(shell-mode): Use define-derived-mode.

[emacs.git] / etc / ONEWS

GNU Emacs NEWS -- history of user-visible changes.  1992.
Copyright (C) 1995 Free Software Foundation, Inc.
See the end for copying conditions.

For older news, see the file ONEWS.4.
\f
* Editing Changes in Emacs 19.30.

** Be sure to recompile your byte-compiled Emacs Lisp files
if you last compiled them with Emacs 19.28 or earlier.
You can use M-x byte-force-recompile to recompile all the .elc files
in a specified directory.

** Emacs now provides multiple-frame support on Windows NT
and Windows 95.

** M-x column-number-mode toggles a minor mode which displays
the current column number in the mode line.

** Line Number mode is now enabled by default.

** M-x what-line now displays the line number in the accessible
portion of the buffer as well as the line number in the full buffer,
when narrowing is in effect.

** If you type a M-x command that has an equivalent key binding,
the equivalent is shown in the minibuffer before the command executes.
This feature is enabled by default for the sake of beginning users.
You can turn the feature off by setting suggest-key-bindings to nil.

** The menu bar is now visible on text-only terminals.  To choose a
command from the menu bar when you have no mouse, type M-`
(Meta-Backquote) or F10.  To turn off menu bar display,
do (menu-bar-mode -1).

** Whenever you invoke a minibuffer, it appears in the minibuffer
window that the current frame uses.

Emacs can only use one minibuffer window at a time.  If you activate
the minibuffer while a minibuffer window is active in some other
frame, the outer minibuffer window disappears while the inner one is
active.

** Echo area messages always appear in the minibuffer window that the
current frame uses.  If a minibuffer is active in some other frame,
the echo area message does not hide it even temporarily.

** The minibuffer now has a menu-bar menu.  You can use it to exit or
abort the minibuffer, or to ask for completion.

** Dead-key and composite character processing is done in the standard
X11R6 manner (through the default "input method" using the
/usr/lib/X11/locale/*/Compose databases of key combinations).  I.e. if
it works in xterm, it should also work in emacs now.

** Mouse changes

*** You can now use the mouse when running Emacs in an xterm.
Use M-x xterm-mouse-mode to let emacs take control over the mouse.

*** C-mouse-1 now once again provides a menu of buffers to select.
S-mouse-1 is now the way to select a default font for the frame.

*** There is a new mouse-scroll-min-lines variable to control the
minimum number of lines scrolled by dragging the mouse outside a
window's edge.

*** Dragging mouse-1 on a vertical line that separates windows
now moves the line, thus changing the widths of the two windows.
(This feature is available only if you don't have vertical scroll bars.
If you do use them, a scroll bar separates two side-by-side windows.)

*** Double-click mouse-1 on a character with "symbol" syntax (such as
underscore, in C mode) selects the entire symbol surrounding that
character.  (Double-click mouse-1 on a letter selects a whole word.)

** When incremental search wraps around to the beginning (or end) of
the buffer, if you keep on searching until you go past the original
starting point of the search, the echo area changes from "Wrapped" to
"Overwrapped".  That tells you that you are revisiting matches that
you have already seen.

** Filling changes.

*** If the variable colon-double-space is non-nil, the explicit fill
commands put two spaces after a colon.

*** Auto-Fill mode now supports Adaptive Fill mode just as the
explicit fill commands do.  The variable adaptive-fill-regexp
specifies a regular expression to match text at the beginning of
a line that should be the fill prefix.

*** Adaptive Fill mode can take a fill prefix from the first line of a
paragraph, *provided* that line is not a paragraph-starter line.

Paragraph-starter lines are indented lines that start a new
paragraph because they are indented.  This indentation shouldn't
be copied to additional lines.

Whether indented lines are paragraph lines depends on the value of the
variable paragraph-start.  Some major modes set this; you can set it
by hand or in mode hooks as well.  For editing text in which paragraph
first lines are not indented, and which contains paragraphs in which
all lines are indented, you should use Indented Text mode or arrange
for paragraph-start not to match these lines.

*** You can specify more complex ways of choosing a fill prefix
automatically by setting `adaptive-fill-function'.  This function
is called with point after the left margin of a line, and it should
return the appropriate fill prefix based on that line.
If it returns nil, that means it sees no fill prefix in that line.

** Gnus changes.

Gnus, the Emacs news reader, has been rewritten and expanded.  Most
things that worked with the old version should still work with the new
version.  Code that relies heavily on Gnus internals is likely to
fail, though.

*** Incompatibilities with the old GNUS.

**** All interactive commands have kept their names, but many internal
functions have changed names.

**** The summary mode gnus-uu commands have been moved from the `C-c
C-v' keymap to the `X' keymap.

**** There can now be several summary buffers active at once.
Variables that are relevant to each summary buffer are buffer-local to
that buffer.

**** Old hilit code doesn't work at all.  Gnus performs its own
highlighting based not only on what's visible in the buffer, but on
other data structures.

**** Old packages like `expire-kill' will no longer work.  

**** `C-c C-l' in the group buffer no longer switches to a different
buffer, but instead lists killed groups in the group buffer.

*** New features.

**** The look of all buffers can be changed by setting format-like
variables.
 
**** Local spool and several NNTP servers can be used at once.

**** Groups can be combined into virtual groups.

**** Different mail formats can be read much the same way as one would
read newsgroups.  All the mail backends implement mail expiry schemes.

**** Gnus can use various strategies for gathering threads that have
lost their roots (thereby gathering loose sub-threads into one thread)
or it can go back and retrieve enough headers to build a complete
thread.

**** Killed groups can be read.

**** Gnus can do partial group updates - you do not have to retrieve
the entire active file just to check for new articles in a few groups.

**** Gnus implements a sliding scale of subscribedness to groups.

**** You can score articles according to any number of criteria.  You
can get Gnus to score articles for you using adaptive scoring.

**** Gnus maintains a dribble buffer that is auto-saved the normal
Emacs manner, so it should be difficult to lose much data on what you
have read if your machine should go down.

**** Gnus now has its own startup file (`.gnus.el') to avoid
cluttering up the `.emacs' file.

**** You can set the process mark on both groups and articles and
perform operations on all the marked items.

**** You can grep through a subset of groups and create a group from
the results.

**** You can list subsets of groups using matches on group names or
group descriptions.

**** You can browse foreign servers and subscribe to groups from those
servers.

**** Gnus can pre-fetch articles asynchronously on a second connection
to the servers.

**** You can cache articles locally.

**** Gnus can fetch FAQs to and descriptions of groups.

**** Digests (and other files) can be used as the basis for groups.

**** Articles can be highlighted and customized.

** Changes to Version Control (VC)

*** General changes (all backends).

VC directory listings (C-x v d) are now kept up to date when you do a
vc-next-action (C-x v v) on the marked files.  The `g' command updates
the buffer properly.  `=' in a VC dired buffer produces a version
control diff, not an ordinary diff.

*** CVS changes.

Under CVS, you no longer need to type C-x C-q before you can edit a
file.  VC doesn't write-protect unmodified buffers anymore; you can
freely change them at any time.  The mode line keeps track of the
file status.

If you do want unmodified files to be write-protected, set your
CVSREAD environment variable.  VC sees this and behaves accordingly;
that will give you the behaviour of Emacs 19.29, similar to that under
RCS and SCCS.  In this mode, if the variable vc-mistrust-permissions
is nil, VC learns the modification state from the file permissions.
When setting CVSREAD for the first time, you should check out the
whole module anew, so that the file permissions are set correctly.

VC also works with remote repositories now.  When you visit a file, it
doesn't run "cvs status" anymore, so there shouldn't be any long delays.

Directory listings under VC/CVS have been enhanced.  Type C-x v d, and
you get a list of all files in or below the current directory that are
not up-to-date.  The actual status (modified, merge, conflict, ...) is
displayed for each file.  If you give a prefix argument (C-u C-x v d),
up-to-date files are also listed.  You can mark any number of files,
and execute the next logical version control command on them (C-x v v).

*** Starting a new branch.

If you try to lock a version that is not the latest on its branch, 
VC asks for confirmation in the minibuffer.  If you say no, it offers
to lock the latest version instead.

*** RCS non-strict locking.

VC can now handle RCS non-strict locking, too.  In this mode, working
files are always writable and you needn't lock the file before making
changes, similar to the default mode under CVS.  To enable non-strict
locking for a file, use the "rcs -U" command.

*** Sharing RCS master files.

If you share RCS subdirs with other users (through symbolic links),
and you always want to work on the latest version, set
vc-consult-headers to nil and vc-mistrust-permissions to `t'.
Then you see the state of the *latest* version on the mode line, not
that of your working file.  When you do a check out, VC overwrites
your working file with the latest version from the master.

*** RCS customization.

There is a new variable vc-consult-headers.  If it is t (the default),
VC searches for RCS headers in working files (like `$Id: ONEWS,v 1.3 2000/11/02 13:34:50 gerd Exp $') and
determines the state of the file from them, not from the master file.
This is fast and more reliable when you use branches.  (The variable
was already present in Emacs 19.29, but didn't get mentioned in the
NEWS.)

** Calendar changes.

*** New calendars supported: Chinese, Coptic, Ethiopic

Here are the commands for converting to and from these calendars:

   gC: calendar-goto-chinese-date
   gk: calendar-goto-coptic-date
   ge: calendar-goto-ethiopic-date

   pC: calendar-print-chinese-date
   pk: calendar-print-coptic-date
   pe: calendar-print-ethiopic-date

*** Printed calendars

Calendar mode now has commands to produce fancy printed calendars via
LaTeX.  You can ask for a calendar for one or more days, weeks, months
or years.  The commands all start with `t'; see the manual for a list
of them.

*** New sexp diary entry type

Reminders that apply in the days leading up to an event. 

** The CC-mode package now provides the default C and C++ modes.
See the manual for documentation of its features.

** The uniquify package chooses buffer names differently when you
visit multiple files with the same name (in different directories).

** RMAIL now always uses the movemail program when it renames an
inbox file, so that it can interlock properly with the mailer
no matter where it is delivering mail.

** tex-start-of-header and tex-end-of-header are now regular expressions,
not strings.

** To enable automatic uncompression of compressed files,
type M-x auto-compression-mode.  (This command used to be called
toggle-auto-compression, but was not documented before.)  In Lisp,
you can do

   (auto-compression-mode 1)

to turn the mode on.

** The new pc-select package emulates the key bindings for cutting and
pasting, and selection of regions, found in Windows, Motif, and the
Macintosh.

** Help buffers now use a special major mode, Help mode.  This mode
normally turns on View mode; it also provides a hook, help-mode-hook,
which you can use for other customization.

** Apropos now uses faces for enhanced legibility.  It now describes
symbol properties as well as their function definitions and variable
values.  You can use Mouse-2 or RET to get more information about a
function definition, variable, or property.

** Font Lock mode

*** Supports Scheme, TCL and Help modes

For example, to automatically turn on Font Lock mode in the *Help*
buffer, put:

 (add-hook 'help-mode-hook 'turn-on-font-lock)

in your ~/.emacs.

*** Enhanced fontification

The structure of font-lock-keywords is extended to allow "anchored" keywords.
Typically, a keyword item of font-lock-keywords comprises a regexp to search
for and information to specify how the regexp should be highlighted.  However,
the highlighting information is extended so that it can be another keyword
item.  This keyword item, its regexp and highlighting information, is processed
before resuming with the keyword item of which it is part.

For example, a typical keyword item might be:

 ("\\<anchor\\>" (0 anchor-face))

which fontifies each occurrence of the discrete word "anchor" in the value of
the variable anchor-face.  However, the highlighting information can be used to
fontify text that is anchored to the word "anchor".  For example:

 ("\\<anchor\\>" (0 anchor-face) ("\\=[ ,]*\\(item\\)" nil nil (1 item-face)))

which fontifies each occurrence of "anchor" as above, but for each occurrence
of "anchor", each occurrence of "item", in any following comma separated list,
is fontified in the value of the variable item-face.  Thus the "item" text is
anchored to the "anchor" text.  See the variable documentation for further
information.

This feature is used to extend the level and quality of fontification in a
number of modes.  For example, C/C++ modes now have level 3 decoration that
includes the fontification of variable and function names in declaration lists.
In this instance, the "anchor" described in the above example is a type or
class name, and an "item" is a variable or function name.

*** Fontification levels

The variables font-lock-maximum-decoration and font-lock-maximum-size are
extended to specify levels and sizes for specific modes.  The variable
font-lock-maximum-decoration specifies the preferred level of fontification for
modes that provide multiple levels (typically from "subdued" to "gaudy").  The
variable font-lock-maximum-size specifies the buffer size for which buffer
fontification is suppressed when Font Lock mode is turned on (typically because
it would take too long).

These variables can now specify values for individual modes, by supplying
lists of mode names and values.  For example, to use the above mentioned level
3 decoration for buffers in C/C++ modes, and default decoration otherwise, put:

 (setq font-lock-maximum-decoration '((c-mode . 3) (c++-mode . 3)))

in your ~/.emacs.  Maximum buffer size values for individual modes are
specified in the same way with the variable font-lock-maximum-size.

*** Font Lock configuration

The mechanism to provide default settings for Font Lock mode are the variables
font-lock-defaults and font-lock-maximum-decoration.  Typically, you should
only need to change the value of font-lock-maximum-decoration.  However, to
support Font Lock mode for buffers in modes that currently do not support Font
Lock mode, you should set a buffer local value of font-lock-defaults for that
mode, typically via its mode hook.

These variables are used by Font Lock mode to set the values of the variables
font-lock-keywords, font-lock-keywords-only, font-lock-syntax-table,
font-lock-beginning-of-syntax-function and font-lock-keywords-case-fold-search.

You need not set these variables directly, and should not set them yourself
since the underlining mechanism may change in future.

** Archive mode is now the default mode for various sorts of
archive files (files whose names end with .arc, .lzh, .zip, and .zoo).

** You can automatically update the years in copyright notice by
means of (add-hook 'write-file-hooks 'copyright-update).
Optionally it can update the GPL version as well.

** Scripts of various languages (Shell, AWK, Perl, makefiles ...) can
be automatically provided with a magic number and be made executable
by their respective modes under control of various user variables.
The mode must call (executable-set-magic "perl") or
(executable-set-magic "make" "-f").  The latter for example has no
effect on [Mm]akefile.

** Shell script mode now supports over 15 different shells.  The new
command C-c ! executes the region, and optionally beginning of script
as well, by passing them to the shell.

Cases such as `sh' being a `bash' are now accounted for.
Fontification now also does variables, the magic number and all
builtin commands.  Shell script mode no longer mingles `tab-width' and
indentation style.  The variable `sh-tab-width' has been renamed to
`sh-indentation'.  Empty lines are now indented like previous
non-empty line, rather than just previous line.

The annoying $ variable prompting has been eliminated.  Instead, shell
script mode uses `comint-dynamic-completion' for commands, variables
and filenames.

** Two-column mode now automatically scrolls both buffers together,
which makes it possible to eliminate the special scrolling commands
that used to do so.

The commands that operate in two-column mode are no longer bound to
keys outside that mode.  f2 o will now position at the same point in
associated buffer.

the new command f2 RET inserts a newline in both buffers, at point and
at the corresponding position in the associated buffer.

** Skeleton commands now work smoothly as abbrev definitions.  The
element < no longer exists, ' is a new element.

** The autoinsert insert facility for prefilling empty files as soon
as they are found has been extended to accommodate skeletons or calling
functions.  See the function auto-insert.

** TPU-edt Changes

Loading tpu-edt no longer turns on tpu-edt mode.  In fact, it is no
longer necessary to explicitly load tpu-edt.  All you need to do to
turn on tpu-edt is run the tpu-edt function.  Here's how to run
tpu-edt instead of loading the file:
  
  Running Emacs:   Type      emacs -f tpu-edt
                    not      emacs -l tpu-edt

  Within Emacs:    Type      M-x tpu-edt <ret>
                    not      M-x load-library <ret> tpu-edt <ret>
  
  In .emacs:       Use       (tpu-edt)
                   not       (load "tpu-edt")
  
The default name of the tpu-edt X key definition file has changed from
~/.tpu-gnu-keys to ~/.tpu-keys.  If you don't rename the file yourself,
tpu-edt will offer to rename it the first time you invoke it under
x-windows.

** MS-DOS Enhancements:

*** Better mouse control by adding the following functions [in dosfns.c]
msdos-mouse-enable, msdos-mouse-disable, msdos-mouse-init.

*** If another foreground/background color than the default is setup in
your ~/_emacs, then the screen briefly flickers with the default
colors before changing to the colors you have specified.  To avoid
this, the EMACSCOLORS environment variable exists.  It shall be
defined as a string with the following elements:
  
    set EMACSCOLORS=fb;fb
  
The first set of "fb" defines the initial foreground and background
colors using standard dos color numbers (0=black,.., 7=white).
If specified, the second set of "fb" defines the colors which are
restored when you leave emacs.
  
*** The new SUSPEND environment variable can now be set as the shell to
use when suspending emacs.  This can be used to override the stupid
limitation on the environment of sub-shells in MS-DOS (they are just
large enough to hold the currently defined variables, not leaving
room for more); to overcome this limitation, add this to autoexec.bat:
  
    set SUSPEND=%COMSPEC% /E:2000

** The escape character can now be displayed on X frames.  Try
this:
    (aset standard-display-table 27 (vector 27))
after first creating a display table (you can do that by loading
the disp-table library).

** The new command-line option --eval specifies an expression to evaluate
from the command line.

** etags has now the ability to tag Perl files.  They are recognised
either by the .pm and .pl suffixes or by a first line which starts
with `#!' and specifies a Perl interpreter.  The tagged lines are
those beginning with the `sub' keyword.

New suffixes recognised are .hpp for C++; .f90 for Fortran; .bib,
.ltx, .TeX for TeX (.bbl, .dtx removed); .ml for Lisp; .prolog for
prolog (.pl is now Perl).

** The files etc/termcap.dat and etc/termcap.ucb have been replaced
with a new, merged, and much more comprehensive termcap file.  The
new file should include all the special entries from the old one.
This new file is under active development as part of the ncurses
project.  If you have any questions about this file, or problems with
an entry in it, email terminfo@ccil.org.
\f
* Lisp changes in Emacs 19.30.

** New Data Types

*** There is a new data type called a char-table which is an array
indexed by a character.  Currently this is mostly equivalent to a
vector of length 256, but in the future, when a wider character set is
in use, it will be different.  To create one, call
   (make-char-table SUBTYPE INITIAL-VALUE)

SUBTYPE is a symbol that identifies the specific use of this
character table.  It can be any of these values:

  syntax-table
  display-table
  keyboard-translate-table
  case-table

The function `char-table-subtype' returns the subtype of a char-table.
You cannot alter the subtype of an existing char-table.

A char-table has an element for each character code.  It also has some
"extra slots".  The number of extra slots depends on the subtype and
their use depends on the subtype.  (Each subtype symbol has a
`char-table-extra-slots' property that says how many extra slots to
make.)  Use (char-table-extra-slot TABLE N) to access extra slot N and
(set-char-table-extra-slot TABLE N VALUE) to store VALUE in slot N.

A char-table T can have a parent, which should be another char-table
P.  If you look for the value in T for character C, and the table T
actually holds nil, P's element for character C is used instead.
The functions `char-table-parent' and `set-char-table-parent'
let you read or set the parent of a char-table.

To scan all the values in a char-table, do not try to loop through all
possible character codes.  That would work for now, but will not work
in the future.  Instead, call map-char-table.  (map-char-table
FUNCTION TABLE) calls FUNCTION once for each character or character
set that has a distinct value in TABLE.  FUNCTION gets two arguments,
RANGE and VALUE.  RANGE specifies a range of TABLE that has one
uniform value, and VALUE is the value in TABLE for that range.

Currently, RANGE is always a vector containing a single character
and it refers to that character alone.  In the future, other kinds
of ranges will occur.  You can set the value for a given range
with (set-char-table-range TABLE RANGE VALUE) and examine the value
for a range with (char-table-range TABLE RANGE).

*** Syntax tables are now represented as char-tables.
All syntax tables other than the standard syntax table
normally have the standard syntax table as their parent.
Their subtype is `syntax-table'.

*** Display tables are now represented as char-tables.
Their subtype is `display-table'.

*** Case tables are now represented as char-tables.
Their subtype is `case-table'.

*** The value of keyboard-translate-table may now be a char-table
instead of a string.  Normally the char-tables used for this purpose
have the subtype `keyboard-translate-table', but that is not required.

*** A new data type called a bool-vector is a vector of values
that are either t or nil.  To create one, do
   (make-bool-vector LENGTH INITIAL-VALUE)

** You can now specify, for each marker, how it should relocate when
text is inserted at the place where the marker points.  This is called
the "insertion type" of the marker.

To set the insertion type, do (set-marker-insertion-type MARKER TYPE).
If TYPE is t, it means the marker advances when text is inserted.  If
TYPE is nil, it means the marker does not advance.  (In Emacs 19.29,
markers did not advance.)

The function marker-insertion-type reports the insertion type of a
given marker.  The function copy-marker takes a second argument TYPE
which specifies the insertion type of the new copied marker.

** When you create an overlay, you can specify the insertion type of
the beginning and of the end.  To do this, you can use two new
arguments to make-overlay: front-advance and rear-advance.

** The new function overlays-in returns a list of the overlays that
overlap a specified range of the buffer.  The returned list includes
empty overlays at the beginning of this range, as well as within the
range.

** The new hook window-scroll-functions is run when a window has been
scrolled.  The functions in this list are called just before
redisplay, after the new window-start has been computed.  Each function
is called with two arguments--the window that has been scrolled, and its
new window-start position.

This hook is useful for on-the-fly fontification and other features
that affect how the redisplayed text will look when it is displayed.

The window-end value of the window is not valid when these functions
are called.  The computation of window-end is byproduct of actual
redisplay of the window contents, which means it has not yet happened
when the hook is run.  Computing window-end specially in advance for
the sake of these functions would cause a slowdown.

The hook functions can determine where the text on the window will end
by calling vertical-motion starting with the window-start position.

** The new hook redisplay-end-trigger-functions is run whenever
redisplay in window uses text that extends past a specified end
trigger position.  You set the end trigger position with the function
set-window-redisplay-end-trigger.  The functions are called with two
arguments: the window, and the end trigger position.  Storing nil for
the end trigger position turns off the feature, and the trigger value
is automatically reset to nil just after the hook is run.

You can use the function window-redisplay-end-trigger to read a
window's current end trigger value.

** The new function insert-file-contents-literally inserts the
contents of a file without any character set translation or decoding.

** The new function safe-length computes the length of a list.
It never gets an error--it treats any non-list like nil.
If given a circular list, it returns an upper bound for the number
of elements before the circularity.

** replace-match now takes a fifth argument, SUBEXP.  If SUBEXP is
non-nil, that says to replace just subexpression number SUBEXP of the
regexp that was matched, not the entire match.  For example, after
matching `foo \(ba*r\)' calling replace-match with 1 as SUBEXP means
to replace just the text that matched `\(ba*r\)'.

** The new keymap special-event-map defines bindings for certain
events that should be handled at a very low level--as soon as they
are read.  The read-event function processes these events itself,
and never returns them.

Events that are handled in this way do not echo, they are never
grouped into key sequences, and they never appear in the value of
last-command-event or (this-command-keys).  They do not discard a
numeric argument, they cannot be unread with unread-command-events,
they may not appear in a keyboard macro, and they are not recorded
in a keyboard macro while you are defining one.

These events do, however, appear in last-input-event immediately after
they are read, and this is the way for the event's definition to find
the actual event.

The events types iconify-frame, make-frame-visible and delete-frame
are normally handled in this way.

** encode-time now supports simple date arithmetic by means of
out-of-range values for its SEC, MINUTE, HOUR, DAY, and MONTH
arguments; for example, day 0 means the day preceding the given month.
Also, the ZONE argument can now be a TZ-style string.

** command-execute and call-interactively now accept an optional third
argument KEYS.  If specified and non-nil, this specifies the key
sequence containing the events that were used to invoke the command.

** The environment variable NAME, if set, now specifies the value of
(user-full-name), when Emacs starts up.
\f
* User Editing Changes in Emacs 19.29

** If you run out of memory.

If you get the error message "Virtual memory exhausted", type C-x s.
That way of saving files has the least additional memory needs.  Emacs
19.29 keeps a reserve of memory which it makes available when this
error happens; that is to ensure that C-x s can complete its work.

Once you have saved your data, you can exit and restart Emacs, or use
M-x kill-some-buffers to free up space.  If you kill buffers
containing a substantial amount of text, you can go on editing.

Do not use M-x buffer-menu to save or kill buffers when you are out of
memory, because that needs a fair amount memory itself and you may not
have enough to get it started.

** The format of compiled files has changed incompatibly.

Byte-compiled files made with Emacs 19.29 normally use a new format
that will not work in older Emacs versions.  You can compile files
in the old format if you wish; see "Changes in compilation," below.

** Emacs 19.29 supports the DEC Alpha.

** Emacs runs on Windows NT.

This port does not yet support windowing features.  It works like a
text-only terminal, but it does support a mouse.

In general, support for non-GNU-like operating systems is not a high
priority for the GNU project.  We merged in the support for Windows NT
because that system is expected to be very widely used.

** Emacs supports Motif widgets.

You can build Emacs with Motif widgets by specifying --with-x-toolkit=motif
when you run configure.

Motif defines collections of windows called "tab groups", and uses the
tab key and the cursor keys to move between windows in a tab group.
Emacs naturally does not support this--it has other uses for the tab
key and cursor keys.  Emacs does not support Motif accelerators either,
because it uses its normal keymap event binding features.

We give higher priority to operation with a free widget set than to
operation with a proprietary one.

** If Emacs or the computer crashes, you can recover all the files you
were editing from their auto save files by typing M-x recover-session.
This first shows you a list of recorded interrupted sessions.  Move
point to the one you choose, and type C-c C-c.

Then recover-session asks about each of the files that were being
edited during that session, asking whether to recover that file.  If
you answer y, it calls recover-file, which works in its normal
fashion.  It shows the dates of the original file and its auto-save
file and asks once again whether to recover that file.

When recover-session is done, the files you've chosen to recover
are present in Emacs buffers.  You should then save them.
Only this--saving them--updates the files themselves.

** Menu bar menus now stay up if you click on the menu bar item and
release the mouse button within a certain amount of time.  This is in
the X Toolkit version.

** The menu bar menus have been rearranged and split up to make for a
better organization.  Two new menu bar menus, Tools and Search,
contain items that were formerly in the Files and Edit menus, as well
as some that did not exist in the menu bar menus before.

** Emacs can now display on more than one X display at the same time.
Use the command make-frame-on-display to create a frame, specifying
which display to use.

** M-x talk-connect sets up a multi-user talk connection
via Emacs.  Specify the X display of the person you want to talk to.
You can talk to any number of people (within reason) by using
this command repeatedly to specify different people.

Emacs does not make a fuss about security; the people who you talk to
can use all Emacs features, including visiting and editing files.  If
this frightens you, don't use M-x talk-connect.

** The range of integer values is now at least 2**28 on all machines.
This means the maximum size of a buffer is at least 2**27-1,
or 134,217,727.

** When you start Emacs, you can now specify option names in
long GNU form (starting with `--') and you can abbreviate the names.

You can now specify the options in any order.
The previous requirements about the order of options
have been eliminated.

The -L or --directory option lets you specify an additional
directory to search for Lisp libraries (including libraries
that you specify with the -l or --load options).

** Incremental search in Transient Mark mode, if the mark is already
active, now leaves the mark active and does not change its position.
You can make incremental search deactivate the mark once again with
this expression.

    (add-hook 'isearch-mode-hook 'deactivate-mark)

** C-delete now deletes a word backwards.  This is for compatibility
with some editors in the PC world.  (This key is not available on
ordinary ASCII terminals, because C-delete is not a distinct character
on those terminals.)

** ESC ESC ESC is now a command to escape from various temporary modes
and states.

** M-x pc-bindings-mode sets up bindings compatible with many PC editors.
In particular, Delete and its variants delete forward instead of backward.
Use Backspace to delete backward.

C-Backspace kills backward a word (as C-Delete normally would).
M-Backspace does undo.
Home and End move to beginning and end of line
C-Home and C-End move to beginning and end of buffer.

** The key sequence for evaluating a Lisp expression using the minibuffer
is now ESC :.  It used to be ESC ESC, but we moved it to make way for
the ESC ESC ESC feature, on the grounds that people who evaluate Lisp
expressions are experienced users and can cope with a change.
If you prefer the old ESC ESC binding, put in your `~/.emacs':

        (global-set-key "\e\e" 'eval-expression)

** The f1 function key is now equivalent to the help key.  This is
done with key-translation-map; delete the binding for f1 in that map
if you want to use f1 for something else.

** Mouse-3, in the simplest case, still sets the region.  But now, it
places the mark where point was, and sets point where you click.
(It used to set the mark where you click and leave point alone.)

If you position point with Mouse-1, then scroll with the scroll bar
and use Mouse-3, Mouse-3 uses the position you specified with Mouse-1
even if it has scrolled off the screen (and point is no longer there).
This makes it easier to select a region with the mouse which is bigger
than a screenful.

Any editing of the buffer, and any cursor motion or scrolling for any
reason other than the scroll bar, cancels the special state set up by
Mouse-1--so that a subsequent Mouse-3 click will use the actual value
of point.

** C-mouse-3 now pops up a mode-specific menu of commands--normally
the same ones available in the mode's own menu bar menus.

** C-mouse-2 now pops up a menu of faces, indentation, justification,
and certain other text properties.  This menu is also available
through the menu-bar Edit menu.  It is meant for use with Enriched
mode.

*** You can use this menu to change the face of the region.
You can also set the face of the region with the new M-g command.

*** The menu also includes commands for indenting the region, 
which locally changes the values of left-margin and fill-column that
are used.

*** All fill functions now indent every line to the left-margin.  If
there is also a fill-prefix, that goes after the margin indentation.

*** Open-line and newline also make sure that the lines they create
are indented to the left margin.

*** It also allows you to set the "justification" of the region:
whether it should be centered, flush right, and so forth.  The fill
functions (including auto-fill-mode) will maintain the justification
and indentation that you request.

*** The new function `list-colors-display' shows you what colors are
available.  This is also accessible from the C-mouse-2 menu.

** You can now save and load files including their faces and other
text-properties by using Enriched-mode.  Files are saved in an
extended version of the MIME text/enriched format.  You can use the
menus described above, or M-g and other keyboard commands, to
alter the formatting information.

** C-mouse-1 now pops up the menu for changing the frame's default font.

** You can input Hyper, Super, Meta, and Alt characters, as well as
non-ASCII control characters, on an ASCII-only terminal.
To do this, use

  C-x @ h  --  hyper
  C-x @ s  --  super
  C-x @ m  --  meta
  C-x @ a  --  alt
  C-x @ S  --  shift
  C-x @ c  --  control

These are not ordinary key sequences; they operate through
function-key-map, which means they can be used even in the
middle of an ordinary key sequence.

** Outline minor mode and Hideif mode now use C-c @ as their prefix
character.

** Echo area messages are now logged in the "*Messages*" buffer.  The
size of this buffer is limited to message-log-max lines.

** RET in various special modes for read-only buffers that contain
lists of items now selects the item point is on.  These modes include
Dired, Compilation buffers, Buffer-menu, Tar mode, and Occur mode.
(In Info, RET follows the reference near point; in completion list
buffers, RET chooses the completion around point.)

** set-background-color now updates the modeline face in a special
way.  If that face was previously set up to be reverse video, the
reverse of the default face, then set-background-color updates it so
that it remains the reverse of the default face.

** The functions raise-frame and lower-frame are now commands.
When used interactively, they apply to the selected frame.

** M-x buffer-menu now displays the buffer list in the selected window.
Use M-x buffer-menu-other-window to display it in another window.

** M-w followed by a kill command now *does not* append the text in
the kill ring.  In consequence, M-w followed by C-w works as you would
expect: it leaves the top of the kill ring matching the region that
you killed.

** In Lisp mode, the C-M-x command now executes defvar forms in a
special way: it unconditionally sets the variable to the specified
default value, if there is one.  Normal execution of defvar does not
alter the variable if it already has a non-void value.

** In completion list buffers, the left and right arrow keys run the
new commands previous-completion and next-completion.  They move one
completion at a time.

** While doing completion in the minibuffer, the `prior' or `pageup'
key switches to the completion list window.

** When you exit the minibuffer with empty contents, the empty string
is not put in the minibuffer history.

** The default buffer for insert-buffer is now the "first" buffer
other than the current one.  If you have more than one window, this
is a buffer visible in another window.  (Usually it is the buffer
that C-M-v would scroll.)

** The etags program is now capable of recording tags based on regular
expressions provided on the command line.

This new feature allows easy support for constructs not normally
handled by etags, such as the macros frequently used in big C/C++
projects to define project-specific structures.  It also enables the
use of etags and TAGS files for languages not supported by etags.

The Emacs manual section on Tags contains explanations and examples
for Emacs's DEFVAR, VHDL, Cobol, Postscript and TCL.

** Various mode-specific commands that used to be bound to C-c LETTER
have been moved.

*** In gnus-uu mode, gnus-uu-interactive-scan-directory is now on C-c C-d,
and gnus-uu-interactive-save-current-file is on C-c C-z.

*** In Scribe mode, scribe-insert-environment is now on C-c C-v,
scribe-chapter is on C-c C-c, scribe-subsection is on C-c C-s,
scribe-section is on C-c C-t, scribe-bracket-region-be is on C-c C-e,
scribe-italicize-word is on C-c C-i, scribe-bold-word is on C-c C-b,
and scribe-underline-word is on C-c C-u.

*** In Gomoku mode, gomoku-human-takes-back is now on C-c C-b,
gomoku-human-plays is on C-c C-p, gomoku-human-resigns is on C-c C-r,
and gomoku-emacs-plays is on C-c C-e.

*** In the Outline mode defined in allout.el,
outline-rebullet-current-heading is now on C-c *.

** M-s in Info now searches through the nodes of the Info file,
just like s.  The alias M-s was added so that you can use the same
command for searches in both Info and Rmail.

** iso-acc.el now lets you enter inverted-! and inverted-?
with the sequences ~! and ~?.

** M-x compare-windows now pushes mark in both windows before
it starts moving point.

** There are two new commands in Dired, A (dired-do-search)
and Q (dired-do-query-replace).  These are similar to tags-search and
tags-query-replace, but instead of searching the list of files that
appears in a tags table, they search all the files marked in Dired.

** Changes to dabbrev.

A new function, `dabbrev-completion' (bound to M-C-/), expands the
unique part of an abbreviation.

Dabbrev now looks for expansions in other buffers, looks for symbols
instead of words and it works in the minibuffer.

Dabbrev can be customized to work for shell scripts, with variables
that sometimes have and sometimes haven't a leading "$".  See the
variable 'dabbrev-abbrev-skip-leading-regexp'.

** In Rmail, the command rmail-input-menu has been eliminated.  The
feature of selecting an Rmail file from a menu is now implemented in
another way.

** Bookmarks changes.

*** It now works to set bookmarks in Info nodes.

*** Bookmarks can have annotations; type "C-h m" after doing 
"M-x list-bookmarks", for more information on annotations.

*** The bookmark-jump popup menu function is now `bookmark-menu-jump', for
those who bind it to a mouse click.

*** The default bookmarks file name is now "~/.emacs.bmk".  If you
already have a bookmarks file, it will be renamed automagically when
you next load it.

** New package, ps-print.

The ps-print package generates PostScript printouts of buffers or
regions, and includes face attributes such as color, underlining,
boldface and italics in the printed output.

** New package, msb.

The msb package provides a buffer-menu in the menubar with separate
menus for different types of buffers.

** `cpp.el' is a new library that can highlight or hide parts of a C
file according to C preprocessor conditionals.  To try it, run the
command M-x cpp-highlight-buffer.

** Changes in CC mode.

*** c-set-offset and related functions and variables can now accept
variable symbols.  Also ++ and -- which mean 2* positive and negative
c-basic-offset respectively.

*** New variable, c-recognize-knr-p, which controls whether K&R C
constructs will be recognized.  Trying to recognize K&R constructs is a
time hog so if you're programming strictly in ANSI C, set this
variable to nil (it should already be nil in c++-mode).

*** New variable, c-hanging-comment-ender-p for controlling
c-fill-paragraph's behavior.

*** New syntactic symbol: statement-case-open.  This is assigned to lines
containing an open brace just after a case/default label.

*** New variable, c-progress-interval, which controls minibuffer update
message displays during long re-indention.  This is a new feature
which prints percentage complete messages at specified intervals.

** Makefile mode changes. 

*** The electric keys are not enabled by default.

*** There is now a mode-specific menu bar menu.

*** The mode supports font-lock, add-log, and imenu.

*** The command M-TAB does completion of target names and variable names.

** icomplete.el now works more like a minor mode.  Use M-x icomplete-mode
to turn it on and off.

Icomplete now supports an `icomplete-minibuffer-setup-hook', which is
run on minibuffer setup whenever icompletion will be occurring.  This
hook can be used to customize interoperation of icomplete with other
minibuffer-specific packages, eg rsz-mini.  See the doc string for
more info.

** Ediff change.

Use ediff-revision instead of vc-ediff.  It also replaces rcs-ediff,
for those who use that; if you want to use a version control package
other than vc.el, you must set the variable
ediff-version-control-package to specify which package.

** VC now supports branches with RCS.

You can use C-u C-x C-q to select any branch or version by number.
It reads the version number or branch number with the minibuffer,
then checks out the file unlocked.

Type C-x C-q again to lock the selected branch or version.
When you check in changes to that branch or version, there are two
possibilities:

-- If you've selected a branch, or a version at the tip of a branch,
then the new version adds to that branch.  If you wish to create a
new branch, use C-u C-x C-q to specify a version number when you check
in the new version.

-- If you've selected an inner version which is not the latest in its
branch, then the new version automatically creates a new branch.

** VC now supports CVS as well as RCS and SCCS.

Since there are no locks in CVS, some things behave slightly
different when the backend is CVS.  When vc-next-action is invoked
in a directory handled by CVS, it does the following:

   If the file is not already registered, this registers it for version
control.  This does a "cvs add", but no "cvs commit".
   If the file is added but not committed, it is committed.
   If the file has not been changed, neither in your working area or
in the repository, a message is printed and nothing is done.
   If your working file is changed, but the repository file is
unchanged, this pops up a buffer for entry of a log message; when you
finish the log message with C-c C-c, that checks in the resulting
changes along with the log message as change commentary.  A writable
file remains in existence.

   If vc-next-action changes the repository file, it asks you
whether to merge in the changes into your working copy.

vc-directory, when started in a CVS file hierarchy, reports
all files that are modified (and thus need to be committed).
(When the backend is RCS or SCCS vc-directory reports all
locked files).

VC has no support for running the initial "cvs checkout" to get a
working copy of a module.  You can only use VC in a working copy of
a module.

You can disable the CVS support as follows:

  (setq vc-master-templates (delq 'vc-find-cvs-master vc-master-templates))

or by setting vc-handle-cvs to nil.

This may be desirable if you run a non-standard version of CVS, or
if CVS was compiled with FORCE_USE_EDITOR or (possibly)
RELATIVE_REPOS.

** Comint and shell mode changes:

*** Completion works with file names containing quoted characters.

File names containing special characters (such as " ", "!", etc.) that are
quoted with a "\" character are recognised during completion.  Special
characters are quoted when they are inserted during completion.

*** You can use M-x comint-truncate-buffer to truncate the buffer.

When this command is run, the buffer is truncated to a maximum number
of lines, specified by the variable comint-buffer-maximum-size.  Just
like the command comint-strip-ctrl-m, this can be run automatically
during process output by doing this:

(add-hook 'comint-output-filter-functions
          'comint-truncate-buffer)

** Telnet mode buffer name changed.

The buffer name for a Telnet buffer is now *telnet-HOST*, not
 *HOST-telnet*.  This is for consistency with other Emacs packages.

** M-x man (man) is now faster and more robust.  On systems where the
entire man page is indented, the indentation is removed.

The user option names that used to end in -p now end in -flag.  The
new names are: Man-reuse-okay-flag, Man-downcase-section-letters-flag,
Man-circular-pages-flag.  The Man-notify user option has been renamed to
Man-notify-method and accepts one more value, `pushy', that just
switches the current buffer to the manpage buffer, without switching
frames nor changing your windows configuration.

A new user option Man-fontify-manpage-flag disables fontification
(thus speeding up man) when set to nil.  Default is to fontify if a
window system is used.  Two new user options Man-overstrike-face
(default 'bold) and Man-underline-face (default 'underline) can be set
to the preferred faces to be used for the words that man overstrikes
and underlines.  Useful for those who like coloured man pages.

Two new interactive functions are provided: Man-cleanup-manpage and
Man-fontify-manpage.  Both can be used on a buffer that contains the
output of a `rsh host man manpage' command, or the output of an
`nroff -man -Tman manpage' command to make them readable.
Man-cleanup-manpage is faster, but does not fontify.

** The new function modify-face makes it easy to specify
all the attributes of a face, all at once.

** Faces now support background stippling.

Use the command set-face-stipple to specify the stipple-pattern for a
face.  Use face-stipple to access the specified stipple pattern.  The
existing face functions now handle the stipple pattern when
appropriate.

If you specify one of the standard gray colors as a face background
color, and your display doesn't handle gray, Emacs automatically uses
stipple instead to get the same effect.

** Changes in Font Lock mode.

*** Fontification

Two new default faces are provided; `font-lock-variable-name-face' and
`font-lock-reference-face'.  The face `font-lock-doc-string-face' has
been removed since it is the same as the existing
`font-lock-string-face'.  Where appropriate, fontification
automatically uses these new faces.

Fontification via commands `font-lock-mode' and
`font-lock-fontify-buffer' is now cleanly interruptible (i.e., with
C-g).  If you interrupt during the fontification process, the buffer
remains in its previous modified state and all highlighting is removed
from the buffer.

For C/C++ modes, Font Lock mode is much faster but highlights much
more.  Other modes are faster/more extensive/more discriminatory, or a
combination of these.

To enable Font Lock mode, add the new function `turn-on-font-lock' in
one of the following ways:

 (add-hook 'c-mode-hook 'turn-on-font-lock)

Or for any visited file with:

 (add-hook 'find-file-hooks 'turn-on-font-lock)

*** Supports color and grayscale displays

Font Lock mode supports different ways of highlighting, depending on
the type of display and background shade.  Attributes (face color,
bold, italic and underline, and display type and background mode) can
be controlled either from Emacs Lisp or X resources.

See the new variables `font-lock-display-type' and
`font-lock-face-attributes'.

*** Supports more modes

The following modes are directly supported:

ada-mode, asm-mode, bibtex-mode, c++-c-mode, c++-mode, c-mode,
change-log-mode, compilation-mode, dired-mode, emacs-lisp-mode,
fortran-mode, latex-mode, lisp-mode, mail-mode, makefile-mode,
outline-mode, pascal-mode, perl-mode, plain-tex-mode, rmail-mode,
rmail-summary-mode, scheme-mode, shell-mode, slitex-mode, tex-mode,
texinfo-mode.

See the new variables `font-lock-defaults-alist' and
`font-lock-defaults'.

Some modes support different levels of fontification.  You can choose
to use the minimum or maximum available decoration by changing the
value of the new variable `font-lock-maximum-decoration'.

Programmers are urged to make available to the community their own
keywords for modes not yet supported.  See font-lock.el for
information about efficiency.

*** fast-lock

The fast-lock package speeds up Font Lock mode by saving font choices
in associated cache files.  When you visit a file with Font Lock mode
and Fast Lock mode turned on for the first time, the file's buffer is
fontified as normal.  When certain events occur (such as exiting
Emacs), Fast Lock saves the highlighting in a cache file.  When you
subsequently visit this file, its cache is used to restore the
highlighting.

To use this package, put in your `~/.emacs':

 (add-hook 'font-lock-mode-hook 'turn-on-fast-lock)

To control the use of caches, see the documentation for `fast-lock-mode'.

** You can tell pop-to-buffer to display certain buffers in the selected
window rather than finding some other window to display them in.
There are two variables you can use to specify these buffers.

same-window-buffer-names holds a list of buffer names; if a buffer's
name appears in this list, pop-to-buffer puts it in the selected window.

same-window-regexps holds a list of regexps--if any one of them
matches a buffer's name, then pop-to-buffer puts that buffer in the
selected window.

The default values of these variables are not nil: they list various
buffers that normally appear, when you as for them, in the selected
window.  These include shell buffers, mail buffers, telnet buffers,
and others.  By removing elements from these variables, you can ask
Emacs to display those buffers in separate windows.

** The special-display-buffer-names and special-display-regexps lists
have been generalized.  An element may now be a list.  The car of the list
is the buffer name or regular expression for matching buffer names.

The cdr of the list can be an alist specifying additional frame
parameters for use in constructing the special display frame.

Alternatively, the cdr can have this form:

  (FUNCTION ARGS...)

where FUNCTION is a symbol.  Then the frame is constructed by calling
FUNCTION; its first argument is the buffer, and its remaining
arguments are ARGS.

** If the environment variable REPLYTO is set, its value is the default
for mail-default-reply-to.

** When you send a message in Emacs, if you specify an Rmail file with
the FCC: header field, Emacs converts the message to Rmail format
before writing it.  Thus, the file never contains anything but Rmail
format messages.

** The new variable mail-from-style controls whether the From: header
should include the sender's full name, and if so, which format to use.

** The new variable mail-personal-alias-file specifies the name of the
user's personal aliases.  This defaults to the file ~/.mailrc.
mailabbrev.el used to have its own variable for this purpose
(mail-abbrev-mailrc-file).  That variable is no longer used.

** In Buffer-Menu mode, the d and C-d commands (which mark buffers for
deletion) now accept a prefix argument which serves as a repeat count.

** Changes in BibTeX mode.

*** Reference keys can now be entered with TAB completion.  All
reference keys defined in that buffer and all labels that appear in
crossreference entries are object to completion.

*** Braces are supported as field delimiters in addition to quotes.
BibTeX entries may have brace-delimited and quote-delimited fields
intermixed.  The delimiters generated for new entries are specified by
the variables bibtex-field-left-delimiter and
bibtex-field-right-delimiter on a buffer-local basis. Those variables
default to braces, since it is easier to put quote accented characters
(as the german umlauts) into a brace-delimited entry.

*** The function bibtex-clean-entry can now be invoked with a prefix
argument.  In this case, a label is automatically generated from
various fields in the record.  If bibtex-clean-entry is invoked on a
record without label, a label is also generated automatically.
Various variables (all beginning with `bibtex-autokey-') control the
creation of that key.  The variable bibtex-autokey-edit-before-use
determines, if the user is allowed to edit auto-generated reference
keys before they are used.

*** A New function bibtex-complete-string completes strings with
respect to the strings defined in this buffer and a set of predefined
strings (initialized to the string macros defined in the standard
BibTeX style files) in the same way in which ispell-complete-word
works with respect to words in a dictionary.  Candidates for
bibtex-complete-string are initialized from variable
bibtex-predefined-strings and by parsing the files found in
bibtex-string-files for @String definitions.

*** Every reference/field pair has now attached a comment which
appears in the echo area when this field is edited.  These comments
should provide useful hints for BibTeX usage, especially for BibTeX
beginners.  New variable bibtex-help-message determines if these help
messages are to appear in the minibuffer when moving to a text entry.

*** Inscriptions of menu bar changed from "Entry Types" to
"Entry-Types" and "Bibtex Edit" to "BibTeX-Edit".

*** The variable bibtex-include-OPTcrossref is now not longer a binary
switch but a list of reference names which should contain a crossref
field.  E.g., you can tell bibtex-mode you want a crossref field for
@InProceedings and @InBook entries but for no other.

*** The function validate-bibtex-buffer was completely rewritten to
validate if a buffer is syntactically correct.  find-bibtex-duplicates
is no longer a function itself but was moved into
validate-bibtex-buffer.

*** Cleaning a BibTeX entry tests, if necessary fields are there.
E.g., if you tell bibtex-mode to include a crossref entry, some fields
are optional which would be required without the crossref entry.  If
you now leave the crossref entry empty and do a bibtex-clean-entry
with some now required fields left empty, version 2.0 of bibtex.el
complains about the absence of these fields, whereas version 1.3
didn't.

*** Default value for variables bibtex-maintain-sorted-entries and
bibtex-sort-ignore-string-entries is now t.

*** All interactive functions are renamed to begin with `bibtex-'.

*** Keybindings with \C-c\C-e entry changed for unification.  Often
used reference types are now on control-modified keys, mediocre used
types are on unmodified keys, seldom used types are on shift-modified
keys and almost never used types on meta-modified keys.
\f
* Configuration Changes in Emacs 19.29

** Emacs now uses directory /usr/local/share for most of its installed
files.  This follows a GNU convention for directory usage.

** The option --with-x11 is no longer supported.
X11 is the only version of X that Emacs 19.29 supports;
use --with-x if you need to request X support explicitly.
(Normally this should not be necessary, since configure should
automatically enable X support if X is installed on your machine.)

** If you use the site-init.el file to set the variable
mail-host-address to a string in the dumped Emacs, that string becomes
the default host address for initializing user-mail-address.
It is used instead of the value of (system-name).
\f
* Lisp-Level Changes in Emacs 19.29

** Basic Lisp

*** The range of integer values is now at least 2**28 on all machines.
This means the maximum size of a buffer is at least 2**27-1,
or 134,217,727.

*** You can now use Common Lisp syntax for the backquote and comma
macros.  Thus, you can now write `(x ,y z) instead of (` (x (, y) z)).

The old syntax is still accepted.

*** The new function rassoc is like assoc, except that it compares the
key against the cdr of each alist element, where assoc would compare
it against the car of each alist element.

*** The new function unintern deletes a symbol from an obarray.  The
first argument can be the symbol to delete, or a string giving its
name.  The second argument specifies the obarray (nil means the
current default obarray).

If the specified symbol is not in the obarray, or if there's no symbol
in the obarray matching the specified string, unintern does nothing
and returns nil.  If it does delete a symbol, it returns t.

*** You can specify an alternative read function for use by load and
eval-region by binding the variable load-read-function to some other
function.  This function should accept one argument just like read.
If load-read-function is nil, load and eval-region use ordinary read.

*** The new function `type-of' takes any object as argument, and
returns a symbol identifying the type of that object--one of `symbol',
`integer', `float', `string', `cons', `vector', `marker', `overlay',
`window', `buffer', `subr', `compiled-function',
`window-configuration', `process'.

*** When you use eval-after-load for a file that is already loaded, it
executes the FORM right away.  As before, if the file is not yet
loaded, it arranges to execute FORM if and when the file is loaded
later.  The result is: if you have called eval-after-load for a file,
and if that file has been loaded, then regardless of the order of
these two events, the specified form has been evaluated.

*** The Lisp construct #@NUMBER now skips the next NUMBER characters,
treating them as a comment.

You would not want to use this in a file you edit by hand, but it is
useful for commenting out parts of machine-generated files.

*** Two new functions, `plist-get' and `plist-put',
allow you to modify and retrieve values from lists formatted as property-lists.
They work like `get' and `put', but operate on any list.
`plist-put' returns the modified property-list; you must store it
back where you got it.

*** The new function add-to-list is called with two elements,
a variable that holds a list and a new element.
It adds the element to the list unless it is already present.
It compares elements using `equal'.  Here is an example:

(setq foo '(a b)) => (a b)

(add-to-list 'foo 'c) => (c a b)

(add-to-list 'foo 'b) => (c a b)

foo => (c a b)

** Changes in compilation.

Functions and variables loaded from a byte-compiled file
now refer to the file for their doc strings.

This has a few consequences:

-- Loading the file is faster and uses less memory.
-- Reference to doc strings is a little slower (the same speed
   as reference to the doc strings of primitive and preloaded functions).
-- The compiled files will not work in old versions of Emacs.
-- If you move the compiled file after loading it, Emacs can no longer
   find these doc strings.
-- If you alter the compiled file (such as by compiling a new
   version), then further access to documentation strings will get
   nonsense results.

The byte compiler now optionally supports lazy loading of compiled
functions' definitions.  If you enable this feature when you compile,
loading the compiled file does not actually bring the function
definitions into core.  Instead it creates references to the compiled
file, and brings each function's definition into core the first time
you call that function, or when you force it with the new function
`fetch-bytecode'.

Using the lazy loading feature has a few consequences:

-- Loading the file is faster and uses less memory.
-- Calling any function in the file for the first time is slower.
-- If you move the compiled file after loading it, Emacs can no longer
   find the function definitions.
-- If you alter the compiled file (such as by compiling a new
   version), then further access to functions not already loaded
   will get nonsense results.

To enable the lazy loading feature, set up a non-nil file local
variable binding for the variable `byte-compile-dynamic' in the Lisp
source file.  For example, put this on the first line:

    -*-byte-compile-dynamic: t;-*-

It's a good idea to use the lazy loading feature for a file that
contains many functions, most of which are not actually used by a
given user in a given session.

To turn off the basic feature of referring to the file for doc
strings, set byte-compile-dynamic-docstrings to nil.  You can do this
globally, or for one source file by adding this to the first line:

    -*-byte-compile-dynamic-docstrings: nil;-*-

** Strings

*** Do not pass integer arguments to `concat' (or `vconcat' or
`append').  We are phasing out the old unrecommended support for
integers as arguments to these functions, in preparation for treating
numbers as single characters in a future release.  To concatenate
numbers in string form, use `number-to-string' first, or rewrite the
call to use `format' instead of `concat'.

*** The new function match-string returns the string of text matched at
the given parenthesized expression by the last regexp search, or nil
if there was no match.  If the last match was by `string-match' on a
string, the string must be given.  Therefore, this function can be
used in place of `buffer-substring' and `substring', when using
`match-beginning' and `match-end' to find match positions.

   (match-string N)   or   (match-string N STRING)

*** The function replace-match now accepts an optional fourth argument,
STRING.  Use this after performing string-match on STRING, to replace
the portion of STRING that was matched.  When used in this way,
replace-match returns a newly created string which is the same as
STRING except for the matched portion.

*** The new function buffer-substring-no-properties
is like buffer-substring except that the string it returns
has no text properties.

*** The function `equal' now considers two strings to be different
if they don't have the same text properties.

** Completion

*** all-completions now takes an optional fourth argument.
If that argument is non-nil, completions that start with a space
are ignored unless the initial string also starts with a space.
(This used to happen unconditionally.)

** Local Variables

*** Local hook variables.

There is now a clean way to give a hook variable a buffer-local value.
Call the function `make-local-hook' to do this.

Once a hook variable is buffer-local, you can add hooks to it either
globally or locally.  run-hooks runs the local hook functions
of the current buffer, then all the global hook functions.

The functions add-hook and remove-hook take an additional optional
argument LOCAL which says whether to add (or remove) a local hook
function or a global one.

Local hooks use t as an element of the (local) value of the hook
variable as a flag meaning to use the global value also.

*** The new function local-variable-p tells you whether a particular
variable is buffer-local in the current buffer or a specified buffer.

** Editing Facilities

*** The function copy-region-as-kill no longer sets this-command;
as a result, a following kill command will not normally append
to the text saved by copy-region-as-kill.

*** Regular expression searching and matching no longer performs full
Posix backtracking by default.  They now stop with the first match found
instead of looking for the longest match--just as they did in Emacs 18.
The reason for this change is to get higher speed.

There are new functions you can use if you really want to search or
match with Posix behavior: posix-search-forward,
posix-search-backward, posix-looking-at, and posix-string-match.  Call
these just like re-search-forward, re-search-backward, looking-at, and
string-match.

** Files

*** The new variable `format-alist' defines file formats,
which are ways of translating between the data in a file and things
(text, text-properties, and possibly other information) in a buffer.

`format-alist' has one element for each format.  Each element is a
list like this:
  (NAME DOC-STRING REGEXP FROM-FN TO-FN MODIFY MODE-FN)
containing the name of the format, a documentation string, a regular
expression which is used to recognize files in that format, a decoding
function, an encoding function, a flag that indicates whether the
encoding function modifies the buffer, and a mode function.

FROM-FN is called to decode files in that format; it gets two args, BEGIN 
        and END, and can make any modifications it likes, returning the new
        end position.  It must make sure that the beginning of the file no
        longer matches REGEXP, or else it will get called again.
TO-FN   is called to encode a region into that format; it is also passed BEGIN
        and END, and either returns a list of annotations as in
        `write-region-annotate-functions', or modifies the region and returns
        the new end position.
MODIFY, if non-nil, means the TO-FN modifies the region.  If nil, TO-FN may
        not make any changes and should return a list of annotations.

`insert-file-contents' checks the beginning of the file that it is
inserting to see if it matches one of the regexps.  If so, then it
calls the decoding function, and then looks for another match.  When
visiting a file, it also calls the mode function, and sets the
variable `buffer-file-format' to the list of formats that the file
used.

`write-region' calls the encoding functions for each format in
`buffer-file-format' before it writes the file.  To save a file in a
different format, either set `buffer-file-format' to a different
value, or call the new function `format-write-file'.

Since some encoding functions may be slow, you can request that
auto-save use a format different from the buffer's default by setting
the variable `auto-save-file-format' to the desired format.  This will
determine the format of all auto-save files.

*** The new function file-ownership-preserved-p tells you whether
deleting a file and recreating it would keep the file's owner
unchanged.

*** The new function file-regular-p returns t if a file
is a "regular" file (not a directory, symlink, named pipe,
terminal, or other I/O device).

*** The new function file-name-sans-extension discards the extension
of a file name.  You call it with a file name, and returns a string
lacking the extension.

*** The variable path-separator is a string which says which
character separates directories in a search path.  It is ":"
for Unix and GNU systems, ";" for MSDOG and Windows NT.

** Commands and Key Sequences

*** Key sequences consisting of C-c followed by {, }, <, >, : or ; are
now reserved for major modes.  Sequences consisting of C-c followed by
any other punctuation character are now meant for minor modes.  We don't
plan to convert all existing major modes to stop using those sequences,
but we hope to keep them to a minimum.

*** When the post-command-hook or the pre-command-hook gets an error, the error
is silently ignored.  Emacs no longer sets the hook variable to nil when this
happens.  Meanwhile, the hook functions can now alter the hook variable in
a normal fashion; there is no need to do anything special.

*** define-key, lookup-key, and various other functions for changing or
looking up key bindings now let you write an event type with a list
like (ctrl meta newline) or (meta ?d), as in XEmacs.  (ctrl meta newline)
is equivalent to the event type symbol C-M-newline, and (meta ?d)
is equivalent to the character ?\M-d.

*** The function event-convert-list converts a list such as
(meta ?d) into the corresponding event type (a symbol or integer).

*** In an interactive spec, `k' means to read a key sequence.  In this
key sequence, upper case characters and shifted function keys which
have no bindings are converted to lower case if that makes them
defined.

The new interactive code `K' reads a key sequence similarly, but does
not convert the last event.  `K' is useful for reading a key sequence
to be given a binding.

*** The variable overriding-local-map now has no effect on the menu bar
display unless overriding-local-map-menu-flag is non-nil.  This is why
incremental search no longer temporarily changes the menu bars.

Note that overriding-local-map does still affect the execution of key
sequences entered using the menu bar.  So if you use
overriding-local-map, and a menu bar key sequence comes in, you should
make sure to clear overriding-local-map before that key sequence gets
looked up and executed.  But this is what you'd normally do anyway:
programs that use overriding-local-map normally exit and "put back"
any event such as menu-bar that they do not handle specially.

*** The new variable `overriding-terminal-local-map' is like
overriding-local-map, but is specific to a single terminal.

*** delete-frame events.

When you use the X window manager's "delete window" command, this now
generates a delete-frame event.  The standard definition of this event
is a command that deletes the frame that received the event, and kills
Emacs when the last visible or iconified frame is deleted.  You can
rebind the event to some other command if you wish.

*** Two new types of events, iconify-frame and make-frame-visible,
indicate that the user iconified or deiconified a frame with the
window manager.  Since the window manager has already done the work,
the default definition for both event types in Emacs is to do nothing.

** Frames and X

*** Certain Lisp variables are now local to an X terminal (in other
words, all the screens of a single X server).  The value in effect, at
any given time, is the one that belongs to the terminal of the
selected frame.  The terminal-local variables are
default-minibuffer-frame, system-key-alist, defining-kbd-macro, and
last-kbd-macro.  There is no way for Lisp programs to create others.

The terminal-local variables cannot be buffer-local.

*** When you create an X frame, for the `top' and `left' frame
parameters, you can now use values of the form (+ N) or (- N), where N
is an integer.  (+ N) means N pixels to the right of the left edge of
the screen and (- N) means N pixels to the left of the right edge.  In
both cases, N may be zero (exactly at the edge) or negative (putting
the window partly off the screen).

The function x-parse-geometry can return values of these forms
for certain inputs.

*** The variable menu-bar-file-menu has been renamed to
menu-bar-files-menu to match the actual item that appears in the menu.
(All the other such variable names do match.)

*** The new function active-minibuffer-window returns the minibuffer window
currently active, or nil if none is now active.

*** In the functions next-window, previous-window, next-frame,
previous-frame, get-buffer-window, get-lru-window, get-largest-window
and delete-windows-on, if you specify 0 for the last argument,
it means to consider all visible and iconified frames.

*** When you set a frame's cursor type with modify-frame-parameters,
you can now specify (bar . INTEGER) as the cursor type.  This stands
for a bar cursor of width INTEGER.

*** The new function facep returns t if its argument is a face name
(or if it is a vector such as is used internally by the Lisp code
to represent a face).

*** Each frame can now have a buffer-predicate function,
which is the `buffer-predicate' frame parameter.
When `other-buffer' looks for an alternative buffer, it considers
only the buffers that fit the selected frame's buffer predicate (if it
has one).  This is useful for applications that make their own frames.

*** When you create an X frame, you can now specify the frame parameter
`display'.  This says which display to put the frame on.  The value
should be a display name--a string of the form
"HOST:DPYNUMBER.SCREENNUMBER".

The functions x-server-... and x-display-... now take an optional
argument which specifies the display to ask about.  You can use either
a display name string or a frame.  A value of nil stands for the
selected frame.

To close the connection to an X display, use the function
x-close-connection.  Specify which display with a display name.  You
cannot close the connection if Emacs still has frames open on that
display.

x-display-list returns a list indicating which displays Emacs has
connections to.  Its elements are display names (strings).

*** The icon-type frame parameter may now be a file name.
Then the contents of that file specify the icon bitmap to use
for that frame.

*** The title of an Emacs frame, displayed by most window managers, is
set from frame-title-format or icon-title-format.  These have the same
structure as mode-line-format.

*** x-display-grayscale-p is a new function that returns non-nil if
your X server can display shades of gray.  Currently it returns
non-nil for color displays (because they can display shades of gray);
we may change it in the next version to return nil for color displays.

*** The frame parameter scroll-bar-width specifies the width of the
scrollbar in pixels.

** Buffers

*** Creating a buffer with get-buffer-create does not obey
default-major-mode.  That variable is now handled in a separate
function, set-buffer-major-mode.  get-buffer-create and generate-new-buffer
always leave the newly created buffer in Fundamental mode.

Creating a new buffer by visiting a file or with switch-to-buffer,
pop-to-buffer, and similar functions does call set-buffer-major-mode
to select the default major mode specified with default-major-mode.

*** You can now create an "indirect buffer".  An indirect buffer shares
its text, including text properties, with another buffer (the "base
buffer"), but has its own major mode, local variables, overlays, and
narrowing.  An indirect buffer has a name of its own, distinct from
those of the base buffer and all other buffers.  An indirect buffer
cannot itself be visiting a file (though its base buffer can be).
The base buffer cannot itself be indirect.

Use (make-indirect-buffer BASE-BUFFER NAME) to make an indirect buffer
named NAME whose base is BASE-BUFFER.  If BASE-BUFFER is an indirect
buffer, its base buffer is used as the base for the new buffer.

You can make an indirect buffer current, or switch to it in a window,
just as you would a non-indirect buffer.

The function buffer-base-buffer, given an indirect buffer, returns its
base buffer.  It returns nil when given an ordinary buffer (not
indirect).

The library `noutline' has versions of Outline mode and Outline minor
mode which let you display different parts of the outline in different
indirect buffers.

** Subprocesses

*** The functions call-process and call-process-region now allow
you to direct error message output from the subprocess into a
separate destination, instead of mixing it with ordinary output.
To do this, specify for the third argument, BUFFER, a list of the form
  (BUFFER-OR-NAME ERROR-DESTINATION)
BUFFER-OR-NAME specifies where to put ordinary output; it should
be a buffer or buffer name, or t, nil or 0.  This is what would
have been the BUFFER argument, ordinarily.

ERROR-DESTINATION specifies where to put the error output.
nil means discard it, t means mix it with the ordinary output,
and a string specifies a file name to write this output into.

You can't specify a buffer to put the error output in; that is not
easy to implement directly.  You can put the error output into a
buffer by sending it to a temporary file and then inserting the file
into a buffer.

*** Comint mode changes:

**** The variable comint-completion-addsuffix can also be a cons pair
of the form (DIRSUFFIX . FILESUFFIX), where DIRSUFFIX and FILESUFFIX are
strings added on unambiguous or exact completion of directories and file
names, respectively.

** Text properties

*** You can now specify which values of the `invisible' property
make text invisible in a given buffer.  The variable
`buffer-invisibility-spec', which is always local in all buffers,
controls this.

If its value is t, then any non-nil `invisible' property makes
a character invisible.

If its value is a list, then a character is invisible if its
`invisible' property value appears as a member of the list, or if it
appears as the car of a member of the list.

When the `invisible' property value appears as the car of a member of
the `buffer-invisibility-spec' list, then the cdr of that member has
an effect.  If it is non-nil, then an ellipsis appears in place of the
character.  (This happens only for the *last* invisible character in a
series of consecutive invisible characters, and only at the end of a
line.)

If a character's `invisible' property is a list, then Emacs checks each
element of the list against `buffer-invisibility-spec'.  If any element
matches, the character is invisible.

*** The command `list-text-properties-at' shows what text properties
are in effect at point.

*** Frame objects now exist in Emacs even on systems that don't support
X Windows.  You can create multiple frames, and switch between them
using select-frame.  The selected frame is actually displayed on your
terminal; other frames are not displayed at all.  The selected frame
number appears in the mode line after `Emacs', except for frame 1.

Switching frames on ASCII terminals is therefore more or less
equivalent to switching between different window configurations.

*** The new variable window-size-change-functions holds a list of
functions to be called if window sizes change (or if windows are
created or deleted).  The functions are called once for each frame on
which changes have occurred, with the frame as the sole argument.
This takes place shortly before redisplay.

*** The modification hook functions of overlays now work differently.
They are called both before and after each change.  This makes it
possible for the functions to determine exactly what the change was.

This change affects three overlay properties: the modification-hooks
property, a list of functions called for deletions overlapping the
overlay's range and for insertions inside it; the
insert-in-front-hooks, a list of functions called for insertions at
the beginning of the overlay; and the insert-behind-hooks, a list of
functions called for insertions at the end of the overlay.

Each function is called both before and after each change that it
applies to.  Before the change, it is called with four arguments:
    (funcall FUNCTION OVERLAY nil START END)
START and END are the same arguments that the before-change-functions
receive.

After the change, each function is called with five arguments:
    (funcall FUNCTION OVERLAY t START END OLDSIZE)
The last arguments, START and END and OLDSIZE,
are the same arguments that the after-change-functions receive.

This means the function must accept either four or five arguments.

*** You can set defaults for text-properties with the new variable
`default-text-properties'.  Its value is a property list; the values
specified there are used whenever a character (or its category) does
not specify a value.

*** The `face' property of a character or an overlay can now be a list
of face names.  Formerly it had to be just one face name.

*** Changes in handling the `intangible' text property.

**** If inhibit-point-motion-hooks is non-nil, then `intangible' properties
are ignored.

**** Moving to just before a stretch of intangible text
is no longer special in any way.  Point stays at that place.

**** When you move point backwards into the midst of intangible text,
point moves back to the beginning of that text.  (It used to move
forward to the end of that text, which was not very useful.)

**** When moving across intangible text, Emacs stops wherever the
property value changes.  So if you have two stretches of intangible
text, with different non-nil intangible properties, it is possible to
place point between them.

** Overlays

*** Overlay changes.

**** The new function previous-overlay-change returns the position of
the previous overlay start or end, before a specified position.  This
is the backwards-moving counterpart of next-overlay-change.

**** overlay-get now supports category properties on an overlay
the same way get-text-property supports them as text properties.

Specifically, if an overlay does not have the property PROP that you
ask for, but it does have a `category' property which is a symbol,
then that symbol's PROP property is used.

**** If an overlay has a non-nil `evaporate' property, it will be
deleted if it ever becomes empty (i.e., when it spans no characters).

**** If an overlay has a `before-string' and/or `after-string' property,
these strings are displayed at the overlay's endpoints.

** Filling

*** The new variable fill-paragraph-function provides a way for major
modes to override the filling of paragraphs.  If this is non-nil,
fill-paragraph calls it as a function, passing along its sole
argument.  If the function returns non-nil, fill-paragraph assumes it
has done the job and simply passes on whatever value it returned.

The usual use of this feature is to fill comments in programming
language modes.

*** Text filling and justification changes:

**** The new variable use-hard-newlines can be used to make a
distinction between "hard" and "soft" newlines; the fill functions
will then never remove a newline that was manually inserted.  Hard
newlines are marked with a non-nil `hard' text-property.

**** The fill-column and left-margin can now be modified by text-properties.
Most lisp programs should use the new functions (current-fill-column) and
(current-left-margin), which return the proper values to use for the
current line.

**** There are new functions for dealing with margins: 

***** Set-left-margin and set-right-margin (set the value for a region
and re-fill).  These functions take three arguments: two to specify
a region, and the desired margin value.

***** Increase-left-margin, decrease-left-margin, increase-right-margin, and
decrease-right-margin (change settings relative to current values, and
re-fill).

***** move-to-left-margin moves point there, optionally adding
indentation or changing tabs to spaces in order to make that possible.
beginning-of-line-text also moves past the fill-prefix and any
indentation added to center or right-justify a line, to the beginning
of the text that the user actually typed.

***** delete-to-left-margin removes any left-margin indentation, but
does not change the property.

**** The paragraph-movement functions look for the paragraph-start and
paragraph-separate regexps at the current left margin, not at the
beginning of the line.  This means that those regexps should NOT use ^
to anchor the search.  However, for backwards compatibility, a ^ at
the beginning of the regexp will be ignored, so most packages won't break.

**** justify-current-line is now capable of doing left, center, or
right justification as well as full justification.

**** The fill functions can do any kind of justification based on the new
`justification' text-property and `default-justification' variable,
or arguments to the functions.  They also have a new option which
defeats the normal removal of extra whitespace.

**** The new function `current-justification' returns the kind of
justification used for the current line.  The new function
`set-justification' can be used to change it, including re-justifying
the text of the region according to the new value.

**** Filling and auto-fill are disabled if justification is `none'.

**** The auto-fill-function is now called regardless of whether 
the fill-column has been exceeded; the function can determine on its
own whether filling (or justification) is necessary.

** Processes

*** process-tty-name is a new function that returns the name of the
terminal that the process itself reads and writes on (not the name of
the pty that Emacs uses to talk with that terminal).

*** Errors in process filters and sentinels are now normally caught
automatically, so that they don't abort other Lisp programs.

Setting debug-on-error non-nil turns off this feature; then errors in
filters and sentinels are not caught.  As a result, they can invoke
the debugger, under the control of debug-on-error.

*** Emacs now preserves the match data around the execution of process
filters and sentinels.  You can use search and match functions freely
in filters and sentinels without explicitly bothering to save the
match data.

** Display

*** The variable message-log-max controls how messages are logged in the
"*Messages*" buffer.  An integer value means to keep that many lines;
t means to log with no limit; nil means disable message logging.  Lisp
code that calls `message' excessively (e.g. isearch.el) should probably
bind this variable to nil.

*** Display tables now have a new element, at index 261, specifying the
glyph to use for the separator between two side-by-side windows.  By
default, this is the vertical bar character `|'.  Probably the only
other useful character to store for this element is a space, to make
less visual separation between two side-by-side windows displaying
related information.

*** The new mode-line-format spec %c displays the current column number.

*** The new variable blink-matching-delay specifies how long to keep
the cursor at the matching open-paren, after you insert a close-paren.
This is useful mainly on systems which can wait for a fraction of a
second--you can then specify fractional values such as 0.5.

*** Faster processing of buffers with long lines

The new variable cache-long-line-scans determines whether Emacs
should use caches to handle long lines more quickly.  This variable is
buffer-local, in all buffers.

Normally, the line-motion functions work by scanning the buffer for
newlines.  Columnar operations (like `move-to-column' and
`compute-motion') also work by scanning the buffer, summing character
widths as they go.  This works well for ordinary text, but if the
buffer's lines are very long (say, more than 500 characters), these
motion functions will take longer to execute.  Emacs may also take
longer to update the display.

If cache-long-line-scans is non-nil, these motion functions cache
the results of their scans, and consult the cache to avoid rescanning
regions of the buffer until the text is modified.  The caches are most
beneficial when they prevent the most searching---that is, when the
buffer contains long lines and large regions of characters with the
same, fixed screen width.

When cache-long-line-scans is non-nil, processing short lines will
become slightly slower (because of the overhead of consulting the
cache), and the caches will use memory roughly proportional to the
number of newlines and characters whose screen width varies.

The caches require no explicit maintenance; their accuracy is
maintained internally by the Emacs primitives.  Enabling or disabling
the cache should not affect the behavior of any of the motion functions;
it should only affect their performance.

** System Interface

*** The function user-login-name now accepts an optional
argument uid.  If the argument is non-nil, user-login-name
returns the login name for that user id.

*** system-name, user-name, user-full-name and user-real-name are now
variables as well as functions.  The variables hold the same values
that the functions would return.  The new variable multiple-frames
is non-nil if at least two non-minibuffer frames are visible.  These
variables may be useful in constructing the value of frame-title-format
or icon-title-format.

*** Changes in time-conversion functions.

**** The new function format-time-string takes a format string and a
time value.  It converts the time to a string, according to the format
specified.  You can specify what kind of conversion to use with
%-specifications.

**** The new function decode-time converts a time value into a list of
specific items of information: the year, month, day of week, day of
month, hour, minute and second.  (A time value is a list of two or
three integers.)

**** The new function encode-time converts specific items of time
information--the second, minute, hour, day, month, year, and time
zone--into a time value.
\f
* Changes in Emacs 19.27

There are no changes; however, here is one bug fix made in 19.26 that users
think should be documented here.

** SPC and DEL in Info now handle menus consistently.

SPC and DEL scroll through an entire subtree an Info manual.  Once you
scroll through a node far enough to reach a menu, SPC begins moving
into the subnodes of the menu, starting with the first one.  When you
reach the end of a subnode, SPC moves into the next subnode, and so
on.

DEL more or less scrolls through the same text in reverse order.
\f
* User Editing Changes in Emacs 19.26

** In the X toolkit version, if you click on a menu bar item and
release the button quickly outside the menu, the menu remains visible
until you click or type something else.  If you click on the menu, you
select from the menu.  Any other mouse click makes the menu disappear.
Keyboard input gets rid of the menu and then is processed normally.

"Quickly" means within double-click-time milliseconds.

** The C-x 5 commands to select a buffer in "another frame" now use an
existing iconified frame, if any, deiconifying it.  They also raise
the frame.

** Region highlighting on a black-and-white-only display now uses
underlining.  Inverse-video had the problem that you couldn't see
the cursor.

** You can now change the height of a window by pressing mouse-1 on
the mode line and dragging it up and down.

** If you set the environment variable LC_CTYPE to iso_8859_1 or
iso-8859-1, Emacs automatically sets up for display and syntactic
handling of the ISO Latin-1 character set.

This does not automatically load any of the packages for input of
these characters, because it's not yet clear what is right to do.
You must still explicitly load either iso-transl or iso-acc.

** For a read-only buffer that is also modified, the mode line now displays
%* instead of %%.

** M-prior (scroll-other-window-down) is a new command that works like
M-next (and C-M-v) but scrolls in the opposite direction.

M-home moves to the beginning of the buffer, in the other window.
M-end moves to the end of the buffer, in the other window.  These two
commands, along with M-next and M-prior, form a series of commands for
moving around in the other window.

** In change logs, the mail address is now delimited with <...> instead
of (...).

This makes it a little more convenient to extract the mail address for
use in mailing a message.

** In Shell mode and other comint modes, C-a has now returned to
its ordinary meaning: move to the beginning of the line.
Use C-c C-a to move to the end of the prompt.

** If you set mail-signature to t to cause automatic insertion of
your .signature file, you now get a -- before the signature.

** Setting rmail-highlighted-headers to nil entirely turns off
highlighting in Rmail.  However, if your motivation for doing this is
that the highlighted text doesn't look good on your display, it might
be better to change the appearance of the `highlight' face.  Once
you've done that, you may find Rmail highlighting is useful.

** In the calendar, mouse-2 is now used only for commands that apply to a date.
If you click it when not on a date, it gives an immediate error.

Mouse-3 in the calendar now gives a menu of commands that do not apply
to a particular date.

The D command displays diary entries from a specified diary file (not
your standard diary file).

** In the gnus-uu package, the binding for gnus-uu-threaded-decode-and-view
is now C-c C-v C-d, not C-c C-v C-h.  Thus, C-c C-v C-h is now available
for asking for a list of the subcommands of C-c C-v.

** You can now specify "who you are" for various Emacs packages by
setting just one variable, user-mail-address.  This currently applies
to posting news with GNUS and to making change log entries.  It may
apply to additional Emacs features in the future.
\f
* Lisp-Level Changes in Emacs 19.26:

** The function insert-char now takes an optional third argument
which, if non-nil, says the inserted characters should inherit sticky
text properties from the surrounding text.

** The `diary' library has been renamed to `diary-lib'.  If you refer
to this library in your Lisp code, you must update the references.

** Sending text to a subprocess can read input from subprocesses if it
has to wait because the destination subprocess's terminal input buffer
is full.

It was already possible in unusual occasions for this operation to
read subprocess input, but it did not happen very often.  It is now
more likely to happen.

** last-nonmenu-event is now bound to t around filter functions and sentinels.
This is to ensure that y-or-n-p and yes-or-no-p use the keyboard by default.

** In mode lines, %+ now displays as % for unmodified read-only
buffers.  It is now the same as %* except in the case of a modified
read-only buffer; in that case, %+ displays as *.

The old meaning of %+ is now available on %&.
It displays * for a modified buffer and - for an unmodified buffer,
regardless of read-only status.

** You can now use `underline' in the color list of a face.
It serves as a last resort, and says to underline the face
(if previous color list elements can't be used).

** The new function x-color-values returns the list of color values
for a given color name (a string).  The list contains three integers
which give the amounts of red, green and blue in the color: (R G B).

** In run-at-time, 0 as the repeat interval means "don't repeat".

** The variable trim-versions-without-asking has been renamed to
delete-old-versions.

** The new function other-window-for-scrolling returns the choice of
other window for C-M-v to scroll.

** Note that the function fceiling was mistakenly documented as fceil before.
\f
* Changes in cc-mode.el in Emacs 19.26:

** A new syntactic symbol has been added: substatement-open.  It
  defines the open brace of a substatement block.  These used to get:
  ((block-open ...) (substatement . ...)).

  Non-block substatement lines still get just ((substatement . ...))

  Note that the custom indent function c-adaptive-block-open has been
  removed as obsolete.

** You can now specify the `hanginess' of closing braces.  See
  c-hanging-braces-alist.

** Recognizes try and catch blocks in C++.  They are given the
  substatement syntactic symbol.

** should be generally more forgiving about non-GNU standard top-level
  construct definition styles (i.e. where the function/class/struct
  opening brace does not start in column zero).

  If you hang the braces that open a top-level construct on the right
  edge, and you find you still need to define defun-open-prompt (Emacs
  19) please let me know.  Note that there may still be performance
  issues related to non-column zero opening braces.

** c-macro-expand is put on C-c C-e

** New style: "Default".  Resets indentation to those shipped with
  cc-mode.el.

** internal defun c-indent-via-language-element has been renamed
  c-indent-line for compatibility with c-mode.el and awk-mode.

** new buffer-local variable c-comment-start-regexp for (potential)
  flexibility in adding new modes based on cc-mode.el
\f
* Changes in Emacs 19.25

The variable x-cross-pointer-shape (which didn't really exist) has
been renamed to x-sensitive-text-pointer-shape, and now does exist.
\f
* Changes in Emacs 19.24

Here is a list of new Lisp packages introduced since 19.22.

derived.el              Define new major modes based on old ones.
dired-x.el              Extra Dired features.
double.el               New mode for conveniently inputting non-beyond chars.
easymenu.el             Create menus easily.
ediff.el                Snazzy diff interface.
foldout.el              A kind of outline mode designed for editing programs.
gnus-uu.el              UUdecode in GNUS buffers.
ielm.el                 Interactively evaluate Lisp.
                         This is a replacement for Lisp Interaction Mode.
iso-cvt.el              Conversion of beyond-ASCII characters between
                         various different representations.
jka-compr.el            Automatic compression/decompression.
mldrag.el               Drag modeline to change heights of windows.
mail-hist.el            Provides history for headers of outgoing mail.
rsz-mini.el             Automatically resizing minibuffers.
s-region.el             Set region by holding shift.
skeleton.el             Templates for statement insertion.
soundex.el              Classifying words by how they sound.
tempo.el                Template insertion with hotspots.
\f
* User Editing Changes in 19.23.

** Emacs 19.23 uses Ispell version 3.

Previous Emacs 19 versions used Ispell version 4.  That version had
improvements in storing the dictionary compactly, but these are not
very important nowadays.  Meanwhile, in parallel to the work on Ispell
4, many useful features were added to Ispell 3.  Until a few months
ago, the terms on Ispell 3 did not let us use it; but they have now
been changed, so now we are using it.  We are dropping Ispell 4.

** Emacs 19.23 can run on MS-DOG.  See the file MSDOS in the same
directory as this file.

** Emacs 19.23 can work with an X toolkit.  You must specify toolkit
operation when you configure Emacs: use the option
--with-x-toolkit=yes.  (This option uses code developed by Lucid;
thanks to Frederic Pierresteguy for helping to adapt it.)

** Emacs now has dialog boxes; yes/no and y/n questions automatically
use them in commands invoked with the mouse.  For more information,
see below under "Lisp programming changes".

** Menus now display the keyboard equivalents (if any) of the menu
commands in parentheses after the menu item.

** Kill commands, used in a read-only buffer, now move point across
the text they would otherwise have killed.  This way, you can use
repeated kill commands to transfer text into the kill ring.

** There is now a global mark ring in addition to the mark ring that is local
to each buffer.  The global mark ring stores positions in any buffer.  Any
time the mark is set and the current buffer is different from the last time
the mark was set, the new mark is pushed on the global mark ring as well.
The new command C-x C-SPC (pop-global-mark) pops the global mark ring and
jumps to the last mark pushed, first switching to that buffer.

** Query Replace is now available in the Edit menu.

** ESC no longer simply exits a Query Replace.  It now exits the Query
Replace and remains pending.  Thus, ESC A and M-A are now equivalent
in Query Replace.

To simply exit a Query Replace, type RET or Period.

** M-mouse-2 now puts point at the end of the yanked secondary selection.

** Mouse-1 in the mode line now simply selects the window above that
mode line.  Mouse-2 in the mode line selects that window and expands
it to fill the frame it is in.

** You can now use mouse-2 in a Dired buffer or Tar mode buffer to find
a file you click on, in a compilation buffer to go to a particular
error message, and in a *Occur* buffer to go to a particular
occurrence.

(It was already possible to do likewise in Info and in completion list
buffers.)

What's more, the sensitive areas of the buffer now highlight when you
move the mouse over them.

** In a completion list buffer, the command RET now chooses the completion
that is around or next to point.

** If you specify the foreground color for the `mode-line' face, and
mode-line-inverse-video is non-nil, then the default background color
is the usual foreground color.

** revert-buffer now preserves markers pointing within the unchanged
text (if any) at the beginning and end of the file.

** Version control checkin and checkout preserve all markers if the
file does not contain any of the magic version header sequences that
are updated automatically by RCS and SCCS.  If such version headers
are present, checkin and checkout preserve a marker unless it comes
between two such sequences.  (So it's a good idea to put all the
header sequences close together.)

** When a large deletion shuts off auto save temporarily in a buffer,
you can now turn it on again by saving the buffer with C-x C-s (as was
possible in Emacs 18).  You can also turn it on again with M-1 M-x
auto-save (as has been possible in Emacs 19).

** C-x r d now runs the command delete-rectangle.

** The new command imenu shows you a menu of interesting places in the
current buffer and lets you select one; then it moves point there.
The definition of interesting places depends on the major mode, but
typically this includes function definitions and such.  Normally,
imenu displays the menu in a buffer; but if you bind it to a mouse
event, it shows a mouse popup menu.

** You can make certain chosen buffers, that normally appear in a
separate window, appear in special frames of their own.  To do this,
set special-display-buffer-names to a list of buffer names; any buffer
whose name is in that list automatically gets a special frame when it
is to be displayed in another window.

A good value to try is ("*compilation*" "*grep*" "*TeX Shell*").

More generally, you can set special-display-regexps to a list of regular
expressions; then each buffer whose name matches any of those regular
expressions gets its own frame.

The variable special-display-frame-alist specifies the frame
parameters for these frames.  It has a default value, so you don't
need to set it.

** If you set sentence-end-double-space to nil, the fill commands
expect just one space at the end of a sentence.  (If you want the
sentence commands to accept single spaces, you must modify the regexp
sentence-end also.)

** You can suppress the startup echo area message by adding text like
this to your .emacs file:

(setq inhibit-startup-echo-area-message "YOUR-LOGIN-NAME")

Simply setting inhibit-startup-echo-area-message to your login name is
not sufficient to inhibit the message; Emacs explicitly checks whether
.emacs contains an expression as shown above.  Your login name must
appear in the expression as a Lisp string constant.

This way, you can easily inhibit the message for yourself if you wish,
but thoughtless copying of your .emacs file will not inhibit the
message for someone else.

** Outline minor mode now uses C-c C-o as a prefix instead of just C-c.

** In Outline mode, hide-subtree is now C-c C-d.  (It was C-c C-h; but
that is now a conventional way to ask for help about C-c commands.)

** There are two additional commands in Outline mode.
M-x hide-sublevels
  hides all headers except the topmost N levels.
M-x hide-other
  hides everything about the body that point is in
  plus the headers leading up from there to the top of the tree.

** In iso-transl and iso-insert, the sequences for entering A-ring and
the AE ligature are now just A and E (plus the initial C-x 8 or Alt).
You used to have to enter AA or AE, after the C-x 8 prefix of course.
Likewise for lower case a-ring and ae.

** iso-transl now defines convenient Alt keys as well as the C-x 8 prefix.
Instead of prefixing a sequence with C-x 8, you can add Alt to the
first character of the sequence.  For example, Alt-" a is now a way
to enter an a-umlaut.

** CC mode is a greatly improved mode for C and C++.
See the following page.

** tcl mode is a new major mode.  It provides features for
editing, indenting and running tcl programs.

** Compilation minor mode lets you parse error messages in any buffer,
not just a normal compilation output buffer.  Type M-x
compilation-minor-mode to enable the minor mode; then C-c C-c jumps to
the source location for the error at point, as in the `*compilation*'
buffer.  If you use compilation-minor-mode in an Rlogin buffer, it
automatically accesses remote source files by ftp.

** Comint and shell mode changes:

*** Comint modes (including Shell mode, GUD modes, etc.) now bind
C-M-l to the command comint-show-output.  This command scrolls the
buffer to show the last batch of output from the subprogram.

*** Completion in Comint modes now truly operates on the string before
point, rather than the word that point is within.

*** Comint mode file name completion ignores those files that end with a
string in the new variable comint-completion-fignore.  This variable's
default value is nil.

*** Shell mode uses the variable shell-completion-fignore to set
comint-completion-fignore.  The default value is nil, but some
people prefer ("~" "#" "%").

*** The function `comint-watch-for-password-prompt' can be used to 
suppress echoing when a subprocess asks for a password.  To use it,
do this:

(add-hook 'comint-output-filter-functions
          'comint-watch-for-password-prompt)

*** You can use M-x shell-strip-ctrl-m to strip ^M characters from
process output.

*** In Shell mode, TAB now completes environment variables, if possible,
and expands directory references.

*** You can use M-x comint-run to execute any program of your choice in
a comint mode.  Some programs such as shells, rlogin, and debuggers
have their own specialized modes; this command is one way to use
comint to run programs for which no such specialized mode exits.  (You
can also run a shell with M-x shell and run the program of your choice
under the shell--but that gives you the specializations of Shell
mode.)

** When you run GUD (M-x gdb, M-x dbx, and so on), you can use TAB
to do file name completion in the minibuffer.

The "Complete" menu includes an item for directory expansion.

** GUD working with future versions of GDB will permit TAB for
GDB-style symbol completion.  This will work with GDB 4.13.

** Rmail no longer gets new mail automatically when you visit an Rmail
file specified by name--not even if it is your primary Rmail file.  To
get new mail, type `g'.  This feature is an advantage because you now
have a choice of whether to get new mail.  (This change actually
occurred in an earlier version, but wasn't listed here then, since it
made the code do what the documentation already said.)

** Rmail now highlights certain fields automatically, when you use X
windows.  The variable rmail-highlighted-headers controls which
fields.

** If you set rmail-summary-window-size to an integer, Rmail uses
a window that many lines high for the summary buffer.

** rmail-input-menu is a new command that visits an Rmail file letting
you choose which file with a mouse menu.  rmail-output-menu is
similar; it outputs the current message, using a mouse menu to choose
which Rmail file.  These commands use the variables
rmail-secondary-file-directory and rmail-secondary-file-regexp.

** The mh-e package has been changed substantially.
See the file ./MH-E-NEWS for details.

** The calendar and diary have new features.

The menu bar for the calendar contains most of the calendar commands,
arranged into logical categories.

Mouse-2 now performs specific-date-related commands when clicked on a
date in the calendar window and common three-month-related commands
when clicked elsewhere in the calendar window.

You can set up colored/shaded highlighting of holidays, diary entry
dates, and today's date, by setting calendar-holiday-marker,
diary-entry-marker, and calendar-today-marker to a face instead of a
character.  Using a special face is now the default if you are using a
window system.

** The appt package for displaying appointment reminders has new
features.

*** The appt alarm window stays for the full duration of
appt-display-duration.  It no longer disappears when you start typing
text.

*** You can change the way the appointment window is created/deleted by
setting the variables appt-disp-window-function and
appt-delete-window-function.

For instance, these variables can be set to functions that display
appointments in pop-up frames, which are lowered or iconified after
appt-display-duration seconds.

** desktop.el can now save a list of buffer-local variables,
and saves more global ones.

** Pascal mode has been completely rewritten.  It now features
completing of function names, variables and type definitions around
current point (like M-TAB does with lisp-symbols).  There's also an
outline mode (M-x pascal-outline) that hides the bodies of all
functions you're not working with.

** Edebug has a number of changes:

*** Edebug syntax error reporting is improved.

*** Top-level forms and defining forms other than defun and defmacro may
now be debugged with Edebug.

*** Edebug specifications may now contain body, &define, name, arg or
arglist, def-body, and def-form, to support definitions.

*** edebug-all-defuns is renamed to edebug-all-defs.
def-edebug-form-spec is replaced by def-edebug-form whose arguments
are unevaluated.  The old names are still available for now.

*** Frequency counts and coverage data may be displayed for functions being
debugged.

*** A global break condition is now checked at every stop point.

*** The previous condition at a breakpoint may now be edited.

*** A new "next" mode stops only after expression evaluation.

*** A new command, top-level-nonstop, does not even stop for unwind-protect,
as top-level would.
\f
* Changes in CC mode in Emacs 19.23.

`cc-mode' provides ANSI C, K&R C, and ARM C++ language editing.  It
represents the merge of c++-mode.el and c-mode.el.  cc-mode provides a
new, more flexible indentation engine so that indentation
customization is more intuitive.  There are two steps to calculating
indentation: first, CC mode analyzes the line for syntactic content,
then based on this content it applies user defined offsets and adds
this offset to the indentation of some previous line.

The syntactic analysis determines if the line describes a `statement',
`substatement', `class-open', `member-init-intro', etc.  These are
described in detail with C-h v c-offsets-alist.  You can change the
offsets interactively with C-c C-o (c-set-offsets), or
programmatically in your c-mode-common-hook, which is run both by
c-mode and c++-mode.  You can also set up "styles" in the same way
that you could with c-mode.el.   The variable c-basic-offset controls
the basic offset given to a level of indentation.

If, for example, you wanted to change this style:

int foo (int i) 
{
    switch (i) {
    case 1:
        printf ("its a foo\n");
        break;
    default:
        printf ("don't know what it is\n");
        break;
    }
}

into this:

int foo (int i) 
{
    switch (i) {
      case 1:
        printf ("its a foo\n");
        break;
      default:
        printf ("don't know what it is\n");
        break;
    }
}

you could add the following to your .emacs file:

(defun my-c-mode-common-hook ()
  (c-set-offset 'case-label 2)
  (c-set-offset 'statement-case-intro 2))
(add-hook 'c-mode-common-hook 'my-c-mode-common-hook)

** New variables:

c-offsets-alist contains an association list of syntactic symbols and
their relative offsets.  Do a "C-h v c-offsets-alist" to get a list of
all syntactic symbols currently defined, and their meanings.  You
should not change this variable directly; use the supplied interface
commands c-set-offset and c-set-style.

c-mode-common-hook is run by both c-mode and c++-mode during their
common initializations.  You should put any customizations that are
the same for both C and C++ into this hook.

The variable c-strict-semantics-p is used mainly for debugging.  When
non-nil, CC mode signals an error if it returns a syntactic symbol
that can't be found in c-offsets-alist.

If you want CC mode to echo the syntactic analysis for a particular
line when you hit the TAB key, set c-echo-semantic-information-p to
non-nil.

c-basic-offset controls the standard amount of offset for a level of
indentation.  You can set a syntactic symbol's offset to + or - as a
short-hand for positive or negative c-basic-offset.

c-comment-only-line-offset lets you control indentation given to lines
which contain only a comment, in the case of C++ line style comments,
or the introduction to a C block comment.  Comment-only lines at
column zero can be anchored there independent of the indentation given
to other comment-only lines.

c-block-comments-indent-p controls the style of C block comment
re-indentation.  If you put leading stars in front of comment
continuation lines, you should set this variable to nil.

c-cleanup-list is a list describing certain C and C++ constructs to be
"cleaned up" as they are typed, but only when the auto-newline feature
is turned on.  In C++, make sure this variable contains at least
'scope-operator so that double colons will not be separated by a
newline.

Colons (`:') and braces (`{` and `}') are special in C and C++.  For
certain constructs, you may like them to hang on the right edge of the
code, or you may like them to start a new line of code.  You can use
the two variables c-hanging-braces-alist and c-hanging-colons-alist
to control whether newlines are placed before and/or after colons and
braces when certain C and C++ constructs are entered.  For example,
you can control whether the colon that introduces a C++ member
initialization list hangs on the right edge, starts a new line, or has
no newlines either before or after it.

c-special-indent-hook is run after a line is indented by CC mode.  You
can perform any custom indentations here.

c-delete-function is the function that is called when a single
character is deleted with the c-electric-delete command (DEL).

c-electric-pound-behavior describes what happens when you enter the
`#' that introduces a cpp macro.

If c-tab-always-indent is neither t nor nil, then TAB inserts a tab
when within strings, comments, and cpp directives, but it reindents
the line unconditionally.

c-inhibit-startup-warnings-p inhibits warnings about any old
version of Emacs you might be running, which could be incompatible
with cc-mode.

** There are two new minor-mode features in CC mode: auto-newline and
hungry-delete.  Auto-newline inserts newlines automatically as you
type certain constructs.  Hungry-delete consumes all preceding
whitespace (spaces, tabs, and newlines) when the delete key is hit.
You can toggle auto-newline on and off on a per-buffer basis by
hitting C-c C-a.  You can toggle hungry-delete on and off by hitting
C-c C-d.   You can toggle them both on and off together with C-c C-t.

** Slash (`/') and star (`*') are now both electric characters.

** New commands:

The new C-c C-o (c-set-offset) command can be used to interactively change
the offset for a particular syntactic symbol.

The new command C-c : (c-scope-operator) inserts the C++ scope operator in
c++-mode only.

The new command C-c C-q (c-indent-defun) indents the entire enclosing
top-level function or class.

The new command C-c C-s (c-show-semantic-information) echos the current
syntactic analysis without re-indenting the current line.

The new commands M-x c-forward-into-nomenclature and M-x
c-backward-into-nomenclature (currently otherwise unbound to a key
sequence), make movement easier when using the C++ variable naming
convention of VariableNamesWithoutUnderscoresButEachWordCapitalized.

** Command from c-mode.el that have been renamed in cc-mode.el:

  electric-c-brace      => c-electric-brace
  electric-c-semi       => c-electric-semi&comma
  electric-c-sharp-sign => c-electric-pound
  mark-c-function       => c-mark-function
  electric-c-terminator => c-electric-colon
  indent-c-exp          => c-indent-exp
  set-c-style           => c-set-style

** Variables from c-mode.el that are obsolete with cc-mode.el:

  c-indent-level
  c-brace-imaginary-offset
  c-brace-offset
  c-argdecl-indent
  c-label-offset
  c-continued-statement-offset
  c-continued-brace-offset
\f
* Lisp programming changes in Emacs 19.23.

** To pop up a dialog box, call x-popup-dialog.
It takes two arguments, POSITION and CONTENTS.

POSITION specifies which frame to place the dialog box over;
the dialog box always goes on the center of the frame.
POSITION may be a mouse event, a window, a frame,
or t meaning use the frame that the mouse is in.

CONTENTS specifies the contents of the dialog box.
It looks like a single pane of a popup menu:
(TITLE ITEM1 ITEM2 ...), where each ITEM has the form (STRING . VALUE).
The return value is VALUE from the chosen item.

An ITEM may also be just a string--that makes a nonselectable item.
An ITEM may also be nil--that means to put all preceding items
on the left of the dialog box and all following items on the right.
(By default, approximately half appear on each side.)

If your Emacs is not using an X toolkit, then it cannot display a
real dialog box; so instead it displays a pop-up menu in the center
of the frame.

** y-or-n-p, yes-or-no-p and map-y-or-n-p now use menus or dialog boxes
to ask their question(s) if the command that is running was reached by
a mouse event.

If you want to control which way these functions work, bind the
variable last-nonmenu-event around the call.  These functions use the
keyboard if that variable holds a keyboard event (actually, any
non-list); they use the mouse if that variable holds a mouse event
(actually, any list).

** The mouse-face property is now implemented, both in overlays and as
a text property.  It specifies a face to use when the mouse is in the
range of text for which the property is specified.

** When text has a non-nil `intangible' property, you cannot move point
within it or right before it.  If you try, point actually moves to the
end of the intangible text.  Note that this means that backward-char
is a no-op when there is an intangible character to the left of point.

** minibuffer-exit-hook is a new normal hook that is run when you
exit the minibuffer.

** The variable x-cross-pointer-shape specifies the cursor shape to use
when the mouse is over text that has a mouse-face property.

** The new variable interpreter-mode-alist specifies major modes to use
for shell scripts that specify a command interpreter.  Its elements
look like (INTERPRETER . MODE); for example, ("perl" . perl-mode) is
one element present by default.  This feature applies only when the
file name doesn't indicate which mode to use.

** If you use a minibuffer-only frame, set the variable
minibuffer-auto-raise to t, and entering the minibuffer will then
raise the minibuffer frame.

** If pop-up-frames is t, display-buffer now looks for an existing
window in any visible frame, showing the specified buffer, and uses
such a window in preference to making a new frame.

** In the functions next-window, previous-window, next-frame,
previous-frame, get-buffer-window, get-lru-window, get-largest-window
and delete-windows-on, if you specify `visible' for the last argument,
it means to consider all visible frames.

** Mouse events now give the X and Y coordinates in pixels, rather than
in characters.  You can convert these values to characters by dividing by
the values of (frame-char-width) and (frame-char-height).

** The new functions mouse-pixel-position and set-mouse-pixel-position
read and set the mouse position in units of pixels.  The existing
functions mouse-position and set-mouse-position continue to work with
units of characters.

** The new function compute-motion is useful for computing the width
of certain text when it is displayed.

** The function vertical-motion now takes an option second argument WINDOW
which says which window to use for the display calculations.

vertical-motion always operates on the current buffer.
It is ok to specify a window displaying some other buffer.
Then vertical-motion uses the width, hscroll and display-table of
the specified window, but still scans the current buffer.

** An error no longer sets last-command to t; the value of last-command
does reflect the previous command (the one that got an error).

If you do not want a particular command to be recognized as the
previous command in the case where it got an error, you must code that
command to prevent this.  Set this-command to t at the beginning of
the command, and set this-command back to its proper value at the end,
like this:

  (defun foo (args...)
    (interactive ...)
    (setq this-command t)
    ...do the work...
    (setq this-command 'foo))

or like this:

  (defun foo (args...)
    (interactive ...)
    (let ((old-this-command this-command))
      (setq this-command t)
      ...do the work...
      (setq this-command old-this-command)))

The undo and yank commands do this.

** If you specify an explicit title for a new frame when you create it,
the title is used as the resource name when looking up X resources to
control the shape of that frame.  If you don't specify the frame title, 
the value of x-resource-name is used, as before.

** The frame parameter user-position, if non-nil, says that the user
has specified the frame position.  Emacs reports this to the window
manager, to tell it not to override the position that the user
specified.

** Major modes can now set change-major-mode-hook to arrange for state
to be cleaned up when the user switches to a new major mode.  The function
kill-all-local-variables runs this hook.  For best results, make the hook a
buffer-local variable so that it will disappear after doing its job and will
not interfere with the subsequent major mode.

** The new variable overriding-local-map, if non-nil, specifies a keymap
that overrides the current local map, all minor mode keymaps, and all
text property keymaps.  Incremental search uses this feature to override
all other keymaps temporarily.

** A key definition in a menu keymap can now have additional structure:
in addition to (ITEMNAME [HELPSTRING] . COMMAND) which was allowed
before, the form (ITEMNAME [HELPSTRING] (...) . COMMAND) is
allowed.  (HELPSTRING is optional, and is not currently used.)

Here (...) represents a sublist containing information about keyboard
key sequences that run the same command COMMAND.  Displaying the menu
automatically creates and updates the sublist when appropriate; you
need never set these up yourself.

lookup-key, key-binding, and similar functions return just COMMAND,
not the whole binding.

To precompute this information for a given keymap, you can do
  (x-popup-menu nil KEYMAP).

** When you specify coordinates for x-popup-menu as a list ((XOFFSET
YOFFSET) WINDOW), the coordinates are now measured in pixels.

** where-is-internal now takes just four arguments:
DEFINITION KEYMAP FIRSTONLY NOINDIRECT.
The single argument KEYMAP replaces two arguments KEYMAP and KEYMAP1.

If KEYMAP is non-nil, where-is-internal searches only KEYMAP and the
global keymap.

If KEYMAP is nil, where-is-internal searches all the currently active
keymaps, but finds the active keymaps as if overriding-local-map were
nil.

If you pass a list of the form (keymap) as KEYMAP, where-is-internal
searches only the global map.  (This is not a special case--it follows
from the specifications above.)

If you pass the value of overriding-local-map as KEYMAP, where-is-internal
searches in exactly the same was as command execution does.

** Use the macro define-derived-mode to define a new major mode that
inherits the definition of another major mode.  Here's how to define a
command named hypertext-mode that inherits from the command text-mode:

    (define-derived-mode hypertext-mode text-mode "Hypertext"
      "Major mode for hypertext.\n\n\\{hypertext-mode-map}"
      (setq case-fold-search nil))

    (define-key hypertext-mode-map [down-mouse-3] 'do-hyper-link)

The new mode has its own keymap, which inherits from that of the
original mode.  It also has its own syntax and abbrev tables, which
are initialized by copying those of the original mode.  It also has
its own mode hook.  All are given names made by appending a suffix
to the name of the new mode.

** A syntax table can now inherit the data for some characters from
standard-syntax-table, while specifying other characters itself.
Syntax code 13 means "inherit this character from the standard syntax
table."  In modify-syntax-entry, the character `@' represents this code.

The function `make-syntax-table' now creates a syntax table which
inherits all letters and control characters (0 to 31 and 128 to 255)
from the standard syntax table, while copying the other characters
from the standard syntax table.  Most syntax tables in Emacs are set
up this way.

This sort of inheritance is useful for people who set up character
sets with additional alphabetic characters in the range 128 to 255.
Just changing the standard syntax for these characters affects all
major modes.

** The new function transpose-regions swaps two regions of the buffer.
It preserves the markers in those two regions, so that they stay with
the surrounding text as it is swapped.

** revert-buffer now runs before-revert-hook at the beginning and
after-revert-hook at the end.  These can be used by minor modes
that need to clean up state variables.

** The new function get-char-property is like get-text-property, but
checks for overlays with properties as well as for text properties.
It checks for overlays first, in order of descending priority, and
text properties last.

get-char-property allows windows as the OBJECT argument, as well
as buffers and strings.  If you specify a window, then only overlays
active on that window are considered.

** Overlays can have the `invisible' property.

** The function insert-file-contents now takes an optional fifth 
argument called REPLACE.  If this is t, it means to replace the
contents of the buffer (actually, just the accessible portion)
with the contents of the file.

This is better than simply deleting and inserting the whole thing
because (1) it preserves some marker positions and (2) it puts less
data in the undo list.

** The variable inhibit-first-line-modes-regexps specifies classes of
file names for which -*- on the first line should not be looked for.

** The variables before-change-functions and after-change-functions
hold lists of functions to call before and after a change in the
buffer's text.  They work much like before-change-function and
after-change-function, except that they hold a list of functions
instead of just one.

These variables will eventually make before-change-function and
after-change-function obsolete.

** The variable kill-buffer-query-functions holds a list of functions
to be called with no arguments when a buffer is about to be killed.
(That buffer is the current buffer when the function is called.)
If any of the functions returns nil, the buffer is not killed
(and the remaining functions in the list are not called).

** The variable kill-emacs-query-functions holds a list of functions
to be called with no arguments when you ask to exit Emacs.
If any of the functions returns nil, the exit is canceled
(and the remaining functions in the list are not called).

** The argument for buffer-disable-undo is now optional,
like the argument for buffer-enable-undo.

** The new variable system-configuration holds the canonical three-part
GNU configuration name for which Emacs was built.

** The function system-name now tries harder to return a fully qualified
domain name.

** The variable emacs-major-version holds the major version number
of Emacs.  (Currently 19.)

** The variable emacs-minor-version holds the minor version number
of Emacs.  (Currently 23.)

** The default value of comint-input-autoexpand is now nil.
However, Shell mode sets it from the value of shell-input-autoexpand,
whose default value is `history'.

** The new function set-process-window-size specifies the terminal window
size for a subprocess.  On some systems it sends the subprocess a signal
to let it know that the size has changed.

** %P is a new way to display a percentage in the mode line.  It
displays the percentage of the buffer text that is above the *bottom*
of the window (which includes the text visible, in the window as well
as the text above the top).  It displays `Top' as well as the
percentage if the top of the buffer is visible on screen.

** %+ in the mode line specs displays `*' if the buffer is modified,
and otherwise `-'.  It never displays `%', as `%*' would do; whether the
buffer is read-only has no effect on %+.

** The new functions ffloor, fceiling, fround and ftruncate take a
floating point argument and return a floating point result whose value
is a nearby integer.  ffloor returns the nearest integer below; fceiling,
the nearest integer above; ftruncate, the nearest integer in the
direction towards zero; fround, the nearest integer.

** Setting `print-escape-newlines' to a non-nil value now also makes
formfeeds print as ``\f''.

** auto-mode-alist now has a new feature.  If an element has the form
(REGEXP FUNCTION t), and REGEXP matches the file name, then after calling
FUNCTION, Emacs deletes the part of the file name that matched REGEXP
and then searches auto-mode-alist again for a new match.

This is useful for uncompression packages.  An entry of this sort for
.gz can uncompress the file and then put the uncompressed file in the
proper mode according to the name sans .gz.

** The new function emacs-pid returns the process ID number of Emacs.

** user-login-name now consistently checks the LOGNAME environment
variable before USER.  user-original-login-name is obsolete, since it
provides the same functionality.  To ignore the environment variables,
use user-real-login-name.

** There is a more general way of handling the system-specific X
keysyms.  Set the variable system-key-alist to an alist containing
elements of the form (CODE . SYMBOL), where CODE is the numeric keysym
code minus the "vendor specific" bit, and symbol is the name for the
function key.

** You can use the variable command-line-functions to set up functions
to process unrecognized command line arguments.  The variable's value
should be a list of functions of no arguments.  The functions are
called successively until one of them returns non-nil.

Each function should access the free variables argi (the current
argument) and command-line-args-left (the remaining arguments).  The
function should return non-nil only if it recognizes and processes the
argument in argi.  If it does so, it may consume following arguments
as well by removing them from command-line-args-left.

** There's a new way for a magic file name handler to run a primitive
and inhibit handling of the file name.  Here is how to do it:

(let ((inhibit-file-name-handlers
       (cons 'ange-ftp-file-handler 
             (and (eq inhibit-file-name-operation operation)
                      inhibit-file-name-handlers)))
      (inhibit-file-name-operation operation))
  (apply this-operation args))

The function find-file-name-handler now takes two arguments.  The
second argument is OPERATION, the operation for which the handler is
being sought.

People have suggested that the second argument should be optional, for
backward compatibility.  It would be nice if that were possible, but
it is not.  There is simply no way for find-file-name-handler to do
the right thing without receiving the proper value for its second
argument.

** The variable completion-regexp-list affects the completion
primitives try-completion and all-completions.  They consider
only the possible completions that match each regexp in the list.

** Case conversion in the function replace-match has been changed.

The old behavior was this: if any word in the old text was
capitalized, replace-match capitalized each word of the replacement
text.

The new behavior is this: if the first word in the old text is capitalized,
replace-match capitalizes the first word of the replacement text.

** You can now specify a case table with CANON non-nil and EQV nil.
Then the EQV part of the case table is deduced from CANON.

** The new function minibuffer-prompt takes no arguments and returns
the current minibuffer prompt string.

The new function minibuffer-prompt-width takes no arguments and
returns the display width of the minibuffer prompt string.

** The new function frame-first-window returns the window at the
upper left corner of a given frame.

** wholenump is a new alias for natnump.

** The variable installation-directory, if non-@code{nil}, names a
directory within which to look for the `lib-src' and `etc'
subdirectories.  This is non-nil when Emacs can't find those
directories in their standard installed locations, but can find them
near where the Emacs executable was found.

** invocation-name and invocation-directory are now variables as well
as functions.  The variable values are the same values that the
functions return: the Emacs program name sans directories, and the
directory it was found in.  (invocation-directory may be nil, if Emacs
can't determine which directory it should be.)

** Installation change regarding version number counting.

The version number of an Emacs executable contains three numbers.
The first two describe the Emacs release and the third increments
each time you build Emacs.

Now the file version.el contains only the first two version numbers.
The third component is now determined on the basis of the names of the
existing executable files.  This means that version.el is not altered
by building Emacs.
\f
* Changes in 19.22.

** The mouse click M-mouse-2 now inserts the current secondary
selection (from Emacs or any other X client) where you click.
It does not move point.
This command is called mouse-yank-secondary.

mouse-kill-secondary no longer has a key binding by default.
Clicking M-mouse-3 (mouse-secondary-save-then-kill) twice
may be a convenient enough way of killing the secondary selection.
Or perhaps there should be a keyboard binding for killing the
secondary selection.  Any suggestions?

** New packages:

*** `icomplete' provides character-by-character information
about what you could complete if you type TAB.

*** `avoid' moves the mouse away from point so that it doesn't hide
your typing.

*** `shadowfile' helps you update files that are supposed to be stored
identically in different places (perhaps on different machines).

** C-h p now knows about four additional keywords: data, faces, mouse,
and matching.

** The key for starting an inferior Lisp process, in Lisp mode,
is now C-c C-z instead of C-c C-l.

** When the VC commands ask whether to save the buffer, if you say no,
they signal an error.  This is so that you won't operate on the wrong
data.

** ISO Accents mode now supports `"s' as a way of typing German sharp s.

** By default, comint buffers (including Shell mode and debuggers)
no longer try to scroll to keep the cursor on the bottom line.
This feature was added in 19.21 but did not work smoothly enough.

** Emacs now handles the window manager "delete window" operation.

** Display of buffers with text properties is much faster now.

** The feature previously announced whereby `insert' does not inherit
text properties from surrounding text was not fully implemented
before; but now it is.  use `insert-and-inherit' if you wish to
inherit sticky properties from the surrounding text.

** The functions next-property-change, previous-property-change,
next-single-property-change, and previous-single-property-change
now take one additional optional argument LIMIT that is a position at
which to stop scanning.  If scan ends without finding the property
change sought, these functions return the specified limit.

The value returned by previous-single-property-change and
previous-property-change, when they do find a change, is now one
greater than what it used to be.  It is the position between the two
characters whose properties differ, which is one greater than the
position of the first character found (while scanning back) with
different properties.
\f
* User editing changes in version 19.21.

** ISO Accents mode supports four additional characters:
A-with-ring (entered as /A), AE ligature (entered as /E),
and their lower-case equivalents.
\f
* User editing changes in version 19.20.
(See following page for Lisp programming changes.)

Note that some of these changes were made subsequent to the Emacs 19.20
editions of the Emacs manual and Emacs Lisp manual; therefore, if you
have those editions, do read this page.

** Dragging with mouse button 1 now puts the selected region
in the kill ring so you can paste it into other X applications.

** Double and triple clicks with button 1 now behave as in xterm,
selecting the word or line surrounding where you click.  If you drag
after the last click, you can select a range of words or lines.

** You can use button 3 to extend a mouse-selected region, as in xterm.
This works for regions selected either by dragging Mouse-1 or by
multiple-clicking Mouse-1.  Clicking Mouse-3 moves the end of the
region that is (initially) nearer to where you click.

If the selection was first made by multiple-clicking Mouse-1, and thus
consists of entire words or lines, Mouse-3 preserves that state.

As before, clicking Mouse-3 again in the same place kills the region
thus selected.

** The secondary selection commands, M-Mouse-1 and M-Mouse-3, have been
likewise modified.

** You can now search for strings and regexps using the Edit menu bar menu.

** You can now access bookmarks using the Bookmark submenu in the File
menu in the menu bar.

** ISO Accents mode, a buffer-local minor mode, provides a convenient
way to type certain non-ASCII characters.  It makes the characters `,
', ", ^, ~ and / serve as modifiers for the following letter.  ` and '
add accents, " adds an umlaut or dieresis, ^ adds a circumflex, ~
adds a tilde, and / adds a slash to the following letter.

If the following character is not a letter, or cannot be modified as
requested, then both characters stand for themselves.  If you
duplicate the modifier accent character, that enters the corresponding
ISO non-spacing accent character (thus, '' enters the ISO acute-accent
character).  To enter a modifier character itself, type it followed by
a space.

This feature can be used whenever a key sequence is expected: for
ordinary insertion, for searching, and for certain command arguments.

A few special combinations:

~c => c with cedilla
~d => d with stroke
~< => left guillemot
~> => right guillemot

** iso-transl.el is a new library that replaces iso-insert.el.
It defines C-x 8 as an insertion prefix for the ISO characters
between 128 and 255, much like iso-insert, except that iso-transl
works even in searches and help commands--wherever a key sequence
is expected.

To define case-conversion for these characters for ISO 8859/1,
load the library iso-syntax.  (This is not new.)

** M-TAB in Text mode now runs the command ispell-complete-word
which performs completion using the spelling dictionary.

The spelling correction submenu now includes this command 
and another command which completes a word fragment (that is,
it doesn't assume that the text to be completed starts at the
beginning of a word.

** In incremental search, you can use M-y to yank the most recent kill
into the search string.

** The new function ispell-message checks the spelling of a message
you are about to send or post.  It ignores text cited from other
messages.

To automatically check all your outgoing messages, include the
following line in your .emacs file:
  (setq news-inews-hook (setq mail-send-hook 'ispell-message))

** There is now a separate minibuffer history list for the names of
extended commands.  This history list is used by M-x when reading
the command name.  The motivation for this is to prevent command
names from appearing in the history used for other minibuffer
arguments.

Note that the history list for entire commands that use the minibuffer
is a separate feature.  That history list records a command with all
its arguments, and you must use C-x ESC ESC to access it.

** You can use the new command C-x v ~ VERSION RET to examine a
specified version of a file that is maintained with version control.

** In Indented Text mode, only blank lines now separate paragraphs.
Indented lines continue the paragraph that is in progress.  This makes
the user option variable adaptive-fill-mode have its intended effect.

** Local variable specifications in files for variables whose names end
in `-hook' and `-function' are now controlled by the variable
`enable-local-eval', just like the `eval' variable.

** C-x r j (jump-to-register) when restoring a frame configuration now
makes all unwanted frames (existing frames not mentioned in the
configuration) invisible.

If you want to delete these unwanted frames, use a prefix argument for
C-x r j.

** You can customize the calendar to display weeks beginning on
Monday: set the variable `calendar-week-start-day' to 1.

** Rmail changes.

If you save messages to a file in Unix format while viewing a message
with its whole header, this now copies to the file the entire header
of each message copied.

** Comint mode changes.

C-c C-e shows as much output as possible in the window.
C-c RET copies an old input (the one at point)
and places the copy after the latest prompt.
C-c C-p and C-c C-n move through the buffer, stopping at places
where the subshell prompted for input.
C-c C-h lists the input history in a `*Help*' buffer.

There are new menu bar items for completion/input/output/signal commands.

Input behaviour is configurable.  Variables control whether some windows
showing the buffer scroll to the bottom before insertion.  These are
`comint-scroll-to-bottom-on-input' and `before-change-function'.  By default,
insertion causes the selected window to scroll to the bottom before insertion
occurs.

Subprocess output now keeps point at the end of the buffer in each
window individually if point was already at the end of the buffer in
that window.

If `comint-scroll-show-maximum-output' is non-nil (which is the
default), then scrolling due to arrival of output tries to place the
last line of text at the bottom line of the window, so as to show as
much useful text as possible.  (This mimics the scrolling behavior of
many terminals.)

By setting `comint-scroll-to-bottom-on-output', you can opt for having
point jump to the end of the buffer whenever output arrives--no matter
where in the buffer point was before.  If the value is `this', point
jumps in the selected window.  If the value is `all', point jumps in
each window that shows the comint buffer.  If the value is `other',
point jumps in all nonselected windows that show the current buffer.
The default value is nil, which means point does not jump to the end.

Input history insertion is configurable.  A variable controls whether only the
first instance of successive identical inputs is stored in the input history.
This is `comint-input-ignoredups'.

Completion (bound to TAB) is now more general.  Depending on context,
completion now operates on the input history, on command names, or (as
before) on filenames.

Filename completion is configurable.  Variables control whether
file/directory suffix characters are added (`comint-completion-addsuffix'),
whether shortest completion is acceptable when no further unambiguous
completion is possible (`comint-completion-recexact'), and the timing of
completion candidate listing (`comint-completion-autolist').

Comint mode now provides history expansion.  Insert input using `!'
and `^', in the same syntax that typical shells use; then type TAB.
This searches the comint input history for a matching element,
performs substitution if necessary, and places the result in the
comint buffer in place of the original input.

History references in the input may be expanded before insertion into
the input ring, or on input to the interpreter (and therefore
visibly).  The variable `comint-input-autoexpand' specifies which.

You can make the SPC key perform history expansion by binding
SPC to the command `comint-magic-space'.

The command `comint-dynamic-complete-variable' does variable name
completion using the environment variables as set within Emacs.  The
variables controlling filename completion apply to variable name
completion too.  This command is normally available through the menu
bar.

** Shell mode

Paragraph motion and marking commands (default bindings M-{, M-}, M-h) operate
on output groups (i.e., shell prompt plus associated shell output).

TAB now completes commands, as well as file names and expand history.
Commands are searched for along the path that Emacs has on startup.

C-c C-f now moves forward a command (`shell-forward-command') and
C-c C-b now moves backward a command (`shell-backward-command').

Command completion is configurable.  The variables controlling
filename completion in comint mode apply, together with a variable
controlling whether to restrict possible completions to only files
that are executable (`shell-command-execonly').

The input history is initialised from the file name given in the
variable `shell-input-ring-file-name'--normally `.history' in your
home directory.

Directory tracking is more robust.  It can cope with command sequences
and forked commands, and can detect the failure of directory changing
commands in most circumstances.  It's still not infallible, of course.

You can now configure the behaviour of `pushd'.  Variables control
whether `pushd' behaves like `cd' if no argument is given
(`shell-pushd-tohome'), pop rather than rotate with a numeric argument
(`shell-pushd-dextract'), and only add directories to the directory
stack if they are not already on it (`shell-pushd-dunique').  The
configuration you choose should match the underlying shell, of course.
\f
* Emacs Lisp programming changes in Emacs 19.20.

** A new function `remove-hook' is now used to remove a hook that you might
have added with `add-hook'.

** There is now a Lisp pretty-printer in the library `pp'.

** The partial Common Lisp support has been entirely reimplemented.

** When you insert text using `insert', `insert-before-markers' or
`insert-buffer-substring', text properties are no longer inherited
from the surrounding text.

When you want to inherit text properties, use the new functions
`insert-and-inherit' or `insert-before-markers-and-inherit'.

The self-inserting character command does do inheritance.

** Frame creation hooks.

The function make-frame now runs the normal hooks
before-make-frame-hook and after-make-frame-hook.

** You can now use function-key-map to make a key an alias for other
key sequences that can vary depending on circumstances.  To do this,
give the key a definition in function-key-map which is a function
rather than a specific expansion key sequence.

If the function reads input itself, it can have the effect of altering
the event that follows.  For example, here's how to define C-c h to
turn the character that follows into a hyper character:

(define-key function-key-map "\C-ch" 'hyperify)

(defun hyperify (prompt)
  (let ((e (read-event)))
    (vector (if (numberp e)
                (logior (lsh 1 20) e)
              (if (memq 'hyper (event-modifiers e))
                  e
                (add-event-modifier "H-" e))))))

(defun add-event-modifier (string e)
  (let ((symbol (if (symbolp e) e (car e))))
    (setq symbol (intern (concat string (symbol-name symbol))))
    (if (symbolp e)
        symbol
      (cons symbol (cdr e)))))

The character translation function gets one argument, which is the
prompt that was specified in read-key-sequence--or nil if the key
sequence is being read by the editor command loop.  In most cases
you can just ignore the prompt value.

** Changes for reading and writing text properties.

New low-level Lisp features make it possible to write Lisp programs to
save text properties in files, and read text properties from files.
You can program any file format you like.

The variable `write-region-annotation-functions' should contain a list
of functions to be run by `write-region' to encode text properties in
some fashion as annotations to the text that is written.

Each function in the list is called with two arguments: the start and
end of the region to be written.  These functions should not alter the
contents of the buffer.  Instead, they should return lists indicating
annotations to write in the file in addition to the text in the
buffer.

Each function should return a list of elements of the form (POSITION
. STRING), where POSITION is an integer specifying the relative
position in the text to be written, and STRING is the annotation to
add there.

Each list returned by one of these functions must be already sorted in
increasing order by POSITION.  If there is more than one function,
`write-region' merges the lists destructively into one sorted list.

When `write-region' actually writes the text from the buffer to the
file, it intermixes the specified annotations at the corresponding
positions.  All this takes place without modifying the buffer.

The variable `after-insert-file-functions' should contain a list of
functions to be run each time a file's contents have been inserted into
a buffer.  Each function receives one argument, the length of the
inserted text; point indicates the start of that text.  The function
should make whatever changes it wants to make, then return the updated
length of the inserted text, as it stands after those changes.  The
value returned by one function is used as the argument to the next.
These functions should always return with point at the beginning of
the inserted text.

The intended use of `after-insert-file-functions' is for converting
some sort of textual annotations into actual text properties.  But many
other uses may be possible.

We now invite users to begin implementing Lisp programs to store and
retrieve text properties in files, using these new primitive features,
and thus to experiment with various data formats and find good ones.

We suggest not trying to handle arbitrary Lisp objects as property
names or property values--because a program that general is probably
difficult to write, and slow.  Instead, choose a set of possible data
types that are reasonably flexible, and not too hard to encode.

** Comint completion.

Currently comint-dynamic-complete-command (and associated variable
comint-after-partial-pathname-command) are set by default to complete a
filename.  Other comint-mode users should have their own functions to achieve
this.  For example, gud-mode could complete debugger commands.  A completion
function is provided solely for this reason (comint-dynamic-simple-complete).

Other comint-mode users should bind comint-dynamic-complete (shell-mode does
already).

** Comint history reference expansion

Currently comint-input-autoexpand is 'history, which means only expand
history on insertion to comint-input-ring.  For non-shell modes, this is
a strange default, since non-shells will not understand history references.
Perhaps it would be better for the variable to be 'input, which means expand
on RET.

The value 'history might possibly be wrong even for shells, since the
expansion will be done both by comint and the underlying shell (except sh, of
course).  It would be better for expansion to be done by one or the other,
not both since they may (ahem) disagree.  Since it is silly to put a literal
history reference into comint-input-ring, perhaps it would be better for the
variable to be 'input too.

The reason the variable is not 'input by default is that I was attempting to
adhere to The Principle of Least Astonishment.  I didn't want to shock users
by having their input change in front of their eyes.

** Argument delimiters and Comint mode.

Currently comint-delimiter-argument-list is '(), which means no strings are
to be treated as delimiters and arguments.  In shell-mode, this variable is
set to shell-delimiter-argument-list, '("|" "&" "<" ">" "(" ")" ";").  Other
comint-mode users should set this variable too.  For example, a lisp-type
mode might want to set this to '("." "(" ")") or some such.

** Comint output hook.

There is now a hook, comint-output-filter-hook, that is run-hooks'ed by the
output filter, comint-output-filter.  This is useful for scrolling (see
below), but also things like processing output for specific text, output
highlighting, etc.

So that such output processing may be done efficiently, there is a new
variable, comint-last-output-start, that records the position of the start of
the lastest output inserted into the buffer (effectively the previous value
of process-mark).  Output processing functions should process the text
between comint-last-output-start (or perhaps the beginning of the line that
the position lies on) and process-mark.

** Comint scrolling.

There is now automatic scrolling of process windows.

Currently comint-scroll-show-maximum-output is t, which means when scrolling
output put process-mark at the bottom of the window.  There is a good case
for it to be t, since the user is likely to want to see as much output as
possible.  But, then again, there is a comint-show-maximum-output command.

** Comint history retrieval.

The input following point is not deleted when moving around the input history
(with M-p etc.).  Emacs maintainers may not like this.  However, I feel this
is a useful feature.  The simple remedy is to put end-of-line in before
delete-region in comint-previous-matching-input.

The input history retrieval commands still wrap-around the input ring, unlike
Emacs command history.
\f
* Changes in version 19.19.

** The new package bookmark.el records named bookmarks: positions that
you can jump to.  Bookmarks are saved automatically between Emacs
sessions.

** Another simpler package saveplace.el records your position in each
file when you kill its buffer (or kill Emacs), and jumps to the same
position when you visit the file again (even in another Emacs
session).  Use `toggle-save-place' to turn on place-saving in a given file;
use (setq-default save-place t) to turn it on for all files.

** In Outline mode, you can now customize how to compute the level of a
heading line.  Set `outline-level' to a function of no arguments which
returns the level, assuming point is at the beginning of a heading
line.

** You can now specify the prefix key to use for Outline minor mode.
(The default is C-c.)  Set the variable outline-minor-mode-prefix to
the key sequence you want to use (as a string or vector).

** In Bibtex mode, C-c e has been changed to C-c C-b.  This is because
C-c followed by a letter is reserved for users.

** The `mod' function is no longer an alias for `%', but is a separate function
that yields a result with the same sign as the divisor.  `floor' now takes an
optional second argument, which divides the first argument before the floor is
taken.

** `%' no longer allows floating point arguments, since the results were often
inconsistent with integer `%'.
\f
* Changes in version 19.18.

** Typing C-z in an iconified Emacs frame now deiconifies it.

** hilit19 is a new library for automatic highlighting of parts of the
text in the buffer, based on its meaning and context.

** Killing no longer sends the killed text to the X clipboard.
And large strings are not put in the cut buffer either.
The variable x-cut-buffer-max specifies the maximum number of characters
to put in the cut buffer.

** The new command C-x 5 o (other-frame) selects different frames,
successively, in cyclic order.  It does for frames what C-x o
does for windows.

** The command M-ESC (eval-expression) has its own command history.

** The commands M-! and M-| for running shell commands have their own
command history.

** If the directory containing the Emacs executable has a sibling named
`lisp', that `lisp' directory is added to the end of `load-path'
(provided you don't override the normal value with the EMACSLOADPATH
environment variable).  This feature may make it easier to move
an installed Emacs from place to place.

** M-x validate-tex-buffer now records the locations of mismatches
found in the `*Occur*' buffer.  You can go to that buffer and type C-c
C-c to visit a particular mismatch.

** There are new commands in Shell mode.

C-c C-n and C-c C-p move point to the next or previous shell input line.

C-c C-d is now another way to send an end-of-file to the subshell.

** Changes to calendar/diary.

Time zone data is now determined automatically, including the
start/stop days and times of daylight savings time.  The code now
works correctly almost anywhere in the world.

The format of the holiday specifications has changed and IS NO LONGER
COMPATIBLE with the old (version 18) format.  See the documentation of
the variable calendar-holidays for details of the new, improved
format.

The hook `diary-display-hook' has been split into two:
diary-display-hook which should be used ONLY for the display and
`diary-hook' which should be used for appointment notification.  If
diary-display-hook is nil (the default), simple-diary-display is
used.  This allows the diary hooks to be correctly set with add-hook.

The forms used for dates in diary entries and general display are no
longer autoloaded, but set at load time; this means they will be set
correctly based on values you assign to various variables.

** The functions x-rebind-key and x-rebind-keys have been deleted,
because you can accomplish the same job by binding keys to keyboard
macros.

** Emacs now distinguishes double and triple drag events and double and
triple button-down events.  These work analogously to double and
triple click events.

Double drag events, if not defined, convert to ordinary click events.
Double down events, if not defined, convert first to ordinary down
events, which are then discarded if not defined.  Triple events that
are not defined convert to the corresponding double event; if that is
also not defined, it may convert further.

** The new function event-click-count returns the number of clicks,
from an event which is a list.  It is 1 for an ordinary click, drag,
or button-down event, 2 for a double event, and 3 or more for a triple
event.

** The new function previous-frame is like next-frame, but moves
around through the set of existing frames in the opposite order.

** The post-command-hook now runs even after commands that get an error
and return to top level.  As a consequence of the same change, this
hook also runs before Emacs reads the first command.  That might sound
paradoxical, as if this hook were the same as the pre-command-hook.
Actually, they are not similar; the latter runs before *execution* of
a command, but after it has been read.

** You can turn off the text property hooks that run when point moves
to certain places in the buffer, by binding inhibit-point-motion-hooks
to a non-nil value.

** Inserting a string with no text properties into the buffer normally
inherits the properties of the preceding character.  You can now
control this inheritance by setting the front-sticky and
rear-nonsticky properties of a character.

If you make a character's front-sticky property t, then insertion
before the character inherits its properties.  If you make the
rear-nonsticky property t, then insertion after the character does not
inherit its properties.  You can regard characters as normally being
rear-sticky and not front-sticky, and this is why insertion normally
inherits from the previous character.

If neither side of an insertion is suitably sticky, then the inserted
text gets no properties.  If both sides are sticky, then the inserted
text gets the properties of both sides, with the previous character's
properties taking precedence when both sides have a property in
common.

You can also specify stickiness for individual properties.  To do so,
use a list of property names as the value of the front-sticky property
or the rear-nonsticky property.  For example, if a character has a
rear-nonsticky property whose value is (face read-only), then
insertion after the character will not inherit its face property or
read-only property (if any), but will inherit any other properties.

The merging of properties when both sides of the insertion are sticky
takes place one property at a time.  If the preceding character is
rear-sticky for the property, and the property is non-nil, it
dominates.  Otherwise, the following character's property value is
used if it is front-sticky for that property.

** If you give a character a non-nil `invisible' text property, the
character does not appear on the screen.  This works much like
selective display.

The details of this feature are likely to change in future Emacs
versions.

** In Info, when you go to a node, it runs the normal hook
Info-selection-hook.

** You can use the new function `invocation-directory' to get the name
of the directory containing the Emacs executable that was run.

** Entry to the minibuffer runs the normal hook minibuffer-setup-hook.

** The new function minibuffer-window-active-p takes one argument, a
minibuffer window, and returns t if the window is currently active.
\f
* Changes in version 19.17.

** When Emacs displays a list of completions in a buffer, 
you can select a completion by clicking mouse button 2
on that completion.

** Use the command `list-faces-display' to display a list of 
all the currently defined faces, showing what they look like.

** Menu bar items from local maps now come after the usual items.

** The Help menu bar item always comes last in the menu bar.

** If you enable Font-Lock mode on a buffer containing a program
(certain languages such as C and Lisp are supported), everything you
type is automatically given a face property appropriate to its
syntactic role.  For example, there are faces for comments, string
constants, names of functions being defined, and so on.

** Dunnet, an adventure game, is now available.

** Several major modes now have their own menu bar items, 
including Dired, Rmail, and Sendmail.  We would like to add 
suitable menu bar items to other major modes.

** The key binding C-x a C-h has been eliminated.
This is because it got in the way of the general feature of typing
C-h after a prefix character.  If you want to run
inverse-add-global-abbrev, you can use C-x a - or C-x a i g instead.

** If you set the variable `rmail-mail-new-frame' to a non-nil value,
all the Rmail commands to send mail make a new frame to do it in.
When you send the message, or use the menu bar command not to send it,
that frame is deleted.

** In Rmail, the o and C-o commands are now almost interchangeable.
Both commands check the format of the file you specify, and append
the message to it in Rmail format if it is an Rmail file, and in
inbox file format otherwise.  C-o and o are different only when you
specify a new file.

** The function `copy-face' now takes an optional fourth argument
NEW-FRAME.  If you specify this, it copies the definition of face
OLD-FACE on frame FRAME to face NEW-NAME on frame NEW-FRAME.

** A local map can now cancel out one of the global map's menu items.
Just define that subcommand of the menu item with `undefined'
as the definition.  For example, this cancels out the `Buffers' item
for the current major mode:

    (local-set-key [menu-bar buffer] 'undefined)

** To put global items at the end of the menu bar, use the new variable
`menu-bar-final-items'.  It should be a list of symbols--event types
bound in the menu bar.  The menu bar items for these symbols are
moved to the end.

** The list returned by `buffer-local-variables' now contains cons-cell
elements of the form (SYMBOL . VALUE) only for buffer-local variables
that have values.  For unbound buffer-local variables, the variable
name (symbol) appears directly as an element of the list.

** The `modification-hooks' property of a character no longer affects
insertion; it runs only for deletion and modification of the character.

To detect insertion, use `insert-in-front-hooks' and
`insert-behind-hooks' properties.  The former runs when text is
inserted immediately preceding the character that has the property;
the latter runs when text is inserted immediately following the
character.

** Buffer modification now runs hooks belonging to overlays as well as
hooks belonging to characters.  If an overlay has a
`modification-hooks' property, it applies to any change to text in the
overlay, and any insertion within the overlay.  If the overlay has a
`insert-in-front-hooks' property, it runs for insertion at the
beginning boundary of the overlay.  If the overlay has an
`insert-behind-hooks' property, it runs for insertion at the end
boundary of the overlay.

The values of these properties should be lists of functions.  Each
function is called, receiving as arguments the overlay in question,
followed by the bounds of the range being modified.

** The new `-name NAME' option directs Emacs to search for its X
resources using the name `NAME', and sets the title of the initial
frame.  This argument was added for consistency with other X clients.

** The new `-xrm DATABASE' option tells Emacs to treat the string
DATABASE as the text of an X resource database.  Emacs searches
DATABASE for resource values, in addition to the usual places.  This
argument was added for consistency with other X clients.

** Emacs now searches for X resources in the files specified by the
XFILESEARCHPATH, XUSERFILESEARCHPATH, and XAPPLRESDIR environment
variables, emulating the functionality provided by programs written
using Xt.  Because of this change, Emacs will now notice system-wide
application defaults files, as other X clients do.

XFILESEARCHPATH and XUSERFILESEARCHPATH should be a list of file names
separated by colons; XAPPLRESDIR should be a list of directory names
separated by colons.

Emacs searches for X resources
  + specified on the command line, with the `-xrm RESOURCESTRING'
    option,
  + then in the value of the XENVIRONMENT environment variable,
    - or if that is unset, in the file named ~/.Xdefaults-HOSTNAME if it exists
      (where HOSTNAME is the hostname of the machine Emacs is running on),
  + then in the screen-specific and server-wide resource properties
    provided by the server,
    - or if those properties are unset, in the file named ~/.Xdefaults
      if it exists,
  + then in the files listed in XUSERFILESEARCHPATH,
    - or in files named LANG/Emacs in directories listed in XAPPLRESDIR
      (where LANG is the value of the LANG environment variable), if
      the LANG environment variable is set,
    - or in files named Emacs in the directories listed in XAPPLRESDIR
    - or in ~/LANG/Emacs (if the LANG environment variable is set),
    - or in ~/Emacs,
  + then in the files listed in XFILESEARCHPATH.

The paths in the variables XFILESEARCHPATH, XUSERFILESEARCHPATH, and
XAPPLRESDIR may contain %-escapes (like the control strings passed to
the the Emacs lisp `format' function or C printf function), which
Emacs expands.

%N is replaced by the string "Emacs" wherever it occurs.
%T is replaced by "app-defaults" wherever it occurs.
%S is replaced by the empty string wherever it occurs.
%L and %l are replaced by the value of the LANG environment variable; if LANG
   is not set, Emacs does not use that directory or file name at all.
%C is replaced by the value of the resource named "customization"
   (class "Customization"), as retrieved from the server's resource
   properties or the user's ~/.Xdefaults file, or the empty string if
   that resource doesn't exist.

So, for example,
  if XFILESEARCHPATH is set to the value
    "/usr/lib/X11/%L/%T/%N%C:/usr/lib/X11/%T/%N%C:/usr/lib/X11/%T/%N",
  and the LANG environment variable is set to
    "english",
  and the customization resource is the string
    "-color",
then, in the last step of the process described above, Emacs checks
for resources in the first of the following files that is present and
readable:
        /usr/lib/X11/english/app-defaults/Emacs-color
        /usr/lib/X11/app-defaults/Emacs-color
        /usr/lib/X11/app-defaults/Emacs
If the LANG environment variable is not set, then Emacs never uses the
first element of the path, "/usr/lib/X11/%L/%T/%N%C", because it
contains the %L escape.

If XFILESEARCHPATH is unset, Emacs uses the default value
"/usr/lib/X11/%L/app-defaults/Emacs%C:\
/usr/lib/X11/app-defaults/Emacs%C:\
/usr/lib/X11/%L/app-defaults/Emacs:\
/usr/lib/X11/app-defaults/Emacs"

This feature was added for consistency with other X applications.

** The new function `text-property-any' scans the region of text from
START to END to see if any character's property PROP is `eq' to
VALUE.  If so, it returns the position of the first such character.
Otherwise, it returns nil.

The optional fifth argument, OBJECT, specifies the string or buffer to
be examined.

** The new function `text-property-not-all' scans the region of text from
START to END to see if any character's property PROP is not `eq' to
VALUE.  If so, it returns the position of the first such character.
Otherwise, it returns nil.

The optional fifth argument, OBJECT, specifies the string or buffer to
be examined.

** The function `delete-windows-on' now takes an optional second
argument FRAME, which specifies which frames it should affect.
  + If FRAME is nil or omitted, then `delete-windows-on' deletes windows
    showing BUFFER (its first argument) on all frames.
  + If FRAME is t, then `delete-windows-on' only deletes windows on the
    selected frame; other frames are unaffected.
  + If FRAME is a frame, then `delete-windows-on' only deletes windows on
    the given frame; other frames are unaffected.

\f
* Changes in version 19.16.

** When dragging the mouse to select a region, Emacs now highlights the
region as you drag (if Transient Mark mode is enabled).  If you
continue the drag beyond the boundaries of the window, Emacs scrolls
the window at a steady rate until you either move the mouse back into
the window or release the button.

** RET now exits `query-replace' and `query-replace-regexp'; this makes it
more consistent with the incremental search facility, which uses RET
to end the search.

** In C mode, C-c C-u now runs c-up-conditional.
C-c C-n and C-c C-p now run new commands that move forward
and back over balanced sets of C conditionals (c-forward-conditional
and c-backward-conditional).

** The Edit entry in the menu bar has a new alternative:
"Choose Next Paste".  It gives you a menu showing the various
strings in the kill ring; click on one to select it as the text 
to be yanked ("pasted") the next time you yank.

** If you enable Transient Mark mode and set `mark-even-if-inactive' to
non-nil, then the region is highlighted in a transient fashion just as
normally in Transient Mark mode, but the mark really remains active
all the time; commands that use the region can be used even if the
region highlighting turns off.

** If you type C-h after a prefix key, it displays the bindings
that start with that prefix.

** The VC package now searches for version control commands in the
directories named by the variable `vc-path'; its value should be a
list of strings.

** If you are visiting a file that has locks registered under RCS,
VC now displays each lock's owner and version number in the mode line
after the string `RCS'.  If there are no locks, VC displays the head
version number.

** When using X, if you load the `paren' library, Emacs automatically
underlines or highlights the matching paren whenever point is
next to the outside of a paren.  When point is before an open-paren,
this shows the matching close; when point is after a close-paren,
this shows the matching open.

** The new function `define-key-after' is like `define-key',
but takes an extra argument AFTER.  It places the newly defined
binding after the binding for the event AFTER.

** `accessible-keymaps' now takes an optional second argument, PREFIX.
If PREFIX is non-nil, it means the value should include only maps for
keys that start with PREFIX.

`describe-bindings' also accepts an optional argument PREFIX which
means to describe only the keys that start with PREFIX.

** The variable `prefix-help-command' hold a command to run to display help
whenever the character `help-char' follows a prefix key and does not have
a key binding in that context.

** Emacs now detects double- and triple-mouse clicks.  A single mouse
click produces a pair events of the form:
        (down-mouse-N POSITION)
        (mouse-N POSITION)
Clicking the same mouse button again, soon thereafter and at the same
location, produces another pair of events of the form:
        (down-mouse-N POSITION)
        (double-mouse-N POSITION 2)
Another click will produce an event pair of the form:
        (down-mouse-N POSITION)
        (triple-mouse-N POSITION 3)
All the POSITIONs in such a sequence would be identical, except for
their timestamps.

To count as double- and triple-clicks, mouse clicks must be at the
same location as the first click, and the number of milliseconds
between the first release and the second must be less than the value
of the lisp variable `double-click-time'.  Setting `double-click-time'
to nil disables multi-click detection.  Setting it to t removes the
time limit; Emacs then detects multi-clicks by position only.

If `read-key-sequence' finds no binding for a double-click event, but
the corresponding single-click event would be bound,
`read-key-sequence' demotes it to a single-click.  Similarly, it
demotes unbound triple-clicks to double- or single-clicks.  This means
you don't have to distinguish between single- and multi-clicks if you
don't want to.

Emacs reports all clicks after the third as `triple-mouse-N' clicks,
but increments the click count after POSITION.  For example, a fourth
click, soon after the third and at the same location, produces a pair
of events of the form:
        (down-mouse-N POSITION)
        (triple-mouse-N POSITION 4)

** The way Emacs reports positions of mouse events has changed
slightly.  If a mouse event includes a position list of the form:
        (WINDOW (PLACE-SYMBOL) (COLUMN . ROW) TIMESTAMP)
this denotes exactly the same position as the list:
        (WINDOW  PLACE-SYMBOL  (COLUMN . ROW) TIMESTAMP)
That is, the event occurred over a non-textual area of the frame,
specified by PLACE-SYMBOL, a symbol like `mode-line' or
`vertical-scroll-bar'.

Enclosing PLACE-SYMBOL in a singleton list does not change the
position denoted, but the `read-key-sequence' function uses the
presence or absence of the singleton list to tell whether or not it
should prefix the event with its place symbol.

Normally, `read-key-sequence' prefixes mouse events occurring over
non-textual areas with their PLACE-SYMBOLs, to select the sub-keymap
appropriate for the event; for example, clicking on the mode line
produces a sequence like
        [mode-line (mouse-1 POSN)]
However, if lisp code elects to unread the resulting key sequence by
placing it in the `unread-command-events' variable, it is important
that `read-key-sequence' not insert the prefix symbol again; that
would produce a malformed key sequence like
        [mode-line mode-line (mouse-1 POSN)]
For this reason, `read-key-sequence' encloses the event's PLACE-SYMBOL
in a singleton list when it first inserts the prefix, but doesn't
insert the prefix when processing events whose PLACE-SYMBOLs are
already thus enclosed.

\f
* Changes in version 19.15.

** `make-frame-visible', which uniconified frames, is now a command,
and thus may be bound to a key.  This makes sense because frames
respond to user input while iconified.

** You can now use Meta mouse clicks to set and use the "secondary
selection".  You can drag M-Mouse-1 across the region you want to
select.  Or you can press M-Mouse-1 at one end and M-Mouse-3 at the
other (this also copies the text to the kill ring).  Repeating M-Mouse-3
again at the same place kills that text.

M-Mouse-2 kills the secondary selection.

Setting the secondary selection does not move point or the mark.  It
is possible to make a secondary selection that does not all fit on the
screen, by using M-Mouse-1 at one end, scrolling, then using M-Mouse-3
at the other end.

Emacs has only one secondary selection at any time.  Starting to set
a new one cancels any previous one.  The secondary selection displays
using a face named `secondary-selection'.

** There's a new way to request use of Supercite (sc.el).  Do this:

    (add-hook 'mail-citation-hook 'sc-cite-original)

Currently this works with Rmail.  In the future, other Emacs based
mail-readers should be modified to understand this hook also.
In the mean time, you should keep doing what you have done in the past
for those other mail readers.

** When a regular expression contains `\(...\)' inside a repetition
operator such as `*' or `+', and you ask about the range that was matched
using `match-beginning' and `match-end', the range you get corresponds
to the *last* repetition *only*.  In Emacs 18, you would get a range
corresponding to all the repetitions.

If you want to get a range corresponding to all the repetitions,
put a `\(...\)' grouping *outside* the repetition operator.  This
is the syntax that corresponds logically to the desired result, and
it works the same in Emacs 18 and Emacs 19.

(This change actually took place earlier, but we didn't know about it
and thus didn't document it.)
\f
* Changes in version 19.14.

** To modify read-only text, bind the variable `inhibit-read-only'
to a non-nil value.  If the value is t, then all reasons that might
make text read-only are inhibited (including `read-only' text properties).
If the value is a list, then a `read-only' property is inhibited
if it is `memq' in the list.

** If you call `get-buffer-window' passing t as its second argument, it
will only search for windows on visible frames.  Previously, passing t
as the secord argument caused `get-buffer-window' to search all
frames, visible or not.

** If you call `other-buffer' with a nil or omitted second argument, it
will ignore buffers displayed windows on any visible frame, not just
the selected frame.

** You can specify a window or a frame for C-x # to use when
selects a server buffer.  Set the variable server-window
to the window or frame that you want.

** The command M-( now inserts spaces outside the open-parentheses in
some cases--depending on the syntax classes of the surrounding
characters.  If the variable `parens-dont-require-spaces' is non-nil,
it inhibits insertion of these spaces.

** The GUD package now supports the debugger known as xdb on HP/UX
systems.  Use M-x xdb.  The variable `gud-xdb-directories' lets you
specify a list of directories to search for source code.

** If you are using the mailabbrev package, you should note that its
function for defining an alias is now called `define-mail-abbrev'.
This package no longer contains a definition for `define-mail-alias';
that name is used only in mailaliases.

** Inserted characters now inherit the properties of the text before
them, by default, rather than those of the following text.

** The function `insert-file-contents' now takes optional arguments BEG
and END that specify which part of the file to insert.  BEG defaults to
0 (the beginning of the file), and END defaults to the end of the file.

If you specify BEG or END, then the argument VISIT must be nil.
\f
* Changes in version 19.13.

** Magic file names can now handle the `load' operation.

** Bibtex mode now sets up special entries in the menu bar.

** The incremental search commands C-w and C-y, which copy text from
the buffer into the search string, now convert it to lower case
if you are in a case-insensitive search.  This is to avoid making
the search a case-sensitive one.

** GNUS now knows your time zone automatically if Emacs does.

** Hide-ifdef mode no longer defines keys of the form
C-c LETTER, since those keys are reserved for users.
Those commands have been moved to C-c M-LETTER.
We may move them again for greater consistency with other modes.
\f
* Changes in version 19.12.

** You can now make many of the sort commands ignore case by setting
`sort-fold-case' to a non-nil value.
\f
* Changes in version 19.11.

** Supercite is installed.

** `write-file-hooks' functions that return non-nil are responsible
for making a backup file if you want that to be done.
To do so, execute the following code:

   (or buffer-backed-up (backup-buffer))

You might wish to save the file modes value returned by
`backup-buffer' and use that to set the mode bits of the file
that you write.  This is what `basic-save-buffer' does when
it writes a file in the usual way.

(This is not actually new, but wasn't documented before.)
\f
* Changes in version 19.10.

** The command `repeat-complex-command' is now on C-x ESC ESC.
It used to be bound to C-x ESC.

The reason for this change is to make function keys work after C-x.

** The variable `highlight-nonselected-windows' now controls whether
the region is highlighted in windows other than the selected window
(in Transient Mark mode only, of course, and currently only when
using X).
\f
* Changes in version 19.8.

** It is now simpler to tell Emacs to display accented characters under
X windows.  M-x standard-display-european toggles the display of
buffer text according to the ISO Latin-1 standard.  With a prefix
argument, this command enables European character display iff the
argument is positive.

** The `-i' command-line argument tells Emacs to use a picture of the
GNU gnu as its icon, instead of letting the window manager choose an
icon for it.  This option used to insert a file into the current
buffer; use `-insert' to do that now.

** The `configure' script now supports `--prefix' and `--exec-prefix'
options.

The `--prefix=PREFIXDIR' option specifies where the installation process
should put emacs and its data files.  This defaults to `/usr/local'.
- Emacs (and the other utilities users run) go in PREFIXDIR/bin
  (unless the `--exec-prefix' option says otherwise).
- The architecture-independent files go in PREFIXDIR/lib/emacs/VERSION
  (where VERSION is the version number of Emacs, like `19.7').
- The architecture-dependent files go in
  PREFIXDIR/lib/emacs/VERSION/CONFIGURATION
  (where CONFIGURATION is the configuration name, like mips-dec-ultrix4.2),
  unless the `--exec-prefix' option says otherwise.

The `--exec-prefix=EXECDIR' option allows you to specify a separate
portion of the directory tree for installing architecture-specific
files, like executables and utility programs.  If specified,
- Emacs (and the other utilities users run) go in EXECDIR/bin, and
- The architecture-dependent files go in
  EXECDIR/lib/emacs/VERSION/CONFIGURATION.
EXECDIR/bin should be a directory that is normally in users' PATHs.

** When running under X, the new lisp function `x-list-fonts'
allows code to find out which fonts are available from the X server.
The first argument PATTERN is a string, perhaps with wildcard characters;
  the * character matches any substring, and
  the ? character matches any single character.
  PATTERN is case-insensitive.
If the optional arguments FACE and FRAME are specified, then
`x-list-fonts' returns only fonts the same size as FACE on FRAME.


\f
* Changes in version 19.

** When you kill buffers, Emacs now returns memory to the operating system,
thus reducing the size of the Emacs process.  All the space that you free
up by killing buffers can now be reused for other buffers no matter what
their sizes, or reused by other processes if Emacs doesn't need it.

** Emacs now does garbage collection and auto saving while it is waiting
for input, which often avoids the need to do these things while you
are typing.

The variable `auto-save-timeout' says how many seconds Emacs should
wait, after you stop typing, before it does an auto save and a garbage
collection.

** If auto saving detects that a buffer has shrunk greatly, it refrains
from auto saving that buffer and displays a warning.  Now it also turns
off Auto Save mode in that buffer, so that you won't get the same
warning again.

If you reenable Auto Save mode in that buffer, Emacs will start saving
it again with no further warnings.

** A new minor mode called Line Number mode displays the current line
number in the mode line, updating it as necessary when you move
point.

However, if the buffer is very large (larger than the value of
`line-number-display-limit'), then the line number doesn't appear.
This is because computing the line number can be painfully slow if the
buffer is very large.

** You can quit while Emacs is waiting to read or write files.

** The arrow keys now have default bindings to move in the appropriate
directions. 

** You can suppress next-line's habit of inserting a newline when
called at the end of a buffer by setting next-line-add-newlines to nil
(it defaults to t).

** You can now get back recent minibuffer inputs conveniently.  While
in the minibuffer, type M-p to fetch the next earlier minibuffer
input, and use M-n to fetch the next later input.

There are also commands to search forward or backward through the
history for history elements that match a regular expression.  M-r
searches older elements in the history, while M-s searches newer
elements.  By special dispensation, these commands can always use the
minibuffer to read their arguments even though you are already in the
minibuffer when you issue them.

The history feature is available for all uses of the minibuffer, but
there are separate history lists for different kinds of input.  For
example, there is a list for file names, used by all the commands that
read file names.  There is a list for arguments of commands like
`query-replace'.  There are also very specific history lists, such
as the one that `compile' uses for compilation commands.

** You can now display text in a mixture of fonts and colors, using the
"face" feature, together with the overlay and text property features.
See the Emacs Lisp manual for details.  The Emacs Users Manual describes
how to change the colors and font of standard predefined faces.

** You can refer to files on other machines using special file name syntax:

/HOST:FILENAME
/USER@HOST:FILENAME

When you do this, Emacs uses the FTP program to read and write files on
the specified host.  It logs in through FTP using your user name or the
name USER.  It may ask you for a password from time to time; this
is used for logging in on HOST.

** Some C-x key bindings have been moved onto new prefix keys.

C-x r is a prefix for registers and rectangles.
C-x n is a prefix for narrowing.
C-x a is a prefix for abbrev commands.

C-x r C-SPC
C-x r SPC       point-to-register    (Was C-x /)
C-x r j         jump-to-register     (Was C-x j)
C-x r s         copy-to-register     (Was C-x x)
C-x r i         insert-register      (Was C-x g)
C-x r r         copy-rectangle-to-register  (Was C-x r)
C-x r k         kill-rectangle
C-x r y         yank-rectangle
C-x r o         open-rectangle
C-x r f         frame-configuration-to-register
                  (This saves the state of all windows in all frames.)
C-x r w         window-configuration-to-register
                  (This saves the state of all windows in the selected  frame.)

(Use C-x r j to restore a configuration saved with C-x r f or C-x r w.)

C-x n n         narrow-to-region        (Was C-x n)
C-x n p         narrow-to-page          (Was C-x p)
C-x n w         widen                   (Was C-x w)

C-x a l         add-mode-abbrev         (Was C-x C-a)
C-x a g         add-global-abbrev       (Was C-x +)
C-x a i l       inverse-add-mode-abbrev (Was C-x C-h)
C-x a i g       inverse-add-global-abbrev  (Was C-x -)
C-x a e         expand-abbrev           (Was C-x ')

(The old key bindings C-x /, C-x j, C-x x and C-x g
have not yet been removed.)

** You can put a file name in a register to be able to visit the file
quickly.  Do this:

   (set-register ?CHAR '(file . NAME))

where NAME is the file name as a string.  Then C-x r j CHAR finds that
file.

This is useful for files that you need to visit frequently,
but that you don't want to keep in buffers all the time.

** The keys M-g (fill-region) and C-x a (append-to-buffer)
have been eliminated.

** The new command `string-rectangle' inserts a specified string on
each line of the region-rectangle.

** C-x 4 r is now `find-file-read-only-other-window'.

** C-x 4 C-o is now `display-buffer', which displays a specified buffer
in another window without selecting it.

** Picture mode has been substantially improved.  The picture editing commands
now arrange for automatic horizontal scrolling to keep point visible
when editing a wide buffer with truncate-lines on.  Picture-mode
initialization now does a better job of rebinding standard commands;
it finds not just their normal keybindings, but any function keys
attached to them.

** If you enable Transient Mark mode, then the mark becomes "inactive"
after every command that modifies the buffer.  While the mark is
active, the region is highlighted (under X, at least).  Most commands
that use the mark give an error if the mark is inactive, but you can
use C-x C-x to make it active again.  This feature is also sometimes
known as "Zmacs mode".

** Outline mode is now available as a minor mode.  This minor mode can
combine with any major mode; it substitutes the C-c commands of
Outline mode for those of the major mode.  Use M-x outline-minor-mode
to enable and disable the new mode.

M-x outline-mode is unchanged; it still switches to Outline mode as a
major mode.

** The default setting of `version-control' comes from the environment
variable VERSION_CONTROL.

** The user option for controlling whether files can set local
variables is now called `enable-local-variables'.  A value of t means
local-variables lists are obeyed; nil means they are ignored; anything
else means query the user.

The user option for controlling use of the `eval' local variable is
now called is `enable-local-eval'; its values are interpreted like
those of `enable-local-variables'.

** X Window System changes:

C-x 5 C-f and C-x 5 b switch to a specified file or buffer in a new
frame.  Likewise, C-x 5 m starts outgoing mail in another frame, and
C-x 5 . finds a tag in another frame.

When you are using X, C-z now iconifies the selected frame.

Emacs can now exchange text with other X applications.  Killing or
copying text in Emacs now makes that text available for pasting into
other X applications.  The Emacs yanking commands now insert the
latest selection set by other applications, and add the text to the
kill ring.  The Emacs commands for selecting and inserting text with
the mouse now use the kill ring in the same way the keyboard killing
and yanking commands do.

The option to specify the title for the initial frame is now `-name NAME'.
There is currently no way to specify an icon title; perhaps we will add
one in the future.

** Undoing a deletion now puts point back where it was before the
deletion.

** The variables that control how much undo information to save have
been renamed to `undo-limit' and `undo-strong-limit'.  They used to be
called `undo-threshold' and `undo-high-threshold'.

** You can now use kill commands in read-only buffers.  They don't
actually change the buffer, and Emacs will beep and warn you that the
buffer is read-only, but they do copy the text you tried to kill into
the kill ring, so you can yank it into other buffers.

** C-o inserts the fill-prefix on the newly created line.  The command
M-^ deletes the prefix (if it occurs) after the newline that it
deletes.

** C-M-l now runs the command `reposition-window'.  It scrolls the
window heuristically in a way designed to get useful information onto
the screen.

** C-M-r is now reverse incremental regexp search.

** M-z now kills through the target character.  In version 18, it
killed up to but not including the target character.

** M-! now runs the specified shell command asynchronously if it
ends in `&' (just as the shell does).

** C-h C-f and C-h C-k are new help commands that display the Info
node for a given Emacs function name or key sequence, respectively.

** The C-h p command system lets you find Emacs Lisp packages by
topic keywords.  Here is a partial list of package categories:

abbrev        abbreviation handling, typing shortcuts, macros
bib           code related to the bib bibliography processor
c             C and C++ language support
calendar      calendar and time management support
comm          communications, networking, remote access to files
docs          support for Emacs documentation
emulations    emulations of other editors
extensions    Emacs Lisp language extensions
games         games, jokes and amusements
hardware      support for interfacing with exotic hardware
help          support for on-line help systems
i14n          internationalization and alternate character-set support
internal      code for Emacs internals, build process, defaults
languages     specialized modes for editing programming languages
lisp          Lisp support, including Emacs Lisp
local         code local to your site
maint         maintenance aids for the Emacs development group
mail          modes for electronic-mail handling
news          support for netnews reading and posting
processes     process, subshell, compilation, and job control support
terminals     support for terminal types
tex           code related to the TeX formatter
tools         programming tools
unix          front-ends/assistants for, or emulators of, UNIX features
vms           support code for vms
wp            word processing

More will be added soon.

** The command to split a window into two side-by-side windows is now
C-x 3.  It was C-x 5.

** M-. (find-tag) no longer has any effect on what M-, will do
subsequently.  You can no longer use M-, to find the next similar tag;
you must use M-. with a prefix argument, instead.

The motive for this change is so that you can more reliably use
M-, to resume a suspended `tags-search' or `tags-query-replace'.

** C-x s (`save-some-buffers') now gives you more options when it asks
whether to save a particular buffer.  In addition to `y' or `n', you
can answer `!' to save all the remaining buffers, `.' to save this
buffer but not save any others, ESC to stop saving and exit the
command, and C-h to get help.  These options are analogous to those 
of `query-replace'.

** M-x make-symbolic-link does not expand its first argument.
This lets you make a link with a target that is a relative file name.

** M-x add-change-log-entry and C-x 4 a now automatically insert the
name of the file and often the name of the function that you changed.
They also handle grouping of entries.

There is now a special major mode for editing ChangeLog files.  It
makes filling work conveniently.  Each bunch of grouped entries is one
paragraph, and each collection of entries from one person on one day
is considered a page.

** The `comment-region' command adds comment delimiters to the lines that
start in the region, thus commenting them out.  With a negative argument,
it deletes comment delimiters from the lines in the region--this cancels
the effect of `comment-region' without an argument.

With a positive argument, `comment-region' adds comment delimiters
but duplicates the last character of the comment start sequence as many
times as the argument specifies.  This is a way of calling attention to
the comment.  In Lisp, you should use an argument at least two, because
the indentation convention for single semicolon comments does not leave
them at the beginning of a line.

** If `split-window-keep-point' is non-nil, C-x 2 tries to avoid
shifting any text on the screen by putting point in whichever window
happens to contain the screen line the cursor is already on.
The default is that `split-window-keep-point' is non-nil on slow
terminals.

** M-x super-apropos is like M-x apropos except that it searches both
Lisp symbol names and documentation strings for matches.  It describes
every symbol that has a match in either the symbol's name or its
documentation.

Both M-x apropos and M-x super-apropos take an optional second
argument DO-ALL which controls the more expensive part of the job.
This includes looking up and printing the key bindings of all
commands.  It also includes checking documentation strings in
super-apropos.  DO-ALL is nil by default; use a prefix arg to make it
non-nil.

** M-x revert-buffer no longer offers to revert from a recent auto-save
file unless you give it a prefix argument.  Otherwise it always
reverts from the real file regardless of whether there has been an
auto-save since thenm.  (Reverting from the auto-save file is no longer
very useful now that the undo capacity is larger.)

** M-x recover-file no longer turns off Auto Save mode when it reads
the last Auto Save file.

** M-x rename-buffer, if you give it a prefix argument,
avoids errors by modifying the new name to make it unique.

** M-x rename-uniquely renames the current buffer to a similar name
with a numeric suffix added to make it both different and unique.

One use of this command is for creating multiple shell buffers.
If you rename your shell buffer, and then do M-x shell again, it
makes a new shell buffer.  This method is also good for mail buffers,
compilation buffers, and any Emacs feature which creates a special
buffer with a particular name.

** M-x compare-windows with a prefix argument ignores changes in whitespace.
If `compare-ignore-case' is non-nil, then differences in case are also
ignored.

** `backward-paragraph' is now bound to M-{ by default, and `forward-paragraph'
to M-}.  Originally, these commands were bound to M-[ and M-], but they were
running into conflicts with the use of function keys.  On many terminals,
function keys send a sequence beginning ESC-[, so many users have defined this
as a prefix key.

** C-x C-u (upcase-region) and C-x C-l (downcase-region) are now disabled by
default; these commands seem to be often hit by accident, and can be
quite destructive if their effects are not noticed immediately.

** The function `erase-buffer' is now interactive, but disabled by default.

** When visiting a new file, Emacs attempts to abbreviate the file's
path using the symlinks listed in `directory-abbrev-alist'.

** When you visit the same file in under two names that translate into
the same name once symbolic links are handled, Emacs warns you that
you have two buffers for the same file.

** If you wish to avoid visiting the same file in two buffers under
different names, set the variable `find-file-existing-other-name'
non-nil.  Then `find-file' uses the existing buffer visiting the file,
no matter which of the file's names you specify.

** If you set `find-file-visit-truename' non-nil, then the file name
recorded for a buffer is the file's truename (in which all symbolic
links have been removed), rather than the name you specify.  Setting
`find-file-visit-truename' also implies the effect of
`find-file-existing-other-name'.

** C-x C-v now inserts the entire current file name in the minibuffer.
This is convenient if you made a small mistake in typing it.  Point
goes after the last slash, before the last file name component, so if
you want to replace it entirely, you can use C-k right away to delete
it.

** Commands such as C-M-f in Lisp mode now ignore parentheses within comments.

** C-x q now uses ESC to terminate all iterations of the keyboard
macro, rather than C-d as before.

** Use the command `setenv' to set an individual environment variable
for Emacs subprocesses.  Specify a variable name and a value, both as
strings.  This command applies only to subprocesses yet to be
started.

** Use `rot13-other-window' to examine a buffer with rot13.

This command does not change the text in the buffer.  Instead, it
creates a window with a funny display table that applies the code when
displaying the text.

** The command `M-x version' now prints the current Emacs version; The
`version' command is an alias for the `emacs-version' command.

** More complex changes in existing packages.

*** `fill-nonuniform-paragraphs' is a new command, much like
`fill-individual-paragraphs' except that only separator lines separate
paragraphs.  Since this means that the lines of one paragraph may have
different amounts of indentation, the fill prefix used is the smallest
amount of indentation of any of the lines of the paragraph.

*** Filling is now partially controlled by a new minor mode, Adaptive
Fill mode.  When this mode is enabled (and it is enabled by default),
if you use M-x fill-region-as-paragraph on an indented paragraph and
you don't have a fill prefix, it uses the indentation of the second
line of the paragraph as the fill prefix.

Adaptive Fill mode doesn't have much effect on M-q in most major
modes, because an indented line will probably count as a paragraph
starter and thus each line of an indented paragraph will be considered
a paragraph of its own.

*** M-q in C mode now runs `c-fill-paragraph', which is designed
for filling C comments.  (We assume you don't want to fill
the code in a C program.)

*** M-$ now runs the Ispell program instead of the Unix spell program.

M-$ starts an Ispell process the first time you use it.  But the process
stays alive, so that subsequent uses of M-$ run very fast.
If you want to get rid of the process, use M-x kill-ispell.

To check the entire current buffer, use M-x ispell-buffer.
Use M-x ispell-region to check just the current region.

Ispell commands often involve interactive replacement of words.
You can interrupt the interactive replacement with C-g.
You can restart it again afterward with C-u M-$.

During interactive replacement, you can type the following characters:

a       Accept this word this time.
DIGIT   Replace the word (this time) with one of the displayed near-misses.
        The digit you use says which near-miss to use.
i       Insert this word in your private dictionary
          so that Ispell will consider it correct it from now on.
r       Replace the word this time with a string typed by you.

When the Ispell process starts, it reads your private dictionary which
is the file `~/ispell.words'.  If you "insert" words with the `i' command,
these words are added to that file, but not right away--only at the end
of the interactive replacement process.

Use M-x reload-ispell to reload your private dictionary from
`~/ispell.words' if you edit it outside of Ispell.

** Changes in existing modes.

*** gdb-mode has been replaced by gud-mode.

The package gud.el (Grand Unified Debugger) replaces gdb.el in Emacs
19.  It provides a gdb.el-like interface to any of three debuggers;
gdb itself, the sdb debugger supported on some Unix systems, or the
dbx debugger on Berkeley systems.

   You start it up with one of the commands M-x gdb, M-x sdb, or
M-x dbx.  Each entry point finishes by executing a hook; gdb-mode-hook,
sdb-mode-hook or dbx-mode-hook respectively.

These bindings have changed:
C-x C-a >       gud-down                        (was M-d)
C-x C-a <       gud-up                          (was M-u)
C-x C-a C-r     gud-cont                        (was M-c)
C-x C-a C-n     gud-next                        (was M-n)
C-x C-a C-s     gud-step                        (was M-s)
C-x C-a C-i     gud-stepi                       (was M-i)
C-x C-a C-l     gud-recenter                    (was C-l)
C-d             comint-delchar-or-maybe-eof     (was C-c C-d)

These bindings have been removed:
C-c C-r         (was comint-show-output; now gud-cont)

Since GUD mode uses comint, it uses comint's input history commands,
superseding C-c C-y (copy-last-shell-input):
  M-p           comint-next-input
  M-n           comint-previous-input
  M-r           comint-previous-similar-input
  M-s           comint-next-similar-input
  M-C-r         comint-previous-input-matching

The C-x C-a bindings are also active in source files.

*** The old TeX mode bindings of M-{ and M-} have been moved to C-c {
and C-c }.  (These commands are `up-list' and `tex-insert-braces';
they are the TeX equivalents of M-( and M-).)  This is because M-{ 
and M-} are now globally defined commands.

*** Changes in Mail mode.

`%' is now a word-separator character in Mail mode.

`mail-signature', if non-nil, tells M-x mail to insert your
`.signature' file automatically.  If you don't want your signature in
a particular message, just delete it before you send the message.

You can specify the text to insert at the beginning of each line when
you use C-c C-y to yank the message you are replying to.  Set
`mail-yank-prefix' to the desired string.  A value of `nil' (the
default) means to use indentation, as in Emacs 18.  If you use just
C-u as the prefix argument to C-c C-y, then it does not insert
anything at the beginning of the lines, regardless of the value of
`mail-yank-prefix'.

If you like, you can expand mail aliases as abbrevs, as soon as you
type them in.  To enable this feature, execute the following:

    (add-hook 'mail-setup-hook 'mail-abbrevs-setup)

This can go in your .emacs file.

Word abbrevs don't expand unless you insert a word-separator character
afterward.  Any mail aliases that you didn't expand at insertion time
are expanded subsequently when you send the message.

*** Changes in Rmail.

Rmail by default gets new mail only from the system inbox file,
not from `~/mbox'.

In Rmail, you can retry sending a message that failed
by typing `M-m' on the failure message.

By contrast, another new command M-x rmail-resend is used for
forwarding a message and marking it as "resent from" you
with header fields "Resent-From:" and "Resent-To:".

`e' is now the command to edit a message.
To expunge, type `x'.  We know this will surprise people
some of the time, but the surprise will not be disastrous--if
you type `e' meaning to expunge, just turn off editing with C-c C-c
and then type `x'.

Another new Rmail command is `<', which moves to the first message.
This is for symmetry with `>'.

Use the `b' command to bury the Rmail buffer and its summary buffer,
if any, removing both of them from display on the screen.

The variable `rmail-output-file-alist' now controls the default
for the file to output a message to.

In the Rmail summary buffer, all cursor motion commands select
the message you move to.  It's really neat when you use
incremental search.

You can now issue most Rmail commands from an Rmail summary buffer.
The commands do the same thing in that buffer that they do in the
Rmail buffer.  They apply to the message that is selected in the Rmail
buffer, which is always the one described by the current summary
line.

Conversely, motion and deletion commands in the Rmail buffer also
update the summary buffer.  If you set the variable
`rmail-redisplay-summary' to a non-nil value, then they bring the
summary buffer (if one exists) back onto the screen.

C-M-t is a new command to make a summary by topic.  It uses regexp
matching against just the subjects of the messages to decide which
messages to show in the summary.

You can easily convert an Rmail file to system mailbox format with the
command `unrmail'.  This command reads two arguments, the name of
the Rmail file to convert, and the name of the new mailbox file.
(This command does not change the Rmail file itself.)

Rmail now handles Content Length fields in messages.

*** `mail-extract-address-components' unpacks mail addresses.
It takes an address as a string (the contents of the From field, for
example) and returns a list of the form (FULL-NAME
CANONICAL-ADDRESS).

*** Changes in C mode and C-related commands.

**** M-x c-up-conditional

In C mode, `c-up-conditional' moves back to the containing
preprocessor conditional, setting the mark where point was
previously.

A prefix argument acts as a repeat count.  With a negative argument,
this command moves forward to the end of the containing preprocessor
conditional.  When going backwards, `#elif' acts like `#else' followed
by `#if'.  When going forwards, `#elif' is ignored.

**** In C mode, M-a and M-e are now defined as
`c-beginning-of-statement' and `c-end-of-statement'.

**** In C mode, M-x c-backslash-region is a new command to insert or
align `\' characters at the ends of the lines of the region, except
for the last such line.  This is useful after writing or editing a C
macro definition.

If a line already ends in `\', this command adjusts the amount of
whitespace before it.  Otherwise, it inserts a new `\'.

*** New features in info.

When Info looks for an Info file, it searches the directories
in `Info-directory-list'.  This makes it easy to install the Info files
that come with various packages.  You can specify the path with
the environment variable INFOPATH.

There are new commands in Info mode.

`]' now moves forward a node, going up and down levels as needed.
`[' is similar but moves backward.  These two commands try to traverse
the entire Info tree, node by node.  They are the equivalent of reading
a printed manual sequentially.

`<' moves to the top node of the current Info file.
`>' moves to the last node of the file.

SPC scrolls through the current node; at the end, it advances to the
next node in depth-first order (like `]').

DEL scrolls backwards in the current node; at the end, it moves to the
previous node in depth-first order (like `[').

After a menu select, the info `up' command now restores point in the
menu.  The combination of this and the previous two changes means that
repeated SPC keystrokes do the right (depth-first traverse forward) thing.

`i STRING RET' moves to the node associated with STRING in the index
or indices of this manual.  If there is more than one match for
STRING, the `i' command finds the first match.

`,' finds the next match for the string in the previous `i' command

If you click the middle mouse button near a cross-reference,
menu item or node pointer while in Info, you will go to the node
which is referenced.

*** Changes in M-x compile.

You can repeat any previous compilation command conveniently using the
minibuffer history commands, while in the minibuffer entering the
compilation command.

While a compilation is going on, the string `Compiling' appears in
the mode line.  When this string disappears, that tells you the
compilation is finished.

The buffer of compiler messages is in Compilation mode.  This mode
provides the keys SPC and DEL to scroll by screenfuls, and M-n and M-p
to move to the next or previous error message.  You can also use C-c
C-c on any error message to find the corresponding source code.

Emacs 19 has a more general parser for compiler messages.  For example, it
can understand messages from lint, and from certain C compilers whose error
message format is unusual.  Also, it only parses until it sees the error
message you want; you never have to wait a long time to see the first
error, no matter how big the buffer is.

*** M-x diff and M-x diff-backup.

This new command compares two files, displaying the differences in an
Emacs buffer.  The options for the `diff' program come from the
variable `diff-switches', whose value should be a string.

The buffer of differences has Compilation mode as its major mode, so you
can use C-x ` to visit successive changed locations in the two
source files, or you can move to a particular hunk of changes and type
C-c C-c to move to the corresponding source.  You can also use the
other special commands of Compilation mode: SPC and DEL for
scrolling, and M-n and M-p for cursor motion.

M-x diff-backup compares a file with its most recent backup.
If you specify the name of a backup file, `diff-backup' compares it
with the source file that it is a backup of.

*** The View commands (such as M-x view-buffer and M-x view-file) no
longer use recursive edits; instead, they switch temporarily to a
different major mode (View mode) specifically designed for moving
around through a buffer without editing it.

*** Changes in incremental search.

**** The character to terminate an incremental search is now RET.
This is for compatibility with the way most other arguments are read.

To search for a newline in an incremental search, type LFD (also known
as C-j).

**** Incremental search now maintains a ring of previous search
strings.  Use M-p and M-n to move through the ring to pick a search
string to reuse.  These commands leave the selected search ring
element in the minibuffer, where you can edit it.  Type C-s or C-r to
finish editing and search for the chosen string.

**** If you type an upper case letter in incremental search, that turns
off case-folding, so that you get a case-sensitive search.

**** If you type a space during regexp incremental search, it matches
any sequence of whitespace characters.  If you want to match just a space,
type C-q SPC.

**** Incremental search is now implemented as a major mode.  When you
type C-s, it switches temporarily to a different keymap which defines
each key to do what it ought to do for incremental search.  This has
next to no effect on the user-visible behavior of searching, but makes
it easier to customize that behavior.

Emacs 19 eliminates the old variables `search-...-char' that used to
be the way to specify the characters to use for various special
purposes in incremental search.  Instead, you can define the meaning
of a character in incremental search by modifying `isearch-mode-map'.

*** New commands in Buffer Menu mode.

The command C-o now displays the current line's buffer in another
window but does not select it.  This is like the existing command `o'
which selects the current line's buffer in another window.

The command % toggles the read-only flag of the current line's buffer.

The way to switch to a set of several buffers, including those marked
with m, is now v.  The q command simply quits, replacing the buffer
menu buffer with the buffer that was displayed previously.

** New major modes and packages.

*** The news reader GNUS is now installed.

*** There is a new interface for version control systems, called VC.
It works with both RCS and SCCS; in fact, you don't really have to
know which one of them is being used, because it automatically deals
with either one.

Most of the time, the only command you have to know about is C-x C-q.
This command normally toggles the read-only flag of the current
buffer.  If the buffer is visiting a file that is maintained with a
version control system, the command still toggles read-only, but does
so by checking the file in or checking it out.

When you check a file in, VC asks you for a log entry by popping up a
buffer.  Edit the entry there, then type C-c C-c when it is ready.
That's when the actual checkin happens.  If you change your mind about
the checkin, simply switch buffers and don't ever go back to the log
buffer.

To start using version control for a file, use the command C-x v v.
This works like C-x C-q (performing the next logical version-control
operation needed to change the file's writability) but it will also
perform initial checkin on an unregistered file.

By default, VC uses RCS if RCS is installed on your machine;
otherwise, SCCS.  If you want to make the choice explicitly, you can do
it by setting `vc-default-back-end' to the symbol `RCS' or the symbol
`SCCS'.

You can tell when a file you visit is maintained with version control
because either `RCS' or `SCCS' appears in the mode line.

*** A new Calendar mode has been added, the work of Edward M. Reingold.
The mode can display the Gregorian calendar and a variety of other
calendars at any date, and interacts with a diary facility similar to
the UNIX `calendar' utility.

*** There is a new major mode for editing binary files: Hexl mode.
To use it, use M-x hexl-find-file instead of C-x C-f to visit the file.
This command converts the file's contents to hexadecimal and lets you
edit the translation.  When you save the file, it is converted 
automatically back to binary.

You can also use M-x hexl-mode to translate an existing buffer into hex.
Do this if you have already visited a binary file.  

Hexl mode has a few other commands:

C-M-d   insert a byte with a code typed in decimal.
C-M-o   insert a byte with a code typed in octal.
C-M-x   insert a byte with a code typed in hex.

C-x [   move to the beginning of a 1k-byte "page".
C-x ]   move to the end of a 1k-byte "page".

M-g     go to an address specified in hex.
M-j     go to an address specified in decimal.

C-c C-c leave hexl mode and go back to the previous major mode.

*** Miscellaneous new major modes include Awk mode, Icon mode, Makefile
mode, Perl mode and SGML mode.

*** Edebug, a new source-level debugger for Emacs Lisp functions.

To use Edebug, use the command M-x edebug-defun to "evaluate" a
function definition in an Emacs Lisp file.  We put "evaluate" in
quotation marks because it doesn't just evaluate the function, it also
inserts additional information to support source-level debugging.

You must also do

    (setq debugger 'edebug-debug)

to cause errors and single-stepping to use Edebug instead of the usual
Emacs Lisp debugger.

For more information, see the Edebug manual, which should be included
in the Emacs distribution.

*** C++ mode is like C mode, except that it understands C++ comment syntax
and certain other differences between C and C++.  It also has a command
`fill-c++-comment' which fills a paragraph made of comment lines.

The command `comment-region' is useful in C++ mode for commenting out
several consecutive lines, or removing the commenting out of such lines.

*** A new package for merging two variants of the same text.

It's not unusual for programmers to get their signals crossed and
modify the same program in two different directions.  Then somebody
has to merge the two versions.  The command `emerge-files' makes this
easier.

`emerge-files' reads two file names and compares them.  Then it
displays three buffers: one for each file, and one for the
differences.

If the original version of the file is available, you can make things
even easier using `emerge-files-with-ancestor'.  It reads three file
names--variant 1, variant 2, and the common ancestor--and uses diff3
to compare them.

You control the merging interactively.  The main loop of Emerge
consists of showing you one set of differences, asking you what to do
about them, and doing it.  You have a choice of two modes for giving
directions to Emerge: "fast" mode and "edit" mode.

In Fast mode, Emerge commands are single characters, and ordinary
Emacs commands are disabled.  This makes Emerge operations fast, but
prevents you from doing more than selecting the A or the B version of
differences.  In Edit mode, all emerge commands use the C-c prefix,
and the usual Emacs commands are available.  This allows editing the
merge buffer, but slows down Emerge operations.  Edit and fast modes
are indicated by `F' and `E' in the minor modes in the mode line.

The Emerge commands are:

        p       go to the previous difference
        n       go to the next difference
        a       select the A version of this difference
        b       select the B version of this difference
        j       go to a particular difference (prefix argument
                specifies which difference) (0j suppresses display of
                the flags)
        q       quit - finish the merge*
        f       go into fast mode
        e       go into edit mode
        l       recenter (C-l) all three windows*
        - and 0 through 9
                prefix numeric arguments
        d a     select the A version as the default from here down in
                the merge buffer*
        d b     select the B version as the default from here down in
                the merge buffer*
        c a     copy the A version of the difference into the kill
                ring
        c b     copy the B version of the difference into the kill
                ring
        i a     insert the A version of the difference at the point
        i b     insert the B version of the difference at the point
        m       put the point and mark around the difference region
        ^       scroll-down (like M-v) the three windows*
        v       scroll-up (like C-v) the three windows*
        <       scroll-left (like C-x <) the three windows*
        >       scroll-right (like C-x >) the three windows*
        |       reset horizontal scroll on the three windows*
        x 1     shrink the merge window to one line (use C-u l to restore it
                to full size)
        x a     find the difference containing a location in the A buffer*
        x b     find the difference containing a location in the B buffer*
        x c     combine the two versions of this difference*
        x C     combine the two versions of this difference, using a
                register's value as the template*
        x d     find the difference containing a location in the merge buffer*
        x f     show the files/buffers Emerge is operating on in Help window
                (use C-u l to restore windows)
        x j     join this difference with the following one
                (C-u x j joins this difference with the previous one)
        x l     show line numbers of points in A, B, and merge buffers
        x m     change major mode of merge buffer*
        x s     split this difference into two differences
                (first position the point in all three buffers to the places
                to split the difference)
        x t     trim identical lines off top and bottom of difference
                (such lines occur when the A and B versions are
                identical but differ from the ancestor version)
        x x     set the template for the x c command*

Normally, the merged output goes back in the first file specified.
If you use a prefix argument, Emerge reads another file name to use
for the output file.

Once Emerge has prepared the buffer of differences, it runs the hooks
in `emerge-startup-hooks'.

*** Asm mode is a new major mode for editing files of assembler code.
It defines these commands:

TAB     tab-to-tab-stop.
LFD     Insert a newline and then indent using tab-to-tab-stop.
:       Insert a colon and then remove the indentation
        from before the label preceding colon.  Then tab-to-tab-stop.
;       Insert or align a comment.

*** Two-column mode lets you conveniently edit two side-by-side columns
of text.  It works using two side-by-side windows, each showing its
own buffer.

Here are three ways to enter two-column mode:

C-x 6 2 makes the current buffer into the left-hand buffer.  In the
right-hand window it puts a buffer whose name is based on the current
buffer's name.

C-x 6 b BUFFER RET makes the current buffer into the left-hand buffer,
and uses buffer BUFFER as the right-hand buffer.

C-x 6 s splits the current buffer, which contains two-column text,
into two side-by-side buffers.  The old current buffer becomes the
left-hand buffer, but the text in the right column is moved into the
right-hand buffer.  The current column specifies the split point.
Splitting starts with the current line and continues to the end of the
buffer.

C-x 6 s takes a prefix argument which specifies how many characters
before point constitute the column separator.  (The default argument
is 1, as usual, so by default the column separator is the character
before point.)  Lines that don't have the column separator at the
proper place remain unsplit; they stay in the left-hand buffer, and
the right-hand buffer gets an empty line to correspond.

You can scroll both buffers together using C-x 6 SPC (scroll up), C-x
6 DEL (scroll down), and C-x 6 RET (scroll up one line).  C-x 6 C-l
recenters both buffers together.

If you want to make a line which will span both columns, put it in
the left-hand buffer, with an empty line in the corresponding place in
the right-hand buffer.

When you have edited both buffers as you wish, merge them with C-x 6
1.  This copies the text from the right-hand buffer as a second column
in the other buffer.  To go back to two-column editing, use C-x 6 s.

Use C-x 6 d to disassociate the two buffers, leaving each as it
stands.  (If the other buffer, the one that was not current when you
type C-x 6 d, is empty, C-x 6 d kills it.)

*** You can supply command arguments such as files to visit to an Emacs
that is already running.  To do this, you must do this in your .emacs
file:
   (add-hook 'suspend-hook 'resume-suspend-hook)
Also you must use the shellscript emacs.csh or emacs.sh, found in the
etc subdirectory.

*** Shell mode has been completely replaced.
The basic idea is the same, but there are new commands available in
this mode.
      
TAB now completes the file name before point in the shell buffer.
To get a list of all possible completions, type M-?.

There is a new convenient history mechanism for repeating previous
commands.  Use the command M-p to recall the last command; it copies
the text of that command to the place where you are editing.  If you
repeat M-p, it replaces the copied command with the previous command.
M-n is similar but goes in the opposite direction towards the present.
When you find the command you wanted, you can edit it, or just
resubmit it by typing RET.

You can also use M-r and M-s to search for (respectively) earlier or 
later inputs starting with a given string.  First type the string, 
then type M-r to yank a previous input from the history which starts
with that string.  You can repeat M-r to find successively earlier
inputs starting with the same string.  You can start moving in the
opposite direction (toward more recent inputs) by typing M-s instead
of M-r.  As long as you don't use any commands except M-r and M-s,
they keep using the same string that you had entered initially.

C-c C-o kills the last batch of output from a shell command.  This is
useful if a shell command spews out lots of output that just gets in
the way.

C-c C-r scrolls to display the beginning of the last batch of output
at the top of the window; it also moves the cursor there.

C-a on a line that starts with a shell prompt moves to the end of the
prompt, not to the very beginning of the line.

C-d typed at the end of the shell buffer sends EOF to the subshell.
At any other position in the buffer, it deletes a character as usual.

If Emacs gets confused while trying to track changes in the shell's
current directory, type M-x dirs to re-synchronize.

M-x send-invisible reads a line of text without echoing it, and
sends it to the shell.

If you accidentally suspend your process, use M-x comint-continue-subjob 
to continue it.
         
*** There is now a convenient way to enable flow control on terminals
where you can't win without it.  Suppose you want to do this on
VT-100 and H19 terminals; put the following in your `.emacs' file:

   (enable-flow-control-on "vt100" "h19")

When flow control is enabled, you must type C-\ to get the effect of a
C-s, and type C-^ to get the effect of a C-q.

The function `enable-flow-control' enables flow control unconditionally.
\f
** Changes in Dired

Dired has many new features which allow you to do these things:

- Rename, copy, or make links to many files at once.

- Make distinguishable types of marks for different operations.

- Display contents of subdirectories in the same Dired buffer as the
parent directory.

*** Setting and Clearing Marks

There are now two kinds of marker that you can put on a file in Dired:
`D' for deletion, and `*' for any other kind of operation.
The `x' command deletes only files marked with `D', and most
other Dired commands operate only on the files marked with `*'.

To mark files with `D' (also called "flagging" the files), you
can use `d' as usual.  Here are some commands for marking with
`*' (and also for unmarking):

**** `m' marks the current file with `*', for an operation other than
deletion.

**** `*' marks all executable files.  With a prefix argument, it
unmarks all those files.

**** `@' marks all symbolic links.  With a prefix argument, it unmarks
all those files.

**** `/' marks all directory files except `.' and `..'.  With a prefix
argument, it unmarks all those files.

**** M-DEL removes a specific or all marks from every file.  With an
argument, queries for each marked file.  Type your help character,
usually C-h, at that time for help.

**** `c' replaces all marks that use the character OLD with marks that
use the character NEW.  You can use almost any character as a mark
character by means of this command, to distinguish various classes of
files.  If OLD is ` ', then the command operates on all unmarked
files; if NEW is ` ', then the command unmarks the files it acts on.

*** Operating on Multiple Files

The Dired commands to operate directly on files (rename them, copy
them, and so on) have been generalized to work on multiple files.
There are also some additional commands in this series.

All of these commands use the same convention to decide which files to
manipulate:

- If you give the command a numeric prefix argument @var{n}, it operates
on the next @var{n} files, starting with the current file.

- Otherwise, if there are marked files, the commands operate on all the
marked files.

- Otherwise, the command operates on the current file only.

These are the commands:

**** `C' copies the specified files.  You must specify a directory to
copy into, or (if copying a single file) a new name.

If `dired-copy-preserve-time' is non-`nil', then copying sets
the modification time of the new file to be the same as that of the old
file.

**** `R' renames the specified files.  You must specify a directory to
rename into, or (if renaming a single file) a new name.

Dired automatically changes the visited file name of buffers associated
with renamed files so that they refer to the new names.

**** `H' makes hard links to the specified files.  You must specify a
directory to make the links in, or (if making just one link) the name
to give the link.

**** `S' makes symbolic links to the specified files.  You must specify
a directory to make the links in, or (if making just one link) the
name to give the link.

**** `M' changes the mode of the specified files.  This calls the
`chmod' program, so you can describe the desired mode change with any
argument that `chmod' would handle.

**** `G' changes the group of the specified files.

**** `O' changes the owner of the specified files.  (On normal systems,
only the superuser can do this.)

The variable `dired-chown-program' specifies the name of the
program to use to do the work (different systems put `chown' in
different places.

**** `Z' compresses or uncompresses the specified files.

**** `L' loads the specified Emacs Lisp files.

**** `B' byte compiles the specified Emacs Lisp files.

**** `P' prints the specified files.  It uses the variables
`lpr-command' and `lpr-switches' just as `lpr-file' does.

*** Shell Commands in Dired

`!' reads a shell command string in the minibuffer and runs the shell
command on all the specified files.  There are two ways of applying a
shell command to multiple files:

- If you use `*' in the command, then the shell command runs just
once, with the list of file names substituted for the `*'.

Thus, `! tar cf foo.tar * RET' runs `tar' on the entire list of file
names, putting them into one tar file `foo.tar'.  The file names are
inserted in the order that they appear in the Dired buffer.

- If the command string doesn't contain `*', then it runs once for
each file, with the file name attached at the end.  For example, `!
uudecode RET' runs `uudecode' on each file.

To run the shell command once for each file but without being limited
to putting the file name inserted in the middle, use a shell loop.
For example, this shell command would run `uuencode' on each of the
specified files, writing the output into a corresponding `.uu' file:

    for file in *; uuencode $file $file >$file.uu; done

The working directory for the shell command is the top level directory
of the Dired buffer.

*** Regular Expression File Name Substitution

**** `% m REGEXP RET' marks all files whose names match the regular
expression REGEXP.

Only the non-directory part of the file name is used in matching.  Use
`^' and `$' to anchor matches.  Exclude subdirs by hiding them.

**** `% d REGEXP RET' flags for deletion all files whose names match
the regular expression REGEXP.

**** `% R', `% C', `% H', `% S'

These four commands rename, copy, make hard links and make soft links,
in each case computing the new name by regular expression substitution
from the name of the old file.  They  effectively perform
`query-replace-regexp' on the selected file names in the Dired buffer.

The commands read two arguments: a regular expression, and a
substitution pattern.  Each selected file name is matched against the
regular expression, and then the part which matched is replaced with
the substitution pattern.  You can use `\&' and `\DIGIT' in the
substitution pattern to refer to all or part of the old file name.

If the regular expression matches more than once in a file name,
only the first match is replaced.

Normally, the replacement process does not consider the directory names;
it operates on the file name within the directory.  If you specify a
prefix argument of zero, then replacement affects entire file name.

To apply the command to all files matching the same regexp that you
use in the command, mark those files with `% m REGEXP RET', then use
the same regular expression in `% R'.  To make this easier, `% R' uses
as a default the last regular expression specified in a `%' command.

*** Dired Case Conversion

**** `% u' renames each of the selected files to an upper case name.

**** `% l' renames each of the selected files to a lower case name.

*** File Comparison with Dired

**** `=' compares the current file with another file (the file at the
mark), by running the `diff' program.  The file at the mark is given
to `diff' first.

**** `M-=' compares the current file with its backup file.  If there
are several numerical backups, it uses the most recent one.  If this
file is a backup, it is compared with its original.

The backup file is the first file given to `diff'.

*** Subdirectories in Dired

You can display more than one directory in one Dired buffer.
The simplest way to do this is to specify the options `-lR' for
running `ls'.  That produces a recursive directory listing showing
all subdirectories, all within the same Dired buffer.

You can also insert the contents of a particular subdirectory with the
`i' command.  Use this command on the line that describes a file which
is a directory.  Inserted subdirectory contents follow the top-level
directory of the Dired buffer, just as they do in `ls -lR' output.

If the subdirectory's contents are already present in the buffer, the
`i' command just moves to it (type `l' to refresh it).  It sets the
Emacs mark before moving, so C-x C-x takes you back to the old
position in the buffer.

When you have subdirectories in the Dired buffer, you can use the page
motion commands C-x [ and C-x ] to move by entire directories.

The following commands move up and down in the tree of directories
in one Dired buffer:

**** C-M-u  Go up to the parent directory's headerline.

**** C-M-d  Go down in the tree, to the first subdirectory's
headerline.

**** C-M-n  Go to next subdirectory headerline, regardless of level.

**** C-M-p  Go to previous subdirectory headerline, regardless of
level.

*** Hiding Subdirectories

"Hiding" a subdirectory means to make it invisible, except for its
headerline.  Files inside a hidden subdirectory are never considered
by Dired.  For example, the commands to operate on marked files ignore
files in hidden directories even if they are marked.

**** `$' hides or unhides the current subdirectory and move to next
subdirectory.  A prefix argument serves as a repeat count.

**** `M-$' hides all subdirectories, leaving only their header lines.
Or, if at least one subdirectory is currently hidden, it makes
everything visible again.  You can use this command to get an overview
in very deep directory trees or to move quickly to subdirectories far
away.

*** Editing the Dired Buffer

**** `l' updates the specified files in a Dired buffer.  This means
reading their current status from the file system and changing the
buffer to reflect it properly.

If you use this command on a subdirectory header line, it updates the
contents of the subdirectory.

**** `g' updates the entire contents of the Dired buffer.  It preserves
all marks except for those on files that have vanished.  Hidden
subdirectories are updated but remain hidden.

**** `k' kills all marked lines (not the files).  With a prefix
argument, it kills that many lines starting with the current line.

This command does not delete files; it just deletes text from the Dired
buffer.

If you kill the line for a file that is a directory, then its contents
are also deleted from the buffer.  Typing `C-u k' on the header line
for a subdirectory is another way to delete a subdirectory from the
Dired buffer.

*** `find' and Dired.

To search for files with names matching a wildcard pattern use
`find-name-dired'.  Its arguments are DIRECTORY and
PATTERN.  It selects all the files in DIRECTORY or its
subdirectories whose own names match PATTERN.

The files thus selected are displayed in a Dired buffer in which the
ordinary Dired commands are available.

If you want to test the contents of files, rather than their names, use
`find-grep-dired'.  This command takes two minibuffer arguments,
DIRECTORY and REGEXP; it selects all the files in
DIRECTORY or its subdirectories that contain a match for
REGEXP.  It works by running `find' and `grep'.

The most general command in this series is `find-dired', which lets
you specify any condition that `find' can test.  It takes two
minibuffer arguments, DIRECTORY and FIND-ARGS; it runs `find' in
DIRECTORY with using FIND-ARGS as the arguments to `find' specifying
which files to accept.  To use this command, you need to know how to
use `find'.
\f
** New amusements and novelties.

*** `M-x mpuz' displays a multiplication puzzle, in which each letter
stands for a digit, and you must determine which digit.  The puzzles
are determined randomly, so they are always different.

*** `M-x gomoku' plays the game Gomoku with you.  It needs more work.

*** `M-x spook' adds a line of randomly chosen keywords to an outgoing
mail message.  The keywords are chosen from a list of words that
suggest you are discussing something subversive.

The idea is that the NSA reads all messages that contain keywords
suggesting they might be interested, and that adding these lines could
help to overload them.  I would guess that they have modified their
program by now to ignore these lines of keywords; perhaps the program
can be updated if some clever hacker can determine what criterion they
actually use now.
\f
** Installation changes

*** The configure script has been provided to help with the
installation process.  It takes the place of editing the Makefiles and
src/config.h, and can often guess the appropriate operating system to
use for a particular machine type.  See INSTALL for a more detailed
description of the steps required for installation.

*** If you create a Lisp file named `site-start.el', Emacs loads the file
whenever it starts up.

*** A new Lisp variable, `data-directory', indicates the directory
containing the DOC file, tutorial, copying agreement, and other
familiar `etc' files.  The value of `data-directory' is a simple string.
The default should be set at build time, and the person installing
Emacs should place all the data files in this directory.  The `help.el'
functions that look for docstrings and information files check this
variable.  All Emacs Lisp packages should also be coded so that they
refer to `data-directory' to find data files.

*** The PURESIZE definition has been moved from config.h to its own
file, puresize.h.  Since almost every file of C source in the
distribution depends on config.h, but only alloc.c and data.c depend
on puresize.h, this means that changing the value of PURESIZE causes
only those two files to be recompiled.

*** The makefile at the top of the Emacs source tree now supports a
`dist' target, which creates a compressed tar file suitable for
distribution, using the contents of the source tree.  Object files,
old file versions, executables, DOC files, and other
architecture-specific or easy-to-recreate files are not included in
the tar file.
\f
* For older news, see the file ONEWS.4.  For Lisp changes in (the first
* release of) Emacs 19, see the file LNEWS.

----------------------------------------------------------------------
Copyright information:

Copyright (C) 1993, 1994, 1995 Free Software Foundation, Inc.

   Permission is granted to anyone to make or distribute verbatim copies
   of this document as received, in any medium, provided that the
   copyright notice and this permission notice are preserved,
   thus giving the recipient permission to redistribute in turn.

   Permission is granted to distribute modified versions
   of this document, or of portions of it,
   under the above conditions, provided also that they
   carry prominent notices stating who last changed them.
\f
Local variables:
mode: outline
paragraph-separate: "[  \f]*$"
end: