Flesh out the get_os_parameters impl for macOS. When running on a
system that provides `NSScreen::safeAreaInsets`, use that to determine
the border required to avoid the "notch" on certain models of mac.
In the GUI layer: when the os parameters include a border, adjust
the render position to account for it.
This is a bit of a speculative change, as I don't have a mac with
a notch.
refs: https://github.com/wez/wezterm/issues/1737
If the WM has never confirmed our size and we're about to paint,
explicitly query it and generate a resize event so that we stand
a better chance of showing the right stuff.
refs: https://github.com/wez/wezterm/issues/1710
When the user switches to another virtual console and then back into a
Wayland session, a number of seat updates are received by the listener
installed by `Environment::listen_for_seats()`. These updates
essentially flip the keyboard and pointer flags off and then back on in
the seat data, which indicates that new handlers need to be assigned to
the keyboard and pointer objects.
However, because one update is received for _each_ flag being toggled,
this means that the listener is fired twice with `has_keyboard` set to
true, which means that two keyboard handlers end up being installed.
Users then experience each keypress being delivered twice to the
terminal.
This commit adds a map to track the keyboard object that has been
assigned on each seat, thereby preventing duplicate assignments and (by
extension) duplicate keypresses being registered. This logic is
essentially the same as what's used in the `kbd_input` example in SCTK,
which doesn't have this issue.
Something similar is required for pointer handling, which also breaks
after switching to another virtual console and back into Wayland, but I
was scared off by the TODO in the `listen_for_seats` callback and didn't
investigate this further.
On Windows in FRA layout, `^^SPACE` results in `^^SPACE`, whereas
on Linux it results in `^SPACE`. We were doing the latter instead
of the former.
Let's just be consistent with other windows apps.
closes: #1729closes: #1730
* Detect mouse leaving the window
* Implement leave
* Use new API
* Fix mouse leave
* Fix mouse leave on Wayland
* Mouse leave on X11
* Detect mouse leaving window on macOS
* Fix example
refs: #1434
This might "break" eg: pasting from apps that just use ATOM_STRING
and that happen to have utf-8 compatible text.
Ideally the get_text_property interface would expose the actual
text encoding in the returned data and allow the application to
"do stuff" with it.
refs: https://github.com/meh/rust-xcb-util/issues/21
Resolves a little bit of the awkward duplication of color types
between some of the crates by factoring them a little bit better.
This is prep for allowing specifying alpha for some colors
in the config.
This puts to final rest #478, wherein ligatured glyphs that span
cells would render portions of the glyph with the wrong fg color,
where wrong was usually the bg color and cause the glyph to turn
invisible when cursoring through the ligature.
The approach used here is to divide the glyph into 7 discrete strips
where each strip either intersects with the cursor, the selection, or
neither. That allows us to render each strip with the appropriate
foreground color.
This change simplifies some of the logic and allows some other code
to be removed, so that feels good!
As is tradition with these renderer changes, there's a good chance that
I overlooked something in testing and that the metrics or alignment
might be slightly off for some font/text combo. Let's see how this
goes!
refs: #784
refs: #478
refs: #1617
This allows unicode_version to be respected again when rendering.
The updated emoji-presentation.sh script now highlights this slightly
better by putting `.` characters after the emoji; unicode version 14
emoji presentation will show the `.` in the 3rd column, while earlier
versions will show it in the 2nd column for glyphs that are sensitive
to the version.
refs: #1607
refs: #1563
This action causes the active workspace for the gui to change.
If the name is omitted a random name will be generated.
If the workspace doesn't exist, it will be be created.
The optional spawn parameter can be used to launch a specific
program into the new workspace; if omitted, the default prog
will be used.
The gui only supports a single active workspace. Switching workspaces
will repurpose existing gui windows and re-assign them to windows
in the new workspace, adjusting their size to fit those windows,
spawning new windows or closing unused windows as required.
The gui now exits when there are no panes in the active workspace,
rather than no panes at all.
refs: #1531
When a dead key is composing, we gave no information about what was
composing. Contrast with Windows and macOS where we show the first
key in the composition as part of that state.
This commit makes an attempt to populate equivalent information.
It's a bit more complex with the xkeyboard stuff as there can be
multiple combining sequences and there's no guarantee that we can
show a meaningful label.
We try our best for the common case of a single dead key, and have
a probably reasonable fall back for other cases where we don't
other get that information.
This allows removing DeadKeyStatus::Holding.
lalt-` is technically a dead key combo, so we entered dead key
processing, then realized that
`send_composed_key_when_left_alt_is_pressed == false` and that
we should ignore the dead-key-ness of the combo, but returned
a composed result, which had the modifiers discarded.
The correct way to handle this is to signal that it wasn't
dead after all and to allow the main flow to build the KeyEvent
as normal.
We need 100% of the info for it to work correctly, so this commit:
* Exposes the keyboard encoding mode via the Pane trait
* Adds the scan code to the RawKeyEvent
* Has the GUI perform the encoding if the keyboard is set that way
* Removes the basic encoder from termwiz in favor of the gui level one
The net result is that we bypass the Pane::key_up/Pane::key_down methods
in almost all cases when the encoding mode is set to win32-input-mode.
There is now a config option: allow_win32_input_mode that can be
used to prevent using this mode.
refs: #1509
This commit causes the terminal to emit win32-input-mode encoded key up
and down events for a limited subset of keys When win32-input-mode is
enabled.
We limit them to keys where we know the VK key code equivalent,
and where those keys are either not representable (eg: modifier
only key events), or may generate ambiguous output (eg: CTRL-SPACE
in different keyboard layouts).
However, in my experiments, modifier only key presses confuse powershell
and cause it to emit `@`, so I've disabled that in the code for now.
refs: https://github.com/wez/wezterm/issues/318
refs: https://github.com/wez/wezterm/issues/1509
refs: https://github.com/wez/wezterm/issues/1510
When we translate a dead key, send the composed event immediately
and don't try to route the current key press via the IME.
Improve rendering when in the composing state: overlay the composing
text at the cursor position to show what the composing text would
look like, even though it hasn't been committed to the input stream
yet.
refs: https://github.com/wez/wezterm/issues/1504
For Korean text input, pressing SHIFT and then typing in certain
keys begins a composition sequence. Our logic for when to route
via the IME got so distracted by ALT that it didn't consider SHIFT
and didn't send this sequence to the IME, so we'd fail to compose
those sequences.
While poking at this, I realized that we should treat this composition
similarly to dead keys, in that we can signal the term window to
highlight the cursor color and report that status.
There's some further work to do: the composing text should be rendered
by us. So far our IME integration assumes that the IME itself will
render over the top of our window, but for this particular input
it doesn't do that.
refs: https://github.com/wez/wezterm/issues/1504
Since we now have RawKeyEvent and a sane way to indicate handling,
we don't need these any more, and it simplifies key dispatch to
remove them.
refs: #1483
Pass down whether we are in a live resize to the gui layer, so that
we don't incorrectly assume that the scale has changed, and fight
with the window manager.
Built this on my mac: will need to fixup for windows and linux.
refs: https://github.com/wez/wezterm/issues/1491
This moves the event dispatching into the keyboard processor,
which will allow for the processor to skip feeding an event
into the dead key/composition stuff if a key assignment is
handled.
It doesn't actually do that bit yet though, as the wayland
key repeat processing was a bit more involved and I wanted
to constrain the scope of this commit.
refs: #877
on macos only (for now), we generate a RawKeyEvent prior to
dead key or IME composition and route it to the window to give it
a chance to handle the event.
RawKeyEvent handling is scoped only to key assignments, not generating
input for the terminal.
A raw key event can be marked as handled to prevent moving on to
performing composition and generating cooked key input.
refs: https://github.com/wez/wezterm/issues/877
Previously, we'd take a couple of guesses at how to map the key
to a utf8 value, but! the keyboard state has a method that can tell
us what to use.
This is important in non-latin keymaps where, for example, the `c`
key generates cyrillic small letter es and we'd end up sending
CTRL + that through to the terminal when CTRL is held down.
If we get the utf8 string from the keyboard layer then we get
CTRL+c instead, and that is what we want.
refs: https://github.com/wez/wezterm/issues/678
I saw this when the IME was active:
```
2021-12-31 00:46:26.941 wezterm-gui[46160:16665949] -[NSConcreteMutableAttributedString UTF8String]: unrecognized selector sent to instance 0x600002b24180
```
the issue is that some of the window callbacks can receive either
NSString or NSAttributedString. The latter doesn't have a UTF8String
method, but does have a string property that returns an NSString
that can be used instead.
Certain keys are "handled" by the IME through it generating a "noop"
command.
That's not super useful for us, so this commit detects the noop case
and then treats it as though the IME didn't handle the input event.
While implementing the above fix, I realized that the same technique
could be used more generally to return processing to our main input
handling for the various selectors that we do recognize: we were
essentially inferring the original key combinations based on the
selector which is not scalable and potentially lossy.
We can't capture CTRL-ESC this same way, as that key combination
is magical and is routed to the callback without generating any
key events.
refs: https://github.com/wez/wezterm/issues/615
refs: https://github.com/wez/wezterm/issues/975
refs: https://github.com/wez/wezterm/pull/1410
This commit adds a CSS box model inspired element / layout
facility, and replaces the hand implemented fancy tab bar
element render.
This makes the code for fancy tab bar much easier to read
and update.
The right status area now expands to the full height of the
tab bar area, and uses a line height of 2.0, which makes
it line up nicely in the tab bar.
On Windows, both EGL and MESA render modes were too dark.
After a bit of hunting around what I found made EGL and MESA
consistent with my default nVidia GPL rendering was:
* Tell glium that our shader outputs srgb
* Add explicit gamma conversion from linear to srgb in the shader
AFAICT, that shouldn't be required, but it seems as though something
deep in glium really wants to apply some kind of gamma conversion,
and it seems to select the wrong kind unless we set things explicitly
to SRGB.
There are some people complaining about this in
https://github.com/glium/glium/issues/1615.
I actually tried to move entirely aware from the glium srgbtexture2d
type in the hope of having explicit control over the gamma, but the
issue is in what happens to the outputs rather than the inputs.
It appears to me as though the text now looks slightly less
intense, so I think this may be what we need for the gamma issue
in https://github.com/wez/wezterm/issues/544 and potentially
also https://github.com/wez/wezterm/issues/1025
refs: https://github.com/wez/wezterm/issues/1373
In the worst case this can cause badly implemented desktop components like panels
to constantly rerender and waste 100% CPU (that's how I found a bug in mine :D)
when anything is constantly updating in the terminal (like viewing a log stream).
When using client-side decorations, we can now skip rendering
the header/title bar.
If you've set NONE decorations, then wezterm will configure
the window that way, but that is only respected when the window
is created, as weston crashed when I tried to change this in
a window config reload event.
The wayland frame now also observes config change events,
so frame color adjustments should now take effect without
restarting.
refs: #1077
Changing surface configuration flags *to* default would never be applied.
This was uncovered when implementing increment step resize: it would stop
working after *one* maximization.
The wl_keyboard definition does not define that the incoming fd is always at seek position 0.
In fact, the spec says the fd "can be memory-mapped" and that's it.
And e.g. smithay client-toolkit uses mmap, but we don't :/
Use pread() to read from zero.
This commit introduces a mechanism for specifying resize increments
for a window, and then arranges for the termwindow to set those
to match the current font cell metrics.
This should help to avoid cases where there is excess padding pixels
resulting from the window being slightly larger than computed number
of cells and the font metrics.
Not sure what happened here: presumably a borked merge or something
similar, but this commit allows `window:get_appearance` to return
the actual appearance value on X11.
Even though this plumbs the call through to Wayland, Wayland doesn't
provide an equivalent concept so still always return Light, as is
mentioned on our docs.
closes: #1098
* Trigger a paint immediately from invalidate if not throttled
* Otherwise defer the other events until we're about to sleep for xcb
events, which should maximize the coalesce around resize/expose events
refs: #1051
The thesis is that some WM's might send a whole bunch of events
that cause us to over draw/over resize.
I'm not convinced that this is a righteous change, but it can't
hurt to try.
refs: #1051
Some users mentioned that there's a lag after selecting text
on X11. Tracing through, I saw that the we invalidate the window
quite a lot when dragging the selection, and the buffer swap could
delay for several ms each time while waiting for the vsync.
Rather than blocking the GUI thread and making it bog down, this
commit adopts a technique similar to the recent Wayland frame sync
changes, except that we enforce a minimum of 33ms between frames
in our own scheduler to avoid blocking for several ms at a time.
This seems to do a decent job of balancing responsiveness during
selection with updating the display, and keeps the buffer swap
delay down to microseconds.
We may want to make this delay configurable.
* WIP: IME support for X11
* Handle text generated by IME.
* Set IME position according to the cursor position.
* Improve IME position handling.
Geometry as well as window focus changes are now handled.
* Dispatch IME strings like it's done on windows.
* Make sure not to silently drop IME errors.
* Respect `use_ime` configuration.
* Add xcb-util as dependency.
* Only update IME position if necessary.
* Formatting.
* Update xcb-imdkit-rs.
* Set IME position under the start of the cursor.
This seems to be the way it is commonly done among gui frameworks.
(Tested with Firefox for GTK and Konsole for QT).
* Update xcb-imdkit-rs.
* Handle systems only providing libxcb-util0-dev.
* Add libxcb to freebsd dependencies.
Required by xcb-imdkit-rs.
* Update xcb-imdkit-rs.
* Try to use more recent gcc on centos7.
* More recent C++ compiler on centos7 as well.
* Also setup correct env on centos7 for tests.
This makes the comparison in https://github.com/wez/wezterm/issues/544
work for me on mac, linux (x11, wayland) and also on Windows but
only using WGL.
It looks like we can use the proper colorspace on all targets
except for ANGLE EGL. For whatever reason, the combination of
glium and ANGLE EGL on windows over-gamma corrects.
AFAICT, the framebuffer and perhaps the surfaces it creates
don't indicate srgb support, and whatever combination of status
they return tickles glium's srgb stuff the wrong way.
I think the "solution" is just to directly use WGL by default.
EGL was on by default because it tended to be more survivable
when graphics card drivers were updated, but the last couple
of times I updated mine it still killed wezterm anyway.
refs: #544
This commit annotates fonts with a boolean that indicates whether
we think it contains glyphs with emoji presentation, and then
passes the cluster.presentation field down to the shaper.
If the presentation doesn't match the current font in the fallback,
then it will be skipped until we exhaust its options.
`wezterm ls-fonts` also shows whether we think a font has emoji
presentation.
refs: #997
This replaces the slightly gnarly subpixel enabled blending in the
shader with Dual Source Blending, which is a technique where the
fragment shader can specify both the primary color (RGBA) as well
as an additional per-channel mask that can be used to alpha blend
with the destination.
This enables artifact-free alpha blending when used together
with a transparent window background.
refs: https://github.com/wez/wezterm/issues/932
This commit ties our invalidation requests together with the surface
frame callback request so that we can throttle our frame rate if
we're busy, but still remain largely idle if we're not changing
any content.
refs: https://github.com/wez/wezterm/issues/884
https://emersion.fr/blog/2018/wayland-rendering-loop/ suggests that
it is best practice to do this so that the compositor doesn't
cause an application to block forever if the window is moved to
an off-screen state.
That article also suggests using the frame callback to schedule
paints; this commit has that code included, but I've left it
disabled because it causes us to repaint at the monitor refresh
rate which is often more frequently than we would anyway;
in our problem scenario we're painting once per second and we
just want to make sure that that doesn't block.
So hopefully this makes the sway/scratchpad experience better!
refs: https://github.com/wez/wezterm/issues/884
WindowState is a bitfield that can represent maximized, full screen
and hidden window states.
WindowState is passed along with resize events, improving on the
prior basic is_full_screen boolean by representing the other states.
Notably, WindowState::MAXIMIZED is used to represent a state where
the window's size is constrained by some window environment function;
it could be due to the window being maximized in either or both the
vertical or horizontal directions, or by the window being in a tiled
state on any edge.
When the window is MAXIMIZED, wezterm will behave as though
`adjust_window_size_when_changing_font_size = false` because it knows
that it cannot adjust the window size in that state.
This potentially helps with #695, depending on whether the window
manager propagates this state information to wezterm. Gnome/mutter
does a good job at this with both X11 and Wayland, but I couldn't get
sway to report these states and I don't know of any other tiling wm
that I can easily install and use on fedora, so there's a question
mark around that.
Switch to using `xterm` rather than `text` for the name of the
xterm style I-beam mouse cursor, as that appears to be more
compatible across themes; the gnome theme aliases text -> xterm
via a symlink.
Improve error diagnostics in the case that no cursor is found.
refs: https://github.com/wez/wezterm/issues/532
Avoid setting the title again if it matches the last thing we set;
this in turn avoids receiving a property update from the window
manager to tell us that we set the title.
refs: #964 (but isn't the root cause)
This commit causes a window-config-reloaded event to trigger
when the appearance (dark/light mode) is changed on macos.
It also arranges to propagate the window level config to newly
spawned panes and tabs, created both via the gui and via the
CLI/mux interface.
refs: https://github.com/wez/wezterm/issues/894
refs: https://github.com/wez/wezterm/issues/806
I added this originally thinking that it would make it easier to resolve
https://github.com/wez/wezterm/issues/695 and to integrate wgpu support,
but it's the cause of https://github.com/wez/wezterm/issues/922 so let's
take it out and more directly connect the window events to those in the
terminal.
This commit likely breaks mac and windows; pushing it so that I can
check it out and verify on those systems.
The main culprit was the calloop feature that is used by default
in the underlying SCTK crate.
This commit:
* Routes keyboard processing via the same keyboard mapping code
that we use for X11
* Implements key repeats directly, and with awareness of elapsed
time in case the repeat rate is quicker than the event dispatching
quantum
* Disables the calloop feature of SCTK and let us do our own polling
of the wayland connection.
Critically, key repeat is sticky and unpredictable while calloop is
enabled.
closes: #669
This simplifies it a bit and exposes the config via the config file;
the following options are possible, each one specifies a color
```lua
return {
window_frame = {
inactive_titlebar_bg = "",
active_titlebar_bg = "",
inactive_titlebar_fg = "",
active_titlebar_fg = "",
inactive_titlebar_border_bottom = "",
active_titlebar_border_bottom = "",
button_fg = "",
button_bg = "",
button_hover_fg = "",
button_hover_bg = "",
}
}
```
refs: https://github.com/wez/wezterm/issues/761
This isn't perfect by any means:
* Should allow configuring a sans-serif font
* Emoji need to be scaled
but it allows us to upgrade SCTK without loosing the titlebar
or any control over client side decorations.
You can run `cargo build --release --no-default-features` to build
without wayland support.
This is useful for systems that do not have wayland (eg: the `slint`
distro).
in https://github.com/wez/wezterm/issues/767 CTRL-Tab was getting
incorrectly normalized to CTRL-i; that normalization is valid (Tab is
actually equivalent to CTRL-i as far as unix terminals are concerned)
but unwanted at this layer.
I suspect that this change will come back to haunt me in the future,
as keyboard input is a bit of a zoo.
At the bottom of https://github.com/wez/wezterm/issues/815 is some
discussion about an apparent hang.
Let's make the self pipe writing a bit more robust and log to see
if that might be related.
In an earlier incarnation we needed to wake up more often to paint,
but we now should have reliable event sources for all of the
invalidation cases and we can thus sleep for much longer in
the main event loop.
refs: #770
When a window is being destroyed we expect the receiver end
to be disconnected, so we don't want to break out of the
message loop if a couple of residual windows fail to notify.
Removes the callbacks type and replaces event dispatch with
an async capable channel.
This makes it a bit simpler to model some of the window internals,
and to prepare for a wgpu enabled future.
This changes have been tested only on linux so far.
This should help us reason about whether a resize respected
our request.
This may break linux and windows builds, will fixup once
this is pushed.
refs: https://github.com/wez/wezterm/issues/695
We were getting abs() of the backing-rect-adjusted value, so we need to
restore the sign to avoid what should be logically cell position "-1"
being treated as cell position "1" and so on.
With this change, the logic in the gui layer successfully clamps the x
coordinate to zero when trying to drag left off the window.
NSCursor hide/unhide must be perfectly balanced otherwise
the cursor can vanish unexpectedly when the mouse leaves
the window, and not come back.
So, let's use NSCursor:setHiddenUntilMouseMoves instead; it's
not 100% fulfilling the promise of the API, but it's close enough
and should avoid the permanent invisibility issue.
refs: https://github.com/wez/wezterm/issues/618
This commit allows the x11 window implementation to detect changes
in the DPI that occur after a window is created.
These can occur when changing desktop resolution or when changing
the accessibility option for "Large Text" in gnome.
In order to avoid continually polling for the value on every resize,
we look for the `_GTK_EDGE_CONSTRAINTS` atom in our property change
notifications. This seems to be sent at least as often as the
dpi/scaling changes.
It's also worth noting that some dpi changes don't generate resize
events, so we can't just read the dpi value on every resize, because
we'd miss some of those changes.
Part of this commit changes the font scaling logic: previously
we'd keep a notion of "dpi scale" to apply. That dates from an
earlier time in wezterm where we didn't think that we knew an
actual dpi value.
The way that worked was that we'd compare our current guestimate
of the DPI against what we though the baseline OS dpi should be to
produce a scaling factor.
On X11 that dpi value is global and we'd effectively always produce
a revised scaling factor of 1 after we'd set up the initial window.
This commit changes that logic to just pass down the actual DPI value
to the font code. That DPI value already accounts for HiDPI scaling
so this is hopefully a NOP change for the other systems.
refs: https://github.com/wez/wezterm/issues/667
On macos, once all panes were closed, the GUI would request that the app
stop, but it wouldn't actually exit until a UI event occurred.
This was mostly noticeable when running debug builds from another
terminal.
Calling NSApp::abortModal is sufficient to knock us out of that state.
One of the default key assignments was registered as `SUPER+SHIFT+{`
which worked on macOS, but on Linux, would never match because the
keypress over there was (correctly) reporting as `SUPER+{`.
I originally thought that the user reported issue was a linux
normalization problem, but in looking deeper, the issue is really
that macos is doing something funky!
On macos we collect the interpreted key event as a string, and also
the interpretation of that event without any modifiers applied.
For letters this means that eg: `ALT-l` reports as `¬` for the
processed string and `l` for the unmodified string. That's good!
However, for punctuation we get a backwards result: SUPER+SHIFT+[
produces `[` for the processed text and `{` for the unmodified
text!
This commit tries to detect this, using a heuristic that is
potentially bad on non-US layouts: if both the processed and
unmodified strings are punctuation then we bias to the unmodified
version.
With that change, that key press is correctly reported as `SUPER+{`,
and we can fix the key assignment registration to reflect that.
I quickly checked the behavior of pressing that same physical key
combination with a DEU layout active, and it appears that the unmodified
stuff is also flipped there; we get a lower-case version of something
that I think should be uppercase. This commit doesn't change that
behavior:
```
key_event KeyEvent { key: Char('ü'), modifiers: NONE,
raw_key: Some(Char('Ü')),
raw_modifiers: SHIFT | SUPER,
raw_code: Some(33),
repeat_count: 1, key_is_down: true }
```
refs: https://github.com/wez/wezterm/issues/601
There's something fishy with colorspaces and blending.
This commit removes the `window::Color` type and replaces
it and the confusing array of color types exposed by the
`palette` crate with a pair of much simpler types:
`LinearRgb` - a tuple of f32 linear color components
`SrgbaPixel` - the u32 sRGBA pixel representation
This doesn't change anything about rendering, it just
makes it a bit simpler and makes the SrgbaPixel -> LinearRgb
conversion happen slightly earlier which shaves off some
ad-hoc conversions.
Refs: https://github.com/wez/wezterm/issues/544
* Make window invalidation more efficient by avoiding spawning a call
that spawns a call to invalidate the window. Just directly mark as
invalidated.
* Suppress default background erase
* hoist the bg_color calc for quads that don't have Cells outside of
its loop.
refs: https://github.com/wez/wezterm/issues/546
I'm not sure what exactly changed (perhaps it was a Windows updated?)
but window_background_opacity was only taking effect for windows
with no title bar.
I found that explicitly configuring a region makes transparency
work again.
refs: #553
Continuing along the same lines as the prior commit, the goal
of this commit is to remove the buffer transformation that was
part of uploading the texture to the GPU provided surface.
In order to do so:
* The sense of our local textures needs to change from bgra32 to rgba32.
bgra32 was a hangover from earlier versions of our window crate that
allowed direct-to-fb writes in software mode. We had to pick bgra32
for that for the broadest OS compatibility. I believe that that
constraint has been totally removed, although there is a chance that
this will flip the colors on macos.
* There was an additional linear-to-srgb conversion inlined in that
buffer transformation. I have no idea where that is needed because
the source data is carefully constructed as SRGB. I don't yet know
how to signal that, but for now I've moved that gamma correction
into the shader when we sample the texture.
With this change, timg playback now has vtparse as the hottest
region of code.
refs: #537
Two issues highlighted by profiling:
* Clearing the texture takes a non-trivial percentage of the profile.
The docs suggest that it is better to create a new texture than
to update large portions of a texture, so add some plumbing so
that we can do that in the first texture-full case.
* Next on the list is the code that translates from linear BGRA to
SRGBA. This is present for reasons that I believe are now legacy,
but for the moment: those two primitives now have faster and
easier implementations, so simplify to those.
This improves the timg video playback performance by ~10% for me.
refs: #537
The leftmost pixel was being set to at least 1 by the scale
function.
Fix that up by computing the x coordinate without calling
the scale function.
refs: https://github.com/wez/wezterm/issues/536
