As pointed out by the Valgrind trace in #363224, delaying the signal causes
the EffectWindow* argument to become invalid before the connected slot is
called (this is because Qt discards only delayed signal->slot calls where
the receiver gets deleted meanwhile, not the sender and definitely not
a random argument.
If the supposed glitches really happen, they should get fixed correctly,
and for all cases (I doubt only desktop number would be involved but
not e.g. shaded or minimized states).
https://phabricator.kde.org/D5164
CCBUG: 363224
Summary:
Change 0bec9ad733 introduced a regrssion
on Wayland. On Wayland getTimestamp always returns 0, thus the X11
timestamp gets reset again and again. This results in broken interaction
for Xwayland windows as the broken unit tests show.
This change addresses the regression by ignoring a value of 0.
It means the addressed bug is still present on Wayland for X11 windows
as the timestamp doesn't get updated properly. This requires further
changes.
Test Plan: TestShade works again, testX11TimestampUpdate still works
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5731
Summary:
So far KWin only updated the x11 timestamp if the new timestamp is larger
than the existing one. While this is a useful thing it creates problems
when the 32 bit msec based time stamp wraps around which happens after
running an X server for 49 days. After the timestamp wrapped around KWin
would not update the timestamp any more and thus some calls might fail.
Most prominent victims are keyboard and pointer grab which fails as the
timestamp is either larger than the server timestamp or smaller than the
last grab timestamp.
Another problem related to timestamp handling is KWin getting broken by
wrong timestamps sent by applications. A prominent example is clusterssh
which used to send a timestamp as unix time which is larger than the
x timestamp and thus our timestamp gets too large.
This change addresses these problems by allowing to reset the timestamp.
This is only used from updateXTime (which is normally invoked before we
do things like grabKeyboard). Thus we make QX11Info::getTimestamp the
ultimate trusted source for timestamps.
BUG: 377901
BUG: 348569
FIXED-IN: 5.8.7
Test Plan: As I cannot wait 50 days: unit tests for the two conditions added.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5704
Summary:
Even though we're using a QSortFilterProxy model, by default it doesn't
actually sort anything until instructed to.
This patch turns sorting on.
Test Plan:
Opened the KCM
Looked at it
Reviewers: #plasma, mart
Reviewed By: mart
Subscribers: plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5407
Summary:
The KCM has a context property of the currently set theme index. This is
set before the decorations model is populated, so it is currently always
-1.
This model is populated after the constructor but before
KCModule::load().
KCModule::load is called from KCModule::showEvent so before
QQuickGridView will start doing anything with delegates.
This fixes the problem simply and also avoid parsing the config file
multiple times.
This bug was introduced in 5.9.4:
Someone made a (tested) change to make sure the view scrolled to the
right place on startup.
I then made a (tested) commit fixing the crash on exit
The author then updated his patch to my changes, but now in a way that
didn't work.
Test Plan:
Opened system settings module with a million decorations.
The correct entry was visible and highlighted.
Reviewers: #plasma, graesslin
Reviewed By: #plasma, graesslin
Subscribers: plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5401
Summary:
Libdrm 2.4.78 introduces a version 2 and if KWin gets built against it
our code would break. Given that this change is for Plasma/5.8 branch.
Closes T5839
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Maniphest Tasks: T5839
Differential Revision: https://phabricator.kde.org/D5380
Summary:
Currently, if user tries to move one of buttons to the left, ending up
dragging one button onto another, crash occurs.
In addition, this patch replaces verbose replacement(remove/insert) with
more elegant QVector<T>::move(int, int)
BUG: 374153
FIXED-IN: 5.8.7
Reviewers: graesslin, #kwin
Reviewed By: graesslin, #kwin
Subscribers: kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D5117
Summary:
using the same trick as elsewhere, set the currentIndex
and move the view to currentIndex right at startup
the only way to be sure is onContentHeightChanged
as there are no signals for when "the view has been
populated and settled up"
Test Plan:
the view is at the right state since the first frame shown,
no more jumping around effect
Reviewers: #plasma, davidedmundson
Reviewed By: #plasma, davidedmundson
Subscribers: davidedmundson, plasma-devel, kwin, #kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D4703
Summary:
So far KWin put a version requirement on Breeze which is the current
version number of KWin. This is technically correct, but distributions
seem to have problems with it. E.g. I noticed twice in Debian testing
that kwin defaults to Plastik which happens when Breeze is not found.
I reported it, it got fixed, now it's again broken.
Thus I think that distributions have problems with the requirement of
building Breeze prior to KWin. As KWin does not depend on the minor
version, let's depend only on the actual required version, which is
5.8 in the 5.8 branch, 5.9 in the 5.9 branch and master.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D4644
Summary:
As described in https://bugreports.qt.io/browse/QTBUG-57714 exposing a
QWidget as a context item to QtQuick can crash. Especially as the
engine's context property gets deleted whilst deleting the parent item.
This patch reworks the code so that the models are exposed to QML
directly rather than going through a QWidget.
CCBUG: 373628
Test Plan:
Pressing back whilst in the decoration KCM used to crash every time, now it doesn't.
I still have the buttons.
Reviewers: #plasma, graesslin
Reviewed By: #plasma, graesslin
Subscribers: graesslin, cfeck, plasma-devel
Tags: #plasma
Differential Revision: https://phabricator.kde.org/D4533
Summary:
So far we didn't trigger modifier-only-shortcuts when capslock was
enabled. In fact we even ensured that the shortcuts did not trigger.
This seems not to be what our users expect. Meta should still trigger
if capslock is on.
This change modifies the logic to determine which modifier is currently
pressed by using the modifiersRelevantForGlobalShortcuts. The difference
to the "normal" modifiers is that this excludes capslock from modifiers
and excludes consumed modifiers. The latter is not really relevant as
modifier-only-shortcuts do not trigger if multiple keys are pressed,
which is required to have a modifier consumed.
BUG: 375355
FIXED-IN: 5.8.6
Test Plan: Only with adjusted autotest
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D4241
Summary:
Xkbcommon does not localize the layout names. We always get the US
English variant. KWin so far passed the strings from xkbcommon directly
to UI visible parts such as the Plasma OSD. This results in not
translated elements in our UI.
As I figured out today the translations exist in the domain
xkeyboard-config. Passing the strings from xkbcommon through this domain
gives us translated layouts.
Unfortunately the xkeyboard-config is not installed by default on all
distributions. E.g. on ubuntu/neon one needs to install xkb-data-i18n
to have the translations available.
Test Plan: Untested as I fail to start Plasma/5.8 KWin on my dev system
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D4243
This extends the testCapsLock to ensure that modifier only shortcut
trigger when capslock is enabled. So far the test only ensured that
capslock doesn't trigger the shift modifier only shortcut.
As the test shows currently when capslock is on, the meta key doesn't
trigger any more.
CCBUG: 375355
Typically after deselecting "Show caps" option, the associated options "Cap color" and "Display image on caps" are grayed out (triggered by signal). The bug was when Desktop Cube settings were closed and opened again then the options were not grayed out.
Succesfully compiled and tested by deselecting "Show caps" option in Desktop Cube settings, closing the settings window and opening it once again. The options "Cap color" and "Display image on caps" were then grayed out.
REVIEW: 129194
In case we need to set the SwitcherItem model in
`TabBoxHandlerPrivate::show()`, remember the current index row,
otherwise it gets reset to the first item.
Reviewers: #kwin, graesslin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D4162
BUG: 333511
Summary:
Currently the configure button is aligned to the win decoration in the
background. This is bound to 40 - the inactive decorations shadow.
This means the configure button ends up in different places across each
item,
looking a bit weird and potentially not even on top of the right
delegate.
This patch aligns the configure button relative to the overall delegate
so that it's in the same place on every item
Test Plan: Looked at KCM
Reviewers: #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3526
BUG: 372685
Summary:
The method is supposed to return the AbstractClient's size for a given
client size. That is the size including the window decoration.
The default implementation returned the passed in client size without
adjusting for the decoration. This resulted in ShellClient getting a
wrong size especially when AbstractClient::adjustedClientSize (which
calls sizeForClientSize) was called.
The result of the incorrect size was for example a shrinking of the
window when starting to resize a window.
BUG: 370345
FIXED-IN: 5.8.4
Reviewers: #kwin, #plasma_on_wayland, broulik, subdiff
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D3414
Summary:
When windows get added some effects grab the window and want to be the
only one animating this window. For this the grab roles exists. An
effect being notified later on evaluates the grab state and does not
start the animation.
This process failed due to being dependent on the order the effects are
loaded. Window Added/Closed are signals emitted by EffectsHandler, thus
first come, first serve. The requested effect order does not play into
it.
Due to that it could happen that an Effect which should not animate,
started to animate as the grab was still there.
This change adds the possibility to be notified whenever the window data
changes. A new signal is added to EffectsHandler which is emitted
whenever the windowData changes. The interested effects connect to it
and cancel their (just started) animation for the window.
Adjusted effects are:
* ScaleIn
* Fade
* WobblyWindows
In case of WobblyWindows an additional logical error was fixed that the
animations were only run when an effect grabbed instead of the other way
around.
BUG: 336866
FIXED-IN: 5.8.4
Reviewers: #kwin, #plasma, broulik
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3211
Summary:
The QStyle might interact with our internal Wayland connection. If the
cleanup happens after destroying our internal Wayland connection KWin
will crash at tear-down.
With this change the QStyle can perform cleanup already prior to the
deconstruction of the Wayland connection.
CCBUG: 372001
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D3241
Summary:
The killer can only kill X11 windows. It is also only invoked for Client
and not for ShellClient. The tool interacts with QX11Info which means it
would crash if tried to use on Wayland.
Thus force platform xcb on the tool.
Test Plan: Killer shows for Xwayland windows.
Reviewers: #kwin, #plasma_on_wayland
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D3228
Summary:
So far KWin used the window title provided from the window directly
without any sanitizing. This could result in broken window decorations
if the title included line breaks. Those were passed to the decoration
and depending on the way how the decoration renders the title, it could
result in visual breakage.
Having line breaks in a window title doesn't make sense. Given that KWin
now simplifies the title when copying it to it's own structure. This
also ensures that the title passed to e.g. task manager does not have
any line breaks on Wayland.
BUG: 323798
FIXED-IN: 5.8.4
Test Plan: Opened the web page in a nested KWin, properly rendered now.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3215
Summary:
Desktop Grid has a feature that right or middle clicking a window toggles
the on all desktop state. If that is done for a desktop window, things
obviously start to fail.
KWin internally allows to change virtual desktop for all windows, even
for windows like the desktop or docks. Changing in KWin core is not a
solution.
Thus a special casing in desktop grid is added to disallow the selection
for desktop windows.
BUG: 372037
FIXED-IN: 5.8.4
Test Plan: Tried right clicking desktop window and other windows.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3253
Summary:
For whatever reason Qt scrolled the wrong way on internal windows. As a
new QWheelEvent is constructed anyway, just multiply by -1 to get the
direction fixed.
BUG: 371999
FIXED-IN: 5.8.4
Test Plan: Tried debug console in nested setup, scrolled correctly now.
Reviewers: #kwin, #plasma_on_wayland, broulik
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D3247
Summary:
This change ensures that the DebugConsole does not have a minimize
button.
BUG: 372000
FIXED-IN: 5.8.4
Reviewers: #kwin, #plasma_on_wayland, broulik
Subscribers: plasma-devel, kwin
Tags: #plasma_on_wayland, #kwin
Differential Revision: https://phabricator.kde.org/D3244
Summary:
When moving windows we don't want to snap against not visible windows
like auto-hidden panels.
BUG: 365892
FIXED-IN: 5.8.4
Test Plan: So far only auto-test, manual test will follow.
Reviewers: #kwin, #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3225
The export of the WorkspaceWrapper was changed to be a subclass with all
the elements still being in the parent class. But the "workspace" was
exported with QScriptEngine::ExcludeSuperClassContents. Thus all usages
of workspace were broken and our tests started to fail on build.kde.org.
This change removes the ExcludeSuperClassContents which means that also
the QObject properties and slots are now exposed which was previously
not the case.
CCMAIL: github@chilon.net
Summary:
The version provided is only compatible with QtScript so it became
necessary to split WorkspaceWrapper into a base class and two child
classes, one for QtScript and one for QmlScript.
BUG: 340125
FIXED-IN: 5.8.4
REVIEW: D3185
Summary:
QSpinBox can't handle plural suffixes. Something previously done by
KIntSpinBox.
Using setSuffix(ki18np("pixel", "pixels")).toString() does nothing, as
at the time of conversion we don't know which one to use.
This patch uses KPluralHandlingSpinBox and correct ki18np.
Note, "new" dependency was already linked implicitly in other kwin, but
we need to add it for this KCM.
Test Plan:
Opened KCM (in English) set counter to 1 pixel and 2 pixels.
No longer had a big warning. Also appropriate number of s's appeared.
Reviewers: #plasma
Subscribers: plasma-devel, kwin
Tags: #kwin
Differential Revision: https://phabricator.kde.org/D3222
