The windows tagged builds currently fail some tests since the
winpty-agent.exe is inside the release directory instead of the debug
directory with tagged builds.
This release introduces some config to automatically build deploy a
binaries on the github release page using travis. The build only happens
when a commit is tagged and it uses the stable version of rust.
The main travis sections (install/script/before_deploy) have been
moved out of the .travis.yml to make it easier to read, maintain and
extend the different steps of the CI process.
Since checking for the Rust version in CI is enough to know if clippy
should be used or not, the environment variable `CLIPPY` has also been
removed, which further allowed simplifying the CI process.
Besides the executables, some auxillary files are now also published as
part of a release when they have changed since the last tagged Alacritty
release. This should make it clear for returning users when a new
version of a specific auxillary file is required.
Instead of using the 14.04 image which travis provides by default, an
18.04 docker image is used to build the output binaries for Linux.
This affects both the .deb and the .tar.gz binary.
The advantage of this is that while binaries compiled on 14.04, do not
work on 18.04, it does work the other way around. The generated .tar.gz
binary has been tested on 18.04, Debian, Fedora and Archlinux and all
systems were able to run it without any warnings or errors.
Properly update an active selection while scrolling the main scrollback buffer.
This does not affect the alternate screen buffer, since no scrollback buffer is
available.
The 2c37da48b5 change introduced some
changes to the way cursor configuration is handled. However it did not
properly handle the default behavior of the hollow cursor when the
`cursor` field was not specified at all.
By implementing the `Default` trait for the `Cursor` struct in
`config.rs` manually, the default value of the `unfocused_hollow` field
has been corrected back to `true` when the `cursor` struct isn't present
at all.
It is now possible to paste in mouse mode again by making use of the
`shift` key while pressing the mouse button reserved for PasteSelection.
All mouse bindings are now also matching the modifiers in a relaxed way,
so extra modifiers are ignored.
There are a couple of cursor-related options in the Alacritty config
file now, however they aren't grouped together in any way.
To resolve this a new `cursor` field has been added where all cursor
configuration options (besides colors) have been moved.
The `custom_cursor_colors` option has also been removed, since it's not
necessary anymore. Simply making the `colors.cursor.*` fields optional,
allows overriding the cursor colors whenever one of them is present.
Like that the user doesn't have to think about a relation between two
separate configuration options.
This PR initially put the `hide_cursor_when_typing` variable under
`cursor.hide_when_typing`. However this field is completely unrelated to
the cursor, but instead relates to the mouse cursor.
Since the word `cursor` is already used for the active cell in the grid
of a terminal emulator, all occurences of the word `cursor` when talking
about the mouse have been replaced with the word `mouse`.
The configuration option has also been moved to
`mouse.hide_when_typing`, to make it clear what this option is changing.
This fixes#1080.
Since cargo-deb builds all members of the workspace by default, it is
necessary that the winpty subcrate can be built on all operating
systems, since it's not possible to have OS-specific workspace members.
To achieve this the crate has been changed to be empty by default on
non-windows systems. It might make sense to do something similar with
winpty-sys, but it's not strictly necessary at this point since we don't
directly depend on it.
This fixes#1716.
The Windows configuration file was outdated compared to the other
configuration files, since there were some major changes to the config
during the development of the Windows PR.
All changes between the Linux and Windows configuration file have been
removed, preserving platform-specific configuration defaults that were
intentionally changed.
Some minor changes have also been made to the Linux and macOS
configuration files, since there were some inconsistencies left.
This fixes#1692.
The Windows configuration file was outdated compared to the other
configuration files, since there were some major changes to the config
during the development of the Windows PR.
All changes between the Linux and Windows configuration file have been
removed, preserving platform-specific configuration defaults that were
intentionally changed.
Some minor changes have also been made to the Linux and macOS
configuration files, since there were some inconsistencies left.
This fixes#1692.
To click on links in the alternate screen buffer, it is necessary that
the `shift` button is held down, otherwise mouse events are captured by
the application. However this would also require that `Shift` is added
to the `mouse.url.modifiers` option. Thus it is not possible anymore to
click on URLs unless the shift button is always held down.
To resolve this issue, the matching of modifiers has been relaxed. If a
modifier is specified in the config, it is always required to be held
down. However if a modifier is held down which is not specified, it is
still accpeted. This one-way relaxed matching allows clicking on links
with or without shift held down without having to make use of the
`mouse.url.modifiers` setting at all.
This adds the option to automatically launch URLs with a specified
program when clicking on them.
The config option `mouse.url_launcher` has been added to specify which
program should be used to open the URL. The URL is always passed as the
last parameter to the specified command.
It is not always desired for URLs to open automatically when clicking on
them. To resolve this a new `modifiers` field has been introduced to the
config, which allows specifying which keyboard modifiers need to be held
down to launch URLs in the specified launcher.
Some tests have been added to make sure that the edge-cases of the URL
parsing are protected against future regressions. To make testing easier
the parsing method has been moved into the `SemanticSearch` trait. The
name of the trait has also been changed to just `Search` and it has been
moved to `src/term/mod.rs` to fit the additional functionality.
This fixes#113.
Whenever the viewport is scrolled, the selection is rotated to make sure
that it moves with the viewport. However this did not correctly handle
the underflow that happens when the selection goes below 0.
This resolves that problem for the selection by moving the internal line
representation to an isize, thus correctly keeping track of the
selection start/end points even when they have a negative index. Once
the selection is converted to a span, the lines are clamped to the
visible region.
This fixes#1640 and fixes#1643.
This change should allow the usage of scancodes in the configuration
file.
When a VirtualKeyCode for glutin is not present, this should now allow
the user to use the scancodes instead. If the user specifiecs a key with
its scancode even though the key has a VirtualKeyCode, it should still
work.
The behavior of directly specifying a VirtualKeyCode should be unchanged
by this.
This fixes#1265.
Initial support for Windows is implemented using the winpty translation
layer. Clipboard support for Windows is provided through the `clipboard`
crate, and font rasterization is provided by RustType.
The tty.rs file has been split into OS-specific files to separate
standard pty handling from the winpty implementation.
Several binary components are fetched via build script on windows
including libclang and winpty. These could be integrated more directly
in the future either by building those dependencies as part of the
Alacritty build process or by leveraging git lfs to store the artifacts.
Fixes#28.
Alacritty's manual installation instructions take up a lot of space in
the README and they are hard to navigate because there's a lot of
information for specific distros which are irrelevant to most users.
To improve this situation, the manual installation section has been
moved to the `INSTALL.md` file. This allows keeping the readme to a
minimum which just includes the straight-forward instructions required
to install it on systems which provide packages for it.
Having a separate file also makes it possible to add a full index just
for the installation instructions, which makes it possible to skip
sections which are irrelevant.
Serde has problems deserializing yaml files which contain sections
without any values. Since the `TERM` setting has been removed recently,
the `env` section was completely empty leading to deserialization
errors.
To resolve this the `env` section has been commented-out by default, if
the user wants to set a variable, it is now necessary to uncomment that
section.
Some minor tweaks have also been made to the existing `TERM` comments,
to clearly indicate these are value examples instead of comments.
On macOS 10.12.6, running `make app` creates a bundle that refuses to
start (it complains application is only compatible with more recent
versions of macOS). The binary works great when running it directly.
To resolve this, the `MACOSX_DEPLOYMENT_TARGET` environment
variable is now set automatically from the makefile. This allows
building on macOS 10.12 and earlier, without negatively impacting the
later versions of macOS.
The terminal lock is now dropped before rendering by storing
all grid cells before clearing the screen.
This frees the terminal to do other things since the lock is now
free, which lead to a performance benefit with high throughput
applications.
The patch uses the Cargo.toml patch section to force a single downstream
choice of freetype-sys instead of relying on forks of other crates. It
also bumps the fontconfig/freetype dependencies in the process.
Using the `bench` feature, `cargo test` was failing since one of the
benchmarks was running into a debug assertion for attempting to access a
line with an index beyond the grid length.
Since this issue was caused by the `len` property not being serialized
and deserialized, the `#[serde(skip)]` attribute has been changed to
`#[serde(default)]`. The ref-test has been edited to include the correct
grid length for proper deserialization.
This fixes#1604.
When rendering selections with both start and end outside of the visible
area, Alacritty would assume that both start and end are either above or
below the viewport and not render the selection at all.
To fix this the `buffer_line_to_visible` method now returns a
`ViewportPosition` instead of an `Option<Line>`, this allows giving more
feedback about where outside of the visible region the line is using the
`ViewportPosition::Above` and `ViewportPosition::Below` variants.
Using these newly introduced variants, a selection spanning the whole
screen is now rendered if the selection should go from above the visible
area to below it.
This fixes#1557.
The deprecated `window.decoration` values `true` and `false` were using
the `visit_bool` visitor for serde. However, only the `str` visitor was ever
called.
To print the correct deprecation notice, the bool visitor has been
removed and the warning has been added for the `"true"` and `"false"`
str visitor.
Since the mouse start position has been the first movement event after
the mouse button was held down, there have been some issues with the
start point lagging behind the cursor because movement events were not
reported from the initial position but there was a gap until movement
starts reporting.
To fix this whenever the mouse button is pressed, the position and cell
side is stored on the `Mouse` struct. Because of this it does not matter
anymore if the movement events are all reported and we can just start a
selection using the stored position/side whenever there currently is no
selection present.
This fixes#1366
When selecting multiple lines in Alacritty, there was an issue with
empty lines not being copied. This behavior has been chanaged so empty
lines should be correctly copied now.
When copying content which ends with an empty line, Alacritty would copy
an additional empty line.
This has been resolved by only adding empty lines when the empty line
was not in the last selected line.
The reset cursor terminal escape sequnce specified in the terminfo file
was using xterm's `\E[2 q`. However this just resets the cursor to the
block cursor shape.
Since Alacritty supports the `\E[0 q` escape sequence for resetting the
cursor shape to the original shape specified in the configuration file,
the terminfo file has been altered to make use of this sequence instead.
This replaces the current definitions, which depend on the system's
'xterm-256color' terminfo definition with the `alacritty` and
`alacritty-direct` definitions.
The new definitions are completely standalone.
The default `TERM` value has been changed to be dynamically
set based on the definitions installed on the system. Alacritty will
try to use the `alacritty` definition first and fall back to
`xterm-256color` if the `alacritty` definition is not present.
Previously Alacritty has initialized all lines in the buffer as soon as
it is started. This had the effect that terminals which aren't making
use of the scrollback buffer yet, would still consume large amounts of
memory, potentially even freezing the system at startup.
To resolve this problem, the grid is now dynamically resized in chunks
of `1000` rows. The initial size is just the visible area itself, then
every time lines are written to the terminal emulator, the grid storage
is grown when required.
With the worst-case scenario of having 100_000 lines scrollback
configured, this change improves startup performance at the cost of
scrolling performance.
On my machine the startup changes from ~0.3 to ~0.2 seconds.
The scrolling performance with large throughput is not affected, however
it is slowed down when the number of lines scrolled are close to the
100_000 configured as scrollback. The most taxing benchmark I've found
for this was running `yes | dd count=500 > 500.txt` (note the relatively
small file size). This will cause a slowdown on the first run from 0.05s
to 0.15s. While this is significant, it lines up with the time saved at
startup.
This fixes#1236.
A new configuration option `save_to_clipboard` has been added
to the `selection` section of the configuration. This allows writing
every selection to the primary system clipboard when it is set
to `true`.
On linux the selection is still written to the selection clipboard,
independent of the value of the `save_to_clipboard` setting.