Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

In Alacritty 0.13+ some keys (such as the Pause key) are now spewing "weird" text, when previously nothing happened #7623

Closed
jfly opened this issue Jan 15, 2024 · 25 comments · Fixed by #7677
Assignees

Comments

@jfly
Copy link

jfly commented Jan 15, 2024

System

OS: Linux (X11, xmonad with no compositor)
Version: 0.13.1 (I've confirmed this happens in 0.13.0 and does not happen in 0.12.3)

What

On alacritty 0.12.3, I could press the Pause key, and nothing would happen. Now, on Alacritty 0.13+, pressing the Pause key (and Scroll Lock, and presumably other keys I have not discovered yet) causes some weird spew to get emitted:

On v0.12.3:

$ target/debug/alacritty --version
alacritty 0.12.3 (5efb0690)
$ xdotool key Pause
$

And on v0.13.1 (note the ^[[57362u spew to the terminal and some keys (362u) getting typed without me actually typing them):

$ target/debug/alacritty --version
alacritty 0.13.1 (fe2a3c56)
$ xdotool key Pause; echo "done"
^[[57362udone
$ 362u

I performed a git bisect, and this behavior was introduced in cb03806. I tried reading through that diff, but it's beyond me. cc-ing the author (@kchibisov), and hoping he knows what's up.

Misc

I was (sort of*) able to work around this by adding a keyboard.binding to my alacritty conf file to ignore these keys:

[keyboard]
bindings = [
    { key = "Pause", action = "None" },
    { key = "Pause", mods = "Shift", action = "None" },
]

*I checked, and this causes these keypresses to be ignored by the kitty keyboard protocol implementation, which I don't want. Hopefully it doesn't matter, but if it's relevant, I can explain why I sometimes press keys such as Pause that I expect to have no visual output, yet I still want them to show up in the keyboard kitty protocol implementation.

@kchibisov
Copy link
Member

Pause is defined as part of the kitty keyboard protocol https://sw.kovidgoyal.net/kitty/keyboard-protocol/#functional-key-definitions

I remember reading that we should send them, since they are not naturally exposed, but I can't find the text anymore saying so, should probably just not send until the mode is set.

@jfly
Copy link
Author

jfly commented Jan 15, 2024

Interesting! Can you help me understand why I see this behavior with the Pause key but not an F key like F4?

@kchibisov
Copy link
Member

F4 has the original meaning of ^[OS, however those media never had them.

Do you have this issue in kitty terminal when you press the Pause key? If you don't see that I'll update to not emit the keys for these characters, but I think I've tested and it emitted the said chars.

@jfly
Copy link
Author

jfly commented Jan 15, 2024

F4 has the original meaning of ^[OS

Sorry, I know very little about terminal emulators and escape sequences. I don't really understand this. Can you direct me to some reading?

Do you have this issue in kitty terminal when you press the Pause key?

I just checked, and I do see this same behavior in kitty (v0.31.0).

I'm a bit confused though: AFAIK, I'm not using the keyboard kitty protocol here, right? So just because Alacritty got a new feature, why would the behavior of Alacritty change? Or does zsh (my shell) actually opt into this CSI 1 keyboard handling behavior and that's why I'm seeing a change in behavior?

@kchibisov
Copy link
Member

I think the point was that it defined key bindings for keys never used before and that are dedicated.

The other part of the spec was to define keys for keys not usually having a special meaning, like media keys or higher F keys, because they generally don't exist. Like there's no key for F13 for example, because it's a Shift+F1, some apps allow you to define it via F13 but it's still Shift+F1.

F4 could be just suppressed by zsh so you don't see it inputted in the prompt, but you can still see what it is if you press ctrl + v followed by f4 in zsh. And if you try to input into cat you'll get that sort of input.

@chrisduerr do we want to do anything with it given that kitty behaves the same? Like we just defined the values for rare keys.

@jfly
Copy link
Author

jfly commented Jan 15, 2024

F4 could be just suppressed by zsh so you don't see it inputted in the prompt, but you can still see what it is if you press ctrl + v followed by f4 in zsh. And if you try to input into cat you'll get that sort of input.

Ah, that makes sense! I also see it get inserted in insert mode in vim. Why don't I see any F4 stuff get inputted in other random repls (I tried both the python repl and the gdb repl. Do they both have special handling to ignore F4?)

It makes sense to me if you want to close this issue. I do now have a question (or possibly a feature request): is there some keycode that Alacritty would ignore/drop in "regular" modes (like how Pause used to behave), but would pass through when this new kitty keyboard protocol is enabled? Or maybe a different question: can you point me towards how I would go about reconfiguring zsh and vim to ignore these new rare keys (similar to how they ignore F4 right now)?

@kchibisov
Copy link
Member

I'm not sure you can ignore them in general without config in each app, I'd simply map them to None the way you did if you don't want bindings for them to appear in general. And keep in mind that F5 is not ignored because it has a bit different payload, and it's not even related to kitty keyboard since the value for F1-F12 are based on terminfo.

but would pass through when this new kitty keyboard protocol is enabled?

Numpad behaves differently, ESC alters its behavior once the mode is enabled, and some other keys. You can read the kitty keyboard protocol spec and get idea from it.

@chrisduerr
Copy link
Member

do we want to do anything with it given that kitty behaves the same?

How about other terminals that implement the protocol?

@jfly
Copy link
Author

jfly commented Jan 21, 2024

I went through the list of terminal emulators on https://sw.kovidgoyal.net/kitty/keyboard-protocol/ and generated a complete table of behavior.

For the record:

  • By "spew": I mean the ^[[57362u + 362u output described in the original post.
  • I don't know what the right term is for "normal" mode. By this, I just mean the behavior of the terminal emulator when zsh is running. If it would be useful, I'd be happy to gather more data for other modes (like whatever mode the terminal is in when vim/neovim is running, if that's even any different than zsh).
Terminal Emulator Pause key behavior in "normal" mode Pause key behavior when kitty keyboard protocol is enabled Testing Notes
Kitty spew key up, down, repeat all work
Alacritty spew key up, down, repeat all work
foot nothing happens key up, down, repeat all work only runs on wayland. i ran it under weston
WezTerm nothing happens nothing happens keyboard kitty protocol support is disabled by default, run with wezterm --config enable_kitty_keyboard=true
rio nothing happens nothing happens keyboard kitty protocol support is disabled by default. need to edit the config file to enable the kitty keyboard protocol (https://raphamorim.io/rio/docs/configuration-file/#keyboard)

So, there seem to be the "camps" of behavior, and rio alone behaves as I want.

I'm not sure you can ignore them in general without config in each app

@kchibisov, could you explain this a little more? This does feel like something alacritty could choose to do if it wanted: if the kitty keyboard protocol is enabled, then send events for these types of keys, and if it's not enabled, then do not. That feels like what rio is doing.

I certainly don't know what the "right" behavior should be here. I personally would prefer for things to behave like rio (again, happy to explain why, but the thing I'm trying to do is so me-specific, I don't think you'd find it to be a particularly useful data point). IIUC, since this is about the terminal emulator's behavior when the kitty keyboard protocol is not active then the specification does not apply?, perhaps it's just up to the terminal emulator to decide what should happen here (or perhaps there's some other specification that applies here?)

@kchibisov
Copy link
Member

The right behavior is what kitty does, since it's kitty protocol. Alacritty follows what kitty does and that's about it, you better of clarifying this with kitty author, I just implement the protocol to behave consistency as much as I can with its desired implementation, which is kitty implementation.

rio for example copied my initial impl until I adjusted it to match what actually kitty does.

@kchibisov
Copy link
Member

Like I can certainly exclude, but I'd rather bring this upstream.

@jfly
Copy link
Author

jfly commented Jan 21, 2024

I can definitely bring this discussion upstream, but before I do, can you help me understand how this is a kitty protocol question? Since this is about the terminal emulator's behavior when the kitty keyboard protocol is not active then how does the kitty keyboard protocol specification or the behavior of kitty even matter?

@kchibisov
Copy link
Member

kovidgoyal/kitty#7037

@chrisduerr
Copy link
Member

@kchibisov Arguably outside of Kitty protocol Alacritty should match XTerm.

@kchibisov
Copy link
Member

I don't think we do so already, unless you consider some subset of xterm.

I mean, the issue is mostly about "I pressed keys that don't do anything just because I want, but suddenly they insert something", so I'm not even sure what are we fixing here.

@chrisduerr
Copy link
Member

I don't think we do so already, unless you consider some subset of xterm.

We should mostly match URxvt, that's true. But it should always be either URxvt or XTerm. As far as I understand this new behavior matches neither and thus is somewhat inconsistent?

I mean, the issue is mostly about "I pressed keys that don't do anything just because I want, but suddenly they insert something", so I'm not even sure what are we fixing here.

Yeah I agree with that. I personally wouldn't really mind just closing this with "just don't press that button" as a solution. But I do think it makes some sense to consider our stance on intended character output in general. I'm personally not too fond of matching Kitty for anything other than the "Kitty protocol".

@kchibisov
Copy link
Member

Yeah I agree with that. I personally wouldn't really mind just closing this with "just don't press that button" as a solution. But I do think it makes some sense to consider our stance on intended character output in general. I'm personally not too fond of matching Kitty for anything other than the "Kitty protocol".

The thing is that users who don't want media key to work can just unbind them, but on the other side they can have them working. The thing about xterm/urxvt is probably that they haven't ever though about these keys in the first place?

@kchibisov
Copy link
Member

kovidgoyal/kitty@a9c7a85

So we can decide not to send.

@jfly
Copy link
Author

jfly commented Jan 22, 2024

From now on I will refer to "normal" mode as "legacy" mode as per today's change to the kitty keyboard protocol spec.

How do you feel about exposing this as a config option for alacritty? Perhaps via the keyboard.bindings action setting?

  • ReceiveChar: alacritty's current behavior, that is: sends the "spew" in legacy mode and sends the key events when kitty's keyboard protocol is enabled
  • None: the same behavior it currently has, that is: no action in both "legacy" and "kitty keyboard protocol" modes
  • LegacyModeNone: new feature. no action in legacy mode, sends key down/repeat/up in kitty keyboard protocol mode. In other words, this is rio's current behavior described above.
  • KittyKeyboardModeNone: new feature. I don't think this needs to exist, but I've listed it here for academic completeness. I'm not sure about the name. In legacy mode, the key would get sent, and in kitty mode, it would not.

I have a little Rust experience (and am looking for more), so I would be willing to try to implement something like this feature if you're open to it.

@kchibisov
Copy link
Member

Like a binding for extra mode? I don't really care, whatever @chrisduerr says here, given that not emitting bindings in normal mode is correct as well.

@chrisduerr
Copy link
Member

I don't think it makes any sense to partially support CSI u like this. As Kitty states, there is no encoding in the "legacy" mode. It's not like XTerm/URxvt didn't bother implementing, there's nothing to implement.

We skipped CSI u and went straight for the Kitty protocol because there's not much point going for some half-baked solution. So I don't see why we'd introduce that half-baked CSI u mode into our "legacy" handling.

@kchibisov
Copy link
Member

@chrisduerr besides, do we mask CSI u bindings from normal mode?

@chrisduerr
Copy link
Member

You mean never sending any CSI u stuff outside of Kitty mode? I think it would make sense to go for pre-protocol behavior yeah.

@kchibisov kchibisov added this to the Version 0.13.2 milestone Jan 23, 2024
@taliaferro
Copy link

Adding to this -- I found that NumpadEnter is also being interpreted incorrectly as control-C in 0.13.0 and above, is that related?

kchibisov added a commit to kchibisov/alacritty that referenced this issue Jan 30, 2024
Originally kitty defined that functional keys, which are not encoded
by default, like `Pause` should be encoded with `CSI u`. However
the specification was clarified and now it says that terminal
may ignore that part. Given that Alacritty tries to follow xterm/urxvt
when it comes to bindings, CSI u bindings are not send for consistency
reasons.

This also brings back F13-F20 bindings used by Alacritty in 0.12.3, as
well as explicitly defines `NumpadEnter` like it was before.

Closes alacritty#7623.
kchibisov added a commit that referenced this issue Jan 30, 2024
Originally kitty defined that functional keys, which are not encoded
by default, like `Pause` should be encoded with `CSI u`. However
the specification was clarified and now it says that terminal
may ignore that part. Given that Alacritty tries to follow xterm/urxvt
when it comes to bindings, CSI u bindings are not send for consistency
reasons.

This also brings back F13-F20 bindings used by Alacritty in 0.12.3, as
well as explicitly defines `NumpadEnter` like it was before.

Closes #7623.
@jfly
Copy link
Author

jfly commented Jan 30, 2024

Thanks, @kchibisov!

chrisduerr pushed a commit to chrisduerr/alacritty that referenced this issue Mar 15, 2024
Originally kitty defined that functional keys, which are not encoded
by default, like `Pause` should be encoded with `CSI u`. However
the specification was clarified and now it says that terminal
may ignore that part. Given that Alacritty tries to follow xterm/urxvt
when it comes to bindings, CSI u bindings are not send for consistency
reasons.

This also brings back F13-F20 bindings used by Alacritty in 0.12.3, as
well as explicitly defines `NumpadEnter` like it was before.

Closes alacritty#7623.
chrisduerr pushed a commit that referenced this issue Mar 19, 2024
Originally kitty defined that functional keys, which are not encoded
by default, like `Pause` should be encoded with `CSI u`. However
the specification was clarified and now it says that terminal
may ignore that part. Given that Alacritty tries to follow xterm/urxvt
when it comes to bindings, CSI u bindings are not send for consistency
reasons.

This also brings back F13-F20 bindings used by Alacritty in 0.12.3, as
well as explicitly defines `NumpadEnter` like it was before.

Closes #7623.
xiaogaozi added a commit to xiaogaozi/alacritty that referenced this issue May 24, 2024
* Bump development version to 0.13.0-dev

This is only an update to the development version and does not represent
a stable release.

* Update INSTALL.md Ubuntu instructions

* Add missing `ToggleMaximized` docs in the config

* Add `window.resize_increments` config option

Given how bugged the resize increments are on X11, it's better to
disable it by default.

* Relax horizontal scrolling

Apply horizontal scrolling when the angle between the axis X
and (x, y) vector is lower than 25 degrees.

Fixes alacritty#6711.

* Fix bracketed paste with EOT payload

This works around an issue in many (all?) shells where the bracketed
paste logic would only strip out `\r` but interpret EOT (`\x03`) as a
termination of the bracketed paste.

* Send D&D input through paste action

Treating D&D like paste allows using D&D to input text into areas other
than the PTY, like the search bar.

* Fix busy rendering when the same warning persists

When the same warning is thrown on the each rendering iteration, it'll
force alacritty to always render.

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Bump winit to 0.28.2

This also bumps copypasta and sctk-adwaita.

Fixes alacritty#6744.
Fixes alacritty#6702.
Fixes alacritty#6696.
Fixes alacritty#2741.

* Bump x11-clipboard to v0.7.1

Fixes alacritty#6760.

* Use `thin` instead of `fat` LTO

The performance between `thin` and `fat` is in the margin of error,
however `thin` LTO is way faster to build.

* Improve fish completions installation instructions

Co-authored-by: Kirill Chibisov <contact@kchibisov.com>

* Bump winit to 0.28.3

Fix macOS leak when closing the window.

* Fix `;` character in URI OSC 8 payload

The special character `;` can be not URL-encoded, thus it'll add
extra parameter in the payload. Handle it joining extra parameters
with the `;` as a separator.

* Bump CHANGELOG development version to 0.13.0-dev

This is only an update to the development version and does not represent
a stable release.

The original bump to 0.13.0-dev was made in 9cf2c89, but did not make
any changes to the changelog. This adds a 0.13.0-dev changelog entry to
master so new changes are properly added to the correct place.

* Fix selection rotation on the last line

This fixes an issue with terminal resizes when the selection is on the
last line. Alacritty would fail to rotate lines and keep the selection
in the same line index whenever the terminal line count was grown or
shrunk.

This issue occurred due to the range passed to the selection's rotate
function still being based on the old terminal size, which caused the
initial or target state of the rotation to be outside of the terminal
bounds.

Closes alacritty#6698.

* Add missing terminfo capabilities

This adds capabilities for focus in/out (XF, kxIN, kxOUT) and bracketed
paste (BD, BE, PE, PS).

* Reset char suppression for every key binding

Previously the character suppression was only reset whenever a key was
released. However this did not take key repetition into account.

Now every key down also resets the character suppression. This should
work since the `ReceivedCharacter` is always received immediately after
the `KeyboardInput` without the chance of a racing condition where
another keyboard event interrupts the two.

* Use paste for ESC action and IME commit

Route string terminal input through 'ActionContext::paste'
instead of char by char write improving performance by
utilizing bracketed paste mode when it's reasonable.

* Allow mode-exclusive bindings in any mode

This patch enables binding chains that go beyond mode changes by
allowing bindings to be defined for modes they do not usually have an
effect in.

* Omit urgency hint focused window

* Bump winit to 0.28.4

Fixes alacritty#6817.

* Add `ALACRITTY_EXTRA_LOG_TARGETS` env variable

This should help with troubleshooting the dependency crates like winit.

* Bump winit to 0.28.5

The 0.28.4 was yanked.

* Fix very long startup time on some Wayland systems

This is not a real fix for the issue given that dbus method sctk-adwaita
is using will being called anyway. The proper fix will be available
with the winit's 0.29.0 release.

Right now the delay reduced from around 20 seconds to 100ms on a
systems with broken xdg-desktop-portal setup.

* Bump `actions/checkout` to v3

Fix Node 12 deprecation warnings in action runs.

* Update glutin to 0.30.8 and winit to 0.28.6

Fixes alacritty#6920.
Fixes alacritty#6824.

* Fix upload from CI on windows

* Update bitflags to 2.2.1

* Update xdg and dirs crate

The recent version on xdg crate moved to home crate which is used
by cargo. Thus to query for home directory we can use the home
dependency instead which is smaller.

* Update clap to v4.2.7

Fixes alacritty#6879.
Fixes alacritty#6874.

* Update patch dependencies

* Improve renderer debuggability

Make the renderer more debuggable by adding extra logging and using
`GL_KHR_debug` when running with the debug log level.

* Add version 0.12.1 to CHANGELOG

* Switch to VTE's built-in ansi feature

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Fix 216-color cube generation

This fixes a regression introduced in cb7ad5b which swapped the green
and blue values when constructing the 216-color RGB cube.

Fixes alacritty#6951.

* Fix hyperlink preview for 2 lines terminal (alacritty#6953)

The intention was to show it, however it was hidden due to wrong
comparisson check.

* Move manpages to scdoc

This rewrites the existing manpages to use the `scdoc` format, making it
simpler to read and edit the manpages without intricate roff knowledge.

Some minor changes to the manpages were made in the process, mostly
focusing on correcting some of the wording. The list of maintainers has
also changed to ensure people not involved in the project anymore aren't
unnecessarily contacted for support.

* Fix Makefile manpage generation

Signed-off-by: Dennis Maier <d.maier94@web.de>
Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Fix crash on ScaleFactorChange on Windows

Windows is known to send zero sizes from winit in Risezed and now
in ScaleFactorChanged events. They were handled in Resized, but
not in ScaleFactorChanged.

Fixes alacritty#6949.

* Bump dependencies

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Add resolver to workspace manifest

This explicitly specifies resolver "2" in the root manifest, to fix a
warning introduced in a recent nightly release.

* Switch to TOML configuration format

This switches Alacritty's default configuration format from yaml to
toml. While yaml is still supported, it is done by converting it to toml
and should be removed entirely in the future.

All existing features were persisted based on my testing. Behavior
should not change much, though `--option` might have slightly different
behavior since the entire line is not interpreted as one line of toml.

A new `alacritty migrate` subcommand has been added which allows
automatic migration from yaml to toml. This also could be used as a
facility to automatically fix configuration file changes in the future.

Closes alacritty#6592.

* Change the way we refer to IRC chat

* Fix warnings for config imports

Closes alacritty#6996.

* Add option to persist hints after selection

Closes alacritty#6976.

* Fix `alacritty msg config` toml replacement

This fixes a regression introduced in bd49067 which broke the override
of configuration file variables using `alacritty msg config`.

To fix this the `replace` functionality was rewritten to behave more
like the `serde_utils::merge` where entire values are inserted into the
existing structure rather than separating the keys from the values.

Fixes: bd49067 (Switch to TOML configuration format)

* Fix freeze with some drivers when using GLX

Fixes alacritty#6972.

* Fix parsing of integer font sizes

Config file conversion broke parsing of the font size value if it's
written as an integer, since TOML integers are always signed.

* Add errors for deserializing with unknown fields

Currently there are still some places where `Deserialize` is used rather
than `ConfigDeserialize`, which means that the built-in warning for
unused fields is not emitted automatically.

To ensure users don't have invalid configurations, the
`#[serde(deny_unknown_fields)]` annotation has been added to these
structs, making it a hard error when an unknown field is present.

* Fix the crash when shrinking scrolled terminal

display_offset was adjusted unconditionally, thus it could go
beyound the history limits, so clamp it to history like we do
in grow_colums.

Fixes alacritty#6862.

* Add support for loading conpty.dll

Co-Authored-By: @fredizzimo
Co-Authored-By: @Grueslayer

* Update wix on the release CI

Fixes alacritty#6939.

* Add TOML migration recommendation to warning

This adds a little recommendation to use `alacritty migrate` to
automatically transition configuration files from YAML to TOML.

* Remove mouse double_click/triple_click options

Fixes alacritty#6962.

* Change the default colorscheme

The new colorscheme is base16 classic dark with the bright colors
generated with oklab toolkits. The base16 classic dark is less washed
out and represents the current maintainers preference. The motivation
to change it was subjective, though it does look like generic dark
theme.

On a side note, this colorscheme was used for alacritty.org web page for
a long time, however it used different foreground color.

* Add manpage links to "see also" section

* Add version 0.12.2 to CHANGELOG

This is only an update to the development version and does not represent
a stable release.

* Add a note in README about older versions

Unfortunately it's not clear to everyone that they are looking at the
development documentation.

* Fix legacy bindings taking precedence over new ones

They were compared by len, but it's was wrong from the start, since
a user provided binding could remove more than one builtin binding,
so it was impossible for users to use their own bindings.

The most reliable way to do so is to use `Option`, given that we fill
default during deserialization.

Fixes alacritty#7050.

* Update binding sections in config manpage

* Create man5 directory along with man1

* Update to the new winit keyboard API

The main highlight of this update is that alacritty will now use new
keyboard API from the winit, which resolves a lot of issues around
key bindings, such as ability to bind dead keys. It also fixes long
standing issues with the virtual key code bindings and make bindings
in general more predictable. It also makes our default Vi key bindings
fully working.

Given that alacritty was using `VirtualKey` directly in the bindings
from the winit, and winit simply removed the enum, we've added internal
conversions to minimize the fallout, but new way to specify the bindings
should be more intuitive.

Other part of this update fixes some forward compatibility bugs with the
Wayland backend, given that wayland-rs 0.30 is fully forward compatible.
The update also fixes weird Maximized startup issues on GNOME Wayland,
however they were present on any sane compositor.

Fixes alacritty#6842.
Fixes alacritty#6455.
Fixes alacritty#6184.
Fixes alacritty#5684.
Fixes alacritty#3574.
Fixes alacritty#3460.
Fixes alacritty#1336.
Fixes alacritty#892.
Fixes alacritty#458.
Fixes alacritty#55.

* Fix configuration link typo

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Fix `window.option_as_alt=Both`

Fixes alacritty#7077.

* Unset `XDG_ACTIVATION_TOKEN` in alacritty_terminal

This variable is what being used for Wayland's activation stuff.

* Raise double click threshold to 400ms

This should improve the situation with some touchpads. GTK4 is also
using the same value.

* Prefer logical key for named keys

Some keyboard layouts have named logical keys via shift combinations
of some sorts. So prefer them.

Fixes alacritty#7076.

* Mention all the actions in the man page

A lot of actions were left without a notice, so the only way to figure
out some of them was to browse source code. An example of such
actions were `Maximize` and `Minimize`, since we don't have a binding
for them, so they were left even in alacritty-bindings(5).

Explicitly list all the bindings we have. The search bindings were
also not accurately restricted, since we allow them to be executed
inside regardless of mode.

* Use lowercase latters in bindings for h/m/l

They were using uppercase latters, however our config expects everything
as lowercase, because we're using `key_without_modifiers` +
`ModifiersState`.

* Remove unnecessary mutable references

* Fix license syntax

This patch updates the license field to follow the SPDX 2.1 license
expression standard.

* Use bracketed paste only for multi-char IME input

Some IME setups do only `commit` single char input, like fcitx5 when
doing ru input.

* Fix manpage docs for shell section

Closes alacritty#7087.

Co-authored-by: Kirill Chibisov <contact@kchibisov.com>

* Fix keys like `+` not working on neo layouts

The key_without_modifier removes all the modifiers including the
multiple shift levels, which is not desired. In alacritty we
just wanted to treat uppercase and lowercase latters the same,
which we can with the help of builtin functions.

* Mention `Forward`/`Back` mouse buttons in man page

They were mentioned only in the changelog, but forgotten in the man
page.

* Add `terminal` config section to control OSCs 

Some environments demand certain OSC sequences to be disabled or
some escape sequence could require handling which is out of scope
of alacritty, but could be done by external script (OSC 777).

Added section for now just handles the `OSC 52` sequence and changes
its default to be `OnlyCopy`, which is handy for remote copy, but
`Paste` is redundant because normal `Paste` hotkey could be used as
well.

Fixes alacritty#3386.

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Skip whitespaces for wide chars in preedit

While we skip the spacers for the wide characters in the grid due to
them having a proper flags, the draw_string method was generating the
cells with incorrect flags leading to wide chars being cut off.

* Make URL_REGEX more strict

* Use ahash instead of fnv and regular hash function

After evaluation of the ahash with the data alacritty uses it was
discovered that it's 1.5-2x times faster when getting the already
hashed values, which is the primary cases for alacritty's renderer.

Given that ahash is generally faster, all the HashSet and HashMap's
inside the alacritty were changed to use it as a hasher function.

* Ignore scrolling multiplier on touchscreens

* Apply transparent background colors to "UI" cells

* Copy global IPC options for new windows

This patch stores all options set for the Window ID `-1` and
automatically applies them to new windows after their creation.

This in theory makes it possible to have a fully dynamic "default
config" without having to reapply it for every new window.

Closes alacritty#7128.

* Fix decorations_theme_variant spelling in manpage

* Remove winit dependency from alacritty_config

* Update regex-automata to v0.3.6

This seems like a sensible first step before looking into alacritty#7097.

* Fix crash with anchored searches

While this does **not** enable the use of anchors (`^`) in user regexes,
it does prevent Alacritty from crashing when attempting to do so.

* Change `window.padding` from u8 to u16

Fixes alacritty#6900.

* Quote `None` consistently inside the man page

`None` must be inside the `"`, since it's a string value.

* Add examples to more sections

Some sections were provided without a default, so provide an example for
them.

* Update winit to 0.29.1-beta

Make use of new winit frame throttling mechanism used in RedrawRequested,
which removes the need for having Wayland queue to ask for the frame
callbacks.

Fixes alacritty#7011.

* Add bindings for macOS tabs

This doesn't represnet the movement to add tabs on any other platform,
unless winit could add a similar API for them.

* Support startup notify on Wayland/X11

Activate a window to indicate that we want initial focus when the
system uses startup notifications.

Fixes alacritty#6931.

* Add `prefer_egl` debug option

Some systems have rendering issues when using GLX rather than EGL. While this is
usually due to a driver bug, it is helpful to provide a workaround for this by
allowing people to prefer EGL over GLX.

This patch adds the new `debug.prefer_egl` option to provide this workaround.

Closes alacritty#7056.

* Remove obsolete ansicode.txt file

The wiki contains a list of links to ANSI references:
https://github.com/alacritty/alacritty/wiki/ANSI-References

* Test man-pages compilation on the CI

* Add missing position config docs

* Fix regex memory usage

This fixes an issue where regexes with a large number of possible states
would consume excessive memory, since the entire DFA was compiled ahead
of time.

To solve this, the DFA is now built at runtime using `regex-automata`'s
hybrid DFA.

There are however still some checks performed ahead of time, causing
errors with obscenely large regexes (`[0-9A-Za-z]{999999999}`), which
shouldn't cause any issues.

A regex which is large, but not large enough to fail the NFA
construction (like `[0-9A-Za-z]{999999}`) will cause a long search of
the entire grid, but will complete and show the match.

Closes alacritty#7097.

* Underline hint matches during selection

This patch underlines the full regex hint match while the keyboard hint
selection is in process.

While it would be possible to color the entire match, this would only
introduce unnecessary configuration options and be too noisy. The
underline matches the mouse highlighting and has a less drastic visual
impact.

Closes alacritty#6178.

* Bump VTE to 0.12.0

Fixes alacritty#6845.

* Update the escape_support.md for sync updates

Remove the DCS sync updates escape sequence since it's no longer
supported.

Fixes: 47d5007 (Bump VTE to 0.12.0)

* Port from mio to polling

This patch replaces the mio crate with the polling. Now that
smol-rs/polling#96 has been merged, we should be at full feature parity
with mio v0.6 now.

Fixes alacritty#7104.
Fixes alacritty#6486.

* Update dependencies

This patch applies all breaking and non-breaking dependency updates
and bumps MSRV to 1.70.0.

* Fix PTY being closed on creation

The PTY got closed because `OwnedFd` automatically closed it.

Fixes: 59c63d3 (Update dependencies)

* Fix regex matches ending on multiline

This fixes an issue where the reverse search for the regex start would
truncate a character when ending on a newline, since it was omitting the
EOI check in that case.

This also fixes a separate issue which caused regexes which capture
empty strings (e.g.: `.*`) to always report a match.

This is a regression introduced in 73276b6.

* Fix regex patterns allowing for empty strings

This patch changes the mode we search for patterns which allow an empty
string, by anchoring all searches. As a result we will match the longest
possible match when multiple patterns are present (like `;*|rust`),
instead of using the leftmost pattern only.

This is only possible with empty matches since our parser is reset on
every byte anyway, so anchoring the search makes no difference.

Fixes alacritty#7276.

* Use openpty-rustix instead of nix

Follow upstream libraries and use rustix to reduce the amount of
dependencies in the future.

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Fix `window.decorations_theme_variant` reload

The live reload handling wasn't introduced when the option got added.

Fixes alacritty#7295.

* Add inline vi mode search

This patch adds inline search to vi mode using `f`/`F` and `t`/`T` as
default bindings. The behavior matches that of vim.

Fixes alacritty#7203.

* Bump rustix to 0.38.20

* Update winit to 0.29.2 and copypasta to 0.10.0

Fixes alacritty#7236.
Fixes alacritty#7201.
Fixes alacritty#7146.
Fixes alacritty#6848.
Fixes alacritty#3601.
Fixes alacritty#3108.
Fixes alacritty#2453.

* Fix crash due to wrong drop order of clipboard

Fixes alacritty#7309.

* Prefer exact matches for bindings in mouse mode

Only consider bindings without Shift if there are no actions defined for the
actual mouse event.

Closes alacritty#7292.

* Fix typos

* Add version 0.12.3 to CHANGELOG

This is only an update to the development version and does not represent
a stable release.

* Avoid maximizing window when creating new tab

This patch ignores the startup mode when creating a new tab on macOS to
avoid maximizing an existing window.

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Fix clippy warnings

* Add `window.blur` config option

Fixes alacritty#972.

* Add man 5 pages to upload_asset.sh

They are compiled, but not being uploaded.

* Bump glutin to 0.31.1

This fixes a crash on startup with macOS Sonoma.

* Fix crash when leaving search after resize

This fixes a crash which could occur when leaving search with a visible
match after shrinking the terminal height to be lower than the original
line the focused match was in.

Closes alacritty#7054.

* Bump ahash to 0.8.6

* Remove `alacritty_config` from alacritty_terminal

There's no need to force alacritty's user configuration on
other users of the crate, thus provide the options actually used
by alacritty_terminal itself.

* Fix cursor being hidden after reaching timeout

The timeout and blink events could be delivered at the same time,
so canceling blinking won't work and we'll still have an event.

* Use builtin font to draw powerline symbols

In addition to box drawing it was decided to also draw powerline
symbols, since those are quite common and rather simple to draw with
present box drawing infra.

* Unify CLI config override mechanisms

This patch changes the way the `-o` config option works when specified
at startup to function the same way as the IPC mechanism.

While this should technically perform the exact same way, it should
hopefully make it a little easier to understand how CLI config
replacement works.

* Add `--option` argument to `create-window`

This patch adds a new CLI parameter to the `create-window` subcommand,
matching the existing `--option` parameter when creating a new Alacritty
instance.

This parameter allows setting up the initial window configuration from
the CLI without having to call `alacritty msg config`, making sure that
all options are set appropriately right from the start.

Closes alacritty#6238.

* Change default `bell.animation` to `Linear`

The default animation feels really choppy, but it's just how its
function looks.

* Fix visual bell getting stuck on macOS

Fixes alacritty#7325.

* Fix message bar damage

Fixes alacritty#7224.

* Fix Sync capability in terminfo

Alacritty has supported mode 2026 for synchornized updates for a few
months, but the terminfo entry still used the old DCS sequence
originally supported by iTerm2. Since many other terminal emulators and
applications seem to be standardizing around 2026, change the terminfo
entry to use SM instead of DCS.

* Fix Vi cursor not being dirty when scrolling

* Simplify powerline drawing algorithm

Iterate over points in line instead of drawing it right away
and then finding it in the buffer.

Fixes: 4a26667 (Use builtin font to draw powerline symbols)

* Add error handling for OpenGL connection details

* Update VTE to 0.13.0

* Add support for DECRPM/DECRQM

* Damage only terminal inside `alacritty_terminal`

The damage tracking was including selection and vi_cursor which were
rendering viewport related, however all the damage tracking inside
the `alacritty_terminal` was _terminal viewport_ related, meaning that
it should be affected by `display_offset`.

Refactor the damage tracking so `alacritty_terminal` is only tracking
actual terminal updates and properly applying display offset to them,
while `alacritty` pulls this damage into its own UI damage state.

Fixes alacritty#7111.

* Create only one branch per major release

Having a separate branch for each release makes it harder to maintain
without an actual benefit, since every release from the major version
is linear, so creating branches doesn't make any sense.

They also collapse with the tag names leading to ambiguous refs.

* Make sections the same case in man pages

* Fix SelectLastTab in alacritty-bindings(5)

* Make man pages look consistent

Part of the man pages were using _value_, other part were using just
value, thus make everything as _value_.

Also properly dot terminate and add spacing through out the man pages.

* Explicitly use `=` in alacritty(5)

This should help with understanding how to write TOML
without reading too much into specification.

* Improve section to TOML section refs

This should give more hints where to put things.

* Add example for [window] section in alacritty(5)

This should give a hint on how to generally tweak things.

* Drop repo link in *SEE ALSO* man sections

The alacritty documentation is fully present inside the man pages, the
repo provides nothing extra other than a way to report issues, which
lives in its own section anyway.

* Fix default for [[hints.enabled]] in man pages

The default was incomplete, since `command` was missing.

* Add migrate to alacritty(1)

* Damage entire window on font size change

Font size could change without changing the cell dimensions, like
becoming slightly higher/wider.

Fixes: 40160c5 (Damage only terminal inside `alacritty_terminal`)

* Fix IME popup positioning

When setting cursor area, the popup will be placed either above or
below not obscuring the supplied region, however we were still
offsetting line with `+1` putting the cursor at the bottom of the line,
and given that area is from the top-left corner, the wrong area
was marked for not being obscured.

It was also discovered that some compositors, like GNOME, position
IME in the bottom right corner of the supplied region, which is wrong,
but it renders IME popup not very useful, since it's far away from
the place it should be, thus try to not obscure just a few characters
near the cursor.

Given that X11 doesn't support area setting, it uses the old logic
with offsetting.

Co-developed-by: lilydjwg <lilydjwg@gmail.com>
Signed-off-by: lilydjwg <lilydjwg@gmail.com>

* Bump dependencies

Update dependencies with `cargo update` as well as do explicit update
of winit, crossfont, and regex-automata.

* Make AA stronger for undercurl

This improves undercurl rendering preserving its original thickness.
This also makes it look not out-of place when places next to other
lines.

* Document macOS shortcuts for search

* Adjust default Vi's SearchBackwards binding

We have the same policy with Shift for numbers.

* Fix DECRPM reporting

The DECRQM uses `p` to query, but the reply uses `y`.

Fixes alacritty#7397.

* Optimize undercurl shader

This removes the if and lowers amount of operations.

* Fallback to underline shader when dotted fails

Some hardware is just bad.

Fixes alacritty#7404.

* Don't use `::*` for enum variants in bindings

* Fix chars usage inside the mouse bindings

Fixes alacritty#7413.

* Implement kitty's keyboard protocol

The protocol enables robust key reporting for the applications, so
they could bind more keys and the user won't have collisions with
the normal control keys.

Links: https://sw.kovidgoyal.net/kitty/keyboard-protocol
Fixes alacritty#6378.

* Update to crossfont 0.6.0

* Fix trigger of normal bindings in mouse mode

We should ensure that the `Shift` is actually pressed when trying to
prefer regular bindings instead of the ones if we had Shift applied.

Fixes: 500b696 (Prefer exact matches for bindings in mouse mode)
Fixes alacritty#7415.

* Don't emit text for NamedKey without text repr

When the key doesn't have textual representation we shouldn't emit
the text for them, since they are processed via bindings.

Also, fix the logic to handle named keys with disambiguate without
special modes/modifiers.

Fixes alacritty#7423.

* Fix message bar not damaged when the same size

The regression was added due to `y` coordinate in OpenGL differs to
`y` inside the damage rectangles.

Fixes: 40160c5 (Damage only terminal inside `alacritty_terminal`)

* Bump crossfont to 0.7.0

* Alacritty version 0.13.0-rc1

* Add link to rendered config docs

This adds a link to alacritty.org's rendered configuration file
documentation.

* Fix the order of `ClearSelection` in Vi mode

Fixes alacritty#7438.

* Account for option_as_alt when doing kitty protocol

By default `Alt` is not a real `Alt` on macOS, so we shouldn't treat
it as a modifier.

Fixes alacritty#7443.

* Bump winit to 0.29.5

Fixes alacritty#7449.

* Keep IME always enabled on X11

Fixes alacritty#7195.

* Alacritty version 0.13.0-rc2

* Apply modifiers before presses in kitty protocol

While this doesn't handle releases with multiple identical modifiers
pressed, the release can't work reliable anyway, since one modifier
could be pressed before focusing the window, thus tracking modifiers
based on the keysym values won't work as it was suggested by kitty
author.

Links: kovidgoyal/kitty#6913

* Bump winit to 0.29.7

This also bumps other dependencies along the way.

Fixes alacritty#2886.

* Alacritty version 0.13.0

* Mark SimpleFullscreen as macOS only startup mode

* Derive `Clone` for `EventLoopSender`

* Fix `alacritty migrate` with nonexistent imports

Fixes alacritty#7473.

* Remove direct dependency on once_cell

With MSRV 1.70, std now contains the necessary parts.

* Use pre-composed key for `Alt` bindings on macOS

Fixes alacritty#7475.

* Fix inability to bind `Alt+Control` on Windows

Fixes alacritty#7506.

* Passthrough potential errors for `EventLoopSender`

* Don't substitute `\n` in char bindings

This broke unintentionally due to routing paste-like input
via paste function.

Fixes alacritty#7476.

* Send associated text for shifted numbers with kitty

Also fix the wrong ordering of base and shifted keys.

Fixes alacritty#7492.

* Powerline glyphs being cut for narrow fonts

Fixes alacritty#7470.

* Remove note about versions pre 0.13.0

* Bump winit to 0.29.8

Fixes alacritty#7514.
Fixes alacritty#7502.
Fixes alacritty#7494.
Fixes alacritty#7474.
Fixes alacritty#7472.

* Fix number-based mouse bindings

The toml migration introduced a regression which stopped numbered key
binding's from working. This patch implements the required number type
to make things work again.

Fixes alacritty#7527.

* Fix replacing optional fields

This fixes an issue with the default `SerdeReplace` implementation where
it would never recurse through options but always replace the entire
option with the new value.

Closes alacritty#7518.

* Add vi search paste bindings

Closes alacritty#7511.

* Ignore null values in `alacritty migrate`

This should help with broken YAML configurations by throwing nulls
away, which are not representable in toml.

* Bump winit to 0.29.9

Fixes alacritty#7559.
Fixes alacritty#7533.

* Fix serde tests without default features

* Alacritty version 0.13.1-rc1

* Fix manpage string escapes

Co-authored-by: Christian Duerr <contact@christianduerr.com>

* Alacritty version 0.13.1

* Reduce allocations during keyboard input

* Fix `debug.renderer="gles2pure"` documentation

Patch 5685ce8 changed the allowed values of the `debug.renderer` enum,
prohibiting the usage of `_` in the `Gles2Pure` variant. This patch
updates the documentation to correct for that.

Co-authored-by: Kirill Chibisov <contact@kchibisov.com>

* Fix inline search expanding across newlines

Closes alacritty#7587.

* Fix env variable overrides through CLI

This fixes an issue where all CLI environment variables would replace
existing configuration file variables instead of merging the two maps
together.

Fixes alacritty#7618.

* Fix typo in config docs

Closes alacritty#7647.

* Allow specifying all config keys on all platforms

Closes alacritty#7592.

* Use builtin font to draw sextants

Sextants are similar to quadrants and should align with them and
other box drawing, thus use builtin font to align them properly.

Part-of: alacritty#7422.

* Fix invalid bounds in selection range conversion

* Don't report associated text only for C0/C1

This has a side effect that we'll have text reported for Alt+Shift+T
and similar, but only C0/C1 should be excluded and Alt+Shift+T is
emitting neither, thus regular `T` will be reported.

Fixes alacritty#7657.

* Move CHANGELOG entry for sextants to proper section

* Don't use kitty sequences outside protocol

Originally kitty defined that functional keys, which are not encoded
by default, like `Pause` should be encoded with `CSI u`. However
the specification was clarified and now it says that terminal
may ignore that part. Given that Alacritty tries to follow xterm/urxvt
when it comes to bindings, CSI u bindings are not send for consistency
reasons.

This also brings back F13-F20 bindings used by Alacritty in 0.12.3, as
well as explicitly defines `NumpadEnter` like it was before.

Closes alacritty#7623.

* Document `command` field in bindings

Closes alacritty#7594.

* Fix unnecessary explicit panic in PTY 

Closes alacritty#7680.

* Fix hang on startup with some Wayland compositors

Fixes alacritty#7665.

* Fix hang on startup with some Wayland compositors

Fixes alacritty#7665.

* Add default `Home`/`End` bindings for Vi mode

* Remove extra TIOCSWINSZ ioctl on startup

The openpty call already performs it, thus no need to call it one more
with the exact same size since it confuses some applications.

* Fix row indexing with inclusive ranges

* Use dynamic MSRV for oldstable CI

Instead of manually specifying the oldstable version in all our CI
scripts, it is now pulled from the `Cargo.toml` which simplifies the
update process.

The contributing guide has also been updated to not include the explicit
version and its wording has been loosened a bit to correctly represent
current maintenance practices.

* Bump MSRV to 1.72.0

* Fix clippy warnings

* Fix regional scrolling leaking into history

This fixes an issue where a scrolling region that does not start at the
top of the screen would still rotate lines into history when scrolling
the content "upwards".

* Fix typo in config docs

* Bump winit to 0.29.11

Fixes alacritty#7633.
Fixes alacritty#7613.
Fixes alacritty#7607.
Fixes alacritty#7571.
Fixes alacritty#7549.

* Fix feature = "cargo-clippy" deprecation

* Check alternative cursor icon names on Wayland

* Add alias support to `SerdeReplace`

* Bump winit to 0.29.12

The 0.29.11 was yanked.

* Fix build failure on NetBSD

x11-clipboard was unconditionally using eventfd which is not present
on NetBSD.

Links: quininer/x11-clipboard#48

* Fix log typos

* Bump winit to 0.29.14

Fixes alacritty#7806.

* Expose more process info on Windows

* Send exit code events on child process exit

Fixes alacritty#7753.

* Set PTY's pixel size on startup

117719b removed the extra call for TIOCSWINSZ, however the initial
`openpty` call itself did not set the pixel size, which caused issues
with some clients.

* Fix hint `Select` action for hyperlink escape

This fixes an issue where the `Select` action for hyperlink escape text
would select the entire line, instead of selecting only the hyperlink
itself.

It also changes the way hyperlinks with the same ID are highlighted,
removing the restriction of being on consecutive lines and instead
highlighting all visible cells that correspond to the matching
hyperlink.

Closes alacritty#7766.

* Bump winit to 0.29.15

* Bump dependencies

This bumps all dependencies that can be updated without introducing a
build failure.

* Drop MSRV to 1.70.0

* Allow setting terminal env vars via PTY options

Closes alacritty#7778.

* Alacritty version 0.13.2-rc1

* Fix kitty encoding used for char input without text

On Windows some key combinations for regular text input, like Ctrl+1
don't have any text attached, so they were generating the kitty escape
sequence even when they shouldn't.

* Fix msi installer build

This works around an issue where wix was pulling pre-release extensions
and thus breaking compatibility with our used wix version.

* Alacritty version 0.13.2

* Fix build

---------

Signed-off-by: Dennis Maier <d.maier94@web.de>
Signed-off-by: lilydjwg <lilydjwg@gmail.com>
Co-authored-by: Kirill Chibisov <contact@kchibisov.com>
Co-authored-by: Nathan Lilienthal <nathan@nixpulvis.com>
Co-authored-by: Christian Duerr <contact@christianduerr.com>
Co-authored-by: CRATO LABS <127831631+cratolabs@users.noreply.github.com>
Co-authored-by: K.B.Dharun Krishna <kbdharunkrishna@gmail.com>
Co-authored-by: Anhad Singh <62820092+Andy-Python-Programmer@users.noreply.github.com>
Co-authored-by: maierd <d.maier94@web.de>
Co-authored-by: Pavel Roskin <1317472+proski@users.noreply.github.com>
Co-authored-by: Sonu Bardai <67749330+SonuBardai@users.noreply.github.com>
Co-authored-by: Yug Shende <yugshende@gwu.edu>
Co-authored-by: Nick Reilingh <nreiling@bard.edu>
Co-authored-by: Andrea Frigido <frisoft@users.noreply.github.com>
Co-authored-by: Kirill Bulatov <mail4score@gmail.com>
Co-authored-by: tv <tv@krebsco.de>
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
Co-authored-by: Arhum Jain <75106472+ArhumJain@users.noreply.github.com>
Co-authored-by: Rolf Sievert <26115145+RolfSievert@users.noreply.github.com>
Co-authored-by: Simone Ragusa <simoneragusa@duck.com>
Co-authored-by: John Nunley <jtnunley01@gmail.com>
Co-authored-by: Kyle Willmon <kylewillmon@gmail.com>
Co-authored-by: Julian Braha <julianbraha@gmail.com>
Co-authored-by: Gregory Anders <8965202+gpanders@users.noreply.github.com>
Co-authored-by: Cédric Van Rompay <cedric.vanrompay@gmail.com>
Co-authored-by: Hyper <HyperHamster@users.noreply.github.com>
Co-authored-by: Philip Dubé <serprex@users.noreply.github.com>
Co-authored-by: Lars Francke <lars.francke@stackable.tech>
Co-authored-by: Omar Magdy <99906646+omagdy7@users.noreply.github.com>
Co-authored-by: James McCoy <jamessan@jamessan.com>
Co-authored-by: WuerfelDev <WuerfelDev@users.noreply.github.com>
Co-authored-by: Popa Ioan Alexandru <49609151+ALEX11BR@users.noreply.github.com>
Co-authored-by: Mohammad AlSaleh <CE.Mohammad.Alsaleh@gmail.com>
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Co-authored-by: Alexandru Placinta <placintaalexandru1@gmail.com>
Co-authored-by: Yadi Abdalhalim <36752216+Ydot19@users.noreply.github.com>
Co-authored-by: Dimitris Apostolou <dimitris.apostolou@icloud.com>
Co-authored-by: Friz64 <friz64@protonmail.com>
Co-authored-by: huajin tong <137764712+thirdkeyword@users.noreply.github.com>
Co-authored-by: Small White <364772080@qq.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Development

Successfully merging a pull request may close this issue.

4 participants