# What’s New in Emacs 27.1?

As tradition dictates, I have annotated the latest Emacs 27.1 NEWS file with my own comments. It’s a big ’un, this time. The maintainers have done a wonderful job, as always.

If you want to build Emacs 27, you can find instructions for Ubuntu in Speed up Emacs with libjansson and native elisp compilation. Conveniently, that article also shows you how to build libjansson, the native C library for JSON. That feature is mandatory if you use LSP servers.

## Installation Changes in Emacs 27.1

Emacs now uses GMP, the GNU Multiple Precision library.
By default, if 'configure' does not find a suitable libgmp, it
arranges for the included mini-gmp library to be built and used.
The new configure option '--without-libgmp' uses mini-gmp even if a
suitable libgmp is available.

That’s a useful change indeed. M-x calc (an amazing RPN calculator, if you’ve never used it before) used its own numerical library to get around the inherent limitations of the built-in numerical library Emacs used to use. As a result M-: (* 198391258329 984958834) now yields the correct answer compared to older Emacsen.

I’m unsure whether M-x calc now uses the self-same library also. Presumably so?

Emacs can now use HarfBuzz as its shaping engine.
The new configure option '--with-harfbuzz' adds support for the
HarfBuzz text shaping engine.  It is on by default; use './configure
--without-harfbuzz' to build without it.  The HarfBuzz text shaping is
available via new font backend drivers 'xfthb' and 'ftcrhb' for Xft
and Cairo drawings, respectively, and via the 'harfbuzz' backend on
MS-Windows.  The HarfBuzz text shaping is preferred to the previously
supported ones, so the font backends that use older shaping engines
(FLT on GNU and Unix systems and Uniscribe on MS-Windows) are not
enabled by default; they can be enabled via the 'font-backend' frame
parameter or via X resources.

So, this is actually an interesting addition and one that I think will go unnoticed by most people. The first thing is that Emacs is now outsourcing the text shaping to a third-party package; that’s good, in and of itself. It lets the maintainers focus on other areas.

How it benefits you is even more interesting. With this change Emacs now gets “proper” support for ligatures, colorful emojis – you name it, provided the font and the text shaper supports it. That’s a big change: there’s apparently an entire contingent of people for whom ligatures in code is essential. If you’re relying on Emacs’s existing M-x prettify-symbols-mode this might be another way to make use of ligatures in your font, if it supports it.

But more on this feature in a separate article.

The new configure option '--with-json' adds native support for JSON.
This uses the Jansson library.  The option is on by default; use
'./configure --with-json=no' to build without Jansson support.  The
new JSON functions 'json-serialize', 'json-insert',
'json-parse-string', and 'json-parse-buffer' are typically much faster
than their Lisp counterparts from json.el.

This is a game changer to people who use language server tools like eglot or lsp-mode, or any other tool that (de)serializes a lot of JSON. The older implementation is perfecly fine, but it’s slow, as it’s written in elisp. The new one is a C library.

In my article Speed up Emacs with libjansson and native elisp compilation I talk at length about the benefits of both the native, C-compiled JSON support and how important that change is to tools such as eglot and lsp-mode, and also a flashy, experimental feature called native compilation.

The configure option '--with-cairo' is no longer experimental.
This builds Emacs with Cairo drawing, and supports built-in printing
when Emacs is built with GTK+.  Some severe bugs in this build were
fixed, and we can therefore offer this to users without caveats.  Note
that building with Cairo enabled results in using Pango instead of
libXft for font support, and that Pango 1.44 has removed support for
bitmapped fonts.

I know little about Cairo or its primary benefits, beyond the ability to do built-in printing; so if you are someone who would benefit from Cairo over whatever Emacs normally uses – then please, leave a comment.

Emacs now uses a "portable dumper" instead of unexec.
This improves compatibility with memory allocation on modern systems,
and in particular better supports the Address Space Layout
Randomization (ASLR) feature, a security technique used by most modern
operating systems.

When built with the portable dumping support (which is the default),
Emacs looks for the "emacs.pdmp" file, generated during the build, in
its data directory at startup, and loads the dumped state from there.
The new command-line argument '--dump-file=FILE' allows specifying a
non-default ".pdmp" file to load the state from; see the node

An Emacs started via a dump file can create a new dump file only if it
was invoked with the '-batch' option.  (This is a temporary
limitation; we plan on lifting it in a future release.)

Although the portable dumper has been tested, it may have a bug on
unusual platforms.  If you require traditional unexec dumping you can
use the configure-time option '--with-dumping=unexec'; however, please
file a bug report describing the situation, as unexec dumping is
deprecated, and we plan on removing it in some future release.

There’s a story here. Briefly, Emacs would – for performance reasons, which is less of a concern today, than when the feature was introduced – load in all the emacs lisp libraries and other modules it depends on, then dump its memory to a new binary using a rather obscure feature of glibc. Well, eventually, the glibc maintainers wanted to remove support as Emacs was the only known tool to still do this. You can read this excellent LWN article for a summary.

I suppose one advantage of the new portable dumper, for the average user, is that you could in theory load in all your favorite packages and then dump a new binary?

The new configure option '--enable-checking=structs' attempts to
check that the portable dumper code has been updated to match the last
change to one of the data structures that it relies on.
The configure options '--enable-checking=conslist' and
'--enable-checking=xmallocoverrun' have been withdrawn.  The former
option was useful with modern debugging tools such as AddressSanitizer.
(See "etc/DEBUG" for the details of using the modern replacements of the
removed configure options.)
Emacs no longer defaults to using ImageMagick to display images.
This is due to security and stability concerns with ImageMagick.  To
override the default, use 'configure --with-imagemagick'.
Several configure options now accept an option-argument 'ifavailable'.
For example, './configure --with-xpm=ifavailable' now configures Emacs
to attempt to use libxpm but to continue building even if libxpm is
absent.  The other affected options are '--with-gif', '--with-gnutls',
'--with-jpeg', '--with-png', and '--with-tiff'.
The 'etags' program now uses the C library's regular expression matcher.
If it's possible, 'etags' will use the regexp matcher from the
system's standard C library, otherwise it will be linked with a
compatible regex substitute.  This lets developers maintain Emacs's
own regex code without having to also support other programs.  The new
configure option '--without-included-regex' forces 'etags' to use the C
library's regex matcher even if the regex substitute ordinarily would
be used to work around compatibility problems.

ETags is certainly useful, but I find that a lot of people get sucked into using it when they maybe don’t need to. So unless you work on codebases large enough to cause gravitational lensing, I recommend you give dumb-jump a try. It’ll probably do most of what you want without maintaining an TAGS file.

Emacs has been ported to the '-fcheck-pointer-bounds' option of GCC.
This causes Emacs to check bounds of some arrays addressed by its
internal pointers, which can be helpful when debugging the Emacs
interpreter or modules that it uses.  If your platform supports it you
can enable it when configuring, e.g., './configure CFLAGS="-g3 -O2
-mmpx -fcheck-pointer-bounds"' on Intel MPX platforms.
Emacs now normally uses a C pointer type instead of a C integer
type to implement Lisp_Object, which is the fundamental machine word
type internal to the Emacs Lisp interpreter.  This change aims to
catch typos and supports '-fcheck-pointer-bounds'.  The configure
option '--enable-check-lisp-object-type' is therefore no longer as
useful and so is no longer enabled by default in developer builds,
to reduce differences between developer and production builds.
The distribution tarball now has test cases; 'make check' runs them.
This is intended mostly to help developers.
Emacs now requires GTK 2.24 and GTK 3.10 for the GTK 2 and GTK 3
builds respectively.
New make target 'help' shows a summary of common make targets.
Emacs now builds with dynamic module support by default.
Pass '--without-modules' to 'configure' to disable dynamic module
support.

A small change, but likely to result in more people considering dynamic modules knowing that users are more likely to have it enabled by default. It’s a shame dynamic modules never saw much uptick, though.

The ftx font backend driver is now obsolete and will be removed in
Emacs 28.

## Startup Changes in Emacs 27.1

Emacs can now use the XDG convention for init files.
The 'XDG_CONFIG_HOME' environment variable (which defaults to
"~/.config") specifies the XDG configuration parent directory.  Emacs
checks for "init.el" and other configuration files inside the "emacs"
subdirectory of 'XDG_CONFIG_HOME', i.e. "$XDG_CONFIG_HOME/emacs/init.el" However, Emacs will still initially look for init files in their traditional locations if "~/.emacs.d" or "~/.emacs" exist, even if "$XDG_CONFIG_HOME/emacs" also exists.  This means that you must delete
or rename any existing "~/.emacs.d" and "~/.emacs" to enable use of
the XDG directory.

If "~/.emacs.d" does not exist, and Emacs has decided to use it
(i.e. "$XDG_CONFIG_HOME/emacs" does not exist), Emacs will create it. Emacs will never create "$XDG_CONFIG_HOME/emacs".

Whichever directory Emacs decides to use, it will set
'user-emacs-directory' to point to it.

Fans of clutter-free home directories, rejoice. Emacs will prefer ~/.config/... (or XDG_CONFIG_HOME) to the usual ~/.emacs.d.

Emacs can now be configured using an early init file.
The file is called "early-init.el", in 'user-emacs-directory'.  It is
loaded very early in the startup process: before graphical elements
such as the tool bar are initialized, and before the package manager
is initialized.  The primary purpose is to allow customizing how the
package system is initialized given that initialization now happens

We recommend against putting any customizations in this file that
don't need to be set up before initializing installed add-on packages,
because the early init file is read too early into the startup
process, and some important parts of the Emacs session, such as
'window-system' and other GUI features, are not yet set up, which could
make some customization fail to work.

A required change as package management initialization is now automatic and happens before the init file is loaded (see below.)

Installed packages are now activated *before* loading the init file.
As a result of this change, it is no longer necessary to call

Previously, a call to 'package-initialize' was automatically inserted
into the init file when Emacs was started.  This call can now safely
be removed.  Alternatively, if you want to ensure that your init file
is still compatible with earlier versions of Emacs, change it to:

(when (< emacs-major-version 27)
(package-initialize))

or 'package-user-dir', or sets 'package-enable-at-startup' to nil then
it won't work right without some adjustment:
- You can move that code to the early init file (see above), so those
settings apply before Emacs tries to activate the packages.
- You can use the new 'package-quickstart' so activation of packages
does not need to pay attention to 'package-load-list' or
'package-user-dir' any more.
Emacs now notifies systemd when startup finishes or shutdown begins.
Units that are ordered after 'emacs.service' will only be started
after Emacs has finished initialization and is ready for use.
(If your Emacs is installed in a non-standard location and you copied the
emacs.service file to e.g. "~/.config/systemd/user/", you will need to copy
the new version of the file again.)

## Changes in Emacs 27.1

Emacs now supports Unicode Standard version 13.0.
Emacs now supports resizing and rotating images without ImageMagick.
All modern systems support this feature.  (On GNU and Unix systems,
Cairo drawing or the XRender extension to X11 is required for this to
be available; the configure script will test for it and, if found,
enable scaling.)

The new function 'image-transforms-p' can be used to test whether any
given frame supports these capabilities.
The Network Security Manager now allows more fine-grained control
of what checks to run via the 'network-security-protocol-checks'
user option.

In practice this means you can enable or disable support for legacy TLS protocols.

TLS connections have their security tightened by default.
Most of the checks for outdated, believed-to-be-weak TLS algorithms
and ciphers are now switched on by default.  (In addition, several new
TLS weaknesses are now warned about.)  By default, the NSM will
flag connections using these weak algorithms and ask users whether to
allow them.  To get the old behavior back (where certificates are
checked for validity, but no warnings about weak cryptography are
issued), you can either set 'network-security-protocol-checks' to nil,
or adjust the elements in that user option to only happen on the 'high'
security level (assuming you use the 'medium' level).

If you find this change breaks your setup, you may need to follow the instructions above.

New user option 'nsm-trust-local-network'.
Allows skipping Network Security Manager checks for hosts on your
local subnet(s).  It defaults to nil.  Usually, there should be no
need to set this non-nil, and doing that risks opening your local
network connections to attacks.  So be sure you know what you are
doing before changing the value.
Native GnuTLS connections can now use client certificates.
Previously, this support was only available when using the external
'gnutls-cli' or 'starttls' command.  Call 'open-network-stream' with
':client-certificate t' to trigger looking up of per-server
certificates via 'auth-source'.
New user option 'network-stream-use-client-certificates'.
When non-nil, 'open-network-stream' performs lookups of client
certificates using 'auth-source' as if ':client-certificate t' were
specified if there is no explicit ':client-certificate' parameter.
Defaults to nil.
'next/previous-multiframe-window' have been renamed.
The new names are as follows:

'next-multiframe-window'     -> 'next-window-any-frame'
'previous-multiframe-window' -> 'previous-window-any-frame'

The old function names are maintained as aliases for backward
compatibility.

### emacsclient

emacsclient now supports the 'EMACS_SOCKET_NAME' environment variable.
The command-line argument '--socket-name' overrides it.
(The same behavior as for the pre-existing 'EMACS_SERVER_FILE' variable.)
Emacs and emacsclient now default to "XDG_RUNTIME_DIR/emacs". This is used as the directory for client/server sockets, if Emacs is running on a platform or environment that sets the 'XDG_RUNTIME_DIR' environment variable to indicate where session sockets should go. To get the old, less-secure behavior, you can set the 'EMACS_SOCKET_NAME' environment variable to an appropriate value. When run by root, emacsclient no longer connects to non-root sockets. (Instead you can use Tramp methods to run root commands in a non-root Emacs.) 'xft-ignore-color-fonts' now ignores even more color fonts. There are color fonts that managed to bypass the existing checks, causing XFT crashes, they are now filtered out. Setting 'xft-ignore-color-fonts' to nil removes those checks, which might require setting 'face-ignored-fonts' to filter out problematic fonts. Known problematic fonts are "Noto Color Emoji" and "Emoji One". I’m curious why these crashes happen at all. I would have thought the Harfbuzz change would result in greater compatibility and stability. The GTK+ font chooser now respects 'face-ignored-fonts'. When using 'menu-set-font' under GTK3, the available fonts are now matched against 'face-ignored-fonts'. The GTK+ font chooser now remembers the previously selected settings. It now remembers the name, size, style, etc. New user option 'what-cursor-show-names'. When non-nil, 'what-cursor-position' will show the name of the character in addition to the decimal/hex/octal representation. Default nil. If you use C-u C-x = to check for what character point is on, you are already given this information. This only adds the name of character code to the abbreviated echo area version. New function 'network-lookup-address-info'. This does IPv4 and/or IPv6 address lookups on hostnames. A ‘native’ way of looking up hosts. There is already M-x dig and M-x nslookup but they use the underlying programs to carry out this action. 'network-interface-list' can now return IPv4 and IPv6 addresses. IPv4 and IPv6 addresses are now returned by default if available, optionally including netmask/broadcast address information. It’s nice to see these things added, even though few people will ever use them. I can imagine they’re useful if you want Emacs to only talk to certain things if you are on a VPN or something. Control of the threshold for using the 'distant-foreground' color. The threshold for color distance below which the 'distant-foreground' color of the face will be used instead of the foreground color can now be controlled via the new variable 'face-near-same-color-threshold'. The default value is 30000, as the previously hard-coded threshold. The function 'read-passwd' uses "*" as default character to hide passwords. Whereas before it used .. The function 'read-answer' now accepts not only single character answers, but also function keys like 'F1', character events such as 'C-M-h', and control characters like 'C-h'. Lexical binding is now used by default when evaluating interactive Elisp. More specifically, 'lexical-binding' is now used by default for 'M-:' and '--eval' (including in evaluations invoked from 'emacsclient' via its '--eval' command-line option), as well as in 'lisp-interaction-mode' and 'ielm-mode', used in the "*scratch*" and "*ielm*" buffers. We envision that most Lisp code is already either written with lexical-binding in mind, or will work unchanged under lexical-binding. If, for some reason, your code used in 'M-:' or '--eval' doesn't work as result of this change, either modify the code to work with lexical binding, or wrap it in an extra level of 'eval'. For example, --eval "FORM" becomes --eval "(eval 'FORM)" (note the extra quote in 'FORM). The onward march of lexical scoping in Emacs. It’s indeed unlikely you are negatively impacted by this, unless you have explicitly written code that (ab)uses dynamic scoping. The new user option 'tooltip-resize-echo-area' avoids truncating tooltip text on GUI frames when tooltips are displayed in the echo area. Instead, it resizes the echo area as needed to accommodate the full tool-tip text. Show mode line tooltips only if the corresponding action applies. Customize the user option 'mode-line-default-help-echo' to restore the old behavior where the tooltip text is also shown when the corresponding action does not apply. New hook 'server-after-make-frame-hook'. This hook is a convenient place to perform initializations in daemon mode which require GUI features to be available. One example is restoration of the previous session using the desktop.el package: put the call to 'desktop-read' in this hook, if you want the GUI settings to be restored, or if desktop.el needs to interact with you during restoration of the session. The functions 'set-frame-height' and 'set-frame-width' are now commands, and will set the currently selected frame to the height/ width specified by the numeric prefix. New function 'logcount' calculates an integer's Hamming weight. New function 'libxml-available-p'. This function returns non-nil if libxml support is both compiled in and available at run time. Lisp programs should use this function to detect built-in libxml support, instead of testing for that indirectly, e.g., by checking that functions like 'libxml-parse-html-region' return nil. 'libxml-parse-xml-region' and 'libxml-parse-html-region' take a parameter that's called DISCARD-COMMENTS, but it really only discards the top-level comment. Therefore this parameter is now obsolete, and the new utility function 'xml-remove-comments' can be used to remove comments before calling the libxml functions to parse the data. A new DOM (the XML/HTML document structure returned by functions such as 'libxml-parse-html-region') traversal function has been added: 'dom-search', which takes a DOM and a predicate and returns all nodes that match. New function 'fill-polish-nobreak-p', to be used in 'fill-nobreak-predicate'. It blocks line breaking after a one-letter word, also in the case when this word is preceded by a non-space, but non-alphanumeric character. The limit on repetitions in regexps has been raised to 2^16-1. It was previously limited to 2^15-1. For example, the following regular expression was previously invalid, but is now accepted: x\{32768\} The German prefix and postfix input methods now support Capital sharp S. New input methods 'hawaiian-postfix' and 'hawaiian-prefix'. New input methods 'georgian-qwerty' and 'georgian-nuskhuri'. New input methods for several variants of the Sami language. The Sami input methods include: 'norwegian-sami-prefix', 'bergsland-hasselbrink-sami-prefix', 'southern-sami-prefix', 'ume-sami-prefix', 'northern-sami-prefix', 'inari-sami-prefix', 'skolt-sami-prefix', and 'kildin-sami-prefix'. Japanese environments use UTF-8 by default. In Japanese environments that do not specify encodings and are not based on MS-Windows, the default encoding is now utf-8 instead of japanese-iso-8bit. New function 'exec-path'. This function by default returns the value of the corresponding user option, but can optionally return the equivalent of 'exec-path' from a remote host. The function 'executable-find' supports an optional argument REMOTE. This triggers searching for the program on the remote host as indicated by 'default-directory'. Both features are welcome additions to authors wanting to build tooling around programs on remote servers. I guess tools like lsp-mode and eglot or M-x compile would benefit from this. New user option 'auto-save-no-message'. When set to t, no message will be shown when auto-saving (default value: nil). The value of 'make-cursor-line-fully-visible' can now be a function. In addition to nil or non-nil, the value can now be a predicate function. Follow mode uses this to control scrolling of its windows when the last screen line in a window is not fully visible. New variable 'emacs-repository-branch'. It reports the git branch from which Emacs was built. New user option 'switch-to-buffer-obey-display-actions'. When non-nil, 'switch-to-buffer' uses 'pop-to-buffer-same-window' that respects display actions specified by 'display-buffer-alist' and 'display-buffer-overriding-action'. The user option 'switch-to-visible-buffer' is now obsolete. Customize 'switch-to-prev-buffer-skip' instead. New user option 'switch-to-prev-buffer-skip'. This user option allows specifying the set of buffers that may be shown by 'switch-to-prev-buffer' and 'switch-to-next-buffer' more stringently than the now obsolete 'switch-to-visible-buffer'. New 'flex' completion style. An implementation of popular "flex/fuzzy/scatter" completion which matches strings where the pattern appears as a subsequence. Put simply, makes "foo" complete to both "barfoo" and "frodo". Add 'flex' to 'completion-styles' or 'completion-category-overrides' to use it. This is only relevant if you make use of Emacs’s default completion mechanism. It seems like a useful addition, but if you rely on TAB-completing stuff with the default completer, know that it will generate a lot of line noise. The 'completion-common-part' face is now visible by default. New face attribute ':extend' to control face extension at EOL. The new face attribute ':extend' controls whether to use the face for displaying the empty space beyond end of line (EOL) till the edge of the window. By default, this attribute is non-nil only for a small number of faces, notably, 'region'; any other face that crosses end of line will not affect the display of the empty space at EOL. This is to make Emacs behave more like other GUI applications with respect to displaying faces that cross line boundaries. This attribute behaves specially when theme definitions are applied: if the theme doesn't specify an explicit value of this attribute for a face, the value from the original face definition is inherited. Consequently, a theme generally shouldn't specify this attribute unless it has a good reason to do so. ### Connection-local variables Connection-local variables are applied by default like file-local and directory-local variables. The macro 'with-connection-local-variables' has been renamed from 'with-connection-local-profiles'. No argument PROFILES needed any longer. I don’t recall seeing connection-local variables in earlier Emacsen so I could be mistaken, but this means you can configure ‘remote environments’ with TRAMP that take effect when you carry out certain actions that match a profile. See 49.2.6 Per-Connection Local Variables for (a little) more information. New user option 'next-error-verbose' controls when 'next-error' outputs a message about the error locus. When you use next-error to move between errors with M-g M-n/p you are now told the name of the group of errors you are jumping to. That matters a lot if you often end up with multiple competing error loci, as they call them, such as flycheck + grep + compilation all competing to give you information. New user option 'grep-search-path' defines the directories searched for grep hits (this used to be controlled by 'compilation-search-path'). Interesting change. I can imagine this would serve package authors well if they want to collect grep (or grep-like) results from locations that do not match their actual path. New user option 'emacs-lisp-compilation-search-path' defines the directories searched for byte-compiler error messages (this used to be controlled by 'compilation-search-path'). Multicolor fonts such as "Noto Color Emoji" can be displayed on Emacs configured with Cairo drawing and linked with cairo >= 1.16.0. I checked cairo-version-string on Ubuntu 20.04 and I do indeed have that version – so it’s likely that a reasonably modern development environment will give you this out of the box as I do not recall making any effort at all to get this working in my compiled build of Emacs. Emacs now optionally displays a fill column indicator. This is similar to what 'fill-column-indicator' package provides, but much faster and compatible with 'show-trailing-whitespace'. Customize the buffer-local user options 'display-fill-column-indicator' and 'display-fill-column-indicator-character' to activate the indicator. The indicator is not displayed at all in minibuffer windows and in tooltips, as it is not useful there. There are 2 new buffer local variables and 1 face to customize this mode, they are described in the manual "(emacs) Display". Nifty feature for displaying an indicator column, but I had trouble getting it to work right with arbitrary characters. Not sure if it’s a bug or not. But if you want an indicator column then it’s worth looking at setting it up. I also imagine the code might be useful for general in-buffer notifications. 'progress-reporter-update' now accepts an optional suffix string to display. New user option 'xref-file-name-display' controls the display of file names in xref buffers. New user option 'byte-count-to-string-function'. It is used for displaying file sizes and disk space in some cases. Emacs now interprets RGB triplets like HTML, SVG, and CSS do. The X convention previously used differed slightly, particularly for RGB triplets with a single hexadecimal digit per component. The toolbar now shows the equivalent key binding in its tooltips. The File menu-bar menu was re-arranged. Print menu items moved to submenu, and also added the new entries for tabs. 'scroll-lock-mode' is now bound to the 'Scroll_Lock' key globally. Note that this key binding will not work on MS-Windows systems if 'w32-scroll-lock-modifier' is non-nil. 'global-set-key', called interactively, now no longer downcases a key binding with an upper case letter - if you can type it, you can bind it. 'read-from-minibuffer' now works with buffer-local history variables. The HIST argument of 'read-from-minibuffer' now works correctly with buffer-local variables. This means that different buffers can have their own separated input history list if desired. 'backup-by-copying-when-privileged-mismatch' applies to file gid, too. In addition to checking the file owner uid, Emacs also checks that the group gid is not greater than 'backup-by-copying-when-privileged-mismatch'; if so, 'backup-by-copying-when-mismatch' will be forced on. Editing Changes in Emacs 27.1 When asked to visit a large file, Emacs now offers to visit it literally. Previously, Emacs would only ask for confirmation before visiting large files. Now it also offers a third alternative: to visit the file literally, as in 'find-file-literally', which speeds up navigation and editing of large files. Good that you are offered the opportunity to open large files literally, but the fact that long lines (despite efforts by modes like so-long-mode, etc. that tries to alleviate it) still cripple Emacs is one of my biggest frustrations to date. 'zap-to-char' now uses the history of characters you used to zap to. 'zap-to-char' uses the new 'read-char-from-minibuffer' function to allow navigating through the history of characters that have been input. This is mostly useful for characters that have complex input methods where inputting the character again may involve many keystrokes. Useful, I suppose, in that you would probably attempt to navigate the history in the zap command prompt by instinct. 'save-some-buffers' now has a new action in the prompt: 'C-f' will exit the command and switch to the buffer currently being asked about. More commands support noncontiguous rectangular regions, namely 'upcase-dwim', 'downcase-dwim', 'capitalize-dwim', 'capitalize-region', 'upcase-initials-region', 'replace-string', 'replace-regexp', and 'delimit-columns-region'. The new 'amalgamating-undo-limit' variable can be used to control how many changes should be amalgamated when using the 'undo' command. The 'newline-and-indent' command (commonly bound to 'RET' in many modes) now takes an optional numeric argument to specify how many times is should insert newlines (and indent). New command 'make-empty-file'. New variable 'x-wait-for-event-timeout'. This controls how long Emacs will wait for updates to the graphical state to take effect (making a frame visible, for example). New user option 'electric-quote-replace-double'. This option controls whether '"' is replaced in 'electric-quote-mode', in addition to other quote characters. If non-nil, ASCII double-quote characters that quote text "like this" are replaced by double typographic quotes, “like this”, in text modes, and in comments in non-text modes. Electric quote mode fixes quote symbols so you have the typographical start and end quote symbols. I suppose it works the same for languages that use guillemets. New user option 'flyspell-case-fold-duplications'. This option controls whether Flyspell mode considers consecutive words to be duplicates if they are not in the same case. If non-nil, the default, words are considered to be duplicates even if their letters' case does not match. 'write-abbrev-file' now includes special properties. 'write-abbrev-file' now writes special properties like ':case-fixed' for abbrevs that have them. 'write-abbrev-file' skips empty tables. 'write-abbrev-file' now skips inserting a 'define-abbrev-table' form for tables which do not have any non-system abbrevs to save. The new functions and commands 'text-property-search-forward' and 'text-property-search-backward' have been added. These provide an interface that's more like functions like 'search-forward'. 'add-dir-local-variable' now uses dotted pair notation syntax to write alists of variables to ".dir-locals.el". This is the same syntax that you can see in the example of a ".dir-locals.el" file in the node "(emacs) Directory Variables" of the user manual. Network connections using 'local' can now use IPv6. 'make-network-process' now uses the correct loopback address when asked to use ':host 'local' and ':family 'ipv6'. The new function 'replace-region-contents' replaces the current region using a given replacement-function in a non-destructive manner (in terms of 'replace-buffer-contents'). The command 'replace-buffer-contents' now has two optional arguments mitigating performance issues when operating on huge buffers. Dragging 'C-M-mouse-1' now marks rectangular regions. Great if you don’t mind reaching for your mouse to make rectangular selections. The command 'delete-indentation' now operates on the active region. If the region is active, the command joins all the lines in the region. When there's no active region, the command works on the current and the previous or the next line, as before. You can now change the font size with the mouse wheel. Scrolling the mouse wheel with the Ctrl key pressed will now act the same as the 'C-x C-+' and 'C-x C--' commands. ## Changes in Specialized Modes and Packages in Emacs 27.1 New HTML mode skeleton 'html-id-anchor'. This new command (which inserts an <a id="foo">_</a> skeleton) is bound to 'C-c C-c #'. New command 'font-lock-refontify'. This is an interactive convenience function to be used when developing font locking for a mode. It recomputes the font locking data and then re-fontifies the buffer. It may not seem it, but that is a very useful change. Tweaking font lock rules is not fun even at the best of times, and having to deal with stale font lock rules even less so. Font Lock is smarter about fontifying unterminated strings and comments. When you type a quote that starts a string, or a comment delimiter that starts a comment, font-lock will not immediately refontify the following characters in 'font-lock-string-face' or 'font-lock-comment-face'. Instead, it will delay the fontification beyond the current line to give you a chance to close the string or comment. This is controlled by the new user option 'jit-lock-antiblink-grace', which specifies the delay in seconds. The default is 2 seconds; set to nil to get back the old behavior. I suppose this is mostly to stop the font locker from going hog-wild and bleed into multiple, possibly thousands, of lines. As a lot of people use either M-x electric-pair-mode or similar, the delay before resuming normal behavior seems sensible. It’s one of those things that you’re unlikely to notice, yet silently makes your life a little bit better. The 'C' command in 'tar-mode' will now preserve the timestamp of the extracted file if the new user option 'tar-copy-preserve-time' is non-nil. 'autoconf-mode' is now used instead of 'm4-mode' for the "acinclude.m4" / "aclocal.m4" / "acsite.m4" files. On GNU/Linux, 'M-x battery' will now list all batteries, no matter what they're named, and the 'battery-linux-sysfs-regexp' variable has been removed. The 'list-processes' command now includes port numbers in the network connection information (in addition to the host name). The 'cl' package is now officially deprecated in favor of 'cl-lib'. ### desktop When called interactively with a prefix arg 'C-u', 'desktop-read' now prompts the user for the directory containing the desktop file. #### display-line-numbers-mode New faces 'line-number-major-tick' and 'line-number-minor-tick', and user options 'display-line-numbers-major-tick' and 'display-line-numbers-minor-tick' can be used to highlight the line numbers of lines multiple of certain numbers. New variable 'display-line-numbers-offset', when non-zero, adds an offset to absolute line numbers. ### winner A new user option, 'winner-boring-buffers-regexp', has been added. You can amend the variable above to tell M-x winner-mode, a useful tool to restore window configurations with simple key binds, to not restore certain buffers matching that regexp. ### table 'table-generate-source' now supports wiki and mediawiki. This command can now output wiki and mediawiki format tables. ### telnet-mode Reverting a buffer in 'telnet-mode' will restart a closed connection. Yep, of course there’s a telnet-mode. There is also a M-x serial-term for serial connections. Useful for talking to hardware devices! ### goto-addr A way to more conveniently specify what URI address schemes should be ignored has been added via the 'goto-address-uri-schemes-ignored' variable. ### tex-mode 'latex-noindent-commands' controls indentation of certain commands. You can use this new user option to control indentation of arguments of \emph, \footnote, and similar commands. ### byte compiler 'byte-compile-dynamic' is now obsolete. This is because on the one hand it suffers from misbehavior in corner cases that have plagued it for years, and on the other hand experience indicates that it doesn't bring any measurable benefit. The 'g' keystroke in "*Compile-Log*" buffers has been bound to a new command that will recompile the file previously compiled with 'M-x byte-compile-file' and the like. ### compile.el The M-x compile command is a hidden gem of Emacs. It comes with a million different error highlighters and, despite its rather specific name and default (it defaults to make -k) it will run anything you type in the prompt, and comes with handy commands like M-x recompile to repeat the last command. If you don’t use it already, start using it now. In 'compilation-error-regexp-alist', 'line' (and 'end-line') can be functions. 'compilation-context-lines' can now take the value t; this is like nil, but instead of scrolling the current line to the top of the screen when there is no left fringe, it inserts a visible arrow before column zero. Despite my best efforts, I cannot parse the docstring or the description here, nor explain what it does. It seems useful, but it talks about the fringe, which is rather specific. If you know what it does, let me know in the comments. The new 'compilation-transform-file-match-alist' user option can be used to transform file name matches compilation output, and remove known false positives being recognized as warnings/errors. ### cl-lib.el 'cl-defstruct' has a new ':noinline' argument to prevent inlining its functions. 'cl-defstruct' slots accept a ':documentation' property. 'cl-values-list' will now signal an error if its argument isn't a list. ### doc-view.el New commands 'doc-view-presentation' and 'doc-view-fit-window-to-page'. Huh. The M-x doc-view-presentation command full screens your Emacs and shows you just the page. In other words, it’s a slide show mode. Neat. Now you don’t probably don’t need evince or whatever. The second command resizes the window to the size of the PDF/image. Useful for documentation. If the image/PDF quality looks horrible, you can configure the DPI with (setq doc-view-resolution 400). Added support for password-protected PDF files. A new user option 'doc-view-pdftotext-program-args' has been added to allow controlling how the conversion to text is done. I played around with this and it does seem to work but the UX is a bit poor. You have to type C-s – like isearch – and then it stores the search results and subsequent calls to C-s then goes forward and C-r for backward. However, to do another search you must type C-u C-s, which is not super intuitive. In the PDFs I tested it did not highlight the search terms, so keep that in mind. The prefix key 's' was changed to 'c' for slicing commands to avoid conflicts with 'image-mode' key 's'. The new key 'c' still has good mnemonics of "cut", "clip", "crop". ### Ido New user option 'ido-big-directories' to mark directories whose names match certain regular expressions as big. Ido won't attempt to list the contents of such directories when completing file names. ### Minibuffer New user option 'minibuffer-beginning-of-buffer-movement'. This option allows control of how the 'M-<' command works in the minibuffer. If non-nil, point will move to the end of the prompt (if point is after the end of the prompt). The default is nil, which preserves the original behavior of 'M-<' moving to the beginning of the prompt. When the minibuffer is active, echo-area messages are displayed at the end of the minibuffer instead of hiding the minibuffer by the echo area display. The new user option 'minibuffer-message-clear-timeout' controls how messages displayed in this situation are removed from the minibuffer. To revert to previous behavior, where echo-area messages temporarily overwrote the minibuffer contents until the user typed something, set 'set-message-function' and 'clear-message-function' to nil. This is perhaps one of the first new features users of Emacs 27 will encounter. Echo area messages, if the minibuffer is active, are now shown alongside the minibuffer contents. I quite like the feature, though it’s taken a bit of getting used to. Minibuffer now uses 'minibuffer-message' to display error messages at the end of the active minibuffer. To disable this, remove 'minibuffer-error-initialize' from 'minibuffer-setup-hook'. 'y-or-n-p' now uses the minibuffer to read 'y' or 'n' answer. Some commands that previously used 'read-char-choice' now read a character using the minibuffer by 'read-char-from-minibuffer'. ### map.el Now also understands plists. Now defined via generic functions that can be extended via 'cl-defmethod'. Deprecate the 'map-put' macro in favor of a new 'map-put!' function. 'map-contains-key' now returns a boolean rather than the key. Deprecate the 'testfn' args of 'map-elt' and 'map-contains-key'. New generic function 'map-insert'. The 'type' arg can be a list '(hash-table :key1 VAL1 :key2 VAL2 ...)'. ### seq.el New convenience functions 'seq-first' and 'seq-rest' give easy access to respectively the first and all but the first elements of sequences. The new predicate function 'seq-contains-p' should be used instead of the now obsolete 'seq-contains'. Follow mode In the current follow group of windows, "ghost" cursors are no longer displayed in the non-selected follow windows. To get the old behavior back, customize 'follow-hide-ghost-cursors' to nil. New variable 'warning-fill-column' for 'display-warning'. ### Windmove 'windmove-create-window' when non-nil makes a new window. This happens upon moving off the edge of the frame. Windmove supports directional window display and selection. The new command 'windmove-display-default-keybindings' binds default keys with provided modifiers (by default, Shift-Meta) to the commands that display the next buffer in the window at the specified direction. This is like 'windmove-default-keybindings' that binds keys to commands that select the window in the specified direction, but additionally it displays the buffer from the next command in that window. For example, 'S-M-right C-h i' displays the "*Info*" buffer in the right window, creating the window if necessary. A special key can be customized to display the buffer in the same window, for example, 'S-M-0 C-h e' displays the "*Messages*" buffer in the same window. 'S-M-t C-h r' displays the Emacs manual in a new tab. Windmove also supports directional window deletion. The new command 'windmove-delete-default-keybindings' binds default keys with provided prefix (by default, 'C-x') and modifiers (by default, 'Shift') to the commands that delete the window in the specified direction. For example, 'C-x S-down' deletes the window below. With a prefix arg 'C-u', also kills the buffer in that window. With 'M-0', deletes the selected window and selects the window that was in the specified direction. New command 'windmove-swap-states-in-direction' binds default keys to the commands that swap the states of the selected window with the window in the specified direction. Windmove code no longer used is now obsolete. That includes the user option 'windmove-window-distance-delta' and the functions 'windmove-coord-add', 'windmove-constrain-to-range', 'windmove-constrain-around-range', 'windmove-frame-edges', 'windmove-constrain-loc-for-movement', 'windmove-wrap-loc-for-movement', 'windmove-reference-loc' and 'windmove-other-window-loc'. A great many useful additions to windmove. Windmove’s basic premise is that it lets you switch to other windows by moving in cardinal directions. It’s a very useful package, though the default bindings will have you reach for the arrow keys to use it. Octave mode The mode is automatically enabled in files that start with the 'function' keyword. ### project.el New commands 'project-search' and 'project-query-replace-regexp'. New user option 'project-read-file-name-function'. Good. But about 20 years too late. Emacs never had proper project support (directory-local files are not a project management solution!) and despite project.el representing the second project management tool that ships with Emacs by default – the other being CEDET’s EDE – they are not widely used. Emacs had a glorious opportunity to solve this very common problem before people, and packages, standardized on third-party solutions years ago. It’s a shame. If everyone used a skeleton framework for project management, that would leave enough scope for individual packages to overwrite or introduce additional features or workflows, in much the same way completion frameworks tries to. Truly, a missed opportunity. ### Etags 'next-file' is now an obsolete alias of 'tags-next-file'. 'tags-loop-revert-buffers' is an obsolete alias of 'fileloop-revert-buffers'. The 'tags-loop-continue' function along with the 'tags-loop-operate' and 'tags-loop-scan' variables are now obsolete; use the new 'fileloop-initialize' and 'fileloop-continue' functions instead. etags is now able to read Zstandard-compressed files. ### bibtex New commands 'bibtex-next-entry' and 'bibtex-previous-entry'. In 'bibtex-mode-map', 'forward-paragraph' and 'backward-paragraph' are remapped to these, respectively. ### Dired New command 'dired-create-empty-file'. New command 'dired-number-of-marked-files'. It is by default bound to '* N'. The marking commands now report how many files were marked by the command itself, not how many files are marked in total. The new user option 'dired-create-destination-dirs' controls whether 'dired-do-copy' and 'dired-rename-file' should create non-existent directories in the destination. 'dired-dwim-target' can be customized to prefer either the next window, or one of the most recently visited windows with a Dired buffer. The latter option, though useful, seems like a recipe for mistakes. I’ve moved stuff around accidentally to the wrong dired buffer simply because I forgot I already had one open. When the new user option 'dired-vc-rename-file' is non-nil, Dired performs file renaming using underlying version control system. Zstandard compression is now supported for 'dired-do-compress' and 'dired-do-compress-to'. On systems that support suid/guid files, Dired now fontifies the permissions of such files with a special face 'dired-set-id'. A new face, 'dired-special', is used to highlight sockets, named pipes, block devices and character devices. ### Find-Dired A fantastically underrated part of Emacs. See Working with multiple files in Dired why it’s a game changer. New user option 'find-dired-refine-function'. The default value is 'find-dired-sort-by-filename'. New sorting options for the user option 'find-ls-option'. ### Change Logs and VC Despite the proliferation of Magit and Git everywhere, the built-in VC is still a great package that seamlessly works across almost all version control systems. Indeed, C-x v g, C-x v v and C-x v = are far more useful and quicker to use than their magit equivalents, I think. New user option 'vc-tor'. When non-nil, this user option causes the VC commands to communicate with the repository via Tor's proxy, using the 'torsocks' wrapper script. The default is nil. New command 'log-edit-generate-changelog-from-diff', bound to 'C-c C-w'. This generates ChangeLog entries from the VC fileset diff. Recording ChangeLog entries doesn't require an actual file. If a ChangeLog file doesn't exist, and if the new user option 'add-log-dont-create-changelog-file' is non-nil (which is the default), commands such as 'C-x 4 a' will add log entries to a suitable named temporary buffer. (An existing ChangeLog file will still be used if it exists.) Set the user option to nil to get the previous behavior of always creating a buffer that visits a ChangeLog file. The new 'd' command ('vc-dir-clean-files') in 'vc-dir-mode' buffers will delete the marked files (or if no files are marked, the file under point). This command does not notify the VC backend, and is mostly useful for unregistered files. 'vc-dir-ignore' now takes a prefix argument to ignore all marked files. New user option 'vc-git-grep-template'. This new user option allows customizing the default arguments passed to 'git-grep' when 'vc-git-grep' is used. Command 'vc-git-stash' now respects marks in the "*vc-dir*" buffer. When some files are marked, only those are stashed. When no files are marked, all modified files are stashed, as before. 'vc-dir' now shows a button allowing you to hide the stash list. Controlled by user option 'vc-git-show-stash'. Default t means show the entire list as before. An integer value limits the list length (but still allows you to show the entire list via the button). 'vc-git-stash' is now bound to 'C' in the stash headers. Some stash keybindings are now available in the stash button. 'vc-git-stash' and 'vc-git-stash-snapshot' can now be run using 'C' and 'S' respectively, including when there are no stashes. The new hook 'vc-retrieve-tag-hook' runs after retrieving a tag. 'vc-hg' now invokes 'smerge-mode' when visiting files. Code that attempted to invoke 'smerge-mode' when visiting an Hg file with conflicts existed in earlier versions of Emacs, but incorrectly never detected a conflict due to invalid assumptions about cached values. The Hg (Mercurial) back-end now supports 'vc-region-history'. The 'C-x v h' command now works in buffers that visit files controlled by Hg. The Hg (Mercurial) back-end now prompts for revision to merge when you invoke 'C-x v m' ('vc-merge'). The Hg (Mercurial) back-end now uses tags, branches and bookmarks instead of revision numbers as completion candidates when it prompts for a revision. New user option 'vc-hg-revert-switches'. It specifies switches to pass to Hg's 'revert' command. 'C-u C-x v D' ('vc-root-version-diff') prompts for two revisions and compares their entire trees. 'C-x v M D' ('vc-diff-mergebase') and 'C-x v M L' ('vc-log-mergebase') print diffs and logs between the merge base (common ancestor) of two given revisions. New command 'vc-log-search' asks for a pattern, searches it in the revision log, and displays matched log entries in the log buffer. For example, 'M-x vc-log-search RET bug#36644 RET' displays all entries whose log messages match the bug number. With a prefix argument asks for a command, so for example, 'C-u M-x vc-log-search RET git log -1 f302475 RET' will display just one log entry found by its revision number. That’s a very useful command. It is now possible to display a specific revision given by its ID. If you invoke 'C-x v L' ('vc-print-root-log') with a numeric argument of 1, as in 'C-1 C-x v L' or 'C-u 1 C-x v L', it asks for a revision ID, and shows its log entry together with the diffs introduced by the revision's commit. (For some less capable VCSes, only the log entry is shown.) New user option 'vc-find-revision-no-save'. With non-nil, 'vc-find-revision' doesn't write the created buffer to file. 'C-x v =' can now mimic Magit's diff format. Set the new user option 'diff-font-lock-prettify' to t for that, see below under "Diff mode". That’s another one of those set and forget variables. The 'diff' function arguments OLD and NEW may each be a buffer rather than a file, in non-interactive calls. This change was made in Emacs 24.1, but wasn't documented until now. New command 'diff-buffers' interactively diffs two buffers. Of course there is already M-x ediff-buffers, but if you want the output of the diff command line tool with highlighting, then this will do that just fine. ### Diff mode Hunks are now automatically refined by font-lock. To disable refinement, set the new user option 'diff-refine' to nil. To get back the old behavior where hunks are refined as you navigate through a diff, set 'diff-refine' to the symbol 'navigate'. 'diff-auto-refine-mode' is deprecated in favor of 'diff-refine'. It is no longer enabled by default and binding it no longer has any effect. Better syntax highlighting of Diff hunks. Fragments of source in Diff hunks are now by default highlighted according to the appropriate major mode. Customize the new user option 'diff-font-lock-syntax' to nil to disable this. You now get font locking – syntax highlighting – in diff hunks. If you don’t like it, or find it distracting, customize diff-refine. Prefix arg of 'diff-goto-source' means jump to the old revision of the file under version control if point is on an old changed line, or to the new revision of the file otherwise. ### Texinfo New function for inserting '@pxref', '@xref', or '@ref' commands. The function 'texinfo-insert-dwim-@ref', bound to 'C-c C-c r' by default, inserts one of three types of references based on the text surrounding point, namely '@pxref' near a parenthesis, '@xref' at the start of a sentence or at '(point-min)', else '@ref'. ### Browse-url The function 'browse-url-emacs' can now visit a URL in selected window. It now treats the optional 2nd argument to mean that the URL should be shown in the currently selected window. A new function, 'browse-url-add-buttons' can be used to add clickable links to most ordinary special-mode buffers that display text that have URLs embedded. 'browse-url-button-regexp' controls what's considered a button. New user option 'browse-url-secondary-browser-function'. It can be set to a function that invokes an alternative browser. ### Comint 'send-invisible' is now an obsolete alias for 'comint-send-invisible'. Also, 'shell-strip-ctrl-m' is declared obsolete. 'C-c .' ('comint-insert-previous-argument') no longer interprets '&'. This feature caused problems when '&&' was present in the previous command. Since this command emulates 'M-.' in Bash and zsh, neither of which treats '&' specially, the feature was removed for compatibility with these shells. This is a very useful command to know about. More information in my article Shell & Comint Secrets: History Commands. 'comint-insert-previous-argument' can now count arguments from the end. By default, invoking 'C-c .' with a numeric argument N would copy the Nth argument, counting from the first one. But if the new user option 'comint-insert-previous-argument-from-end' is non-nil, it will copy the Nth argument counting from the last one. Thus 'C-c .' can now better emulate 'M-.' in both Bash and zsh, since the former counts from the beginning of the arguments, while the latter counts from the end. 'comint-run' can now accept a list of switches to pass to the program. 'C-u M-x comint-run' will prompt for the switches interactively. That is a very useful addition for ad hoc program execution. Abnormal hook 'comint-password-function' has been added. This hook permits a derived mode to supply a password for the underlying command interpreter without prompting the user. For example, in 'sql-mode', the password for connecting to the database may be stored in the connection wallet and may be passed on the command line to start the SQL interpreter. This is a potential security flaw that could expose user's database passwords on the command line through the use of a process list (Bug#8427). With this hook, it is possible to not pass the password on the command line and wait for the program to prompt for the password. When it does so, the password can be supplied to the SQL interpreter without involving the user just as if it had been supplied on the command line. That’s a great change if you use Emacs’s authsource and Emacs to keep and manage encrypted secrets. (See Keeping Secrets in Emacs with GnuPG and Auth Sources.) ### SQL SQL Indent Minor Mode SQL Mode now supports the ELPA 'sql-indent' package for assisting sophisticated SQL indenting rules. Note, however, that SQL is not like other programming languages like C, Java, or Python where code is sparse and rules for formatting are fairly well established. Instead SQL is more like COBOL (from which it came) and code tends to be very dense and line ending decisions driven by syntax and line length considerations to make readable code. Experienced SQL developers may prefer to rely upon existing Emacs facilities for formatting code but the 'sql-indent' package provides facilities to aid more casual SQL developers layout queries and complex expressions. 'sql-use-indent-support' (default t) enables SQL indention support. The 'sql-indent' package from ELPA must be installed to get the indentation support in 'sql-mode' and 'sql-interactive-mode'. 'sql-mode-hook' and 'sql-interactive-mode-hook' changed. Both hook variables have had 'sql-indent-enable' added to their default values. If you have existing customizations to these variables, you should make sure that the new default entry is included. Ah yes, the intractable problem of SQL indentation. I have not yet tried out the sql-indent package but I wish them good luck. Connection Wallet Database passwords can now by stored in NETRC or JSON data files that may optionally be encrypted. When establishing an interactive session with the database via 'sql-connect' or a product specific function, like 'sql-mysql' or 'sql-postgres', the password wallet will be searched for the password. The 'sql-product', 'sql-server', 'sql-database', and the 'sql-username' will be used to identify the appropriate authorization. This eliminates the discouraged practice of embedding database passwords in your Emacs initialization. See the 'auth-source' module for complete documentation on the file formats. By default, the wallet file is expected to be in the 'user-emacs-directory', named "sql-wallet" or ".sql-wallet", with ".json" (JSON) or no (NETRC) suffix. Both file formats can optionally be encrypted with GPG by adding an additional ".gpg" suffix. See Keeping Secrets in Emacs with GnuPG and Auth Sources for more information on how this works in general. I will cover auth-source in greater detail below. ### Term 'term-read-noecho' is now obsolete, use 'read-passwd' instead. 'serial-term' now takes an optional parameter to leave the emulator in line mode. ### Flymake If you gave up on flymake years ago, know that it has received a large number of improvements to make it more competitive. The variable 'flymake-diagnostic-types-alist' is obsolete. You should instead set properties on known diagnostic symbols, like ':error' and ':warning', as demonstrated in the Flymake manual. New user option 'flymake-start-on-save-buffer'. Control whether Flymake starts checking the buffer on save. Flymake and backend functions may exchange hints about buffer changes. This enables more efficient backends. See the docstring of 'flymake-diagnostic-functions' or the Flymake manual for details. 'flymake-start-syntax-check-on-newline' is now obsolete, use 'post-self-insert-hook' to check on newline. ### Ruby The Rubocop Flymake diagnostic function will only run Lint cops if it can't find the config file. Rubocop is called with 'bundle exec' if Gemfile mentions it. New command 'ruby-find-library-file' bound to 'C-c C-f'. ### Package Warn if "footer line" is missing, but still install package. package.el used to refuse to install a package without the so-called "footer line", which appears at the very end of the file: ;;; FILENAME ends here package.el will now install packages without this line, but it will issue a warning. To avoid this warning, packages should keep the "footer line". Note that versions of Emacs older than 27.1 will not only refuse to install packages without such a line -- they will be unable to parse package data. It is therefore recommended to keep this line. Change of 'package-check-signature' for packages with multiple sigs. In previous Emacsen, t checked that all signatures are valid. Now t only checks that at least one signature is valid and the new 'all' value needs to be used if you want to enforce that all signatures are valid. This only affects packages with multiple signatures. The meaning of 'allow-unsigned' in 'package-check-signature' has changed slightly: If a usable OpenPGP configuration can't be found (for instance, if gpg isn't installed), it now has the same meaning as nil. New function 'package-get-version' lets packages query their own version. Example use in auctex.el: '(defconst auctex-version (package-get-version))' New 'package-quickstart' feature. When 'package-quickstart' is non-nil, package.el precomputes a big autoloads file so that activation of packages can be done much faster, which can speed up your startup significantly. It also causes user options like 'package-user-dir' and 'package-load-list' to be consulted when 'package-quickstart-refresh' is run rather than at startup so you don't need to set them in your early init file. Worth a try if you restart Emacs frequently. New function 'package-activate-all'. New functions for filtering packages list. A new function has been added which allows users to filter the packages list by name: 'package-menu-filter-by-name'. By default, it is bound to '/ n'. Additionally, the function 'package-menu-filter-by-keyword' has been renamed from 'package-menu-filter'. Its keybinding has also been changed to '/ k' (from 'f'). To clear any of the two filters, the user can now call the 'package-menu-clear-filter' function, bound to '/ /' by default. Imenu support has been added to 'package-menu-mode'. The package list can now be sorted by version or description. In Package Menu, 'g' now updates package data from archives. Previously, 'g' invoked 'tabulated-list-revert' which did not update the cached archive data. It is now bound to 'revert-buffer', which will update the data. 'package-menu-refresh' is an obsolete alias for 'revert-buffer'. ### Info Clicking on the left/right arrow icon in the Info tool-bar while holding down the Ctrl key pops up a menu of previously visited Info nodes where you can select a node to go back (like in browsers). Info can now follow 'file://' protocol URLs. The 'file://' URLs in Info documents can now be followed by passing them to the 'browse-url' function, like the other protocols: 'ftp', 'http', and 'https'. This allows having references to local HTML files, for example. Display of man pages now limits the width for formatting pages. The new user option 'Man-width-max' (80 by default) limits the number of columns passed to the 'man' program for formatting man pages. This is to enhance readability when man pages are displayed in very wide windows (which are customary with today's large displays). Useful. There’s also the Emacs-only M-x woman (it stands for with/out man, obviously…) ### Xref New command 'xref-find-definitions-at-mouse'. This command finds definitions of the identifier at the place of a mouse click event, and is intended to be bound to a mouse event. Changing 'xref-marker-ring-length' works after xref.el is loaded. Previously, setting 'xref-marker-ring-length' would only take effect if set before xref.el was loaded. 'xref-find-definitions' now sets the mark at the buffer position where it was invoked. New xref faces 'xref-file-header', 'xref-line-number', 'xref-match'. New user option 'xref-show-definitions-function'. It encapsulates the logic pertinent to showing the result of 'xref-find-definitions'. The user can change it to customize its behavior and the display of results. I guess the intent is that you can give the variable your own function and xref will use that to render the results. Search results show the buffer even for one hit. The search-type Xref commands (e.g. 'xref-find-references' or 'project-find-regexp') now show the results buffer even when there is only one hit. This can be altered by changing 'xref-show-xrefs-function'. Xref buffers support refreshing the search results. A new command 'xref-revert-buffer' is bound to 'g'. Much awaited, that. Imenu support has been added to 'xref--xref-buffer-mode'. New generic method 'xref-backend-identifier-completion-ignore-case'. Using it, the etags backend now honors 'tags-case-fold-search' during identifier completion. ### Checkdoc Checkdoc can now optionally spell-check doc strings. Invoking 'checkdoc-buffer' with a non-nil TAKE-NOTES argument (interactively, with a prefix arg) will now spell-check the doc strings and report all the spelling mistakes. ### Icomplete New minor mode Fido mode. This mode is based on Icomplete, and its name stands for "Fake Ido". The point of this mode is to be an 'ido-mode' workalike, providing most of the functionality present in 'ido-mode' that is not in Icomplete, which is much more compatible with all of Emacs's completion facilities. Yet more completion frameworks. This is an honest attempt to make icomplete more like ido. It seems to do a reasonable job of it. It chokes on /ssh: completion though. Not sure if it’s entirely ready for prime time yet, but an upgraded IDO mode is welcome. ### Ecomplete The Ecomplete sorting has changed to a decay-based algorithm. This can be controlled by the new 'ecomplete-sort-predicate' user option. The 'ecomplete-database-file' file is now placed in "~/.emacs.d/ecompleterc" by default. Of course it will still find it if you have it in "~/.ecompleterc". ### Gnus 'mm-uu-diff-groups-regexp' now defaults to matching all groups, which means that "git am" diffs are recognized everywhere. Two new Gnus summary mode navigation commands have been added, bound to the '[' and ']' keys: 'gnus-summary-prev-unseen-article' and 'gnus-summary-next-unseen-article'. These take you (respectively) to the previous unseen or next unseen article. (These are the ones that are marked with "." in the summary mode lines.) The Gnus user variable 'nnimap-expunge' supports three new values: 'never' for never expunging messages, 'immediately' for immediately expunging deleted messages, and 'on-exit' to expunge deleted articles when exiting the group's summary buffer. Setting 'nnimap-expunge' to nil or t is still supported but not recommended, since it may result in Gnus expunging all messages that have been flagged as deleted by any IMAP client (rather than just those that have been deleted by Gnus). New user option 'gnus-use-atomic-windows' makes Gnus window layouts atomic. See the "(elisp) Atomic Windows" node of the Elisp manual for details. There's a new value for 'gnus-article-date-headers', 'combined-local-lapsed', which will show both the time (in the local timezone) and the lapsed time. Gnus now maps imaps to 993 only on old MS-Windows versions. The nnimap backend used to do this unconditionally to work around problems on old versions of MS-Windows. This is now done only for Windows XP and older. The nnimap backend now has support for IMAP namespaces. This feature can be enabled by setting the new 'nnimap-use-namespaces' server variable to non-nil. A prefix argument to 'gnus-summary-limit-to-score' will limit in reverse. Limit to articles with score "at or below" the SCORE argument rather than "at or above". The function 'gnus-score-find-favorite-words' has been renamed from 'gnus-score-find-favourite-words'. Gmane has been removed as an nnir backend, since Gmane no longer has a search engine. Splitting mail on common mailing list headers has been added. See the concept index in the Gnus manual for the 'match-list' entry. nil is no longer an allowed value for 'mm-text-html-renderer'. The default value of 'mm-inline-large-images' has changed from nil to 'resize', which means that large images will be resized instead of displayed with an external program by default. A new Gnus summary mode command, 'S A' ('gnus-summary-attach-article') can be used to attach the current article(s) to a pre-existing Message buffer, or create a new Message buffer with the article(s) attached. A new Gnus summary mode command, 'w' ('gnus-summary-browse-url') scans the article buffer for URLs, and offers them to the user to open with 'browse-url'. New user option 'nnir-notmuch-filter-group-names-function'. This option controls whether and how to use Gnus search groups as 'path:' search terms to 'notmuch'. The buttons in the Gnus article buffer were formerly widgets (i.e., buttons from widget.el). This has now changed, and they are now buttons (from button.el), and commands like 'TAB' now search for buttons instead of widgets. There should be no user-visible changes, but out-of-tree code that relied on widgets being present might now fail. ### erc New hook 'erc-insert-done-hook'. This hook is called after strings have been inserted into the buffer, and is free to alter point and window configurations, as it's not called from inside a 'save-excursion', as opposed to 'erc-insert-post-hook'. 'erc-button-google-url' has been renamed to 'erc-button-search-url' and its value has been changed to Duck Duck Go. 'erc-send-pre-hook' and 'erc-send-this' have been obsoleted. The user option to use instead to alter text to be sent is now 'erc-pre-send-functions'. Improve matching/highlighting of nicknames. Open and close parenthesis and apostrophe are not considered valid nick characters anymore, matching the given grammar in RFC 2812 section 2.3.1. This enables correct matching and highlighting of nicks when they are surrounded by parentheses, like "(nick)", and when adjacent to an apostrophe, like "nick's". Set 'erc-button-url-regexp' to 'browse-url-button-regexp' which better handles surrounding pair of parentheses. New function 'erc-switch-to-buffer-other-window' which is like 'erc-switch-to-buffer', but opens the buffer in another window. New function 'erc-track-switch-buffer-other-window' which is like 'erc-track-switch-buffer', but opens the buffer in another window. ### EUDC XEmacs support has been removed. ### eww/shr EWW is the Emacs Web Wowser. The new user option 'shr-cookie-policy' can be used to control when to use cookies when fetching embedded images. The default is to use them when the images are from the same domain as the main HTML document. The 'eww' command can now create a new EWW buffer. Invoking the command with a prefix argument will cause it to create a new EWW buffer for the URL instead of reusing the default one. Clicking with the Ctrl key or 'C-u RET' on a link opens a new tab when tab-bar-mode is enabled. The 'd' ('eww-download') command now falls back to current page's URL. If this command is invoked with no URL at point, it now downloads the current page instead of signaling an error. When opening external links in eww/shr (typically with the 'C-u RET' keystroke on a link), the link will be flashed with the new 'shr-selected-link' face to give the user feedback that the command has been executed. New user option 'shr-discard-aria-hidden'. If set, shr will not render tags with attribute 'aria-hidden="true"'. This attribute is meant to tell screen readers to ignore a tag. 'shr-external-browser' has been made into an obsolete alias of 'browse-url-secondary-browser-function'. 'shr-tag-ol' now respects the ordered list 'start' attribute. The following tags are now handled: '<code>', '<abbr>', and '<acronym>'. ### Htmlfontify The functions 'hfy-color', 'hfy-color-vals' and 'hfy-fallback-color-values' and the variables 'hfy-fallback-color-map' and 'hfy-rgb-txt-color-map' have been renamed from names that used 'colour' instead of 'color'. Enriched mode supports the 'charset' text property. You can add or modify the 'charset' text properties of text using the 'Edit->Text Properties->Special Properties' menu, or by invoking the 'facemenu-set-charset' command. Documents in Enriched mode will be saved with the charset properties, and those properties will be restored when the file is visited. ### Smtpmail Authentication mechanisms can be added via external packages, by defining new 'cl-defmethod' of 'smtpmail-try-auth-method'. To always force smtpmail to send credentials over on the first attempt when communicating with the SMTP server(s), the 'smtpmail-servers-requiring-authorization' user option can be used. smtpmail will now try resending mail when getting a transient "4xx" error message from the SMTP server. The new 'smtpmail-retries' user option says how many times to retry. ### Footnote mode Support Hebrew-style footnotes. Footnote text lines are now aligned. Can be controlled via the new user option 'footnote-align-to-fn-text'. ### CSS mode A new command 'css-cycle-color-format' for cycling between color formats (e.g. "black" => "#000000" => "rgb(0, 0, 0)") has been added, bound to 'C-c C-f'. That is very handy. CSS mode, SCSS mode, and Less CSS mode now have support for Imenu. ### SGML mode 'sgml-quote' now handles double quotes and apostrophes when escaping text and in addition all numeric entities when unescaping text. ### Python mode Python mode supports three different font lock decoration levels. The maximum level is used by default; customize 'font-lock-maximum-decoration' to tone down the decoration. New user option 'python-pdbtrack-kill-buffers'. If non-nil, the default, buffers opened during pdbtracking session are killed when pdbtracking session is finished. New function 'python-shell-send-statement. It sends the statement delimited by 'python-nav-beginning-of-statement' and 'python-nav-end-of-statement' to the inferior Python process. ### Help Descriptions of variables and functions give an estimated first release where the variable or function appeared in Emacs. Output format of 'C-h l' ('view-lossage') has changed. For convenience, 'view-lossage' now displays the last keystrokes and commands in the same format as the edit buffer of 'edit-last-kbd-macro'. This makes it possible to copy the lines from the buffer generated by 'view-lossage' to the "*Edit Macro*" buffer created by 'edit-last-kbd-macro', and to save the macro by 'C-c C-c'. This is a very welcome change and a surprisingly useful one. The list of help commands produced by 'C-h C-h' ('help-for-help') can now be searched via 'C-s'. ### Ibuffer New filter 'ibuffer-filter-by-process'; bound to '/ E'. All mode filters can now accept a list of symbols. This means you can now easily filter several major modes, as well as a single mode. ### Search and Replace Isearch supports a prefix argument for 'C-s' ('isearch-repeat-forward') and 'C-r' ('isearch-repeat-backward'). With a prefix argument, these commands repeat the search for the specified occurrence of the search string. A negative argument repeats the search in the opposite direction. This makes possible also to use a prefix argument for 'M-s .' ('isearch-forward-symbol-at-point') to find the next Nth symbol. Also a prefix argument is supported for 'isearch-yank-until-char', 'isearch-yank-word-or-char', 'isearch-yank-symbol-or-char'. That could be a useful feature in macros. To go to the first/last occurrence of the current search string is possible now with new commands 'isearch-beginning-of-buffer' and 'isearch-end-of-buffer' bound to 'M-s M-<' and 'M-s M->' in Isearch. With a numeric argument, they go to the Nth absolute occurrence counting from the beginning/end of the buffer. This complements 'C-s'/'C-r' that searches for the next Nth relative occurrence with a numeric argument. Useful, but a finger-twisting combo. C-s M-s M-<. 'isearch-lazy-count' shows the current match number and total number of matches in the Isearch prompt. User options 'lazy-count-prefix-format' and 'lazy-count-suffix-format' define the format of the current and the total number of matches in the prompt's prefix and suffix, respectively. That’s another set and forget, I think. I do wonder how it impacts performance in large buffers. 'lazy-highlight-buffer' highlights matches in the full buffer. It is useful in combination with 'lazy-highlight-cleanup' customized to nil to leave matches highlighted in the whole buffer after exiting isearch. Also when 'lazy-highlight-buffer' prepares highlighting in the buffer, navigation through the matches without flickering is more smooth. 'lazy-highlight-buffer-max-at-a-time' controls the number of matches to highlight in one iteration while processing the full buffer. Isearch already highlights though I interpret this optional change to mostly be about performance optimization plus the ability to set lazy-highlight-cleanup to nil to keep the highlighted results after isearch exit. I tried it out and I would find that setting annoying; but you may not. New isearch bindings. 'C-M-z' invokes new function 'isearch-yank-until-char', which yanks everything from point up to but not including the specified character into the search string. This is especially useful for keyboard macros. I could see this being useful and yet you have to enter isearch first to trigger it – that is far less useful. 'C-M-w' in isearch changed from 'isearch-del-char' to the new function 'isearch-yank-symbol-or-char'. 'isearch-del-char' is now bound to 'C-M-d'. 'M-s h l' invokes 'highlight-lines-matching-regexp' using the search string to highlight lines matching the search string. This is similar to the existing binding 'M-s h r' ('highlight-regexp') that highlights JUST the search string. Very useful for log files and whatnot. New user option 'isearch-yank-on-move' provides options t and 'shift' to extend the search string by yanking text that ends at the new position after moving point in the current buffer. 'shift' extends the search string by motion commands while holding down the shift key. More esoteric isearch workflows. 'isearch-allow-scroll' provides a new option 'unlimited' to allow scrolling any distance off screen. Isearch now remembers the regexp-based search mode for words/symbols and case-sensitivity together with search strings in the search ring. Isearch now has its own tool-bar and menu-bar menu. 'flush-lines' prints and returns the number of deleted matching lines. I use flush-lines a lot, and that is definitely helpful 'char-fold-to-regexp' now matches more variants of a base character. The table used to check for equivalence of characters is now built using the complete chain of unicode decompositions of a character, rather than stopping after one level, such that searching for e.g. "GREEK SMALL LETTER IOTA" will now also find "GREEK SMALL LETTER IOTA WITH OXIA". Char folding is a hidden gem of a feature. New char-folding options: 'char-fold-include' lets you add ad hoc foldings, 'char-fold-exclude' to remove foldings from default decomposition, and 'char-fold-symmetric' to search for any of an equivalence class of characters. For example, with a nil value of 'char-fold-symmetric' you can search for "e" to find "é", but not vice versa. With a non-nil value you can search for either, for example, you can search for "é" to find "e". ### Debugger The Lisp Debugger is now based on 'backtrace-mode'. Backtrace mode adds fontification and commands for changing the appearance of backtrace frames. See the node "(elisp) Backtraces" in the Elisp manual for documentation of the new mode and its commands. ### Edebug 'edebug-eval-last-sexp' and 'edebug-eval-print-last-sexp' interactively now take a zero prefix analogously to the non-Edebug counterparts. New faces 'edebug-enabled-breakpoint' and 'edebug-disabled-breakpoint'. When setting breakpoints in Edebug, an overlay with these faces are placed over the point in question, depending on whether they are enabled or not. Nifty; and they come with sensible defaults already. New command 'edebug-toggle-disable-breakpoint'. This command allows you to disable a breakpoint temporarily. This is mainly useful with breakpoints that are conditional and would take some time to recreate. New command 'edebug-unset-breakpoints'. To clear all breakpoints in the current form, the 'U' command in 'edebug-mode', or 'M-x edebug-unset-breakpoints' can be used. Re-instrumenting a function with Edebug will now try to preserve previously-set breakpoints. However, if the code has changed substantially, this may not be possible. That’s very useful. I look forward to how practical it is in real life. New command 'edebug-remove-instrumentation'. This command removes Edebug instrumentation from all functions that have been instrumented. The runtime behavior of Edebug's instrumentation can be changed using the new variables 'edebug-behavior-alist', 'edebug-after-instrumentation-function' and 'edebug-new-definition-function'. Edebug's behavior can be changed globally or for individual definitions. Edebug's backtrace buffer now uses 'backtrace-mode'. Backtrace mode adds fontification, links and commands for changing the appearance of backtrace frames. See the node "(elisp) Backtraces" in the Elisp manual for documentation of the new mode and its commands. The binding of 'd' in Edebug's keymap is now 'edebug-pop-to-backtrace' which replaces 'edebug-backtrace'. Consequently Edebug's backtrace windows now behave like those of the Lisp Debugger and of ERT, in that when they appear they will be the selected window. The new 'backtrace-goto-source' command, bound to 's', works in Edebug's backtraces on backtrace frames whose source code has been instrumented by Edebug. ### Enhanced xterm support New user option 'xterm-set-window-title' controls whether Emacs sets the XTerm window title. This feature is experimental and is disabled by default. ### Grep 'rgrep', 'lgrep' and 'zrgrep' now hide part of the command line that contains a list of ignored directories and files. Clicking on the button with ellipsis unhides it. The abbreviation can be disabled by the new user option 'grep-find-abbreviate'. The new command 'grep-find-toggle-abbreviation' toggles it interactively. A welcome change, as you’re otherwise stuck staring at a laundry list of directories for long-dead source control systems. 'grep-find-use-xargs' is now customizable with sorting options. It used to be much more of a hassle configuring grep, but Emacs has greatly improved the ease with which we can change these things. ### ERT ERT is the Emacs Lisp Regression Testing tool. New variable 'ert-quiet' allows making ERT output in batch mode less verbose by removing non-essential information. ERT's backtrace buffer now uses 'backtrace-mode'. Backtrace mode adds fontification and commands for changing the appearance of backtrace frames. See the node "(elisp) Backtraces" in the Elisp manual for documentation of the new mode and its commands. ### Gamegrid Gamegrid now determines its default glyph size based on display dimensions, instead of always using 16 pixels. As a result, Tetris, Snake and Pong are better playable on HiDPI displays. 'gamegrid-add-score' can now sort scores from lower to higher. This is useful for games where lower scores are better, like time-based games. ### Filecache Completing file names in the minibuffer via 'C-TAB' now uses the styles as configured by the user option 'completion-styles'. New macros 'thunk-let' and 'thunk-let*'. These macros are analogue to 'let' and 'let*', but create bindings that are evaluated lazily. ### next-error New user option 'next-error-find-buffer-function'. The value should be a function that determines how to find the next buffer to be used by 'next-error' and 'previous-error'. The default is to use the last buffer that navigated to the current error. New command 'next-error-select-buffer'. It can be used to set any buffer as the next one to be used by 'next-error' and 'previous-error'. Going back to what I said earlier about showing the error loci: this feature should prove useful for package authors who use multiple error reporting mechanisms – like lsp-mode and eglot. ### nxml-mode The default value of 'nxml-sexp-element-flag' is now t. This means that pressing 'C-M-SPACE' now selects the entire tree by default, and not just the opening element. ### Eshell TAB completion uses the standard 'completion-at-point' rather than 'pcomplete'. Its UI is slightly different but can be customized to behave similarly, e.g. Pcomplete's default cycling can be obtained with '(setq completion-cycle-threshold 5)'. pcomplete is actually a pretty cool completion tool (see PComplete: Context-Sensitive Completion in Emacs) that works with a wide array of “shell-style” modes including Shell-mode. Expansion of history event designators is disabled by default. To restore the old behavior, use (add-hook 'eshell-expand-input-functions #'eshell-expand-history-references) The function 'eshell-uniquify-list' has been renamed from 'eshell-uniqify-list'. The function 'eshell/kill' is now able to handle signal switches. Previously 'eshell/kill' would fail if provided a kill signal to send to the process. It now accepts signals specified either by name or by its number. Emacs now follows symlinks in history-related files. The files specified by 'eshell-history-file-name' and 'eshell-last-dir-ring-file-name' can include symlinks; these are now followed when Emacs writes the relevant history variables to the disk. ### Shell M-x shell is awesome. It’s not a terminal emulator, but Emacs can do 80% of most curses-based terminal emulation without it anyway, so it’s usually not a problem. Program name completion inside remote shells works now as expected. The user option 'shell-file-name' can be set now as connection-local variable for remote shells. It still defaults to "/bin/sh". ### Single shell commands New values of 'shell-command-dont-erase-buffer'. This user option can now have the value 'erase' to force to erase the output buffer before execution of the command, even if the output goes to the current buffer. Additional values 'beg-last-out', 'end-last-out', and 'save-point' control where to put point in the output buffer after inserting the 'shell-command' output. The new functions 'shell-command-save-pos-or-erase' and 'shell-command-set-point-after-cmd' control how point is handled between two consecutive shell commands in the same output buffer. 'async-shell-command-width' defines the number of display columns available for output of asynchronous shell commands. Prompt for shell commands can now show the current directory. Customize the new user option 'shell-command-prompt-show-cwd' to enable it. ### Pcomplete The 'pcomplete' command is now obsolete. The Pcomplete functionality can be obtained via 'completion-at-point' instead, by adding 'pcomplete-completions-at-point' to 'completion-at-point-functions'. Note this is only for pcomplete the interactive command, as pcomplete the library is still in use. The function 'pcomplete-uniquify-list' has been renamed from 'pcomplete-uniqify-list'. 'pcomplete/make' now completes on targets in included files, recursively. To recover the previous behavior, set new user option 'pcmpl-gnu-makefile-includes' to nil. ### Auth-source Auth-source is a great feature in Emacs. See Keeping Secrets in Emacs with GnuPG and Auth Sources. The Secret Service backend supports the ':create' key now. ".authinfo" and ".netrc" files now use a new mode: 'authinfo-mode'. This is just like 'fundamental-mode', except that it hides passwords under a "****" display property. When the cursor moves to this text, the real password is revealed (via 'reveal-mode'). The new 'authinfo-hidden' user option can be used to control what to hide. This is very nice. Previously, you could never open your auth source file in front of anybody without the risk of them seeing your passwords. That “worry” is now gone as passwords are now hidden with an overlay. ### Tramp New connection method "nextcloud", which allows accessing OwnCloud or NextCloud hosted files and directories. New connection method "rclone", which allows accessing system storages via the 'rclone' program. This feature is experimental. If you’ve never used rclone, think of it as a supercharged file syncing mechanism that works with just about all possible cloud-style file storage services. Having TRAMP work with that is a useful thing indeed. You can now edit files in Amazon S3 buckets (via rclone) New connection method "sudoedit", which allows editing local files with different user credentials. Contrary to the "sudo" method, no session is run permanently in the background. This is for security reasons. I tried it out and seemed to work fine as a drop-in replacement for sudo. Connection methods "obex" and "synce" have been removed, because they are obsoleted in GVFS. Validated passwords are saved by auth-source backends which support this. During user and host name completion in the minibuffer, results from auth-source search are taken into account. This can be disabled by setting the user option 'tramp-completion-use-auth-sources' to nil. The user option 'tramp-ignored-file-name-regexp' allows disabling Tramp for some look-alike remote file names. For some connection methods, like "su" or "sudo", the host name in multi-hop file names must match the previous hop. Default host names are adjusted to the host name from the previous hop. This one caught out a lot of people using multi-hop, including me. I’m keen to see if it’ll save me from my own mistakes! A timeout has been added for the connection methods "sudo" and "doas". The underlying session is disabled when the timeout expires. This is for security reasons. For some connection methods, like "sshx" or "plink", it is possible to configure the remote login shell. This avoids problems with remote hosts, where "/bin/sh" is a link to a shell which cooperates badly with Tramp. New commands 'tramp-rename-files' and 'tramp-rename-these-files'. They allow saving remote files somewhere else when the corresponding host is not reachable anymore. ### Rcirc New user option 'rcirc-url-max-length'. Setting this option to an integer causes URLs displayed in Rcirc buffers to be truncated to that many characters. The default '/quit' and '/part' reasons are now configurable. Two new user options are provided for this: 'rcirc-default-part-reason' and 'rcirc-default-quit-reason'. ### Register The return value of method 'register-val-describe' includes the names of buffers shown by the windows of a window configuration. ### Message Completion of email addresses can use the standard completion UI. This is controlled by 'message-expand-name-standard-ui'. With the standard UI the different sources (ecomplete, bbdb, and eudc) are matched together and try to obey 'completion-styles'. It should work for other completion front ends like Company. 'message-mode' now supports highlighting citations of different depths. This can be customized via the new user option 'message-cite-level-function' and the new 'message-cited-text-*' faces. Messages can now be systematically encrypted when the PGP keyring contains a public key for every recipient. To achieve this, add 'message-sign-encrypt-if-all-keys-available' to 'message-send-hook'. When replying a message that have addresses on the form '"foo@bar.com" <foo@bar.com>', Message will elide the repeated "name" from the address field in the response. The default of 'message-forward-as-mime' has changed from t to nil as it has been reported that many recipients can't read forwards that are formatted as MIME digests. 'message-forward-included-headers' has changed its default to exclude most headers when forwarding. 'mml-secure-openpgp-sign-with-sender' sets also "gpg --sender". When 'mml-secure-openpgp-sign-with-sender' is non-nil, message sender's email address (in addition to its old behavior) will also be used to set gpg's "--sender email@domain" option. The option is useful for two reasons when verifying the signature: 1. GnuPG's TOFU statistics are updated for the specific user id (email) only. See gpg(1) man page about "--sender". 2. GnuPG's "--auto-key-retrieve" functionality can use WKD (web key directory) method for finding the signer's key. You need GnuPG 2.2.17 to fully benefit from this feature. See gpg(1) man page for "--auto-key-retrieve". The 'mail-from-style' variable is now obsolete. According to RFC 5322, only the 'angles' value is valid. ### EasyPG 'epa-pinentry-mode' is renamed to 'epg-pinentry-mode'. It now applies to epg functions as well as epa functions. The alias functions 'epa--encode-coding-string', 'epa--decode-coding-string', and 'epa--select-safe-coding-system' have been removed. Use 'encode-coding-string', 'decode-coding-string', and 'select-safe-coding-system' instead. 'epg-context' structure supports now 'sender' slot. The value of the new 'sender' slot (if a string) is used to set gpg's "--sender" option. This feature is used by 'mml-secure-openpgp-sign-with-sender'. See gpg(1) manual page about "--sender" for more information. ### Rmail New user option 'rmail-output-reset-deleted-flag'. If this option is non-nil, messages appended to an output file by the 'rmail-output' command have their Deleted flag reset. The command 'rmail-summary-by-senders' with an empty argument selects the messages to summarize with a regexp that matches the sender of the current message. ### Threads New variable 'main-thread' holds Emacs's main thread. This is handy in Lisp programs that run on a non-main thread and want to signal the main thread, e.g., when they encounter an error. 'thread-join' now returns the result of the finished thread. 'thread-signal' does not propagate errors to the main thread. Instead, error messages are just printed in the main thread. 'thread-alive-p' is now obsolete, use 'thread-live-p' instead. New command 'list-threads' shows Lisp threads. See the current list of live threads in a tabulated-list buffer which automatically updates. In the buffer, you can use 's q' or 's e' to signal a thread with quit or error respectively, or get a snapshot backtrace with 'b'. ### thingatpt.el 'thing-at-point' supports a new "thing" called 'uuid'. A symbol 'uuid' can be passed to 'thing-at-point' and it returns the UUID at point. 'number-at-point' will now recognize hex numbers like 0xAb09 and #xAb09 and return them as numbers. 'word-at-point' and 'sentence-at-point' accept NO-PROPERTIES. Just like 'thing-at-point' itself. ### Interactive automatic highlighting 'highlight-regexp' can now highlight subexpressions. The new command accepts a prefix numeric argument to choose the subexpression. That’s very useful - something I would use a lot myself. ### Mouse display of minor mode menu 'minor-mode-menu-from-indicator' now displays full minor mode name. When there is no menu for a mode, display the mode name after the indicator instead of just the indicator (which is sometimes cryptic). ### rx: regular expression rx now handles raw bytes in character alternatives correctly, when given in a string. Previously, '(any "\x80-\xff")' would match characters U+0080...U+00FF. Now the expression matches raw bytes in the 128...255 range, as expected. The rx 'or' and 'seq' forms no longer require any arguments. '(or)' produces a regexp that never matches anything, while '(seq)' matches the empty string, each being an identity for the operation. This also works for their aliases: '|' for 'or'; ':', 'and' and 'sequence' for 'seq'. The symbol 'unmatchable' can be used as an alternative to '(or)'. 'regexp' and new 'literal' accept arbitrary lisp as arguments. In this case, 'rx' will generate code which produces a regexp string at run time, instead of a constant string. New rx extension mechanism: 'rx-define', 'rx-let', 'rx-let-eval'. These macros add new forms to the rx notation. I feel rx perfectly captures why LISP is such a powerful language. That you can now globally extend rx with your own constructs, or scoped locally with rx-let is cool. 'anychar' is now an alias for 'anything'. Both match any single character; 'anychar' is more descriptive. New 'intersection' form for character sets. With 'or' and 'not', it can be used to compose character-matching expressions from simpler parts. 'not' now accepts more argument types. The argument can now also be a character, a single-character string, an 'intersection' form, or an 'or' form whose arguments each match a single character. Nested 'or' forms of strings guarantee a longest match. For example, '(or (or "IN" "OUT") (or "INPUT" "OUTPUT"))' now matches the whole string "INPUT" if present, not just "IN". Previously, this was only guaranteed inside a single 'or' form of string literals. ### Frames New command 'make-frame-on-monitor' makes a frame on the specified monitor. New value of 'minibuffer' frame parameter 'child-frame'. This allows creating and immediately parenting a minibuffer-only child frame when making a frame. New predicates 'display-blink-cursor-p' and 'display-symbol-keys-p'. These predicates are to be preferred over 'display-graphic-p' when testing for blinking cursor capability and the capability to have symbols (e.g., '[return]', '[tab]', '[backspace]') as keys respectively. ### Tabulated List mode Tabulated list mode is a great way of representing tabular data. It’s used in the Emacs package list for instance. New user options for tabulated list sort indicators. You can now customize which sorting indicator character to display near the current column in Tabulated Lists (see user options 'tabulated-list-gui-sort-indicator-asc', 'tabulated-list-gui-sort-indicator-desc', 'tabulated-list-tty-sort-indicator-asc', and 'tabulated-list-tty-sort-indicator-desc'). Two new commands and keystrokes have been added to the tabulated list mode: 'w' (which widens the current column) and 'c' which makes the current column contract. New function 'tabulated-list-clear-all-tags'. This function clears all tags from the padding area in the current buffer. Tags are typically added by calling 'tabulated-list-put-tag'. ### Text mode 'text-mode-variant' is now obsolete, use 'derived-mode-p' instead. ### CUA mode New user option 'cua-rectangle-terminal-modifier-key'. This user option allows for the customization of the modifier key used in a terminal frame. ### JS mode JSX syntax is now automatically detected and enabled. If a file imports Facebook's 'React' library, or if the file uses the extension ".jsx", then various features supporting XML-like syntax will be supported in 'js-mode' and derivative modes. ('js-jsx-mode' no longer needs to be enabled.) Great alternatively if you don’t want to use the package web-mode. New user option 'js-jsx-detect-syntax' disables automatic detection. This is turned on by default. New user option 'js-jsx-syntax' enables JSX syntax unconditionally. This is off by default. New variable 'js-jsx-regexps' controls JSX detection. JSX syntax is now highlighted like SGML. JSX code is properly indented in many more scenarios. Previously, JSX indentation usually only worked when an element was wrapped in parenthesis (e.g. in a 'return' statement or a function call). It would also fail in many intricate cases. Now, indentation should work anywhere without parenthesis; many more intricacies are supported; and, indentation conventions align more closely with those of the React developer community (see 'js-jsx-align->-with-<'), otherwise still adhering to SGML conventions. New user option 'js-jsx-align->-with-<' controls '>' indents. Commonly in JSX code, a '>' on its own line is indented at the same level as its opening '<'. This is the new default for JSX. This behavior is slightly different than that used by SGML in Emacs, where '>' is indented at the same level as attributes, which was also the old default for JSX. This is turned on by default. To get back the old default indentation behavior of aligning '>' with attributes, set 'js-jsx-align->-with-<' to nil. Indentation uses 'js-indent-level' instead of 'sgml-basic-offset'. Since JSX is a syntax extension of JavaScript, it makes the most sense for JSX expressions to be indented the same number of spaces as other JS expressions. This is a breaking change, but it probably aligns with how you'd expect this indentation to behave. If you want JSX to be indented like JS, you won't need to change your config. The old behavior can be emulated by controlling JSX indentation independently of JS, by setting 'js-jsx-indent-level'. New user option 'js-jsx-indent-level' for different JSX indentation. If you wish to indent JSX by a different number of spaces than JS, set this user option to the desired number. New user option 'js-jsx-attribute-offset' for JSX attribute indents. New variable 'js-syntactic-mode-name' controls mode name display. Previously, the mode name was simply 'JavaScript'. Now, when a syntax extension like JSX is enabled, the mode name is 'JavaScript[JSX]'. Set this variable to nil to disable the new behavior. New function 'js-use-syntactic-mode-name' for deriving modes. Packages deriving from 'js-mode' with 'define-derived-mode' should call this function to add enabled syntax extensions to their mode name, too. ### Autorevert New user option 'auto-revert-avoid-polling' for saving power. When set to a non-nil value, buffers in Auto Revert mode are no longer polled for changes periodically. This reduces the power consumption of an idle Emacs, but may fail on some network file systems; set 'auto-revert-notify-exclude-dir-regexp' to match files where notification is not supported. The default value is nil. New variable 'buffer-auto-revert-by-notification'. A major mode can declare that notification on the buffer's default directory is sufficient to know when updates are required, by setting the new variable 'buffer-auto-revert-by-notification' to a non-nil value. Auto Revert mode can use this information to avoid polling the buffer periodically when 'auto-revert-avoid-polling' is non-nil. 'global-auto-revert-ignore-buffer' can now also be a predicate function that can be used for more fine-grained control of which buffers to auto-revert. ### auth-source-pass New user option 'auth-source-pass-filename'. Allows setting the path to the password-store, defaults to "~/.password-store". New user option 'auth-source-pass-port-separator'. Specifies separator between host and port, defaults to colon ":". Minimize the number of decryptions during password lookup. This makes the package usable with physical tokens requiring touching a sensor for every decryption. 'auth-source-pass-get' is now autoloaded. ### Bookmarks 'bookmark-file' and 'bookmark-old-default-file' are now obsolete aliases of 'bookmark-default-file'. New user option 'bookmark-watch-bookmark-file'. When non-nil, watch whether the bookmark file has changed on disk. The old bookmark file format is no longer supported. This bookmark file format has not been used in Emacs since at least version 19.34, released in 1996, and will no longer be automatically converted to the new bookmark file format. The following functions are now declared obsolete: 'bookmark-grok-file-format-version', 'bookmark-maybe-upgrade-file-format', 'bookmark-upgrade-file-format-from-0', and 'bookmark-upgrade-version-0-alist'. The mantemp.el library is now marked obsolete. This library generates manual C++ template instantiations. It should no longer be useful on modern compilers, which do this automatically. ### Ispell New hook 'ispell-change-dictionary-hook'. This runs after changing the dictionary and could be used to automatically spellcheck a buffer when changing language without needing to advice 'ispell-change-dictionary'. ### scroll-lock New command 'scroll-lock-next-line-always-scroll'. This command is bound to 'S-down' and scrolls the buffer up in particular when the end of the buffer is visible in the window. ### mwheel.el 'mwheel-install' is now obsolete. Use 'mouse-wheel-mode' instead. Note that 'mouse-wheel-mode' is already enabled by default on most graphical displays. ### Gravatar 'gravatar-cache-ttl' is now a number of seconds. The previously used timestamp format of a list of integers is still supported, but is deprecated. The default value has not changed. 'gravatar-size' can now be nil. This results in the use of Gravatar's default size of 80 pixels. The default fallback gravatar is now configurable. This is possible using the new user options 'gravatar-default-image' and 'gravatar-force-default'. ### ada-mode The built-in ada-mode is now deleted. The GNU ELPA package is a good replacement, even in very large source files. ### time-stamp New '%5z' conversion for 'time-stamp-format' gives time zone offset. Specifying '%5z' in 'time-stamp-format' or 'time-stamp-pattern' expands to the time zone offset, e.g., '+0100'. The time zone used is specified by 'time-stamp-time-zone'. Because this feature is new in Emacs 27.1, do not use it in the local variables section of any file that might be edited by an older version of Emacs. Some conversions recommended for 'time-stamp-format' have changed. The new documented/recommended %-conversions are closer to those used by 'format-time-string' and are compatible at least as far back as Emacs 22.1 (released in 2007). Uppercase abbreviated day name of week: was %3A, now %#a Full day name of week: was %:a, now %:A Uppercase abbreviated month name: was %3B, now %#b Full month name: was %:b, now %:B Four-digit year: was %:y, now %Y Lowercase timezone name: was %z, now %#Z Fully-qualified host name: was %s, now %Q Unqualified host name: (was none), now %q Login name: was %u, now %l User's full name: was %U, now %L Merely having '(add-hook 'before-save-hook 'time-stamp)' in your Emacs init file does not expose you to this change. However, if you set 'time-stamp-format' or 'time-stamp-pattern' with a file-local variable, you may need to update the value. ### mode-local Declare 'define-overload' and 'define-child-mode' as obsolete. Rename several internal functions to use a 'mode-local-' prefix. ### CC Mode You can now flag "wrong style" comments with 'font-lock-warning-face'. To do this, use 'c-toggle-comment-style', if needed, to set the desired default comment style (block or line); then set the user option 'c-mark-wrong-style-of-comment' to non-nil. ### Mailcap The new function 'mailcap-file-name-to-mime-type' has been added. It's a simple convenience function for looking up MIME types based on file name extensions. The default way the list of possible external viewers for MIME types is sorted and chosen has changed. Earlier, the most specific viewer was chosen, even if there was a general override in "~/.mailcap". For instance, if "/etc/mailcap" has an entry for "image/gif", that one will be chosen even if you have an entry for "image/*" in your "~/.mailcap" file. But with the new method, entries from "~/.mailcap" overrides all system and Emacs-provided defaults. To get the old method back, set 'mailcap-prefer-mailcap-viewers' to nil. ### MH-E The hook 'mh-show-mode-hook' is now called before the message is inserted. Functions that want to affect the message text (for example, to change highlighting) can no longer use 'mh-show-mode-hook', because the message contents will not yet have been inserted when the hook is called. Such functions should now be attached to 'mh-show-hook'. ### URL The 'file:' handler no longer looks for "index.html" in directories if you ask it for a "file:///dir" URL. Since this is a low-level library, such decisions (if they are to be made at all) are left to higher-level functions. ### New Modes and Packages in Emacs 27.1 ### Tab Bars Emacs finally acquires a native tab bar, much like XEmacs of old. They’ve opted for a cleverer, I think, way of looking at tabs – as a graphical representation of window configurations. That is something that Emacs never really did before (if you exclude third-party packages) and is I think an important milestone. Window configurations have been around for a long time and few use them directly. The real take-away from the tab bar feature is what people end up doing with it: it’s quite feature rich out of the box, but I am sure someone will build something really cool with it. It also combines with Tab line mode (see below), a more traditional take on “tabs” as you’d find in other IDEs. The new command 'tab-bar-mode' enables the tab bar at the top of each frame (including TTY frames), where you can use tabs to switch between named persistent window configurations. The 'C-x t' sequence is the new prefix key for tab-related commands: 'C-x t 2' creates a new tab; 'C-x t 0' deletes the current tab; 'C-x t b' switches to buffer in another tab; 'C-x t f' and 'C-x t C-f' edit file in another tab; and 'C-TAB' and 'S-C-TAB' switch to the next or previous tab. You can also switch between tabs and create/delete tabs with a mouse. Tab-related commands are available even when 'tab-bar-mode' is disabled: by default, they enable 'tab-bar-mode' in that case. The X resource "tabBar", class "TabBar" enables the tab bar when its value is "on", "yes" or "1". The user option 'tab-bar-position' specifies where to show the tab bar. Tab-related commands can be used even without the tab bar when 'tab-bar-mode' is disabled by a nil value of the user option 'tab-bar-show'. Without the tab bar you can switch between tabs using completion on tab names, or using 'tab-switcher'. Read the new Info node "(emacs) Tab Bars" for full description of all related features. Pay close attention to the symmetry between window creation; frame creation; etc: C-x 2, to split a window; C-x 5 2 to create a new frame. Tab Line mode The new command 'global-tab-line-mode' enables the tab line above each window, which you can use to switch buffers in the window. Selecting the previous window-local tab is the same as typing 'C-x <LEFT>' ('previous-buffer'), selecting the next tab is the same as 'C-x <RIGHT>' ('next-buffer'). Both commands support a numeric prefix argument as a repeat count. Clicking on the plus icon adds a new buffer to the window-local tab line of buffers. Using the mouse wheel on the tab line scrolls tabs. Read the new Info node "(emacs) Tab Line" for full description of all related features. This is what people think of when they think of “tabs”. There’s a myriad of switches you can play around with in the Customize interface: M-x customize-group tab-bar RET. By default the tabs will use whatever the default Emacs faces are. Either upgrade your theme or, if you’re like me and old enough to have customized your color scheme long before themes were even a thing: M-x customize-group tab-bar-faces RET. fileloop.el lets one setup multifile operations like search&replace. Emacs can now visit files in archives as if they were directories. This feature uses Tramp and works only on systems which support GVFS, i.e. GNU/Linux, roughly spoken. See the node "(tramp) Archive file names" in the Tramp manual for full documentation of these facilities. I don’t need to do this often, but I’m happy the feature is there, even though I found the older “read-only archive view” reasonable. New library for writing JSONRPC applications (https://jsonrpc.org). The 'jsonrpc' library enables writing Emacs Lisp applications that rely on this protocol. Since the protocol is designed to be transport-agnostic, the library provides an API to implement new transport strategies as well as a separate API to use them. A transport implementation for process-based communication, such as is used by the Language Server Protocol (LSP), is readily available. People have talked up the C-library JSON implementation as a big deal for LSP, and it definitely is, but this library may result in particularly stable implementations getting reimplemented in C while keeping the API the same. That may be yet another avenue of performance for lsp libraries in particular. Backtrace mode improves viewing of Elisp backtraces. Backtrace mode adds pretty printing, fontification and ellipsis expansion to backtrace buffers produced by the Lisp debugger, Edebug and ERT. See the node "(elisp) Backtraces" in the Elisp manual for documentation of the new mode and its commands. A much appreciated change as reading large backtraces can be a difficult proposition. so-long.el helps to mitigate performance problems with long lines. When 'global-so-long-mode' has been enabled, visiting a file with very long lines will (subject to configuration) cause the user's preferred 'so-long-action' to be automatically invoked (by default, the buffer's major mode is replaced by 'so-long-mode'). In extreme cases this can prevent delays of several minutes, and make Emacs responsive almost immediately. Type 'M-x so-long-commentary' for full documentation. I am happy this made it in, but I am sad (as mentioned previously) that we need such ugly hacks to work around limitations in Emacs. ### Incompatible Lisp Changes in Emacs 27.1 Incomplete destructive splicing support has been removed. Support for Common Lisp style destructive splicing (",.") was incomplete and broken for a long time. It has now been removed. This means that backquote substitution now works for identifiers starting with a period ("."). Consider the following example: (let ((.foo 42)) ,.foo) In the past, this would have incorrectly evaluated to '(\,\. foo)', but will now instead evaluate to '42'. The REGEXP in 'magic-mode-alist' is now matched case-sensitively. Likewise for 'magic-fallback-mode-alist'. Not sure there are too many people who make changes to Emacs’s magic mode (the system that guesses the major mode based on buffer contents) but as it changes case sensitivity, it may affect some people. 'add-hook' does not always add to the front or the end any more. The replacement of 'append' with 'depth' implies that the function is not always added to the very front (when append/depth is nil) or the very end (when append/depth is t) any more because other functions on the hook may have specified higher/lower depths. This makes it possible to control the ordering of functions more precisely, as was already possible in 'add-function' and 'advice-add'. I’ve never found myself in the situation where this has caused issues, but I can see how package authors would want to decide ahead of time when they want their hooks to run. That is much harder to do with a binary flag. In 'compilation-error-regexp-alist' the old undocumented feature where 'line' could be a function of 2 arguments has been dropped. 'define-fringe-bitmap' is always defined, even when Emacs is built without any GUI support. Just loading a theme's file no longer activates the theme's settings. Loading a theme with 'M-x load-theme' still activates the theme, as it did before. However, loading the theme's file with 'M-x load-file', or using 'require' or 'load' in a Lisp program, doesn't actually apply the theme's settings until you either invoke 'M-x enable-theme' or type 'M-x load-theme'. (In a Lisp program, calling 'enable-theme' or invoking 'load-theme' with NO-ENABLE argument omitted or nil has the same effect of activating a theme whose file has been loaded.) The special case of the 'user' theme is an exception: it is frequently used for ad-hoc customizations, so the settings of that theme are by default applied immediately. The variable 'custom--inhibit-theme-enable' controls this behavior; its default value changed in Emacs 27.1. That may catch out a few people relying on the side-effect that loading a theme also applies it. The REPETITIONS argument of 'benchmark-run' can now also be a variable. Interpretation of relative 'HOME' directory has changed. If "HOME" is set to a relative file name, 'expand-file-name' now
interprets it relative to the directory where Emacs was started, not
relative to the 'default-directory' of the current buffer.  We recommend
always setting "\$HOME" to an absolute file name, so that its meaning is
independent of where Emacs was started.
'file-name-absolute-p' no longer considers "~foo" to be an absolute
file name if there is no user named "foo".
The FILENAME argument to 'file-name-base' is now mandatory and no
longer defaults to 'buffer-file-name'.
File metadata primitives now signal an error if I/O, access, or
other serious errors prevent them from determining the result.
Formerly, these functions often (though not always) silently returned
nil.  For example, if there is an access error, I/O error or low-level
integer overflow when getting the attributes of a file F,
'(file-attributes F)' now signals an error instead of returning nil.
These functions still behave as before if the only problem is that the
file does not exist.  The affected primitives are
'directory-files-and-attributes', 'file-acl', 'file-attributes',
'file-system-info', and 'set-visited-file-modtime'.
The function 'eldoc-message' now accepts a single argument.
Programs that called it with multiple arguments before should pass
them through 'format' first.  Even that is discouraged: for ElDoc
support, you should set 'eldoc-documentation-function' instead of
calling 'eldoc-message' directly.
Old-style backquotes now generate an error.
They have been generating warnings for a decade.  To interpret
old-style backquotes as new-style, bind the new variable
'force-new-style-backquotes' to t.
Defining a Common Lisp structure using 'cl-defstruct' or
'cl-struct-define' whose name clashes with a builtin type (e.g.,
'integer' or 'hash-table') now signals an error.
When formatting a floating-point number as an octal or hexadecimal
integer, Emacs now signals an error if the number is too large for the
implementation to format.
'logb' now returns infinity when given an infinite or zero argument,
and returns a NaN when given a NaN.  Formerly, it returned an extreme
fixnum for such arguments.
Some functions and variables obsolete since Emacs 22 have been removed:
'archive-mouse-extract', 'assoc-ignore-case', 'assoc-ignore-representation',
'custom-face-save-command', 'cvs-display-full-path', 'cvs-fileinfo->full-path',
'delete-frame-hook', 'derived-mode-class', 'describe-char-after',
'describe-project', 'desktop-basefilename', 'desktop-buffer-handlers',
'desktop-buffer-misc-functions', 'desktop-buffer-modes-to-save',
'disabled-command-hook', 'dungeon-mode-map', 'electric-nroff-mode',
'electric-nroff-newline', 'electric-perl-terminator', 'executing-macro',
'focus-frame', 'forward-text-line', 'generic-define-mswindows-modes',
'generic-define-unix-modes', 'generic-font-lock-defaults',
'ibuffer-elide-long-columns', 'ibuffer-hooks', 'ibuffer-mode-hooks',
'icalendar-convert-diary-to-ical', 'icalendar-extract-ical-from-buffer',
'ipconfig-program-options', 'isearch-lazy-highlight-cleanup',
'isearch-lazy-highlight-initial-delay', 'isearch-lazy-highlight-interval',
'isearch-lazy-highlight-max-at-a-time', 'iswitchb-use-fonts',
'latin1-char-displayable-p', 'mouse-wheel-click-button',
'mouse-wheel-down-button', 'mouse-wheel-up-button', 'new-frame',
'pascal-outline', 'process-kill-without-query',
'speedbar-buffers-line-path', 'speedbar-ignored-path-expressions',
'speedbar-ignored-path-regexp', 'speedbar-line-path', 'speedbar-path-line',
'timer-set-time-with-usecs', 'tooltip-gud-display', 'tooltip-gud-modes',
'vc-comment-search-forward', 'vc-comment-search-reverse',
'vc-comment-to-change-log', 'vc-diff-switches-list', 'vc-next-comment',
'vc-previous-comment', 'view-todo', 'x-lost-selection-hooks',
'x-sent-selection-hooks'.
Further functions and variables obsolete since Emacs 24 have been removed:
'default-directory-alist', 'dired-default-directory',
'dired-default-directory-alist', 'dired-enable-local-variables',
'dired-hack-local-variables', 'dired-local-variables-file',
'dired-omit-here-always'.
Garbage collection no longer treats miscellaneous objects specially;
they are now allocated like any other pseudovector.  As a result, the
'garbage-collect' and 'memory-use-count' functions no longer return a
'misc' component, and the 'misc-objects-consed' variable has been
removed.
Reversed character ranges are no longer permitted in 'rx'.
Previously, ranges where the starting character is greater than the
ending character were silently omitted.
For example, '(rx (any "@z-a" (?9 . ?0)))' would match '@' only.
Now, such 'rx' expressions generate an error.
Internal 'rx' functions and variables have been removed,
as a consequence of an improved implementation.  Packages using
these should use the public 'rx' and 'rx-to-string' instead.
'rx-constituents' is still available for compatibility, but the new
extension mechanism is preferred: 'rx-define', 'rx-let' and
'rx-let-eval'.
'text-mode' no longer sets the value of 'indent-line-function'.
The global value of 'indent-line-function', which defaults to
'indent-relative', will no longer be reset locally when turning on
'text-mode'.

To get back the old behavior, add a function to 'text-mode-hook' which
performs '(setq-local indent-line-function #'indent-relative)'.
'make-process' no longer accepts a non-nil ':stop' key.  This has
never worked reliably, and now causes an error.
'eventp' no longer returns non-nil for lists whose car is nil.
This is consistent with the fact that nil, though a symbol, is not a
valid event type.
The obsolete package xesam.el (since Emacs 24) has been removed.
The XBM image handler now accepts a ':stride' argument, which should
be specified in image specs representing the entire bitmap as a single
bool vector.
'regexp-quote' may return its argument string.
If the argument needs no quoting, it can be returned instead of a copy.
Mouse scroll up and down with control key modifier changes font size.
Previously, the control key modifier was used to scroll up or down by
an amount which was close to near a full screen.  This is now instead
available by scrolling with the meta modifier key.

To get the old behavior back, customize the user option

(customize-set-variable 'mouse-wheel-scroll-amount
'(5 ((shift) . 1) ((control) . nil)))

By default, the font size will be changed in the window that the mouse
pointer is over.  To change this behavior, you can customize the user
option 'mouse-wheel-follow-mouse'.  Note that this will also affect
scrolling.
Mouse scroll up and down with control key modifier also works on images
where it scales the image under the mouse pointer.
'help-follow-symbol' now signals 'user-error' if point (or the
position pointed to by the argument POS) is not in a symbol.
The options.el library has been removed.
It was obsolete since Emacs 22.1, replaced by customize.
The tls.el and starttls.el libraries are now marked obsolete.
Use of built-in libgnutls based functionality (described in the Emacs
GnuTLS manual) is recommended instead.
The url-ns.el library is now marked obsolete.
This library is used to open configuration files for the long defunct
web browser Netscape, and is no longer relevant.

### Lisp Changes in Emacs 27.1

Emacs Lisp integers can now be of arbitrary size.
Emacs uses the GNU Multiple Precision (GMP) library to support
integers whose size is too large to support natively.  The integers
supported natively are known as "fixnums", while the larger ones are
"bignums".  The new predicates 'bignump' and 'fixnump' can be used to
distinguish between these two types of integers.

All the arithmetic, comparison, and logical (a.k.a. "bitwise")
operations where bignums make sense now support both fixnums and
bignums.  However, note that unlike fixnums, bignums will not compare
equal with 'eq', you must use 'eql' instead.  (Numerical comparison
with '=' works on both, of course.)

Since large bignums consume a lot of memory, Emacs limits the size of
the largest bignum a Lisp program is allowed to create.  The
nonnegative value of the new variable 'integer-width' specifies the
maximum number of bits allowed in a bignum.  Emacs signals an integer
overflow error if this limit is exceeded.

Several primitive functions formerly returned floats or lists of
integers to represent integers that did not fit into fixnums.  These
functions now simply return integers instead.  Affected functions
include functions like 'encode-char' that compute code-points, functions
like 'file-attributes' that compute file sizes and other attributes,
functions like 'process-id' that compute process IDs, and functions like
'user-uid' and 'group-gid' that compute user and group IDs.

'overflow-error' is now documented as a subcategory of 'range-error'.
Formerly it was undocumented, and was (incorrectly) a subcategory
of 'domain-error'.

### Time values

New function 'time-convert' converts Lisp time values to Lisp
timestamps of various forms, including a new timestamp form '(TICKS
. HZ)' where TICKS is an integer and HZ a positive integer denoting a
clock frequency.
Although the default timestamp format is still '(HI LO US PS)',
it is planned to change in a future Emacs version, to exploit bignums.
The documentation has been updated to mention that the timestamp
format may change and that programs should use functions like
'format-time-string', 'decode-time', and 'time-convert' rather than
probing the innards of a timestamp directly, or creating a timestamp
by hand.
Decoded (calendrical) timestamps now have subsecond resolution.
This affects 'decode-time', which generates these timestamps, as well
as functions like 'encode-time' that accept them.  The subsecond info
is present as a '(TICKS . HZ)' value in the seconds element of a
decoded timestamp, and 'decode-time' has a new optional FORM argument
specifying the form of the seconds member.  For example, if X is the
timestamp '(1566009571321878186 . 1000000000)', which represents
"2019-08-17 02:39:31.321878186 UTC", '(decode-time X t t)' returns
'((31321878186 . 1000000000) 39 2 17 8 2019 6 nil 0)' instead of the
traditional '(31 39 2 17 8 2019 6 nil 0)' returned by plain
'(decode-time X t)'.  Although the default FORM is currently
'integer', which truncates the seconds to an integer and is the
traditional behavior, this default may change in future Emacs
versions, so callers requiring an integer should specify FORM
explicitly.
'encode-time' supports a new API '(encode-time TIME)'.
The old 'encode-time' API is still supported.
A new package to parse ISO 8601 time, date, durations and
intervals has been added.  The main function to use is
'iso8601-parse', but there's also 'iso8601-parse-date',
'iso8601-parse-time', 'iso8601-parse-duration' and
'iso8601-parse-interval'.  All these functions return decoded time
structures, except the final one, which returns three of them (start,
end and duration).

Useful for JSON-related stuff especially.

'time-add', 'time-subtract', and 'time-less-p' now accept
infinities and NaNs too, and propagate them or return nil like
floating-point operators do.  If both arguments are finite, these
functions now return exact results instead of rounding in some cases,
and they also avoid excess precision when that is easy.
New function 'time-equal-p' compares time values for equality.
'format-time-string' supports a new conversion specifier flag '+'
that acts like the '0' flag but also puts a '+' before nonnegative
years containing more than four digits.  This is for compatibility
with POSIX.1-2017.
To access (or alter) the elements of a decoded time value, the
'decoded-time-second', 'decoded-time-minute', 'decoded-time-hour',
'decoded-time-day', 'decoded-time-month', 'decoded-time-year',
'decoded-time-weekday', 'decoded-time-dst' and 'decoded-time-zone'
accessors can be used.
The new functions 'date-days-in-month' (which will say how many
days there are in a monthin a specific year), 'date-ordinal-to-time'
(that computes the date of an ordinal day), 'decoded-time-add' (for
doing computations on a decoded time structure), 'make-decoded-time'
(for making a decoded time structure with only the given keywords
filled out), and 'encoded-time-set-defaults' (which fills in nil
elements as if it's midnight January 1st, 1970) have been added.
In the DST slot, 'encode-time' and 'parse-time-string' now return -1
if it is not known whether daylight saving time is in effect.
Formerly they were inconsistent: 'encode-time' returned t in this
situation, whereas 'parse-time-string' returned nil.  Now they
consistently use nil to mean that DST is not in effect, and use -1
to mean that it is not known whether DST is in effect.
New macro 'benchmark-progn'.
This macro works like 'progn', but messages how long it takes to
evaluate the body forms.  The value of the last form is the return
value.
New function 'read-char-from-minibuffer'.
to read a character, so it maintains a history that can be navigated
via usual minibuffer keystrokes 'M-p'/'M-n'.
New variables 'set-message-function' and 'clear-message-function'
can be used to specify functions to show and clear messages that
normally are displayed in the echo area.
'setq-local' can now set an arbitrary number of variables, which
makes the syntax more like 'setq'.

Useful little change if you frequently set lots of variables to buffer-local, particularly in mode initialization.

'reveal-mode' can now also be used for more than to toggle between
invisible and visible: It can also toggle 'display' properties in
overlays.  This is only done on 'display' properties that have the
'reveal-toggle-invisible' property set.
'process-contact' now takes an optional NO-BLOCK argument to allow
not waiting for a process to be set up.
New variable 'read-process-output-max' controls sub-process throughput.
This variable determines how many bytes can be read from a sub-process
in one read operation.  The default, 4096 bytes, was previously a
hard-coded constant.  Setting it to a larger value might enhance
throughput of reading from sub-processes that produces vast
(megabytes) amounts of data in one go.

This I suspect is another tweak that is likely to affect the performance of many packages - quite possibly the language server stuff.

The new user option 'quit-window-hook' is now run first when
executing the 'quit-window' command.
The user options 'help-enable-completion-auto-load',
without the hyphen in their names.  Obsolete aliases from the old
names have been added.
Buttons (created with 'make-button' and related functions) can
now use the 'button-data' property.  If present, the data in this
property will be passed on to the 'action' function instead of the
button itself in 'button-activate'.
'defcustom' now takes a ':local' keyword that can be either t or
'permanent', which mean that the variable should be automatically
buffer-local.  'permanent' also sets the variable's 'permanent-local'
property.
The new macro 'with-suppressed-warnings' can be used to suppress
specific byte-compile warnings.
The new macro 'ignore-error' is like 'ignore-errors', but takes a
specific error condition, and will only ignore that condition.  (This
can also be a list of conditions.)
The new function 'byte-compile-info-message' can be used to output
informational messages that look pleasing during the Emacs build.
New 'help-fns-describe-variable-functions' hook.
It makes it possible to add metadata information to 'describe-variable'.
'ngettext' can be used now to return the right plural form
according to the given numeric value.
'inhibit-null-byte-detection' is renamed to 'inhibit-nul-byte-detection'.
'self-insert-command' takes the char to insert as (optional) argument.
'lookup-key' can take a list of keymaps as argument.
'condition-case' now accepts t to match any error symbol.
New function 'proper-list-p'.
Given a proper list as argument, this predicate returns its length;
otherwise, it returns nil.  'format-proper-list-p' is now an obsolete
alias for the new function.
'define-minor-mode' automatically documents the meaning of ARG.
The function 'recenter' now accepts an additional optional argument.
By default, calling 'recenter' will not redraw the frame even if
'recenter-redisplay' is non-nil.  Call 'recenter' with the new second
argument non-nil to force redisplay per 'recenter-redisplay's value.
New functions 'major-mode-suspend' and 'major-mode-restore'.
Use them when switching temporarily to another major mode, e.g. for
'hexl-mode', or to switch between 'c-mode' and 'image-mode' in XPM.
New macro 'dolist-with-progress-reporter'.
This works like 'dolist', but reports progress similar to
'dotimes-with-progress-reporter'.
New hook 'after-delete-frame-functions'.
This works like 'delete-frame-functions', but runs after the frame to
be deleted has been made dead and removed from the frame list.
The function 'provided-mode-derived-p' was extended to support aliases.
The function now returns non-nil when the argument MODE is derived
from any alias of any of MODES.
New frame focus state inspection interface.
The hooks 'focus-in-hook' and 'focus-out-hook' are now obsolete.
and inspect the focus state of each frame using 'frame-focus-state'.
Emacs now requests and recognizes focus-change notifications from TTYs.
On terminal emulators that support the feature, Emacs can now support
'focus-in-hook' and 'focus-out-hook' for TTY frames.
Window-specific face remapping.
Face specifications (of the kind used in 'face-remapping-alist')
now support filters, allowing faces to vary between different windows
displaying the same buffer.  See the node "(elisp) Face Remapping"
of the Emacs Lisp Reference manual for more detail.
Window change functions have been redesigned.
Hooks reacting to window changes run now only when redisplay detects
that a change has actually occurred.  Six hooks are now provided:
'window-buffer-change-functions' (run after window buffers have
changed), 'window-size-change-functions' (run after a window was
assigned a new buffer or size), 'window-configuration-change-hook'
(like the former but run also when a window was deleted),
'window-selection-change-functions' (run when the selected window
changed) and 'window-state-change-functions' and
'window-state-change-hook' (run when any of the preceding ones is
run).  Applications can enforce running the latter two using the new
function 'set-frame-window-state-change'.  'window-scroll-functions'
are unaffected by these changes.

In addition, a number of functions now allow the caller to detect what
has changed since last redisplay: 'window-old-buffer' returns for any
window the buffer it showed at that time.  'old-selected-window' and
'old-selected-frame' return the window and frame that were selected
during last redisplay.  'window-old-pixel-width' (renamed from
'window-pixel-width-before-size-change'), 'window-old-pixel-height'
(renamed from 'window-pixel-height-before-size-change'),
'window-old-body-pixel-width' and 'window-old-body-pixel-height'
return the total and body sizes of any window during last redisplay.

Also 'run-window-configuration-change-hook' is declared obsolete.

See the section "(elisp) Window Hooks" in the Elisp manual for a
detailed explanation of the new behavior.
Scroll bar and fringe settings can now be made persistent for windows.
The functions 'set-window-scroll-bars' and 'set-window-fringes' now
have a new optional argument that makes the settings they produce
reliably survive subsequent invocations of 'set-window-buffer'.
New user option 'resize-mini-frames'.
This option allows automatically resizing minibuffer-only frames
similarly to how minibuffer windows are resized on "normal" frames.
New buffer display action function 'display-buffer-in-direction'.
This function allows specifying the location of the window chosen by
'display-buffer' in various ways.
New buffer display action alist entry 'dedicated'.
Such an entry allows specifying the dedicated status of a window
created by 'display-buffer'.
New buffer display action alist entry 'window-min-height'.
Such an entry allows specifying a minimum height of the window used
for displaying a buffer.  'display-buffer-below-selected' is the only
action function to respect it at the moment.
New buffer display action alist entry 'direction'.
This entry is used to specify the location of the window chosen by
'display-buffer-in-direction'.
Additional meaning of display action alist entry 'window'.
A 'window' entry can now also specify a reference window for
'display-buffer-in-direction'.
The function 'assoc-delete-all' now takes an optional predicate argument.
New function 'string-distance' to calculate the Levenshtein distance
between two strings.
'print-quoted' now defaults to t, so if you want to see
'(quote x)' instead of 'x you will have to bind it to nil where applicable.
Numbers formatted via '%o' or '%x' are now formatted as signed integers.
This avoids problems in calls like '(read (format "#x%x" -1))', and is
more compatible with bignums.  To get the traditional machine-dependent
behavior, set the experimental variable 'binary-as-unsigned' to t,
<32252@debbugs.gnu.org>.  Because '%o' and '%x' can now format signed
integers, they now support the '+' and space flags.
In Emacs Lisp mode, symbols with confusable quotes are highlighted.
For example, the first character in '‘foo' would be highlighted in
'font-lock-warning-face'.
Omitting variables after '&optional' and '&rest' is now allowed.
For example '(defun foo (&optional))' is no longer an error.  This is
sometimes convenient when writing macros.  See the ChangeLog entry
titled "Allow '&rest' or '&optional' without following variable
(Bug#29165)" for a full listing of which arglists are accepted across
versions.
Internal parsing commands now use 'syntax-ppss' and disregard
'open-paren-in-column-0-is-defun-start'.  This affects mostly things like
'forward-comment', 'scan-sexps', and 'forward-sexp' when parsing backward.
The new variable 'comment-use-syntax-ppss' can be set to nil to recover
the old behavior if needed.
This also means that there is no longer any need to precede opening
brackets at the start of a line inside documentation strings with a
backslash, although there is no harm in doing so to make the code
easier to edit with an older Emacs version.
New symbolic accessor functions for a parse state list.
The new accessor functions 'ppss-depth', 'ppss-list-start',
'ppss-last-sexp-start', 'ppss-string-terminator', 'comment-depth',
'quoted-p', 'comment-style', 'comment-or-string-start', 'open-parens',
and 'two-character-syntax' can be used on the list value returned by
'parse-partial-sexp' and 'syntax-ppss'.

The syntax-ppss and related functions are very useful for building your own, simplified selection mechanisms that can operate on strings and balanced expressions.

The 'server-name' and 'server-socket-dir' variables are set when a
socket has been passed to Emacs.
The 'file-system-info' function is now available on all platforms.
instead of just Microsoft platforms.  This fixes a 'get-free-disk-space'
bug on OS X 10.8 and later.
The function 'get-free-disk-space' returns now a non-nil value for
remote systems, which support this check.
'memory-limit' now returns a better estimate of memory consumption.
When interpreting 'gc-cons-percentage', Emacs now estimates the
heap size more often and (we hope) more accurately.  E.g., formerly
'(progn (let ((gc-cons-percentage 0.8)) BODY1) BODY2)' continued to use
the 0.8 value during BODY2 until the next garbage collection, but that
is no longer true.  Applications may need to re-tune their GC tricks.
New macro 'combine-change-calls' arranges to call the change hooks
('before-change-functions' and 'after-change-functions') just once
each around a sequence of lisp forms, given a region.  This is
useful when a function makes a possibly large number of repetitive
changes and the change hooks are time consuming.
'eql', 'make-hash-table', etc. now treat NaNs consistently.
Formerly, some of these functions ignored signs and significands of
NaNs.  Now, all these functions treat NaN signs and significands as
significant.  For example, '(eql 0.0e+NaN -0.0e+NaN)' now returns nil
because the two NaNs have different signs; formerly it returned t.
Also, Emacs now reads and prints NaN significands; e.g., if X is a
NaN, '(format "%s" X)' now returns "0.0e+NaN", "1.0e+NaN", etc.,
depending on X's significand.
The function 'make-string' accepts an additional optional argument.
If the optional third argument is non-nil, 'make-string' will produce
a multibyte string even if its second argument is an ASCII character.
'(format "%d" X)' no longer mishandles a floating-point number X that
does not fit in a machine integer.
New coding-system 'ibm038'.
This is the International EBCDIC encoding, also available as aliases
'ebcdic-int' and 'cp038'.
New JSON parsing and serialization functions 'json-serialize',
'json-insert', 'json-parse-string', and 'json-parse-buffer'.  These
are implemented in C using the Jansson library.

As mentioned these form part of the C library support for JSON

New function 'ring-resize'.
'ring-resize' can be used to grow or shrink a ring.
New function 'flatten-tree'.
'flatten-list' is provided as an alias.  These functions take a tree
and 'flatten' it such that the result is a list of all the terminal
nodes.
'zlib-decompress-region' can partially decompress corrupted data.
If the new optional ALLOW-PARTIAL argument is passed, then the data
that was decompressed successfully before failing will be inserted
into the buffer.

### Image mode

New library Exif.
An Exif library has been added that can parse JPEG files and output
data about creation times and orientation and the like.
'exif-parse-file' and 'exif-parse-buffer' are the main interface
functions.
'image-mode' now uses this library to automatically rotate images
according to the orientation in the Exif data, if any.
The command 'image-rotate' now accepts a prefix argument.
With a prefix argument, 'image-rotate' now rotates the image at point
90 degrees counter-clockwise, instead of the default clockwise.
In 'image-mode' the image is resized automatically to fit in window.
By default, the image will resize upon first display and whenever the
window's dimensions change.  Two user options 'image-auto-resize' and
'image-auto-resize-on-window-resize' control the resizing behavior
(including the possibility to disable auto-resizing).  A new prefix
key 's' contains the commands that can be used to fit the image to the
window manually.
Some 'image-mode' variables are now buffer-local.
The image parameters 'image-transform-rotation',
'image-transform-scale' and 'image-transform-resize' are now declared
buffer-local, so each buffer could have its own values for these
parameters.
Three new 'image-mode' commands have been added: 'm', which marks
the file in the dired buffer(s) for the directory the file is in; 'u',
which unmarks the file; and 'w', which pushes the current buffer's file
name to the kill ring.

These commands are more useful than they look:

• You can now mark/unmark in an image-mode buffer and Emacs will open a dired in the directory of that image if it does not exist already.
• You can also navigate between images in the same directory with n and p. In other words, you have a simple workflow for marking or unmarking images for further operation in dired.

Pretty nifty.

New library image-converter.
If you need to view exotic image formats for which Emacs doesn't have
native support, customize the new user option
'image-use-external-converter' to t.  If your system has
GraphicsMagick, ImageMagick or 'ffmpeg' installed, they will then be
used to convert images automatically before displaying them.
'auto-mode-alist' now includes many of the types typically
supported by the external image converters, like WEPB, BMP and ICO.
These now default to using 'image-mode'.
'imagemagick-types-inhibit' disables using ImageMagick by default.
'image-mode' started using ImageMagick by default for all images
some years back.  It now respects 'imagemagick-types-inhibit' as a way
to disable that.

### Modules

The function 'load' now behaves correctly when loading modules.
Specifically, it puts the module name into 'load-history', prints
loading messages if requested, and protects against recursive loads.
New module environment function 'process_input' to process user
input while module code is running.
New module environment functions 'make_time' and 'extract_time' to
convert between timespec structures and Emacs Lisp time values.
New module environment functions 'make_big_integer' and
'extract_big_integer' to create and extract arbitrary-size integer
values.
emacs-module.h now defines a macro 'EMACS_MAJOR_VERSION' that expands
to the major version of the latest Emacs supported by the header.
The function 'read-variable' now uses its own history list.
the new variable 'custom-variable-history'.
The functions 'string-to-unibyte' and 'string-to-multibyte' are no
longer declared obsolete.  We have found that there are legitimate use
cases for these functions, where there's no better alternative.  We
believe that the incorrect uses of these functions all but disappeared
by now, so we are un-obsoleting them.
New function 'group-name' returns a group name corresponding to GID.
'make-process' now takes a keyword argument ':file-handler'; if
that is non-nil, it will look for a file name handler for the current
buffer's 'default-directory' and invoke that file name handler to make
the process.  That way 'make-process' can start remote processes.
'(locale-info 'paper)' now returns the paper size on systems that support it.
This is currently supported on GNUish hosts and on modern versions of
MS-Windows.
The function 'regexp-opt', when given an empty list of strings, now
returns a regexp that never matches anything, which is an identity for
this operation.  Previously, the empty string was returned in this
case.
New constant 'regexp-unmatchable' contains a never-matching regexp.
It is a convenient and readable way to specify a regexp that should
not match anything, and is as fast as any such regexp can be.
New functions to handle the URL variant of base-64 encoding.
New functions 'base64url-encode-string' and 'base64url-encode-region'
implement the url-variant of base-64 encoding as defined in RFC4648.

The functions 'base64-decode-string' and 'base64-decode-region' now
accept an optional argument to decode the URL variant of base-64
encoding.
The function 'file-size-human-readable' accepts more optional arguments.
The new third argument is a string put between the number and unit; it
defaults to the empty string.  The new fourth argument is a string
representing the unit to use; it defaults to "B" when the second
argument is 'iec' and the empty string otherwise.  We recommend a
space or non-breaking space as third argument, and "B" as fourth
argument, circumstances allowing.
'format-spec' has been expanded with several modifiers to allow
greater flexibility when customizing variables.  The modifiers include
zero-padding, upper- and lower-casing, and limiting the length of the
interpolated strings.  The function has now also been documented in
the Emacs Lisp manual.
'directory-files-recursively' can now take an optional PREDICATE
parameter to control descending into subdirectories, and a
other directories should be followed.
New function 'xor' returns the boolean exclusive-or of its args.
The function was previously defined in array.el, but has been moved to
subr.el so that it is available by default.  It now always returns the
non-nil argument when the other is nil.  Several duplicates of 'xor'
in other packages are now obsolete aliases of 'xor'.
'define-globalized-minor-mode' now takes BODY forms.
New text property 'help-echo-inhibit-substitution'.
Setting this on the first character of a help string disables
conversions via 'substitute-command-keys'.
New text property 'minibuffer-message'.
Setting this on a character of the minibuffer text will display the
temporary echo messages before that character, when messages need to
be displayed while minibuffer is active.
'undo' can be made to ignore the active region for a command
by setting 'undo-inhibit-region' symbol property of that command to
non-nil.  This is used by 'mouse-drag-region' to make the effect
easier to undo immediately afterwards.
When called interactively, 'next-buffer' and 'previous-buffer' now
signal 'user-error' if there is no buffer to switch to.

### Changes in Emacs 27.1 on Non-Free Operating Systems

Battery status is now supported in all Cygwin builds.
Previously it was supported only in the Cygwin-w32 build.
Emacs now handles key combinations involving the macOS "command"
and "option" modifier keys more correctly.
MacOS modifier key behavior is now more adjustable.
The behavior of the macOS "Option", "Command", "Control" and
"Function" keys can now be specified separately for use with
ordinary keys, function keys and mouse clicks.  This allows using them
in their standard macOS way for composing characters.
The special handling of 'frame-title-format' on NS where setting it
to t would enable the macOS proxy icon has been replaced with a
separate variable, 'ns-use-proxy-icon'.  'frame-title-format' will now
work as on other platforms.
New primitive 'w32-read-registry'.
This primitive lets Lisp programs access the MS-Windows Registry by
retrieving values stored under a given key.  It is intended to be used
for supporting features such as XDG-like location of important files
and directories.
The default value of 'w32-pipe-read-delay' is now zero.
This speeds up reading output from sub-processes that produce a lot of
data.

This variable may need to be non-zero only when running DOS programs
as Emacs subprocesses, which by now is not supported on modern
versions of MS-Windows.  Set this variable to 50 if for some reason
you need the old behavior (and please report such situations to Emacs
developers).
New variable 'w32-multibyte-code-page'.
This variable holds the value of the multibyte code page used by the
system.  It is usually zero, which indicates that 'w32-ansi-code-page'
is being used, except in Far Eastern locales.  When this variable is
non-zero, Emacs at startup sets 'locale-coding-system' to the
corresponding encoding, instead of using 'w32-ansi-code-page'.
The default value of 'inhibit-compacting-font-caches' is t on MS-Windows.
Experience shows that compacting font caches causes more trouble on
MS-Windows than it helps.
Font lookup on MS-Windows was improved to support rare scripts.
To activate the improvement, run the new function
'w32-find-non-USB-fonts' once per Emacs session, or assign to the new
variable 'w32-non-USB-fonts' the list of scripts and the corresponding
fonts.  See the documentation of this function and variable in the
Emacs manual for more details.
On NS the behavior of drag and drop can now be modified by use of
modifier keys in line with Apples guidelines.  This makes the drag and
drop behavior more consistent, as previously the sending application
was able to 'set' modifiers without the knowledge of the user.
On NS multicolor font display is enabled again since it is also
implemented in Emacs on free operating systems via Cairo drawing.