Summary:
If the pointer is constrained all mouse events should go to the window.
Also our Alt+click. To use Alt+click nevertheless one can just unconfine
the window.
CCBUG: 399375
Test Plan: Run the adjusted autotest before and after change
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15982
Summary:
A window maximising is an async operation. We work out what size we want
the client to be, then request the client to update. The window isn't
really maximised until we get that new buffer with the new size.
This patch splits the requested, pending and current state, updating as
appropriate.
Things are a bit complex with things like borders. Technically we
shouldn't update them till we get a response, but we also need to have
the correct geometry of the full size window in our request. For now
they behave as before, updating when we request the change.
X code is untouched.
This hopefully fixes maximise animations on wayland as now we update the
geometry before emitting maximisedChanged.
Test Plan:
Maximised a window with the button and double clicking title bar.
I get only the following events on maximise/restore:
19:51:39.156 KWin::EffectsHandlerImpl::slotGeometryShapeChanged geometry
shape changed QRect(47,24 640x509) QRect(0,0 716x573)
19:51:39.157 KWin::EffectsHandlerImpl::slotClientMaximized slot client
maximised true true
19:51:40.522 KWin::EffectsHandlerImpl::slotGeometryShapeChanged geometry
shape changed QRect(0,0 716x573) QRect(47,24 640x509)
19:51:40.522 KWin::EffectsHandlerImpl::slotClientMaximized slot client
maximised false false
BUG: 382698
Reviewers: #kwin, romangg
Reviewed By: #kwin, romangg
Subscribers: romangg, zzag, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15150
Summary:
When we want to change a client's size and position together we have to
request the client becomes a new size and only then move the window to
the new location.
Currently we process the new position the next time the buffer updates,
but with no guarantee that it has actually tried to resize/whatever yet.
The client could be providing a new buffer just because the contents
have changed.
XDGShell has an acked serial designed to keep everything precisely in
sync. A surface represents the last configure that was acked.
This patch tracks the pending position for each configure and applies it
accordingly.
WL_shell does not have this mechanism, so behaviour is kept the same as
before.
----
This is a pre-requisite to syncing maximisedState/isFullScreen with the
configure request.
Potentially we could remove the isWaitingForResizeSync checks when
resizing and it will still resize smoothly.
Test Plan:
Relevant unit test still passes with the client responding
Resized a window from the left edge with WLShell and XDGShellV6
Reviewers: #kwin, romangg
Reviewed By: #kwin, romangg
Subscribers: romangg, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15135
Summary:
Check that the Sliding Popups effect grabs windows no matter in what
order it and the Scale effect are loaded.
Test Plan: Ran the test.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15729
Summary:
Getter is exposed as a property on scripted effect in a way that hides
pointers from the scripting side.
Setter is implicitly handled as a property of newly created animations
and holds the activeFullScreenEffect whilst any of them are active. Like
existing effects it remains up to the effect author to avoid the
problems of multiple full screen effects. The RAII lock pattern is
somewhat overkill currently, but it's the direction I hope we can take
EffectsHandler in next API break.
BUG: 396790
--
This patch is against the QJSEngine port, though it's not conceptually a
requirement.
Test Plan: Unit test
Reviewers: #kwin, zzag
Reviewed By: #kwin, zzag
Subscribers: zzag, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14688
Summary:
QSignalSpy is not happy about KWin::Deleted*:
QWARN : SlidingPopupsTest::testWithOtherEffectWayland(wobblywindows, slide)
QSignalSpy: Unable to handle parameter 'deleted' of type 'KWin::Deleted*' of
method 'windowClosed', use qRegisterMetaType to register it.
Test Plan: Ran the test, haven't noticed any warnings.
Reviewers: #kwin, broulik
Reviewed By: broulik
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15730
Test Plan:
Compiled latest GTK
Ran gtk4-demo and used WAYLAND_DEBUG to confirm it used the correct
shell
tested a top level and a popup
Reviewers: #plasma, romangg
Reviewed By: #plasma, romangg
Subscribers: romangg, mart, graesslin, zzag, kwin
Tags: #kwin
BUG: 398614
FIXED-IN: 5.15.0
Differential Revision: https://phabricator.kde.org/D13530
Summary:
With 2694839099, 0bd5eff862 and f0ba436c72 it is now possible to
unconstrain a pointer by killing the client or deactivating the window, for
example by switching the window via the TabBox.
A user should always be able to do it therefore without needing to explicitly
break the pointer constrain as in the past with the Esc key.
Therefore remove the functionality and also remove the OSD announcing it,
which was often shown at the wrong time.
Test Plan: Manually and auto test adapted.
Reviewers: #kwin, #vdg, hein
Reviewed By: hein
Subscribers: ngraham, hein, kwin
Tags: #kwin
Maniphest Tasks: T8923
Differential Revision: https://phabricator.kde.org/D15234
Summary:
* effects.h includes client.h and forward declares Client, both at the
same time. Thus, delete the include;
* the blur effect includes effects.h. That, most likely, is a leftover
after 3f5bf65a9e.
Test Plan: Compiles.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15191
Summary:
If the corner shadow tiles(top-left, top-right, and so on) tiles are missing,
then the left/top/right/bottom shadow tiles will overlap.
This diff addresses that problem by changing how the shadow texture
atlas is rendered:
* corner tiles will be drawn in the corners of the atlas(buildQuads
method expects them to be at the corners);
* top, right, bottom, and left tile will be aligned to the top-left
corner of the inner shadow rect.
For majority of desktop themes, the shadow texture atlas looks the same.
For example, here's for Aether:
Before:
{F6190484, layout=center, size=full}
After:
{F6190488, layout=center, size=full}
Depends on D14783
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, abetts, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14784
Summary:
Current implementation of buildQuads assumes that corner shadow tiles
are always present:
const QRectF leftRect(
topLeftRect.bottomLeft(),
bottomLeftRect.topRight());
but that assumption is wrong. For example, if the default panel is on
the bottom screen edge, then the calendar popup won't have the
bottom-left shadow tile(at least on Wayland). Which means that the left
shadow tile won't be visible because
topLeftRect.left() == bottomLeftRect.right().
Corner rectangles only have to influence height of the left/right tile
and width of the top/bottom tile. Width of the left/right tile and
height of the top/bottom tile should not be controlled by corner tiles.
Overall, this is how shadow quads are computed:
* Compute the outer rectangle;
* Compute target rectangle for each corner tile. If some corner tile is
missing, move the target rectangle to the corresponding corner of the
inner shadow rect and set its width and height to 0. We need to do
that to prevent top/right/bottom/left tiles from spanning over
corners:
{F6190219, layout=center, size=full}
We would rather prefer something like this if the top-left tile is
missing:
{F6190233, layout=center, size=full}
* Fix overlaps between corner tiles;
* Compute target rectangles for top, right, bottom, and left tiles;
* Fix overlaps between left/right and top/bottom shadow tiles.
Test Plan:
* Ran tests;
* Resized Konsole to its minimimum size(on X11 and Wayland);
* Opened the calendar popup(on X11 and Wayland):
Before:
{F6190344, layout=center, size=full}
After:
{F6190346, layout=center, size=full}
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: abetts, davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14783
Summary:
Let VirtualOutput be a child class of the new generic class Output.
This allows code sharing and a very similar behavior of the Virtual backend
in comparision to the Drm backend.
Test Plan:
Autotests succesful with two exceptions: The decoration input test fails on
testDoubleTap, row topLeft. This is to be expected because now the
ScreenEdgeInputFilter captures the event at position (0,0) before the
DecorationEventFilter can capture it. The autotest was adapted to take this
special case into account.
Also the lockscreen test fails, because the virtual output is currently missing
the physical size yet.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11789
Summary:
In loading tests. Effects are deleted in a deleteLater potentially
outside the scope of our test. Our MockEffectsHandler (which contains
the global static "effects") has the lifespan of the test.
Fixes failing unit test.
Test Plan: Ran test
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D15167
stackingOrder is an interesting property which needed
special handling in the port. Add an explicit test.
Test Plan: #kwin
Reviewers: broulik
Reviewed By: broulik
Subscribers: zzag, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14533
Summary:
It's superseded by the new scale effect(D13461).
Existing users of this effect will be migrated to the new scale effect.
Depends on D13461
Reviewers: #kwin, #plasma, #vdg, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13462
Summary:
The new effect scales windows as they appear and disappear.
As the the most of window animation effects, it is a monolithic effect,
i.e., if you enable scale effect, it will animate *both* the appearing and
disappearing.
The main difference between the Scale effect and the Scale in effect is
that the Scale in effect only animates windows as they appear. There is
no corresponding "the Scale out" effect, which is odd. Other points that
differentiate the Scale effect from the Scale in effect:
* it is more subtle;
* it doesn't animate the log out screen;
* it doesn't conflict with the Fade effect, etc.
... and overall, the Scale effect supersedes the Scale in effect.
{F5904947}
//Window open animation.//
{F5904948}
//Window close animation.//
{F5905283, layout=center, size=full}
//KCM.//
Test Plan:
* Enabled this effect
* Opened/closed System Settings
Reviewers: #kwin, #plasma, #vdg, davidedmundson
Reviewed By: #kwin, #plasma, #vdg, davidedmundson
Subscribers: ngraham, davidedmundson, fvogt, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13461
Summary:
One cannot use a non metatype frrom an external class inside an
invokable. https://bugreports.qt.io/browse/QTBUG-58454
End result is the script engine arguments wouldn't match up and MOC
would just use the default value.
As far as I can tell this has been broken for 6 years.
The global animate method that unboxes a QJSValue as an object is
unaffected.
No shipped kwin effect actually used it.
To some extent we didn't even actually want to enforce the enum as we
also accept custom value of ScriptedEffect::GuassianCurve, so it has
been switched for an int.
Test Plan: Unit test
Reviewers: #kwin, broulik
Reviewed By: broulik
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14669
Summary:
Ready for QJSEngine port and upcoming other fixes.
Split as it makes it easier to do any before/after testing.
Test Plan:
All tests pass with the current QScriptEngine
Verified expected API against a wiki page and current code.
Reviewers: #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14482
Summary:
If the new pointer position is "off screen", PointerInputRedirection
just ignores that new position. So, pointer remains on its previous
position. In some particular cases, like reaching default panel, it
degrades desktop experience because one have to slowly move pointer in
order to reach what he/she wants.
This change addresses that problem by confining the new pointer position
to screen geometry.
BUG: 374867
FIXED-IN: 5.13.4
Test Plan: Ran tests
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D14036
Summary:
Most effects use QTimeLine in the following manner
```lang=cpp
if (...) {
m_timeline->setCurrentTime(m_timeline->currentTime() + time);
} else {
m_timeline->setCurrentTime(m_timeline->currentTime() - time);
}
```
Because effects do not rely on a timer that QTimeLine has, they can't
toggle direction of the QTimeLine, which makes somewhat harder to write
effects. In some cases that's obvious what condition to use to figure
out whether to add or subtract `time`, but there are cases when it's
not. In addition to that, setCurrentTime allows to have negative
currentTime, which in some cases causes bugs.
And overall, the way effects use QTimeLine is really hack-ish. It makes
more sense just to use an integer accumulator(like the Fall Apart
effect is doing) than to use QTimeLine.
Another problem with QTimeLine is that it's a QObject and some effects
do
```lang=cpp
class WindowInfo
{
public:
~WindowInfo();
QTimeLine *timeLine;
};
WindowInfo::~WindowInfo()
{
delete timeLine;
}
// ...
QHash<EffectWindow*, WindowInfo> m_windows;
```
which is unsafe.
This change adds the TimeLine class. The TimeLine class is a timeline
helper that designed specifically for needs of effects.
Demo
```lang=cpp
TimeLine timeLine(1000, TimeLine::Forward);
timeLine.setEasingCurve(QEasingCurve::Linear);
timeLine.value(); // 0.0
timeLine.running(); // false
timeLine.done(); // false
timeLine.update(420);
timeLine.value(); // 0.42
timeLine.running(); // true
timeLine.done(); // false
timeLine.toggleDirection();
timeLine.value(); // 0.42
timeLine.running(); // true
timeLine.done(); // false
timeLine.update(100);
timeLine.value(); // 0.32
timeLine.running(); // true
timeLine.done(); // false
timeLine.update(1000);
timeLine.value(); // 0.0
timeLine.running(); // false
timeLine.done(); // true
```
Test Plan: Ran tests.
Reviewers: #kwin, davidedmundson, graesslin
Reviewed By: #kwin, davidedmundson, graesslin
Subscribers: romangg, graesslin, anthonyfieroni, davidedmundson, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13740
Summary:
This patch changes KWin's pointer constraining behavior by only allowing
constraints if the surface has keyboard focus. In case the client activation
state changes, it rechecks it.
Test Plan:
Manually with the pointer constraints test application and opening the
launcher by pressing meta. Also amended autotest.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, davidedmundson, kwin
Tags: #kwin
Maniphest Tasks: T8923
Differential Revision: https://phabricator.kde.org/D13492
Summary:
It will then be renderered appropriately when painting to the output
buffer.
Test Plan: Updated unit test, plus used with other relevant patches
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13606
Summary:
Should resolve undefined reference to `bool QTest::qCompare<double, int>
error on 5.9
Test Plan:
Still compiles/passes
Not actually tested on 5.9
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13526
This test is blocking the ability of the CI system to return to service for Extragear projects on some platforms.
This commit may not be reverted without the explicit consent of Sysadmin.
CCMAIL: plasma-devel@kde.org
CCMAIL: kwin@kde.org
Summary:
A client might delete its pointer lock/confinement object. This is supposed to
directly remove the pointer lock/confinement in KWin, but did not explicitly
until now.
BUG: 388885
Test Plan:
Tested manually with Neverball, Nexuiz and the new pointer constraints test
application. The pointer constraints autotest is also appended.
Reviewers: #kwin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, graesslin, kwin
Tags: #kwin
Maniphest Tasks: T8923
Differential Revision: https://phabricator.kde.org/D13466
Summary:
Instead of seeing the cursor <--> on the left edge you now see an icon
that looks like |<- .
This brings kwin decorations in line with GTK CSD icons.
In theory this is also useful to tell which window will resize in the
case of side-by-side windows (regardless of whether borders are on or
not). In practice with the adwaita icon theme I tested with it's not
very intuitive to realise which is which till you learn the icon.
Change is more involved than it should be as Qt::CursorShape doesn't
have these entries, and I don't want to shadow that enum internally or
have
to change kwin effect code.
Specifics depend on cursor icon theme if they are not present it will
fallback to the <--> icon. (Breeze does not have them currently)
Test Plan:
Resized some windows (on X and on Wayland)
Correct icon appeared on Adwaita
Existing icon appeared on Breeze
Reviewers: #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D13396
Summary:
QPainter doesn't render decoration shadows. It renders only
shadows provided through ShadowInterface.
With this change, painting of shadows is done in similar way OpenGL backend is
currently doing.
Before
{F5734867, layout=center, size=full}
After
{F5734870, layout=center, size=full}
Depends on D10811 (dummy decoration with shadows in autotests)
Test Plan:
* start kwin with QPainter backend enabled:
```
KWIN_COMPOSE=Q kwin_wayland --xwayland --windowed
```
* open konsole and kate:
```
DISPLAY=:1 konsole
DISPLAY=:1 kate
```
Reviewers: #kwin, graesslin, davidedmundson
Reviewed By: davidedmundson
Subscribers: abetts, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D10943
Summary:
This problem appears if shadow corner tiles are too big and
some window has size smaller than 2 * shadowTileSize.
This change tries to address the problem above by exclusing
overlapping tile parts. If there are any two overlapping corners
then tile between them(top/right/bottom/left) is not rendered.
Also, because some corner tile parts can be excluded, corner tiles
are expected to be symmetrical(i.e. if we remove right half from
the top-left tile and left half from the top-right tile and
stick them together, they still look fine, there are no misalignments, etc).
Most shadows(e.g. shadows from Breeze) have such behaviour.
No tiles are overlapping
{F5728514, layout=center, size=full}
Overlapping tiles
{F5728516, layout=center, size=full}
And this is how it supposed to be
{F5728517, layout=center, size=full}
Test Plan:
* apply D11069 to Breeze
* in System Settings/Application Style/Window Decorations, choose "Very Large" shadow size
* open Konsole
* resize it to a minimum possible size
Reviewers: #kwin, graesslin, davidedmundson
Reviewed By: #kwin, davidedmundson
Subscribers: davidedmundson, ngraham, anemeth, abetts, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D10811
Summary:
BUG: 393253
FIXED-IN: 5.13.0
Test Plan: manual testing and new unit test
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D12633
Summary:
From Wayland documentation:
"When a seat's focus enters a surface, the pointer image is undefined and
a client should respond to this event by setting an appropriate pointer
image with the set_cursor request."
KWin's interpretation so far for the undefined pointer image was to
remove the pointer image when entering a surface waiting for the client
to set a cursor image. This can result in a short flicker as there might
be a frame without a cursor image.
This patch changes the behavior by keeping the previous image till the
application set a new one. This brings some advantages:
* if the application is not responding a cursor is still shown
* if the same cursor is used as in the previous window we don't have a
flicker
CCBUG: 393639
Test Plan: I cannot see the flicker, so only tested with the adjusted tests
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D12631
Test Plan:
Very minimal expansion of unit tests which uses WaylandScreens
Wrote mini app to debug actual output of xdg-output for testing the DRM code
Main relevant user of this is xwayland > 1.20 which I don't have, so that
part remains untested
Reviewers: #plasma, graesslin
Reviewed By: #plasma, graesslin
Subscribers: romangg, graesslin, bshah, kwin
Tags: #kwin
Maniphest Tasks: T8501
Differential Revision: https://phabricator.kde.org/D12243
Summary:
When using composite key combinations, kwin would do random weird
actions when the first key was pressed (e.g. ` key). This makes sure we
are not trying to match.
BUG: 390110
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D12416
Summary:
When using composite key combinations, kwin would do random weird
actions when the first key was pressed (e.g. ` key). This makes sure we
are not trying to match.
BUG: 390110
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D12416
Summary:
Seems to have regressed in Plasma 5.12 due to code reordering. Now the
property is explicitly updated once the NETRootInfo is created.
BUG: 391034
FIXED-IN: 5.12.5
Test Plan:
Test case exposing the problem added. Fails without the patch,
succeeds with the patch.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D10836
Summary:
This patch implements using EGL_IMG_context_priority to request
high-priority rendering contexts if the extension is available.
EGL_IMG_context_priority is currently used in this fashion by
e.g. Android's SurfaceFlinger (RenderEngine.cpp) and libweston
(gl-renderer.c) and seems promising given this widespread
acceptance.
Reviewers: #kwin, graesslin, romangg, davidedmundson
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11572
Summary:
This matches the DRM backend more closely and allows mid-test removal and
addition of virtual outputs with different properties in the future.
Test Plan: Before and after 93% tests passed.
Reviewers: #kwin, graesslin
Reviewed By: #kwin, graesslin
Subscribers: graesslin, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11351
Summary:
Especially when a window is first mapped it might be that the appId is
not yet set. So window rule matching doesn't happen. This change
evaluates the window rules again after the appId changes, so rules for
the appId match.
Test Plan: added test case
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11268
Summary:
This allows to override the desktop file name.
CCBUG: 351055
Test Plan: Created a window rule for telegram-desktop to fix the icon
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11266
Summary:
The window rules dialog did not properly detect the wayland windows. So
I investigated what ICCCM writes about the WM_CLASS property (which is
the base for window rule matching) and checked how ShellClient maps to
it. Basically name and class was swapped and the reason for the
detection not working properly. As we don't have a proper name, the code
is adjusted to generate a name by using the executable name. This is
also what WM_CLASS should be filled with, according to ICCCM.
Test Plan: Rules dialog detects the name and class correctly
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11023
Summary:
When one uses:
* breeze as of 5.12
* wobbly windows
* shaded window
* a distribution building with assert enabled
and starts to move a shaded window, KWin asserts. The root cause for
this is that WindowQuad::makeSubQuad has an assert for y1 being smaller
than y2. With the combination listed above this is not guaranteed. For
the left shadow quad the y1 and y2 are identical and thus trying to
split it, results in the assert condition.
The problem of the shadow quad having an invalid size might be addressed
as well with D10811. Due to that the generation of the quads is not
touched. Instead a sanity check is introduced to not try to split
already invalid sized quads.
BUG: 390953
FIXED-IN: 5.12.3
Test Plan: Added unit test hit the assert, now doesn't hit it any more
Reviewers: #kwin, #plasma
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D11015
Summary:
If the cursor theme failed to create KWin crashed due to an endless
recursion. There are two reasons for this fault:
1) When the physical size does not exist we perform a division by 0
which results in an invalid size going into wl_cursor_theme_load
2) We emit the signal that the cursor theme changed even if it didn't
change thus creating an endless recursion
This change addresses both problems: it checks that the size is not 0
and changes the handling for theme update to only destroy the previous
theme if the new theme could be created and only emits the signal if
things change.
BUG: 390314
FIXED-IN: 5.12.3
Test Plan: Added a new test case which crashed with old code
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D10549
Martin Flöser