In the config, if `window` is undefined, the derived `Default` for the
String `title` is used, which is an empty String. This was unintended,
and causes issues in gnome-shell (e.g. in the alt-tab dialog) when the
window title is an empty string.
This commit adds a manually implemented default for the `WindowConfig`,
it's the same as the derived `Default`, except for the `title`, which
will now always be "Alacritty" as originally intended.
This commit adds the concept of a "title stack" to the terminal. Some programs
(e.g. vim) send control sequences `CSI 22 ; 0` (push title) and `CSI 23 ; 0`
(pop title).
The title stack is just a history of previous titles. Applications can push
the current title onto the stack, and pop it back off (setting the window title
in the process).
Fixes#2840.
This takes the latest glutin master to port Alacritty to the EventLoop
2.0 rework.
This changes a big part of the event loop handling by pushing the event
loop in a separate thread from the renderer and running both in
parallel.
Fixes#2796.
Fixes#2694.
Fixes#2643.
Fixes#2625.
Fixes#2618.
Fixes#2601.
Fixes#2564.
Fixes#2456.
Fixes#2438.
Fixes#2334.
Fixes#2254.
Fixes#2217.
Fixes#1789.
Fixes#1750.
Fixes#1125.
A semicolon in a title OSC should be interpreted literally, not as a parameter
separator, but the OSC parser is very simple and does not know about arities of
commands.
Therefore, this patch takes all the parameters returned by the OSC parser and
reconstructs the original string by interspersing semicolons. Now an OSC like
'\e]2;hello;world' will set the title to 'hello;world' and not 'hello' like
before.
Commit e964af8 introduced a regression, where if cell's bg color was
equal to NamedColor::Background rgb color it was rendered with transparent
background. However the correct behavior is to render bg transparent
only when bg color is actually a NamedColor::Background.
Fixes#2814.
This fixes an off-by-two error in the renderer which initializes
characters 32 until 128 (inclusive) for each font whenever it is loaded.
The ascii visible range however just goes from 32 until 126 (inclusive).
This commit removes all bindings which are sending escapes from
the default configuration file, adds bindings for F13-F24, adds bindings
for ScrollToTop/ScrollToBottom actions, removes bindings for Super + F1-F12,
fixes bindings for Alt + F1-F12.
Fixes#2688.
This change disabled the mouse cursor and URL highlight (underline)
while a selection is in progress. A click to clear the selection doesn't
trigger a URL action, but will re-enable the URL highlighting to
indicate the next click will trigger the launcher.
The legacy xparsecolor implementation assumed that the \007 ending would
be passed to the parser, however it never is. This caused colors in the
format #rrggbb to be interpreted as #rrggb, leading to incorrect colors
showing up in Alacritty.
Fixes#2759.
If the terminal escape sequences for bold and italic text are active,
the text should be rendered as bold and italic. However, due to missing
support in Alacritty, it would always render this text in bold.
This adds support for combining the bold and italic escapes to render
text in both styles and allows users to override the font for this
scenario using the `font.bold_italic` configuration option.
Escape sequences in xterm are parsed according to xparsecolor.
xparsecolor supports 1, 2, 3, and 4 digit hex colors.
Previously, only 2 digits were supported.
This also fixes a bug where "fX" was parsed as "0xf", where X is an invalid character.
The response to a request for fg/bg must be a valid escape sequence.
The current response uses 4-digit hex, which was previously invalid.
Instead of ignoring unexpected intermediates in CSI escape sequences,
the intermediates are now explicitly checked and the escape sequence is
rejected when an unexpected intermediate is found.
Fixes#2171.
Instead of creating the rectangles for lines after the line is
completed, the rectangle is now initialized as soon as it is started.
Then when following cells also contain the same line type, the rectangle
is updated.
This resolves the problem of having to finish the last line when it ends
in the last non-empty column in the last line, since the render iterator
only returns non-empty cells and we never get the information that the
underline has ended.
Fixes#2680.
This switches to rfind_url for detecting URLs inside the grid. Instead
of expanding at the cursor position, the complete terminal is searched
from the bottom until the visible region is left with no active URL.
Instead of having the field `cur` publicly accessibly on the
`DisplayIterator`, there are the two methods `DisplayIterator::point`
and `DisplayIterator::cell` for accessing the current element of the
iterator now. This allows accessing the current element right after
creating the iterator.
Fixes#2629.
Fixes#2627.
The cursor rework introduced a regression where cursor color was always picked
from a config file, rather then using `ansi::NamedColor::Cursor` for this
purpose.
This commit also removes `CursorText` option from `NamedColor` enum,
since we can't speculate with `CursorText` during runtime.
Cursor rework commits:
cfc20d4f34371d13f8ef0d060d5d80
Since ref tests were only stored whenever winit requested the window
close, they would not get stored properly when the terminal was closed
through Alacritty using `exit`, Ctrl+D or similar.
This moves the ref test code to the and of the main entry point, which
will always be executed regardless of how the terminal was shutdown.
Fixes#2613.
Since ref tests were only stored whenever winit requested the window
close, they would not get stored properly when the terminal was closed
through Alacritty using `exit`, Ctrl+D or similar.
This moves the ref test code to the and of the main entry point, which
will always be executed regardless of how the terminal was shutdown.
This resolves an issue with rows only resetting partially, based on
their `occ` state. However this state is not always accurate, so more
than just the occupied elements need to be cleared.
Fixes#2340.
This should fill window with background color while it is offscreen instead of
showing it with uninitilized surface and then performing `clear`. So, the new
behavior should prevent glitches during startup. e.g. content of the windows
below, garbage from drivers and so on.
This commit fixes regression introduced in cfc20d4f34.
`self.cursor.line` forced the cursor to hold a fixed location while scrolling
until its "original" location (usually the shell prompt) went off the screen.
So cursor position should be keep updated, which can be achieved by using
`self.inner.line()`.
Fixes#2570.
This implements a block selection mode which can be triggered by holding
Control before starting a selection.
If text is copied using this block selection, newlines will be
automatically added to the end of the lines.
This fixes#526.
The dynamic color escape response would answer to requests with
rgb:0/0/0 when the color was completely black, instead of properly
responding with double-digit hex colors. This has been changed so that
Alacritty now always properly responds with the same number of hex
digits for all colors.
The number of digits has also been changed from two to four digits per
color, since that is the more commonly used format.
Using the `write!` macro was also causing problems with NeoVim,
since it caused Alacritty to write the dynamic color escape in multiple
write calls, switching to `write_all` fixed that.
Fixes#2543.
This reworks the selection logic to prevent any possible index out of
bounds exceptions by clamping the start and end points before doing
anything else with them when converting selections to spans.
This also fixes a bug where semantic selections would not automatically
expand across double-width characters.
Fixes#2486.
This is a large refactor of the config parsing structure, attempting to
reduce the size of the file a bit by splitting it up into different
modules with more specific purposes.
This also fixes#2279.
On macOS the clipboard actually returns an error when loading it and it
is empty. However this is not an `Empty` error but all errors are simple
boxed Errors from std.
Since loading the clipboard data usually should not fail, we now do not
log it as error if it fails but just print it to the debug log instead.
This fixes#2389.
This switches our own `copypasta` crate with the more standardized
`clipboard` library, which allows us to get rid of the `xclip`
dependency on X11.
Additionally, this lays the foundation for native Wayland clipboard
support once the clipboard crate is updated (or a fork is created).
Fixes#5.