The ifdefs for have_gbm obfuscate the code unnecessarily - the drm backend
is not a great experience with qpainter, so in practice noone should ship
it without gbm anyways.
The Compositor contains nothing that can potentially get dirty and need
repainting.
As is, the advantages of this move aren't really noticeable, but it
makes sense with multiple scenes.
Backend parts are far from ideal, they can be improved later on as we
progress with the scene redesign.
The main idea behind the render backend is to decouple low level bits
from scenes. The end goal is to make the render backend provide render
targets where the scene can render.
Design-wise, such a split is more flexible than the current state, for
example we could start experimenting with using qtquick (assuming that
the legacy scene is properly encapsulated) or creating multiple scenes,
for example for each output layer, etc.
So far, the RenderBackend class only contains one getter, more stuff will
be moved from the Scene as it makes sense.
The proprietary NVidia driver now supports gbm, which vastly improves the
user experience. For older devices that will not get gbm support dropping
EglStreams will likely not have a big impact as it has several session breaking
issues anyways.
By removing the backend a lot of logic can be simplified, most notably multi-gpu.
The current "Minimize Overlapping" window placement tends to position
windows in locations that seem completely random, typically in a screen
corner. It is doing this because, true to its name, it is trying to
avoid overlapping other windows as much as possible. However in practice
this is rarely helpful. When the user opens a new window, it's because
they want to use it, and positioning the window far from where the
user is likely to be looking is counter-productive. This is even more
true on today's large and wide displays, where placing the window in a
corner may position it entirely outside the user's current field of
vision. We get bug reports about this exact issue for notifications
(which always appear in a corner by default) by users of such screens.
For notifications, this can be justifiable because notifications are
designed to be ignorable; app windows on the other hand, are not.
As a result, I commonly see Plasma users open windows and then
immediately, reflexively grab the window's titlebar and drag it to the
center of the screen. I have seen my wife do this. I have seen every
YouTube reviewer of Plasma do this. I have even see fellow KDE
developers at sprints do this. It seems like quite a common impulse
to want a newly-opened window to appear in the center of the screen,
which is where the user is likely to already be looking.
Thankfully, KWin already has a window placement mode that does this
automatically: "Centered". Accordingly, this commit changes the default
KWin window placement mode from "Minimize Overlapping" to "Centered".
No kconf migration script is provided because this is a better default
for most people in most cases, and existing users are highly likely to
appreciate this change.
The main motivation behind this change is to move management of drm
blobs out of property wrappers in specialized wrappers to simplify state
management with blobs.
Connector mode blobs are created on demand.
When we switch CRTCs it can happen that a CRTC would stay enabled yet has
no connectors anymore. In this case the kernel may reject our atomic commit,
which would cause the modeset to fail. To counteract that, properly disable
unused drm objects
Currently KWin is combining modesets with presentation, which causes problems
when multiple monitors are used and crtcs need to be switched around, because
taking away a CRTC from another output causes the driver to disable the
other output. In order to avoid such problems, delay presentation until
all pipelines are ready to present and then do a modeset with a single atomic
commit. To process the resulting page flip events properly this commit also
ports KWin to page_flip_handler2 and changes how the pageFlipped and
notifyFrameFailed signals are processed.
Hardware constraints limit the number of crtcs and which connector + crtc
combinations can work together. The current code is searching for working
combinations when a hotplug happens but that's not enough, it also needs
to happen when the user enables or disables outputs and when modesets are
done, and the configuration change needs to be applied with a single atomic
commit.
This commit removes the hard dependency of DrmPipeline on crtcs by moving
the pending state of outputs from the drm objects to DrmPipeline itself,
which ensures that it's independent from the set of drm objects currently
used. It also changes requests from KScreen to be applied truly atomically.
The GlStrictBinding flag indicates whether it's okay not to re-bind the X11
pixmap to the OpenGL surface texture if the corresponding window is damaged.
It doesn't really affect the SceneOpenGL, only low level backend stuff.
This ensures that the window will have correct geometry if a maximized
window changes preferred decoration mode. X11Client does something
similar, see X11Client::updateShape().
In hindsight, perhaps, AbstractClient::{create,destroy}Decoration() must
preserve the old frame geometry, but it's not clear how to do that
because it requires decoration updates to be truly async, otherwise
there will be ugly flickering.
Currently, the scene owns the renderer, which puts more
responsibilities on the scene other than painting windows and it also
puts some limitations on what we can do, for example, there can be only
one scene, etc.
This change decouples the scene and the renderer so the scene is more
swappable.
Scenes are no longer implemented as plugins because opengl backend
and scene creation needs to be wrapped in opengl safety points. We
could still create the render backend and then go through the list
of scene plugins, but accessing concrete scene implementation is
much much simpler. Besides that, having scenes implemented as plugins
is not worthwhile because there are only two scenes and each contributes
very small amount of binary size. On the other hand, we still need to
take into account how many times kwin accesses the hard drive to load
plugins in order to function as expected.
This allows using base opengl backends in libkwin, which can be useful
later on for the purpose of moving the ownership of render backends from
the Scene class to the Compositor class.
It has been disabled with Mesa for almost half a decade due to false
positives and even if it weren't disabled, it contributes to the startup
time.
The commit message that added the self test doesn't explain why it was
added, but if it was added to detect unstable drivers, it's not worth it.
With an opaque fullscreen window we can be sure that items under it don't
actually require us to repaint. This should yield some small efficiency
improvements and resolves stutter with adaptive sync.
BUG: 443872
FIXED-IN: 5.23.3
Windows in workspace.clientList() are sorted in the map order. This
means that the minimize all script will try to activate the last mapped
window when unminimizing windows, which is a bit annoying.
This change ensures that the minimize all script doesn't activate wrong
window by minimizing and unminimizing windows in the stacking order.
It's not a bullet-proof solution though, but it should produce good
enough results.
This improves file organization in kwin by putting backends in a single
directory.
It also makes easier to discover kwin's low level components for new
contributors because the plugins directory may come as the last place to
look for. When one hears "plugin", the first thing that comes to mind is
regular plugins, not low level backends.
This ports the nested wayland platform plugin to the InputDevice
abstractions.
Some global handling logic has been simplified to make porting more
straightforward.
The main motivation behind this change is to prepare input abstractions
for virtual input devices so the wl_seat can properly advertise caps or
the cursor getting properly mapped/unmapped when a fake pointer is
added/removed on a system without a hardware mouse connected.
With this, there are three abstractions - InputDevice, InputBackend, and
InputRedirection.
An InputDevice represents an input device such as a mouse, a keyboard, a
tablet, etc. The InputBackend class notifies the InputRedirection about
(dis-)connected devices. The InputRedirection manages the input devices.
Such design allows to unify the event flow for real and virtual input
devices.
There can be several input backends active. For example, the libinput
backend and an input backend that provides virtual input devices, e.g.
libeis or org_kde_kwin_fake_input.
If an output is disconnected, the Workspace will update the
Toplevel.output property for all windows that are on that output, then
it will call AbstractClient::checkWorkspacePosition() to fix window
position.
That may result in some windows partially sticking outside visible area
because AbstractClient::checkWorkspacePosition() has no idea what output
the window was.
This change addresses that problem by delaying updating the
Toplevel.output property so AbstractClient::checkWorkspacePosition()
could pick better window placement.
The role of enabledByDefaultMethod was misinterpreted. It exists only
for kcms so they can display undetermined state. EnabledByDefault should
still be set to a valid value.
Currently, if an output is hotplugged, all windows will be scrambled,
which is highly annoying.
With this change, windows will stick to their outputs if an output has
been connected or disconnected.
BUG: 296673
BUG: 378896
BUG: 412703
BUG: 443698
Currently, AbstractClient::geometryRestore() is abused to put windows
back on their original screen. It makes window placement more complex
and it breaks restoring initially maximized windows.
After the decoration is destroyed, we need to resize XdgToplevelClient
to ensure that the scheduled configure event will have correct size.
As is, xdg-toplevel configure events are delayed. When it's actually the
time to send a configure event, XdgToplevelClient will compute the
requested client size from the moveResizeGeometry().
If the moveResizeGeometry() still includes window decorations, the
window will get bigger.
BUG: 444119
For optimization purposes, kwin will ignore repaint regions scheduled by
invisible windows, e.g. hidden docks, minimized windows, etc.
The problem is that it sort of breaks w->addRepaintFull(). If a lot of
animation frames are dropped, for example due to heavy cpu load, the
sliding popups animation can jump from the middle of animation right up
to the end. It will schedule a repaint but it will be ignored.
In order to work around that issue, this change makes the sliding popups
effect schedule workspace repaints in postPaintScreen() to ensure that
the Scene will always repaint dirty regions.
Hopefully, this should fix bugs where auto-hide panels sometimes flicker
on Wayland.
BUG: 444502
Without spacing it is not immediately evident where a doc comment
ends and a new declaration begins. Adding a line of whitespace
makes everything more readable and easier to scan quickly.
Currently, input events can flow directly to input device handlers and
sometimes they go through the InputRedirection, which simply forwards
them to the corresponding input device handler.
In case the user has just the one display but they don't want to show it
in their main workspace when sharing video, allow creating a virtual
display.
This also will allow using remote devices as support displays.
The WindowThumbnail item uses the GLTexture class. In order to destroy
the thumbnail texture, the item will schedule a destroy job.
This means that the GLTexture can be alive during or after graphics
reset.
On the other hand, as an implementation detail, GLTexture::clear() may
allocate a framebuffer object, which is going to be destroyed together
with the last texture.
Given that window thumbnail textures can be still alive after a graphics
reset and the fact that GLTexturePrivate::s_fbo gets destroyed when the
last texture is destroyed, kwin can end up trying to clear a decoration
texture with now defunct s_fbo.
Since the old old s_fbo is inert, the glBindFramebuffer() function will
fail and the glClear() operation will affect the default framebuffer,
thus leading to black flickering visual artifacts.
In order to fix that issue, this change makes GLTexture destroy s_fbo
unconditionally in GLTexturePrivate::cleanup() which is called whenever
OpenGL stuff is about to tear down, e.g. due to graphics reset, etc.
BUG: 443951
Without this there can be a mismatch between what clients expect and what
KWin actually uses, causing for example stutter in video players.
BUG: 444303
FIXED-IN: 5.23.2
InputDeviceHandler keeps track of which client has focus, and whether
the decoration or main window has focus.
Should the decoration get destroyed whilst the decoration has focus we
don't update correctly. This can happen when
"BorderlessMaximizedWindows" is set, or if you use key presses whilst
above the decoration.
cleanupDecoration (which is an odd name) is called whenever a decoration
gains or loses focus. Here we can connect for the decoration destruction
and force a focus change.
BUG: 411884
The screen count can be retrieved by checking the number of items in the
EffectHandler.screens property.
The replacement for the numberScreensChanged signal are the screenAdded
and the screenRemoved signals.
The main motivation behind this change is to clean up the screens api
and reduce the number of usages of the Screens class.
This provides a way to create, destroy, and rename virtual desktops in
the overview effect as well as switch between desktops.
The mechanics of switching between virtual desktops can be revisited
later though.
This can be useful for QML scripts that deal with virtual desktops.
The virtual desktop model keeps a copy of virtual desktop objects to
avoid hitting asserts in QAbstractItemModel (it has some asserts to
ensure that indices passed to beginInsertRows() or beginRemoveRows()
make sense).
When we're adding the output to the EglGbmBackend pipelines aren't necessarily
setup yet and are thus missing the modifier list. As creating the gbm surface
immediately is useless anyways, delay that until we need it, where the modifier
list is available.
