If the viewport has been scrolled up, it is possible for a
reverse-scroll (rin) to cause the viewport to point to lines outside
the scrollback. This is an issue if the scrollback isn't full, since
in that case, the viewport will contain NULL lines. This will
potentially trigger assertions in a couple of different places.
Example backtrace:
#2 0x555555cd230c in bug ../../debug.c:44
#3 0x555555ad485e in grid_row_in_view ../../grid.h:83
#4 0x555555b15a89 in grid_render ../../render.c:3465
#5 0x555555b3b0ab in fdm_hook_refresh_pending_terminals ../../render.c:5165
#6 0x555555a74980 in fdm_poll ../../fdm.c:435
#7 0x555555ac2b85 in main ../../main.c:676
Detect when this happens, and force-move the viewport to ensure it is
valid.
Closes#2232
Foot likes it when compositor releases buffer immediately, as that
means we only have to re-render the cells that have changed since the
last frame.
For various reasons, not all compositors do this. In this case, foot
is typically forced to switch between two buffers, i.e. double-buffer.
In this case, each frame starts with copying over the damage from the
previous frame, to the new frame. Then we start rendering the updated
cells.
Bringing over the previous frame's damage can be slow, if the changed
area was large (e.g. when scrolling one or a few lines, or on full
screen updates). It's also done single-threaded. Thus it not only
slows down frame rendering, but pauses everything else (i.e. input
processing). All in all, it reduces performance and increases input
latency.
But we don't have to wait until it's time to render a frame to copy
over the previous frame's damage. We can do that as soon as the
compositor has released the buffer (for the frame _before_ the
previous frame). And we can do this in a thread.
This frees up foot to continue processing input, and reduces frame
rendering time since we can now start rendering the modified cells
immediately, without first doing a large memcpy(3).
In worst case scenarios (or perhaps we should consider them best case
scenarios...), I've seen up to a 10x performance increase in frame
rendering times (this obviously does *not* include the time it takes
to copy over the previous frame's damage, since that doesn't affect
neither input processing nor frame rendering).
Implemented by adding a callback mechanism to the shm abstraction
layer. Use it for the grid buffers, and kick off a thread that copies
the previous frame's damage, and resets the buffers age to 0 (so that
foot understands it can start render to it immediately when it later
needs to render a frame).
Since we have certain way of knowing if a compositor releases buffers
immediately or not, use a bit of heuristics; if we see 10 consecutive
non-immediate releases (that is, we reset the counter as soon as we do
see an immediate release), this new "pre-apply damage" logic is
enabled. It can be force-disabled with tweak.pre-apply-damage=no.
We also need to take care to wait for the thread before resetting the
render's "last_buf" pointer (or we'll SEGFAULT in the thread...).
We must also ensure we wait for the thread to finish before we start
rendering a new frame. Under normal circumstances, the wait time is
always 0, the thread has almost always finished long before we need to
render the next frame. But it _can_ happen.
Closes#2188
When we print a character to the grid, we must also update its OSC-8
state if an OSC-8 URI is currently active.
For double-width characters, this was only being done for the first
cell.
This causes the labels in URL mode to be off, as the link was
effectively chopped up into multiple pieces.
Closes#2179
These (non-css) cursor shapes were added to the cursor-shape-v1
protocol in wayland-protocols 1.42.
We don't need (or use them at all) internally, but add them to the
list we use to translate from shape names to shape enums. This allows
users to set a custom shape (via OSC-22), while still using server
side cursors (i.e. no need to fallback to client-side cursors).
If we try to set a shape not implemented by the server, we get a
protocol error and foot exits. This is bad.
So, make sure we don't do that:
1. First, we need to explicitly bind v2 if implemented by the server
2. Track the bound version number in the wayland struct
3. When matching shape enum, skip shapes not supported in the
currently bound version of the cursor-shape protocol
Before this, we only applied custom selection colors, if *both* the
selection bg and fg had been set.
Since the options are already split up into two separate options, and
since it makes sense to at least be able to keep the foreground colors
unchanged (i.e. only setting the selection background), let's allow
only having one of the selection colors set.
Closes#1846
This adds supports for 16-bit surfaces, using the new
PIXMAN_a16b16g16r16 buffer format. This maps to
WL_SHM_FORMAT_ABGR16161616 (little-endian).
Use the new 16-bit surfaces by default, when
gamma-correct-blending=yes.
When set to 'auto', use 10-bit surfaces if gamma-correct blending is
enabled, and 8-bit surfaces otherwise.
Note that we may still fallback to 8-bit surfaces (without disabling
gamma-correct blending) if the compositor does not support 10-bit
surfaces.
Closes#2082
This option selects which color theme to use by default. I.e. at
startup, and after a reset.
This is useful with combined theme files, where a single file defines
e.g. both a dark and light version of the theme.
* color-theme-switch-1: select the primary color theme
* color-theme-switch-2: select the alternative color theme
* color-theme-toggle: toggle between the primary and alternative color themes
Foot doesn't implement RTL, and explicit LTR markers is neither
needed, nor used in anyway. In fact, they cause issues with font
lookup, as fcft often fails to find the marker codepoint in the
primary font, causing a fallback font to be used instead.
Closes#2049
Update tweak.scaling-filter to recognize the new scaling filters added
in fcft-3.3.0.
Since fcft_set_scaling_filter() is deprecated in 3.3.0, don't use it
anymore, and set the scaling filter via fcft_font_options instead.
This way, all lines are treated as having a hard linebreak, until it's
cleared when we do an auto-wrap.
This change alone causes issues when reflowing text, as now all
trailing lines in an otherwise empty window are treated as hard
linebreaks, causing the new grid to insert lots of unwanted, empty
lines.
Fix by doing two things:
* *clear* the linebreak flag when we pull in new lines for the new
grid. We only want to set it explicitly, when an old row has its
linebreak flag set.
* Coalesce empty lines with linebreak=true, and only "emit" them as
new liens in the new grid if they are followed by non-empty lines.
Don't set linebreak on linefeed. Instead, rely on the default value of
true, and that it is only cleared when a character is printed while
LCF=1.
Note that printing to a row that has linebreak cleared, will set the
linebreak flag again.
This implements gamma-correct blending, which mainly affects font
rendering.
The implementation requires compile-time availability of the new
color-management protocol (available in wayland-protocols >= 1.41),
and run-time support for the same in the compositor (specifically, the
EXT_LINEAR TF function and sRGB primaries).
How it works: all colors are decoded from sRGB to linear (using a
lookup table, generated in the exact same way pixman generates it's
internal conversion tables) before being used by pixman. The resulting
image buffer is thus in decoded/linear format. We use the
color-management protocol to inform the compositor of this, by tagging
the wayland surfaces with the 'ext_linear' image attribute.
Sixes: all colors are sRGB internally, and decoded to linear before
being used in any sixels. Thus, the image buffers will contain linear
colors. This is important, since otherwise there would be a
decode/encode penalty every time a sixel is blended to the grid.
Emojis: we require fcft >= 3.2, which adds support for sRGB decoding
color glyphs. Meaning, the emoji pixman surfaces can be blended
directly to the grid, just like sixels.
Gamma-correct blending is enabled by default *when the compositor
supports it*. There's a new option to explicitly enable/disable it:
gamma-correct-blending=no|yes. If set to 'yes', and the compositor
does not implement the required color-management features, warning
logs are emitted.
There's a loss of precision when storing linear pixels in 8-bit
channels. For this reason, this patch also adds supports for 10-bit
surfaces. For now, this is disabled by default since such surfaces
only have 2 bits for alpha. It can be enabled with
tweak.surface-bit-depth=10-bit.
Perhaps, in the future, we can enable it by default if:
* gamma-correct blending is enabled
* the user has not enabled a transparent background
When compiled with grapheme clustering support, zero-width characters
that also are grapheme breaks, were ignored (not stored in the
grid).
When utf8proc says the character is a grapheme break, we try to print
the character to the current cell. But this is only done when width >
0. As a result, zero width grapheme breaks were simply discarded.
This only happens when grapheme clustering is enabled; when disabled,
all zero width characters are appended.
Fix this by also requiring the width to be non-zero when if we should
append the character or not.
Closes#1960
When appending to an existing composed character, "inherit" its forced
width, if set.
Also make sure to actually _use_ the forced width, if set, rather than
the calculated width.
This fixes an issue when appending zero-width codepoints to a
forced-width combining character.
This function "prints" any non-ascii character (i.e. any character
that ends up in the action_utf8_print() function in vt.c) to the
grid. This includes grapheme cluster processing etc.
action_utf8_print() now simply calls this function.
This allows us to re-use the same functionality from other
places (like the text-sizing protocol).
The logic that tries to ensure we don't break a line in the middle of
a multi-cell character was flawed when the number of cells were larger
than 2.
In particular, if the number of cells to copy were limited by the
number of cells left on the current (new) line, and were less than the
length of the multi-cell character, then we failed to insert the
correct number of spacers, and also ended up misplacing the multi-cell
character; instead of pushing it to the next line, it was inserted on
the current line, even though it doesn't fit.
Also change how trailing SPACER cells are rendered (cells that are
"fillers" at then end of a line, when a multi-column character was
pushed over to the next line): don't copy the previous cell's
attributes (which may be wrong anyway), use default attributes
instead.
When the client application emits combining characters, for example
multi-codepoint emojis, in insert-mode, we ended up pushing partial
graphemes to the right, for each codepoint, resulting in too many
cells (and with the wrong content) being inserted.
The fix is fairly simple; don't "insert" when appending characters to
an existing grapheme cluster.
This isn't something we can detect easily in print_insert() (it would
require us to do grapheme clustering again). Fortunately, we do have
the required information in action_utf8_print(). So, pass this
information as a boolean to term_print().
Closes#1947
From the release notes:
system bell - allowing e.g. terminal emulators to hand off system
bell alerts to the compositor for among other things accessibility
purposes
The new protocol is used when the new config option
bell.system=yes (and the compositor implements the protocol,
obviously).
The system bell is rung independent of whether the foot window has
keyboard focus or not (thus relying on compositor configuration to
determine whether anything should be done or not in response to the
bell).
The new option is enabled by default.
If the compositor sends a keyboard enter event before our window has
been mapped, foot crashes; the enter event triggers a cursor
refresh (hollow -> non-hollow block cursor), which crashes since we
haven't yet allocated a grid.
Fix by no-op:ing the refresh if the window hasn't been configured yet.
Closes#1910
If we call render_refresh, that will wait for a callback to the main
surface. In the case of a flash, the main surface might not get callbacks
if the compositor implements fancy culling optimizations like wlroots
wlr_scene compositors such as sway version >=1.10.
Unsure if the protocol imposes a limit (haven't found any
documentation), or if the issue is in the libwayland implementation,
or wlroots (triggers in at least sway+river).
The issue is that setting a too long app-id causes the
compositor (river at least) to peg the CPU at 100%, and stop sending
e.g. frame callbacks to foot.
Closes#1897
In fact, there appears there *is* no escape sequence to set the icon.
Keep most of the logic in place, but in practice, we'll always set the
icon to the app-id. That is, at startup, we set it to the configured
app-id (either from config, or the command line).
OSC-176, which sets the app-id, also updates the icon (to the app-id).
* The toplevel icon is now set to the app-id, unless "overridden" by
OSC-1 or OSC-0.
* Implemented OSC-1
* OSC-0 extended to also set the icon
* Implemented CSI 20 t - report window icon
* Implemented CSI 21 t - report window title
* Implemented CSI 22 ; 1 t - push window icon
* Implemented CS 23 ; 1 t - pop window icon
* Extended CSI 22/23 ; 0 t to also push/pop the icon
* Verify app-id set by OSC-176 is valid UTF-8
* Verify icon set by OSC-0/1 is valid UTF-8
Summary of changes:
* Make xvsnprintf() static
* restrict-qualify pointer arguments (as done by the libc equivalents)
* Make comments and spec references more thorough
* Remove pointless `n <= INT_MAX` assertion (see comment)
* Use FATAL_ERROR() instead of xassert() (since the assertion is inside
a shared util function but the caller is responsible for ensuring the
condition holds true)
* Change some callers to use size_t instead of int for the return value
(negative returns are impossible and all subsequent uses are size_t)
The updated comments and code were taken (and adapted) from:
49260bb154/src/util/xsnprintf.c (L6-50)
This work was entirely authored by me and I hereby license this
contribution under the MIT license (stated explicitly, so that
there's no ambiguity w.r.t. the original license).
Or put more propertly; if the notification daemon, and the
notification helper used by foot has been configured
properly (i.e. they both support XDG activation tokens), notifications
generated by BEL and OSC-777 will now raise/focus the window when the
default action of the notification is activated - typically by
clicking the notification.
Closes#1822
This avoids the need for an unused third argument for most xstrjoin()
calls and replaces the cases where it's needed with a more flexible
function. Code generation is the same in both cases, when there are 2
string params and a compile-time known delimiter.
This commit also converts 4 uses of xasprintf() to use xstrjoin*().
See also: https://godbolt.org/z/xsjrhv9b6
* Don't store a list of unfinished notifications. Use a single one. If
the notification ID of the 'current' notification doesn't match the
previous, unfinished one, the 'current' notification replaces the
previous one, instead of updating it.
* Update xstrjoin() to take an optional delimiter (for example ','),
and use that when joining categories and 'alive IDs'.
* Rename ${action-arg} to ${action-argument}
* Update handling of the 'n' parameter (symbolic icon name); the spec
allows it to be used multiple times, and the terminal is supposed to
pick the first one it can resolve. Foot can't resolve icons at all,
neither can 'notify-send' or 'fyi' (which is what foot typically
executes to display a notification); it's the notification daemon that
resolves icons.
The spec _could_ be interpreted to mean the terminal should lookup
.desktop files, and use the value of the 'Icon' key from the first
matching .desktop files. But foot doesn't read .desktop files, and I
don't intend to implement XDG directory scanning and parsing of
.desktop files just to figure out which icon to use.
Instead, use a simple heuristics; use the *shortest* symbolic
names. The idea is pretty simple: plain icon names are typically
shorter than .desktop file IDs.
First, icons have been finalized in the specification. There were only
three things we needed to adjust:
* symbolic names are base64 encoded
* there are a couple of OSC-99 defined symbolic names that need to be
translated to the corresponding XDG icon name.
* allow in-band icons without a cache ID (that is, allow applications
to use p=icon without having to cache the icon first).
Second, add support for the following new additions to the protocol:
* 'f': custom app-name, overrides the terminal's app-id
* 't': categories
* 'p=alive': lets applications poll for currently active notifications
* 'id' is now 'unset' by default, rather than "0"
* 'w': expire time (i.e. notification timeout)
* "buttons": aka actions. This lets applications add additional (to
the terminal defined "default" action) actions. The 'activated' event
has been updated to report which button/action was used to activate
the notification.
To support button/actions, desktop-notifications.command had to be
reworked a bit.
There's now a new config option:
desktop-notifications.command-action-arg. It has two template
arguments ${action-name} and ${action-label}.
command-action-arg gets expanded for *each* action.
${action-name} and ${action-label} has been replaced by ${action-arg}
in command. This is a somewhat special template, in that it gets
replaced by *all* instances of the expanded actions.
This fixes a regression where closing a terminal instance, or hard- or
soft-resetting a terminal caused FD 0 to be closed.
This meant it became re-usable. Usually, by memfd_create() when
allocating a new surface buffer. So far nothing _really_ bad has
happened.
But what if FD 0 is now used by a memfd, and we close _another_
terminal instance?
This causes our memfd to be closed. And then, when e.g. trying to
scroll the terminal content: fallocate() fails with bad FD.
