Summary:
Client uses a static Xcb::Window helper. This so far didn't get
explicitly destroyed, so the application finalize cleaned it up.
To destroy the window the xcb_connection_t* is used which the
QGuiApplication already destroyed.
This change ensures that the window gets destroyed before the xcb
connection gets destroyed.
In addition an assert is added to KWin::connection() to ensure that
we still have the QGuiApplication::instance() when it's invoked.
This way we'll notice if we have more cases where we call into xcb
after the application went down.
Reviewers: #plasma
Subscribers: plasma-devel
Projects: #plasma
Differential Revision: https://phabricator.kde.org/D1573
The damage event of the Surface does not consider the offset through
the server side decoration. Due to that we need to have a different
repaint and damage area. As Toplevel::addDamage(QRegion) is only used
in the Wayland case the repaint adjustment is removed from the generic
implementation and moved into the specific implementations. While it
wouldn't hurt to have the repaint adjustment in Toplevel, it would
trigger a repaint of an unchanged area.
For Client I'm not sure whether it's correct without considering the
clientPos. My testing shows it's working, but that might also just be
do to Xwayland using OpenGL buffers. Needs further testing.
The only clients that really seem to rely on it are
Java and they're apparently "broken", resp. extremely
picky on the tiemstamp and probably happily refuse focus
if it's *not* equal to the current server time (ie. anything else
happens at this moment) and overmore feel oblieged to act as WM
by juggling around VDs at all.
BUG: 347153
FIXED-IN: 5.6
REVIEW: 126753
Replacement for calls to info->input() which is only valid for the Client
sub class, but not for ShellClient.
In ShellClient the implementation is swapped with wantsInput() and
wantsInput() has a new implementation which properly delegates to rules()
just like Client does.
This includes the methods:
* decoration()
* decoration() const
* isDecorated() const
In addition new protected methods are added to destroy the Decoration
and to set it.
Usage of m_decoration in Client code is adjusted.
-use qstringliteral only when necessary (i.e. not in concat or comparison)
-use qbytearray instead of qstring when dealing with latin1 input and output (glplatform)
-use qstringref to extract numbers from strings (glplatform)
-define qt_use_qstringbuilder to optimize all string concatenations
-anidata: use ctor init lists, add windowType member initialization
REVIEW: 125933
If a rule minus the title match matches,
the captionChanged signal is bound to re-evaluate
the rules for that client, ie. the tracking overhead
only exists for those clients where title matching
is relevant and costs rematching all rules when such
client changes its title (yes, the partial matching
rules could be stored for faster re-check, but that
would make the patch bigger and is probably not worth
it; just some string comparisms)
additional tracking of wm_class or wm_role
(what is iirc a netwm violation anyway) would require
to monitor the resp. property for changes (not done atm.)
BUG: 220227
FIXED-IN: 5.5
REVIEW: 125427
Methods are no longer virtual. The only x11 specific usage in these
methods (resizeInc) is replaced by a virtual method. Default resize
increments is QSize(1,1) for AbstractClient.
Implementation is moved to abstract_client.cpp as so far events.cpp
does not have any code from AbstractClient.
This includes moving the electricMaximizingDelay from Client to
AbstractClient.
Includes moving the m_cursor and Qt::CursorShape cursor() method to
AbstractClient. In addition AbstractClient now emits a signal whenever
the shape changes allowing Client to react on it (update the low level
cursor) and also hopefully the Wayland Backends to react to it, so that
we have the cursor.
Method added so that we can call into TabGroup::updateStates from code
in AbstractClient. Default implementation does nothing, implementation
in Client performs the actuall call.
As a first case added to Client::move in order to be able to move the
implementation to AbstractClient.
Add action type to screen edge show to allow raise/lower as well as
autohide
Add an action type to screen edge show to allow raise/lower as well as
autohide. This uses the same atom, using a mask to separate type and
location.
The logic for handling geometry changes is moved from the screenedge to
the client so that we can handle both types without screenedge needing
to know what the raise is for.
REVIEW: 124272
Merges together the code from ShellClient and Client and removes the
starting differences. Long term it's better to have only one
implementation to prevent diversions in the implementation.
As it doesn't match exactly protected virtual methods are called
which allow more specific implementations for a certain aspect of the
layer resolving.
AbstractClient::mainClients is virtual and overriden in Client,
allMainClients has only a common implementation in AbstractClient.
In activation.cpp we still need one case where a temporary ClientList
needs to be constructed. Once transients are fully migrated that should
be removable again.
The user might just want to move the window from
one screen to another, no point in wasting time to
show the indicator
BUG: 352050
REVIEW: 125024
FIXED-IN: 5.5
notably *after* storing the old values.
Otherwise the old value is polluted because of
m_hints being nullptr, thus a default value is
returned (instead of the actual old value)
BUG: 347818
FIXED-IN: 5.4.2
REVIEW: 125007
The creation of PlasmaWindowInterface is moved from WaylandServer into
AbstractClient. This allows the sub classes to better control when to
create/destroy the Client.
For creation it's bound to becoming visible - that is Windows which are
only created but never shown are not announced at all.
For Client it's destroyed with the normal tear-down of a Client, for
ShellClient it's destroyed on unmapped (which also means a new one
will be created again in case of another mapping of the surface).
As a side effect, this works around the problem that ShellClients do not
yet get destroyed for QtWayland's menus (needs further investigation).
This is a temporary workaround for bug 349992 which causes freezes
during startup as kwin and kamd dead lock each other on DBus.
To workaround we don't call Activities::create and check in every
usage of Activities::self() whether the pointer is valid.
As a result kwin_wayland now starts pretty fast.
CCBUG: 349992
Prime target is to preserve the in-screen
condition of client AND window.i[1]
Atm. when the client is fully in sight
(but the window is not) - regardless of snapping
or screen change - a workspace update (screen change,
resolution change, adding/removing a strutting panel)
would allow the client to partiall escape screen bounds.
This is changed so that if the client is fully in sight,
it's kept fully in sight (but not the decoration)
If the entire window was fully in sight, it's also kept
(as is right now)
The code handles inner screen edges (if the client was in sight,
the entire window will be if we'd bleed to the other screen)
[1] I'd say that handling the client is more relevant,
but foresee more complaints if the window wasn't handled anymore ;-)
During that, i stumbled across some other issues.
- when a window centered on one screen is moved to a screen smaller
than the window, the window is shrinked to the dimensions of that
screen and now randomly touches eg. left AND right edge. When
moved back, the right and bottom edge were preferred
(to the window was "moved" into the lower right corner).
It's now kept centered.
- geom_restore was saved before keeping the window in the new
screen area (causing accidental drops on screen changes)
BUG: 330968
REVIEW: 122517
FIXED-IN: 5.4
Prime target is to preserve the in-screen condition
of client AND window.[1]
Atm. when the client is fully in sight (but the window is not) -
regardless of snapping or screen change - a workspace update
(screen change, resolution change, adding/removing a strutting
panel) would allow the client to partiall escape screen bounds.
This is changed so that if the client is fully in sight,
it's kept fully in sight (but not the decoration)
If the entire window was fully in sight, it's also kept
(as is right now)
The code handles inner screen edges (if the client was in sight,
the entire window will be if we'd bleed to the other screen)
[1] I'd say that handling the client is more relevant,
but foresee more complaints if the window wasn't handled anymore ;-)
During that, i stumbled across some other issues.
- when a window centered on one screen is moved to a screen
smaller than the window, the window is shrinked to the dimensions
of that screen and now randomly touches eg. left AND right edge.
When moved back, the right and bottom edge were preferred
(to the window was "moved" into the lower right corner).
It's now kept centered.
- geom_restore was saved before keeping the window in the new
screen area (causing accidental drops on screen changes)
BUG: 330968
REVIEW: 116029
FIXED-IN: 5.3
a) calculateGravitation on the same borders forth and back is idempotent
b) do not checkWorkspacePosition for shaded windows
may falsely constrain the unshaded geometry into screen bounds
REVIEW: 123882
BUG: 348064
At the same time the functionality for the "original_skip_taskbar" is
splitted out. This removes the weird API with two boolean arguments to
the set method. Instead there is a dedicated method for the orignal
skip taskbar state which delegates to regular skipTaksbar.
So far input events were sent through Xwayland which is not needed as
we have all information available. Even more it had the pointer surface
on the wrong window when interacting with decorations as it was on the
window and not on the decoration.