aliases work for all kinds of
variables, including buffer-local and frame-local variables.

+++
*** The macro define-obsolete-variable-alias combines defvaralias and
make-obsolete-variable.  The macro define-obsolete-function-alias
combines defalias and make-obsolete.

+++
** Enhancements to keymaps.

*** Cleaner way to enter key sequences.

You can enter a constant key sequence in a more natural format, the
same one used for saving keyboard macros, using the macro `kbd'.  For
example,

(kbd "C-x C-f") => "\^x\^f"

*** Interactive commands can be remapped through keymaps.

This is an alternative to using defadvice or substitute-key-definition
to modify the behavior of a key binding using the normal keymap
binding and lookup functionality.

- The new variable `this-original-command' contains the original
  command before remapping.  It is equal to `this-command' when the
  command was not remapped.

*** If text has a `keymap' property, that keymap takes precedence
over minor mode keymaps.

*** The `keymap' property now also works at the ends of overlays and
text-properties, according to their stickiness.  This also means that it
works with empty overlays.  The same hold for the `local-map' property.

*** Dense keymaps now handle inheritance correctly.
Previously a dense keymap would hide all of the simple-char key
bindings of the parent keymap.

*** `define-key-after' now accepts keys longer than 1.

*** New function `current-active-maps' returns a list of currently
active keymaps.

*** New function `describe-buffer-bindings' inserts the list of all
defined keys and their definitions.

*** New function `keymap-prompt' returns the prompt-string of a keymap

*** (map-keymap FUNCTION KEYMAP) applies the function to each binding
in the keymap.

*** New variable emulation-mode-map-alists.

Lisp packages using many minor mode keymaps can now maintain their own
keymap alist separate from minor-mode-map-alist by adding their keymap
alist to this list.

will lead to undesirable results, so don't let it happen; the first
change group you start for any given buffer should be the last one
finished.

+++
** Progress reporters.
The new functions `make-progress-reporter', `progress-reporter-update',
`progress-reporter-force-update', `progress-reporter-done', and
`dotimes-with-progress-reporter' provide a simple and efficient way for
a command to present progress messages for the user.

+++
** New `yank-handler' text property may be used to control how
previously killed text on the kill-ring is reinserted.

  If UNDO is present and non-nil, it is a function that will be called
by `yank-pop' to undo the insertion of the current object.  It is
called with two arguments, the start and end of the current region.
FUNCTION may set `yank-undo-function' to override the UNDO value.

*** The functions kill-new, kill-append, and kill-region now have an
optional argument to specify the yank-handler text property to put on
the killed text.

*** The function yank-pop will now use a non-nil value of the variable
`yank-undo-function' (instead of delete-region) to undo the previous
yank or yank-pop command (or a call to insert-for-yank).  The function
insert-for-yank automatically sets that variable according to the UNDO
element of the string argument's yank-handler text property if present.

*** The function `insert-for-yank' now supports strings where the
`yank-handler' property does not span the first character of the
string.  The old behavior is available if you call
`insert-for-yank-1' instead.

*** The new function insert-for-yank normally works like `insert', but
removes the text properties in the `yank-excluded-properties' list.
However, the insertion of the text may be modified by a `yank-handler'
text property.

+++
** An element of buffer-undo-list can now have the form (apply FUNNAME
. ARGS), where FUNNAME is a symbol other than t or nil.  That stands
for a high-level change that should be undone by evaluating (apply
FUNNAME ARGS).

These entries can also have the form (apply DELTA BEG END FUNNAME . ARGS)
which indicates that the change which took place was limited to the
range BEG...END and increased the buffer size by DELTA.

+++
** If the buffer's undo list for the current command gets longer than
undo-outer-limit, garbage collection empties it.  This is to prevent
it from using up the available memory and choking Emacs.

+++
** Enhancements to process support

*** Function list-processes now has an optional argument; if non-nil,

is handled, suspending output from other processes.  If value is an
integer, also inhibit running timers.  This feature is generally not
recommended, but may be necessary for specific applications, such as
speech synthesis.

*** Adaptive read buffering of subprocess output.

On some systems, when emacs reads the output from a subprocess, the
output data is read in very small blocks, potentially resulting in
very poor performance.  This behavior can be remedied to some extent
by setting the new variable process-adaptive-read-buffering to a
non-nil value (the default), as it will automatically delay reading
from such processes, to allowing them to produce more output before
emacs tries to read it.

*** The new function `call-process-shell-command' executes a shell
command command synchronously in a separate process.

*** The new function `process-file' is similar to `call-process', but
obeys file handlers.  The file handler is chosen based on
default-directory.

*** The new function `set-process-filter-multibyte' sets the
multibyteness of a string given to a process's filter.

*** The new function `process-filter-multibyte-p' returns t if a
string given to a process's filter is multibyte.

*** A filter function of a process is called with a multibyte string
if the filter's multibyteness is t.  That multibyteness is decided by
the value of `default-enable-multibyte-characters' when the process is
created and can be changed later by `set-process-filter-multibyte'.

*** If a process's coding system is raw-text or no-conversion and its
buffer is multibyte, the output of the process is at first converted
to multibyte by `string-to-multibyte' then inserted in the buffer.
Previously, it was converted to multibyte by `string-as-multibyte',
which was not compatible with the behavior of file reading.

+++
** Enhanced networking support.

*** There is a new `make-network-process' function which supports

*** New function network-interface-info.

This function returns the network address, hardware address, current
status, and other information about a specific network interface.

*** The sentinel is now called when a network process is deleted with
delete-process.  The status message passed to the sentinel for a
deleted network process is "deleted".  The message passed to the
sentinel when the connection is closed by the remote peer has been
changed to "connection broken by remote peer".

+++
** New function `force-window-update' can initiate a full redisplay of
one or all windows.  Normally, this is not needed as changes in window
contents are detected automatically.  However, certain implicit
changes to mode lines, header lines, or display properties may require
forcing an explicit window update.

+++
** The line-move, scroll-up, and scroll-down functions will now
modify the window vscroll to scroll through display rows that are
taller that the height of the window, for example in the presense of
large images.  To disable this feature, Lisp code may bind the new
variable `auto-window-vscroll' to nil.

+++
** Function `compute-motion' now calculates the usable window
width if the WIDTH argument is nil.  If the TOPOS argument is nil,
the usable window height and width is used.

+++
** Function pos-visible-in-window-p now returns the pixel coordinates
and partial visiblity state of the corresponding row, if the PARTIALLY
arg is non-nil.

** Changes in using window objects:

+++
*** You can now make a window as short as one line.

A window that is just one line tall does not display either a mode
line or a header line, even if the variables `mode-line-format' and
`header-line-format' call for them.  A window that is two lines tall
cannot display both a mode line and a header line at once; if the
variables call for both, only the mode line actually appears.

+++
*** The new function `window-inside-edges' returns the edges of the
actual text portion of the window, not including the scroll bar or
divider line, the fringes, the display margins, the header line and
the mode line.

+++
*** The new functions `window-pixel-edges' and `window-inside-pixel-edges'
return window edges in units of pixels, rather than columns and lines.

+++
*** The new macro `with-selected-window' temporarily switches the
selected window without impacting the order of buffer-list.

+++
*** `select-window' takes an optional second argument `norecord', like
`switch-to-buffer'.

+++
*** `save-selected-window' now saves and restores the selected window
of every frame.  This way, it restores everything that can be changed
by calling `select-window'.

+++
*** The function `set-window-buffer' now has an optional third argument
KEEP-MARGINS which will preserve the window's current margin, fringe,
and scroll-bar settings if non-nil.

+++
** Customizable fringe bitmaps

*** New function 'define-fringe-bitmap' can now be used to create new
fringe bitmaps, as well as change the built-in fringe bitmaps.

To change a built-in bitmap, do (require 'fringe) and use the symbol
identifing the bitmap such as `left-truncation or `continued-line'.

*** New function 'destroy-fringe-bitmap' may be used to destroy a
previously created bitmap, or restore a built-in bitmap.

*** New function 'set-fringe-bitmap-face' can now be used to set a
specific face to be used for a specific fringe bitmap.  The face is
automatically merged with the `fringe' face, so normally, the face
should only specify the foreground color of the bitmap.

*** There are new display properties, left-fringe and right-fringe,
that can be used to show a specific bitmap in the left or right fringe
bitmap of the display line.

Format is 'display '(left-fringe BITMAP [FACE]), where BITMAP is a
symbol identifying a fringe bitmap, either built-in or defined with
`define-fringe-bitmap', and FACE is an optional face name to be used
for displaying the bitmap instead of the default `fringe' face.
When specified, FACE is automatically merged with the `fringe' face.

*** New function `fringe-bitmaps-at-pos' returns the current fringe
bitmaps in the display line at a given buffer position.

+++
** Controlling the default left and right fringe widths.

The default left and right fringe widths for all windows of a frame
can now be controlled by setting the `left-fringe' and `right-fringe'
frame parameters to an integer value specifying the width in pixels.
Setting the width to 0 effectively removes the corresponding fringe.

The actual default fringe widths for the frame may deviate from the
specified widths, since the combined fringe widths must match an
integral number of columns.  The extra width is distributed evenly
between the left and right fringe.  For force a specific fringe width,
specify the width as a negative integer (if both widths are negative,
only the left fringe gets the specified width).

Setting the width to nil (the default), restores the default fringe
width which is the minimum number of pixels necessary to display any
of the currently defined fringe bitmaps.  The width of the built-in
fringe bitmaps is 8 pixels.

+++
** Per-window fringe and scrollbar settings

*** Windows can now have their own individual fringe widths and
position settings.

To control the fringe widths of a window, either set the buffer-local
variables `left-fringe-width', `right-fringe-width', or call
`set-window-fringes'.

To control the fringe position in a window, that is, whether fringes
are positioned between the display margins and the window's text area,
or at the edges of the window, either set the buffer-local variable
`fringes-outside-margins' or call `set-window-fringes'.

The function `window-fringes' can be used to obtain the current
settings.  To make `left-fringe-width', `right-fringe-width', and
`fringes-outside-margins' take effect, you must set them before
displaying the buffer in a window, or use `set-window-buffer' to force
an update of the display margins.

*** Windows can now have their own individual scroll-bar settings
controlling the width and position of scroll-bars.

To control the scroll-bar of a window, either set the buffer-local
variables `scroll-bar-mode' and `scroll-bar-width', or call
`set-window-scroll-bars'.  The function `window-scroll-bars' can be
used to obtain the current settings.  To make `scroll-bar-mode' and
`scroll-bar-width' take effect, you must set them before displaying
the buffer in a window, or use `set-window-buffer' to force an update
of the display margins.

+++
** When using non-toolkit scroll bars with the default width,
the scroll-bar-width frame parameter value is nil.

+++
** Multiple overlay arrows can now be defined and managed via the new
variable `overlay-arrow-variable-list'.  It contains a list of
varibles which contain overlay arrow position markers, including
the original `overlay-arrow-position' variable.

Each variable on this list may have individual `overlay-arrow-string'
and `overlay-arrow-bitmap' properties that specify an overlay arrow
string (for non-window terminals) or fringe bitmap (for window
systems) to display at the corresponding overlay arrow position.
If either property is not set, the default `overlay-arrow-string' or
'overlay-arrow-fringe-bitmap' will be used.

+++
** New line-height and line-spacing properties for newline characters

A newline may now have line-height and line-spacing text or overlay
properties that control the height of the corresponding display row.

If the line-height property value is t, the newline does not
contribute to the height of the display row; instead the height of the
newline glyph is reduced.  Also, a line-spacing property on this
newline is ignored.  This can be used to tile small images or image
slices without adding blank areas between the images.

If the line-height property value is a positive integer, the value
specifies the minimum line height in pixels.  If necessary, the line
height it increased by increasing the line's ascent.

If the line-height property value is a float, the minimum line height
is calculated by multiplying the default frame line height by the
given value.

If the line-height property value is a cons (FACE . RATIO), the
minimum line height is calculated as RATIO * height of named FACE.
RATIO is int or float.  If FACE is t, it specifies the current face.

If the line-height property value is a cons (nil . RATIO), the line
height is calculated as RATIO * actual height of the line's contents.

If the line-height value is a cons (HEIGHT . TOTAL), HEIGHT specifies
the line height as described above, while TOTAL is any of the forms
described above and specifies the total height of the line, causing a
varying number of pixels to be inserted after the line to make it line
exactly that many pixels high.

If the line-spacing property value is an positive integer, the value
is used as additional pixels to insert after the display line; this
overrides the default frame line-spacing and any buffer local value of
the line-spacing variable.

If the line-spacing property may be a float or cons, the line spacing
is calculated as specified above for the line-height property.

+++
** The buffer local line-spacing variable may now have a float value,
which is used as a height relative to the default frame line height.

+++
** Enhancements to stretch display properties

The display property stretch specification form `(space PROPS)', where
PROPS is a property list now allows pixel based width and height
specifications, as well as enhanced horizontal text alignment.

The value of these properties can now be a (primitive) expression
which is evaluated during redisplay.  The following expressions
are supported:

EXPR ::= NUM | (NUM) | UNIT | ELEM | POS | IMAGE | FORM
NUM  ::= INTEGER | FLOAT | SYMBOL
UNIT ::= in | mm | cm | width | height
ELEM ::= left-fringe | right-fringe | left-margin | right-margin
      |  scroll-bar | text
POS  ::= left | center | right
FORM ::= (NUM . EXPR) | (OP EXPR ...)
OP   ::= + | -

The form `NUM' specifies a fractional width or height of the default
frame font size.  The form `(NUM)' specifies an absolute number of
pixels.  If a symbol is specified, its buffer-local variable binding
is used.  The `in', `mm', and `cm' units specifies the number of
pixels per inch, milli-meter, and centi-meter, resp.  The `width' and
`height' units correspond to the width and height of the current face
font.  An image specification corresponds to the width or height of
the image.

The `left-fringe', `right-fringe', `left-margin', `right-margin',
`scroll-bar', and `text' elements specify to the width of the
corresponding area of the window.

The `left', `center', and `right' positions can be used with :align-to
to specify a position relative to the left edge, center, or right edge
of the text area.  One of the above window elements (except `text')
can also be used with :align-to to specify that the position is
relative to the left edge of the given area.  Once the base offset for
a relative position has been set (by the first occurrence of one of
these symbols), further occurences of these symbols are interpreted as
the width of the area.

For example, to align to the center of the left-margin, use
    :align-to (+ left-margin (0.5 . left-margin))

If no specific base offset is set for alignment, it is always relative
to the left edge of the text area.  For example, :align-to 0 in a
header-line aligns with the first text column in the text area.

The value of the form `(NUM . EXPR)' is the value of NUM multiplied by
the value of the expression EXPR.  For example, (2 . in) specifies a
width of 2 inches, while (0.5 . IMAGE) specifies half the width (or
height) of the specified image.

The form `(+ EXPR ...)' adds up the value of the expressions.
The form `(- EXPR ...)' negates or subtracts the value of the expressions.

+++
** Support for displaying image slices

*** New display property (slice X Y WIDTH HEIGHT) may be used with
an image property to display only a specific slice of the image.

*** Function insert-image has new optional fourth arg to
specify image slice (X Y WIDTH HEIGHT).

*** New function insert-sliced-image inserts a given image as a
specified number of evenly sized slices (rows x columns).

+++
** Images may now have an associated image map via the :map property.

An image map is an alist where each element has the format (AREA ID PLIST).
An AREA is specified as either a rectangle, a circle, or a polygon:
A rectangle is a cons (rect . ((x0 . y0) . (x1 . y1))) specifying the
pixel coordinates of the upper left and bottom right corners.
A circle is a cons (circle . ((x0 . y0) . r)) specifying the center
and the radius of the circle; r may be a float or integer.
A polygon is a cons (poly . [x0 y0 x1 y1 ...]) where each pair in the
vector describes one corner in the polygon.

When the mouse pointer is above a hot-spot area of an image, the
PLIST of that hot-spot is consulted; if it contains a `help-echo'
property it defines a tool-tip for the hot-spot, and if it contains
a `pointer' property, it defines the shape of the mouse cursor when
it is over the hot-spot.  See the variable 'void-area-text-pointer'
for possible pointer shapes.

When you click the mouse when the mouse pointer is over a hot-spot,
an event is composed by combining the ID of the hot-spot with the
mouse event, e.g. [area4 mouse-1] if the hot-spot's ID is `area4'.

+++ (lispref)
??? (man)
** The mouse pointer shape in void text areas (i.e. after the end of a
line or below the last line in the buffer) of the text window is now
controlled by the new variable `void-text-area-pointer'.  The default
is to use the `arrow' (non-text) pointer.  Other choices are `text'
(or nil), `hand', `vdrag', `hdrag', `modeline', and `hourglass'.

+++
** The mouse pointer shape over an image can now be controlled by the
:pointer image property.

+++
** Lisp code can now test if a given buffer position is inside a
clickable link with the new function `mouse-on-link-p'.  This is the
function used by the new `mouse-1-click-follows-link' functionality.

+++
** The mouse pointer shape over ordinary text or images may now be
controlled/overriden via the `pointer' text property.

** Mouse event enhancements:

+++
*** Mouse clicks on fringes now generates left-fringe or right-fringes
events, rather than a text area click event.

+++
*** Mouse clicks in the left and right marginal areas now includes a
sensible buffer position corresponding to the first character in the
corresponding text row.

+++
*** Function `mouse-set-point' now works for events outside text area.

+++
*** Mouse events now includes buffer position for all event types.

+++
*** `posn-point' now returns buffer position for non-text area events.

+++
*** New function `posn-area' returns window area clicked on (nil means
text area).

+++
*** Mouse events include actual glyph column and row for all event types.

+++
*** New function `posn-actual-col-row' returns actual glyph coordinates.

+++
*** Mouse events may now include image object in addition to string object.

+++
*** Mouse events include relative x and y pixel coordinates relative to
the top left corner of the object (image or character) clicked on.

+++
*** Mouse events include the pixel width and height of the object
(image or character) clicked on.

+++
*** New functions 'posn-object', 'posn-object-x-y', and
'posn-object-width-height' return the image or string object of a mouse
click, the x and y pixel coordinates relative to the top left corner
of that object, and the total width and height of that object.

+++
** At the end of a command, point moves out from within invisible
text, in the same way it moves out from within text covered by an
image or composition property.

This makes it generally unnecessary to mark invisible text as intangible.
This is particularly good because the intangible property often has
unexpected side-effects since the property applies to everything
(including `goto-char', ...) whereas this new code is only run after
post-command-hook and thus does not care about intermediate states.

+++
** Normally, the cursor is displayed at the end of any overlay and
text property string that may be present at the current window
position.  The cursor may now be placed on any character of such
strings by giving that character a non-nil `cursor' text property.

+++
** The display space :width and :align-to text properties are now
supported on text terminals.

+++
** Arguments for remove-overlays are now optional, so that you can
remove all overlays in the buffer by just calling (remove-overlay).

+++
** New variable char-property-alias-alist.

This variable allows you to create alternative names for text
properties.  It works at the same level as `default-text-properties',
although it applies to overlays as well.  This variable was introduced
to implement the `font-lock-face' property.

+++
** New function `get-char-property-and-overlay' accepts the same
arguments as `get-char-property' and returns a cons whose car is the
return value of `get-char-property' called with those arguments and
whose cdr is the overlay in which the property was found, or nil if
it was found as a text property or not found at all.

+++
** The new frame parameter `tty-color-mode' specifies the mode to use
for color support on character terminal frames.  Its value can be a
number of colors to support, or a symbol.  See the Emacs Lisp
Reference manual for more detailed documentation.

+++
** The new face attribute `min-colors' can be used to tailor the face
color to the number of colors supported by a display, and define the
foreground and background colors accordingly so that they look best on
a terminal that supports at least this many colors.  This is now the
preferred method for defining default faces in a way that makes a good
use of the capabilities of the display.

+++
** (char-displayable-p CHAR) returns non-nil if Emacs ought to be able
to display CHAR.  More precisely, if the selected frame's fontset has
a font to display the character set that CHAR belongs to.

Fontsets can specify a font on a per-character basis; when the fontset
does that, this value may not be accurate.

+++
** New function display-supports-face-attributes-p may be used to test
whether a given set of face attributes is actually displayable.

A new predicate `supports' has also been added to the `defface' face
specification language, which can be used to do this test for faces
defined with defface.

---
** The special treatment of faces whose names are of the form `fg:COLOR'
or `bg:COLOR' has been removed.  Lisp programs should use the
`defface' facility for defining faces with specific colors, or use
the feature of specifying the face attributes :foreground and :background
directly in the `face' property instead of using a named face.

+++
** The first face specification element in a defface can specify
`default' instead of frame classification.  Then its attributes act as
defaults that apply to all the subsequent cases (and may be overridden
by them).

+++
** The variable `face-font-rescale-alist' specifies how much larger
(or smaller) font we should use.  For instance, if the value is
'((SOME-FONTNAME-PATTERN . 1.3)) and a face requests a font of 10
point, we actually use a font of 13 point if the font matches
SOME-FONTNAME-PATTERN.

---
** `set-fontset-font', `fontset-info', `fontset-font' now operate on
the default fontset if the argument NAME is nil..

---
** The function face-differs-from-default-p now truly checks whether the
given face displays differently from the default face or not (previously
it did only a very cursory check).

+++
** face-attribute, face-foreground, face-background, and face-stipple now
accept a new optional argument, INHERIT, which controls how face
inheritance is used when determining the value of a face attribute.

+++
** New functions face-attribute-relative-p and merge-face-attribute
help with handling relative face attributes.

+++
** The priority of faces in an :inherit attribute face-list is reversed.
If a face contains an :inherit attribute with a list of faces, earlier
faces in the list override later faces in the list; in previous releases
of Emacs, the order was the opposite.  This change was made so that
:inherit face-lists operate identically to face-lists in text `face'
properties.

+++
** New standard font-lock face `font-lock-preprocessor-face'.

+++
** New special text property `font-lock-face'.

This property acts like the `face' property, but it is controlled by
M-x font-lock-mode.  It is not, strictly speaking, a builtin text
property.  Instead, it is implemented inside font-core.el, using the
new variable `char-property-alias-alist'.

---
** jit-lock obeys a new text-property `jit-lock-defer-multiline'.
If a piece of text with that property gets contextually refontified
(see jit-lock-defer-contextually), then all of that text will
be refontified.  This is useful when the syntax of a textual element
depends on text several lines further down (and when font-lock-multiline
is not appropriate to solve that problem).  For example in Perl:

	s{
		foo
	}{
		bar
	}e

Adding/removing the last `e' changes the `bar' from being a piece of
text to being a piece of code, so you'd put a jit-lock-defer-multiline
property over the second half of the command to force (deferred)
refontification of `bar' whenever the `e' is added/removed.

+++
** font-lock can manage arbitrary text-properties beside `face'.
*** the FACENAME returned in font-lock-keywords can be a list
of the form (face FACE PROP1 VAL1 PROP2 VAL2 ...) so you can set
other properties than `face'.
*** font-lock-extra-managed-props can be set to make sure those extra
properties are automatically cleaned up by font-lock.

---
** The precedence of file-name-handlers has been changed.
Instead of blindly choosing the first handler that matches,
find-file-name-handler now gives precedence to a file-name handler
that matches near the end of the file name.  More specifically, the
handler whose (match-beginning 0) is the largest is chosen.
In case of ties, the old "first matched" rule applies.

+++
** A file name handler can declare which operations it handles.

You do this by putting an `operation' property on the handler name
symbol.  The property value should be a list of the operations that
the handler really handles.  It won't be called for any other
operations.

This is useful for autoloaded handlers, to prevent them from being
autoloaded when not really necessary.

+++
** `set-auto-mode' now gives the interpreter magic line (if present)
precedence over the file name.  Likewise an <?xml or <!DOCTYPE declaration
will give the buffer XML or SGML mode, based on the new var
`magic-mode-alist'.

+++
** Major mode functions now run the new normal hook
`after-change-major-mode-hook', at their very end, after the mode hooks.

---
** If a major mode function has a non-nil `no-clone-indirect'
property, `clone-indirect-buffer' signals an error if you use
it in that buffer.

+++
** Major modes can define `eldoc-documentation-function'
locally to provide Eldoc functionality by some method appropriate to
the language.

+++
** define-derived-mode by default creates a new empty abbrev table.
It does not copy abbrevs from the parent mode's abbrev table.

+++
** define-minor-mode now accepts arbitrary additional keyword arguments
and simply passes them to defcustom, if applicable.

+++
** The new function `run-mode-hooks' and the new macro `delay-mode-hooks'
are used by define-derived-mode to make sure the mode hook for the
parent mode is run at the end of the child mode.

+++
** `minor-mode-list' now holds a list of minor mode commands.

+++
** Both the variable and the function `disabled-command-hook' have
been renamed to `disabled-command-function'.  The variable
`disabled-command-hook' has been kept as an obsolete alias.

+++
** The function `eql' is now available without requiring the CL package.

+++
** If optional third argument APPEND to `add-to-list' is non-nil, a
new element gets added at the end of the list instead of at the
beginning.  This change actually occurred in Emacs-21.1, but was not
documented.

+++
** The escape sequence \s is now interpreted as a SPACE character,
unless it is followed by a `-' in a character constant (e.g. ?\s-A),
in which case it is still interpreted as the super modifier.
In strings, \s is always interpreted as a space.

+++
** A hex escape in a string forces the string to be multibyte.
An octal escape makes it unibyte.

+++
** `split-string' now includes null substrings in the returned list if
the optional argument SEPARATORS is non-nil and there are matches for
SEPARATORS at the beginning or end of the string.  If SEPARATORS is
nil, or if the new optional third argument OMIT-NULLS is non-nil, all
empty matches are omitted from the returned list.

+++
** New function `string-to-multibyte' converts a unibyte string to a
multibyte string with the same individual character codes.

+++
** The function `number-sequence' returns a list of equally-separated
numbers.  For instance, (number-sequence 4 9) returns (4 5 6 7 8 9).
By default, the separation is 1, but you can specify a different separation
as the third argument.  (number-sequence 1.5 6 2) returns (1.5 3.5 5.5).

+++
** `sit-for' can now be called with args (SECONDS &optional NODISP).

+++
** A function's docstring can now hold the function's usage info on
its last line.  It should match the regexp "\n\n(fn.*)\\'".

+++
** The `defmacro' form may contain declarations specifying how to
indent the macro in Lisp mode and how to debug it with Edebug.  The
syntax of defmacro has been extended to

   (defmacro NAME LAMBDA-LIST [DOC-STRING] [DECLARATION ...] ...)

DECLARATION is a list `(declare DECLARATION-SPECIFIER ...)'.  The
declaration specifiers supported are:

(indent INDENT)
	Set NAME's `lisp-indent-function' property to INDENT.

(edebug DEBUG)
	Set NAME's `edebug-form-spec' property to DEBUG.  (This is
	equivalent to writing a `def-edebug-spec' for the macro.

+++
** The functions all-completions and try-completion now accept lists
of strings as well as hash-tables additionally to alists, obarrays
and functions.  Furthermore, the function `test-completion' is now
exported to Lisp.  The keys in alists and hash tables may be either
strings or symbols, which are automatically converted with to strings.

+++
** The new macro dynamic-completion-table supports using functions
as a dynamic completion table.

  (dynamic-completion-table FUN)

FUN is called with one argument, the string for which completion is required,
and it should return an alist containing all the intended possible
completions. This alist may be a full list of possible completions so that FUN
can ignore the value of its argument. If completion is performed in the
minibuffer, FUN will be called in the buffer from which the minibuffer was
entered. dynamic-completion-table then computes the completion.

+++
** The new macro lazy-completion-table initializes a variable
as a lazy completion table.

  (lazy-completion-table VAR FUN &rest ARGS)

If the completion table VAR is used for the first time (e.g., by passing VAR
as an argument to `try-completion'), the function FUN is called with arguments
ARGS. FUN must return the completion table that will be stored in VAR. If
completion is requested in the minibuffer, FUN will be called in the buffer
from which the minibuffer was entered. The return value of
`lazy-completion-table' must be used to initialize the value of VAR.

+++
** `load-history' can now have elements of the form (t . FUNNAME),
which means FUNNAME was previously defined as an autoload (before the
current file redefined it).

+++
** `load-history' now records (defun . FUNNAME) when a function is
defined.  For a variable, it records just the variable name.

---
** The variable `recursive-load-depth-limit' has been deleted.
Emacs now signals an error if the same file is loaded with more
than 3 levels of nesting.

+++
** The function symbol-file can now search specifically for function or
variable definitions.

+++
** `provide' and `featurep' now accept an optional second argument
to test/provide subfeatures.  Also `provide' now checks `after-load-alist'
and runs any code associated with the provided feature.

+++
** `declare' is now a macro.  This change was made mostly for
documentation purposes and should have no real effect on Lisp code.

+++
** Byte compiler changes:

*** The byte-compiler now displays the actual line and character
position of errors, where possible.  Additionally, the form of its
warning and error messages have been brought more in line with the
output of other GNU tools.

*** The new macro `with-no-warnings' suppresses all compiler warnings
inside its body.  In terms of execution, it is equivalent to `progn'.

*** You can avoid warnings for possibly-undefined symbols with a
simple convention that the compiler understands.  (This is mostly
useful in code meant to be portable to different Emacs versions.)
Write forms like the following, or code that macroexpands into such
forms:

  (if (fboundp 'foo) <then> <else>)
  (if (boundp 'foo) <then> <else)

In the first case, using `foo' as a function inside the <then> form
won't produce a warning if it's not defined as a function, and in the
second case, using `foo' as a variable won't produce a warning if it's
unbound.  The test must be in exactly one of the above forms (after
macro expansion), but such tests may be nested.  Note that `when' and
`unless' expand to `if', but `cond' doesn't.

*** `(featurep 'xemacs)' is treated by the compiler as nil.  This
helps to avoid noisy compiler warnings in code meant to run under both
Emacs and XEmacs and may sometimes make the result significantly more
efficient.  Since byte code from recent versions of XEmacs won't
generally run in Emacs and vice versa, this optimization doesn't lose
you anything.

*** The local variable `no-byte-compile' in elisp files is now obeyed.

---
** When a Lisp file uses CL functions at run-time, compiling the file
now issues warnings about these calls, unless the file performs
(require 'cl) when loaded.

+++
** New function unsafep returns nil if the given Lisp form can't possibly
do anything dangerous; otherwise it returns a reason why the form might be
unsafe (calls dangerous function, alters global variable, etc).

+++
** There is a new Warnings facility; see the functions `warn'
and `display-warning'.

---
** VC changes for backends:
*** (vc-switches BACKEND OPERATION) is a new function for use by backends.
*** The new `find-version' backend function replaces the `destfile'
parameter of the `checkout' backend function.
Old code still works thanks to a default `find-version' behavior that
uses the old `destfile' parameter.

+++
** Already true in Emacs 21.1, but not emphasized clearly enough:

Multibyte buffers can now faithfully record all 256 character codes
from 0 to 255.  As a result, most of the past reasons to use unibyte
buffers no longer exist.  We only know of three reasons to use them
now:

1. If you prefer to use unibyte text all of the time.

2. For reading files into temporary buffers, when you want to avoid
the time it takes to convert the format.

3. For binary files where format conversion would be pointless and
wasteful.

---
** set-buffer-file-coding-system now takes an additional argument,
NOMODIFY.  If it is non-nil, it means don't mark the buffer modified.

+++
** The new variable `auto-coding-functions' lets you specify functions
to examine a file being visited and deduce the proper coding system
for it.  (If the coding system is detected incorrectly for a specific
file, you can put a `coding:' tags to override it.)

---
** The new function `merge-coding-systems' fills in unspecified aspects
of one coding system from another coding system.

---
** New coding system property `mime-text-unsuitable' indicates that
the coding system's `mime-charset' is not suitable for MIME text
parts, e.g. utf-16.

+++
** New function `decode-coding-inserted-region' decodes a region as if
it is read from a file without decoding.

+++
** Function `translate-region' accepts also a char-table as TABLE
argument.

+++
** The new translation table `translation-table-for-input'
is used for customizing self-insertion.  The character to
be inserted is translated through it.

---
** New CCL functions `lookup-character' and `lookup-integer' access
hash tables defined by the Lisp function `define-translation-hash-table'.

+++
** The flags, width, and precision options for %-specifications in function
`format' are now documented.  Some flags that were accepted but not
implemented (such as "*") are no longer accepted.

---
** New function `redirect-debugging-output' can be used to redirect
debugging output on the stderr file handle to a file.

+++
** `makehash' is now obsolete.  Use `make-hash-table' instead.

+++
** The macro `with-syntax-table' does not copy the table any more.

+++
** New variables `gc-elapsed' and `gcs-done' provide extra information
on garbage collection.

+++
** New function `locale-info' accesses locale information.

+++
** The new variable `print-continuous-numbering', when non-nil, says
that successive calls to print functions should use the same
numberings for circular structure references.  This is only relevant
when `print-circle' is non-nil.

When you bind `print-continuous-numbering' to t, you should
also bind `print-number-table' to nil.

---
** When pure storage overflows while dumping, Emacs now prints how
much pure storage it will approximately need.

+++
** File local variables.

A file local variables list cannot specify a string with text
properties--any specified text properties are discarded.

+++
** The variable `safe-local-eval-forms' specifies a list of forms that
are ok to evaluate when they appear in an `eval' local variables
specification.  Normally Emacs asks for confirmation before evaluating
such a form, but if the form appears in this list, no confirmation is
needed.

---
** If a function has a non-nil `safe-local-eval-function' property,
that means it is ok to evaluate some calls to that function when it
appears in an `eval' local variables specification.  If the property
is t, then any form calling that function with constant arguments is
ok.  If the property is a function or list of functions, they are called
with the form as argument, and if any returns t, the form is ok to call.

If the form is not "ok to call", that means Emacs asks for
confirmation as before.

+++
** Renamed hooks to better follow the naming convention:
find-file-hooks to find-file-hook,
find-file-not-found-hooks to find-file-not-found-functions,
write-file-hooks to write-file-functions,
write-contents-hooks to write-contents-functions,
x-lost-selection-hooks to x-lost-selection-functions,
x-sent-selection-hooks to x-sent-selection-functions.
Marked local-write-file-hooks as obsolete (use the LOCAL arg of `add-hook').

+++
** The new variable `delete-frame-functions' replaces `delete-frame-hook'.
It was renamed to follow the naming conventions for abnormal hooks.  The old
name remains available as an alias, but has been marked obsolete.

+++
** The new function `file-remote-p' tests a file name and returns
non-nil if it specifies a remote file (one that Emacs accesses using
its own special methods and not directly through the file system).
The value in that case is an identifier for the remote file system.

+++
** Functions `get' and `plist-get' no longer signals an error for
a malformed property list.  They also detect cyclic lists.

+++
** The function `atan' now accepts an optional second argument.

When called with 2 arguments, as in `(atan Y X)', `atan' returns the
angle in radians between the vector [X, Y] and the X axis.  (This is
equivalent to the standard C library function `atan2'.)

+++
** New function format-mode-line.

This returns the mode-line or header-line of the selected (or a
specified) window as a string with or without text properties.

+++
** The new mode-line construct `(:propertize ELT PROPS...)' can be
used to add text properties to mode-line elements.

+++
** The new `%i' and `%I' constructs for `mode-line-format' can be used
to display the size of the accessible part of the buffer on the mode
line.

---
** Indentation of simple and extended loop forms has been added to the
cl-indent package.  The new user options
`lisp-loop-keyword-indentation', `lisp-loop-forms-indentation', and
`lisp-simple-loop-indentation' can be used to customize the
indentation of keywords and forms in loop forms.

---
** Indentation of backquoted forms has been made customizable in the
cl-indent package.  See the new user option `lisp-backquote-indentation'.

+++
** field-beginning and field-end now accept an additional optional
argument, LIMIT.

+++
** define-abbrev now accepts an optional argument SYSTEM-FLAG.  If
non-nil, this marks the abbrev as a "system" abbrev, which means that
it won't be stored in the user's abbrevs file if he saves the abbrevs.
Major modes that predefine some abbrevs should always specify this
flag.

---
** describe-vector now takes a second argument `describer' which is
called to print the entries' values.  It defaults to `princ'.

+++
** defcustom and other custom declarations now use a default group
(the last prior group defined in the same file) when no :group was given.

---
** The new customization type `float' specifies numbers with floating
point (no integers are allowed).

+++
** emacsserver now runs pre-command-hook and post-command-hook when
it receives a request from emacsclient.

---
** New function `text-clone-create'.  Text clones are chunks of text
that are kept identical by transparently propagating changes from one
clone to the other.

+++
** Functions `user-uid' and `user-real-uid' now return floats if the
user UID doesn't fit in a Lisp integer.  Function `user-full-name'
accepts a float as UID parameter.

+++
** New vars `exec-suffixes' and `load-suffixes' used when
searching for an executable resp. an elisp file.

+++
** Functions from `post-gc-hook' are run at the end of garbage
collection.  The hook is run with GC inhibited, so use it with care.

+++
** The variables most-positive-fixnum and most-negative-fixnum
hold the largest and smallest possible integer values.

---
** On MS Windows, locale-coding-system is used to interact with the OS.
The Windows specific variable w32-system-coding-system, which was
formerly used for that purpose is now an alias for locale-coding-system.

---
** New function x-send-client-message sends a client message when
running under X.

---
** easy-mmode-define-global-mode has been renamed to
define-global-minor-mode.  The old name remains as an alias.

---
** To manipulate the File menu using easy-menu, you must specify the
proper name "file".  In previous Emacs versions, you had to specify
"files", even though the menu item itself was changed to say "File"
several versions ago.

---
** The dummy function keys made by easymenu are now always lower case.
If you specify the menu item name "Ada", for instance, it uses `ada'
as the "key" bound by that key binding.

This is relevant only if Lisp code looks for the bindings that were
made with easymenu.

---
** `easy-menu-define' now allows you to use nil for the symbol name
if you don't need to give the menu a name.  If you install the menu
into other keymaps right away (MAPS is non-nil), it usually doesn't
need to have a name.

---
** rx.el has new corresponding `symbol-end' and `symbol-start' elements.

** New functions, macros, and commands:

+++
*** The new function `assoc-string' replaces `assoc-ignore-case' and
`assoc-ignore-representation', which are still available, but have
been declared obsolete.

+++
*** The new function `buffer-local-value' returns the buffer-local
binding of VARIABLE (a symbol) in buffer BUFFER.  If VARIABLE does not
have a buffer-local binding in buffer BUFFER, it returns the default
value of VARIABLE instead.

+++
*** The new function `called-interactively-p' does what many people
have mistakenly believed `interactive-p' did: it returns t if the
calling function was called through `call-interactively'.  This should
only be used when you cannot add a new "interactive" argument to the
command.

*** The new function copy-abbrev-table returns a new abbrev table that
is a copy of a given abbrev table.

+++
*** New function copy-tree makes a copy of a tree, recursively copying
both cars and cdrs.

+++
*** New function `delete-dups' destructively removes `equal'
duplicates from a list.  Of several `equal' occurrences of an element
in the list, the first one is kept.

+++
*** The new function `filter-buffer-substring' extracts a buffer
substring, passes it through a set of filter functions, and returns
the filtered substring.  It is used instead of `buffer-substring' or
`delete-and-extract-region' when copying text into a user-accessible
data structure, like the kill-ring, X clipboard, or a register.  The
list of filter function is specified by the new variable
`buffer-substring-filters'.  For example, Longlines mode uses
`buffer-substring-filters' to remove soft newlines from the copied
text.

+++
*** New functions frame-current-scroll-bars and window-current-scroll-bars.

These functions return the current locations of the vertical and
horizontal scroll bars in a frame or window.

+++
*** The new primitive `get-internal-run-time' returns the processor
run time used by Emacs since start-up.

+++
*** The new function insert-buffer-substring-as-yank works like
`insert-buffer-substring', but removes the text properties in the
`yank-excluded-properties' list.

+++
*** The new function insert-buffer-substring-no-properties is like
insert-buffer-substring, but removes all text properties from the
inserted substring.

+++
*** The new functions `lax-plist-get' and `lax-plist-put' are like
`plist-get' and `plist-put', except that they compare the property
name using `equal' rather than `eq'.

+++
*** New function `line-number-at-pos' returns the line number of the
current line in the current buffer, or if optional buffer position is
given, line number of corresponding line in current buffer.

+++
*** New function `looking-back' checks whether a regular expression matches
the text before point.  Specifying the LIMIT argument bounds how far
back the match can start; this is a way to keep it from taking too long.

+++
*** New function `macroexpand-all' expands all macros in a form.
It is similar to the Common-Lisp function of the same name.
One difference is that it guarantees to return the original argument
if no expansion is done, which may be tested using `eq'.

*** The new function `minibufferp' returns non-nil if its optional
buffer argument is a minibuffer.  If the argument is omitted, it
defaults to the current buffer.

+++
*** New function minibuffer-selected-window returns the window which
was selected when entering the minibuffer.

+++
*** The new function `modify-all-frames-parameters' modifies parameters
for all (existing and future) frames.

+++
*** New functions posn-at-point and posn-at-x-y return
click-event-style position information for a given visible buffer
position or for a given window pixel coordinate.

---
*** New function quail-find-key returns a list of keys to type in the
current input method to input a character.

+++
*** The new function `rassq-delete-all' deletes all elements from an
alist whose cdr is `eq' to a specified value.

+++
*** The new function remove-list-of-text-properties is almost the same
as `remove-text-properties'.  The only difference is that it takes a
list of property names as argument rather than a property list.

+++
*** The new primitive `set-file-times' sets a file's access and
modification times.  Magic file name handlers can handle this
operation.

+++
*** New function substring-no-properties returns a substring without
text properties.

+++
*** The new function syntax-after returns the syntax code
of the character after a specified buffer position, taking account
of text properties as well as the character code.

+++
*** `syntax-class' extracts the class of a syntax code (as returned
by syntax-after).

+++
*** New function `tool-bar-local-item-from-menu'

The `tool-bar-add-item-from-menu' must not be used (as previously
recommended) for making entries in the tool bar for local keymaps.
Instead, use the function `tool-bar-local-item-from-menu', which lets
you specify the map to use as an argument.

+++
*** New function window-body-height.
This is like window-height but does not count the mode line
or the header line.

+++
*** (while-no-input BODY...) runs BODY, but only so long as no input
arrives.  If the user types or clicks anything, BODY stops as if a
quit had occurred.  while-no-input returns the value of BODY, if BODY
finishes.  It returns nil if BODY was aborted.

+++
*** New macro with-local-quit temporarily sets inhibit-quit to nil for use
around potentially blocking or long-running code in timers
and post-command-hooks.

** New packages:

*** The new package syntax.el provides an efficient way to find the
current syntactic context (as returned by parse-partial-sexp).

*** The new package bindat.el provides functions to unpack and pack
binary data structures, such as network packets, to and from Lisp
data structures.

---
*** The TCL package tcl-mode.el was replaced by tcl.el.
This was actually done in Emacs-21.1, and was not documented.

+++
*** The new package button.el implements simple and fast `clickable buttons'
in emacs buffers.  `buttons' are much lighter-weight than the `widgets'
implemented by widget.el, and can be used by lisp code that doesn't
require the full power of widgets.  Emacs uses buttons for such things
as help and apropos buffers.

---
*** master-mode.el implements a minor mode for scrolling a slave
buffer without leaving your current buffer, the master buffer.

It can be used by sql.el, for example: the SQL buffer is the master
and its SQLi buffer is the slave.  This allows you to scroll the SQLi
buffer containing the output from the SQL buffer containing the

(add-hook 'sql-set-sqli-hook
   (function (lambda ()
	       (master-set-slave sql-buffer))))

+++
*** New Lisp library testcover.el works with edebug to help you determine
whether you've tested all your Lisp code.  Function testcover-start
instruments all functions in a given file.  Then test your code.  Function
testcover-mark-all adds overlay "splotches" to the Lisp file's buffer to
show where coverage is lacking.  Command testcover-next-mark (bind it to
a key!) will move point forward to the next spot that has a splotch.

Normally, a red splotch indicates the form was never completely
evaluated; a brown splotch means it always evaluated to the same
value.  The red splotches are skipped for forms that can't possibly
complete their evaluation, such as `error'.  The brown splotches are
skipped for forms that are expected to always evaluate to the same
value, such as (setq x 14).

For difficult cases, you can add do-nothing macros to your code to
help out the test coverage tool.  The macro `noreturn' suppresses a
red splotch.  It is an error if the argument to `noreturn' does
return.  The macro 1value suppresses a brown splotch for its argument.
This macro is a no-op except during test-coverage -- then it signals
an error if the argument actually returns differing values.

---
** Support for Mocklisp has been removed.

---
** The function insert-string is now obsolete.

* Installation changes in Emacs 21.3

** Support for GNU/Linux on little-endian MIPS and on IBM S390 has
been added.