xyz -> xy and xy -> xyz conversions have divisions in them, so we need to
handle the edge cases of the divisor being zero. This can happen if an ICC
profile is invalid, or if the XYZ color space is used.
Workspace::cascadeOffset() can be called by placement code before the
frame geometry has a valid position. That is wrong. In order to avoid
that, use the placement area that we are given.
When the user closes the active window, with separate screen focus disabled, a
window on the other screen might get activated, which also switches the active
screen to the other one. As this is quite unintuitive, and in my testing having
separate screen focus enabled didn't have any other unintuitive side effects,
this commit enables separate screen focus by default
This is done by
- fixing isFuzzyScalingOnly to not check the [3, 3] component, which is always 1
- making the comparison between transfer functions fuzzy, so small floating point
errors don't prevent two practically identical functions to be optimized out
- switching manual optimizations to use addMatrix instead, which removes the
matrix or replaces it with a multiplier
and the autotest is expanded to test transformations between color descriptions with
transfer functions and reference luminances that are just scaled versions of each
other
...by removing the keyboard flag, moving keyboard shortcut handling into a separate
method and making the position to tile the window at explicit
This also means KWin doesn't do as many intermediary changes to window geometry on output
hotplugs, which may work around some clients not always reacting to no-op changes.
CCBUG: 479694
Right now it's just a helper to mark items as being affected by some effect,
to prevent direct scanout of the relevant item without needing to block direct
scanout for the whole screen
Rendering intents describe how to handle mapping between different colorspaces,
what to do with out of gamut values and what to do if the whitepoint doesn't match.
This way, clients can choose which behavior their content should get.
Using the custom values for min. and max. luminance in transfer functions, we can reduce the
ranges of values in the shadow buffer to be limited to [0, 1], and with that we can switch
from a floating point buffer back to a normalized format. As gamma 2.2 encoding is much more
efficient at storing color values, this also drops the buffer from 16bpc down to 10bpc.
Furthermore, this offloads the gamma 2.2 -> PQ conversion to KMS when possible, and then uses
the scanout buffer with gamma 2.2 encoding directly. This way the shadow buffer gets completely
skipped and performance and efficiency get improved a lot.
BUG: 491452
CCBUG: 477223
It just tests rec.709 <-> rec.2020 at 0%, 50% and 100% rec.709 luminance, to have
a very simple indicator for when something's gone really wrong while working on
color pipeline changes
This redefines the transfer functions to have a custom luminance at encoded
value zero, and a custom luminance at encoded value 1, neither of which are
tied to the reference luminance, even for relative transfer functions.
The goal of that is that we can use a gamma 2.2 transfer function for the shadow
buffer, with the reference luminance being much lower than the maximum luminance.
For example, on an HDR screen you might have the reference luminance of 600 nits,
while the maximum luminance is 1000 nits. By representing this in gamma 2.2, we
can use a much smaller amount of bits per color to store the values than if we
used a linear transfer function. An additional benefit is that this way the values
in the buffer can be scaled by arbitrary amounts, for example to limit the range of
values to [0, 1], which can be represented in a normalized buffer
Clients can have valid reasons to change the selection when the
same user action that also caused the selection request
to lose keyboard focus. This is notbaly the case for menus
created from a Plasma panel which itself will not take focus
but when clicking on action it only triggers after the menu
is closed.
This also matches what weston and sway do.
BUG:490803
We optionally send some keys to xwayland through the filter when no x11
client has focus. This allows shortcut handling in X11 apps to work.
When kwin is grabbing keys we don't necessarily want X11 to sniff these
keys as things can get out of sync. A key place is the tabbox. The X11
client sill has focus, but xwayland is not active. This means we pass
tab keys to X which then go to application incorrectly.
Part of this patch changes the tabbox filter to not intercept the alt
key release event. This ensures xwayland's concept of pressed modifiers
stays in sync.
BUG: 484992
In some IM backends pre-edit text should be submitted on user interaction, in some it should be discarded.
In TextInputV1 and V2 this was a flag sent to the client along with
the commit string ahead of time.
TextInputV3 does not have a flag for this, so we handle it compositor side.
We flush the text to be committed :
- when we change keyboard focus, before the current client gets wl_keyboard.leave
- when a mouse is pressed in the relevant surface
- when a key is pressed and the InputMethod doesn't have a grab
- when the InputMethod forwards a key to the client
(which includes the InputMethod passing on grabbed keys)
When the tabbox grabs input, the key events won't be sent to the client.
It's not good for a couple of reasons: first, it can fool the client into
repeating a previously pressed key; and the server side and the client
side can be out of sync after the task switcher is dismissed.
In order to handle that properly, this change makes the tabbox reset the
keyboard focus when it's shown or hidden. When the task switcher is
dismissed, an enter event will be sent with the current state of the pressed
keys.
Ideally, the same needs to be done with the pointer focus but it's
challenging to achieve with the current input pipeline design. An input
grab abstraction is needed to handle pointer focus when the task switcher
is active.
This is needed to better support emitting mouse button events when
pressing tablet buttons. It's common in many art programs that an action
is tied to a mouse button + a modifier, such as panning the canvas. So
being able to send keyboard modifiers in tandem with mouse buttons is
very useful when rebinding.
Tests are added for this new feature.
CCBUG: 469232
KeyboardInterface is a multiplexer, it has a global state to kwin
that forwards events the single focussed window.
XWayland also forwards events to clients, but uses the keyboard interface.
It has some overloads that take a specific client, this was used for key events
but not modifiers.
The end result was not only that XWayland could miss a modifier update, but
also that wayland clients would get modifier updates out of order.
Key events must come first.
BUG: 490270
If an output is deleted, the Workspace::desktopResized() is going to
re-assign windows to the new outputs. It is done so so the workspace
re-arrangement procedure is deterministic and has concrete order.
However, with the current Window lifecycle management, it's possible
to encounter the follwing case:
- xdg_toplevel gets created on output A
- xdg_toplevel initial state is committed
- output A is removed
- a wl_buffer is attached to the xdg_toplevel, which results in a
geometry change and an output change
- Window::setMoveResizeOutput() is called, but the previous output
is a dangling pointer
CCBUG: 489632
Since the tablet cursor and the mouse cursor is tracked separately,
rebinding a tablet button to a mouse click is sort of wonky. For
example, if you assign it to Right Click and attempt to open a context
menu it will appear to open in the wrong place.
So before we send the mouse button event, set the mouse position to
the tablet cursor position. A test is added to ensure this functionality
works as intended and doesn't regress.
This is to not destroy the tablet state for the next test (which is
added in the next commit.) We could also explicitly do a tablet tool
up/down dance here, but all we need is the tablet tool to be in
proximity for the tablet button event to fire.
This adds autotests for binding to tablet pad and tool buttons, which
was previously untested. Of note is that we don't explicitly test mouse
buttons, which is already tested in the other functions.
NETWinInfo queries the window state and setState() will do nothing if the
proposed new state matches the cached server side state.
On the other hand, given how the test is structured, there can be pending
fullscreen changes on the kwin side that are yet to be sent to the X server
when the NETWinInfo object is created.
This change fixes that race condition by adding an explicit Xcb::sync().
The effect only modifies the opacity of individual windows that
WorkspaceScene::scanoutCandidate will reject anyways, so there's no reason
for it to block direct scanout.
Once a more proper solution for blocking direct scanout on individual items
is in place, this can be removed again
BUG: 487780
pipewire 1.2 has a known regression where the client's process callback
will not be triggered if the producer has sent only one buffer.
The issue has already been fixed upstream
525360d70a
Until our CI picks up that change, temporarily disable the test to unblock
merging patches.
