This is still pretty rough. It seems to work fine for larger icons, but
smaller ones get misaligned. This will need some corrections, such as
locking the smaller sizes to 16, 22, 32, 48, 64, which should take care
of this issue.
With this change, parts of Plasma scale depending on the DPI. This is
noticeable in Kickoff.
It was an int already anyway, so let's be true about this and also
return an int. This reflects pixel-alignment in the API, not just in the
implementation.
QWindow::setFlags is manipulating the same X property as
KWindowSystem::setType. Thus our setType never succeeded as the
setFlags overwrote the window type again. In the same way we
have to call setFlags when reverting to the default Normal type.
REVIEW: 115258
The last bits have moved out of ThemeProxy, units is used for iconSizes,
which means we can finally directly use Plasma::Theme, and don't need a
ThemeProxy class anymore.
This is means two things:
- icon sizes have moved from theme.iconSizes.* to units.iconSizes.*,
will provide porting script and update docs
- we can kill PlasmaCore's ThemeProxy now. Icon sizing and dpi handling
is kept outside of Plasma::Theme for now, as it pulls in too many
deps.
This change does two things:
- it increases the gridUnit to 11 pixel on 102DPI, 18 on 170DPI. This
means on the one hand larger sizing, but aligns gridUnit and
theme.largeSpacing consistently, and thus makes these types work
together better.
- it scales the gridUnit with the DPI. The width of the font's bounding
box doesn't change on the displays I've tried, the height does.
As a result, gridUnit now scales linearly with the DPI setting.
The outputOnly property allows to specify that the dialog should not
accept any input. Thus it's an output only window which supports click
through. This is obviously platform specific and so far it is only
implemented for the X11 platform using the shape extension.
The input shape needs to be set once the window is visible and thus
the functionality is bound to the visible changed signal. The code
ensures that the required shape extension version is present and only
fetches it once.
REVIEW: 115139
- check some guards as qobject_cast may fail.
- destroyed() is emitted too late, the slot managing it can only access a QObject, Containment members aren't valid anymore
The QQmlPropertyMap isn't moved with this commit, this needs a bit more
changing, so we do a separate one for that, once the more trivial stuff
is moved out.
This removes the last dependency from plasma-framework on kde4support.
This change is a bit more involved than other kde4support removals because
QSharedPointer does not provide a count() method. I therefore reworked the code
to store the SharedSvgRenderer as a QWeakPointer in s_renderers, making it
possible to detect when the last one goes away by creating a QWeakPointer guard
in SvgPrivate::eraseRenderer().
REVIEW: 114912
ToolTipArea.containsMouse allows the use to update the tooltip way
before it's shown, or when the contents of the tooltip should change
within an open dialog.
dismiss() and keepalive() allow handing over the tooltip dialog to
another item, in order to allow for a smoother handover when a new item
recycles the tooltipdialog. This avoids flickering when the tooltip
moves from one item to another.
Instead of registering WidgetExplorer inside the shell, create a new qml
plugin in the org.kde.plasma.private namespace.
This is useful for running widgetexplorer in qmlscene.
This makes testing, debugging and profiling considerably easier.
REVIEW: 114914
Saves quite a lot of roundtrips to the XServer. Most important we
fetch the atom only once which means we don't have to refetch when
checking the property notify events.
REVIEW: 112445
- Derive from MouseEventListener as to let input through
- react to target changes: reparent the MEL accordingly
- kill a bunch of warnings
This "helps" a bit, but still filters out input events since the MEL
only passes events down to children, but the tooltip's target becomes
its parent.
Reparenting the target crashes, and it seems a pretty ugly thing to do,
anyway, since we don't want to mess with the scenegraph structure just
to filter events out for tooltips. Grmblz.
Delete KDeclarativeMouseEvent when emitted
Do not leak KDeclarativeMouseEvent press and hold events when the mouse
moves.
Do not create KDeclarativeMouseEvent when the mouse is hovered. It's
pointless.
... instead of overriding the property.
Removes a warning whenever the implicit width and height changes, and
reduces the API footprint of AppletInterface.
The larger headings are a bit softened, to make their stick out less,
the font is meant to take care of this. Result: visually more consistent
colouring, strangely. Human eyes are beautifully complex things.
same rules of sharing and memory management apply.
The model will be shared between all visualizations and deleted when nobody is connected to the source anymore
Since we are checking if the width() is out of bounds we should also use
width() to correct the position.
Other cases might need change as well, but I think we better go in a case
by case bases to fix all possible issues.
It is up to the Containment to decide if the user is allowed or not
to lock/unluck the widgets, so corona should not add one when there is none
(user is not allow) and it shouldn't add another one when there is already
one.
Fixes duplicated action in toolbox.
I'm not sure for what reason this was changed but it is clearly wrong,
we don't want to create a new action but rather re-add (as the hack says)
the action we took before.
Fixes "Empty action" in toolbox.
The panel would be one pixel above the bottom edge, which prevents us
from using the invisible space beyond the panel. This makes the panel
flush with the bottom edge.
- We would only check for the available screen, but not for the location
when applying borders. Fix that.
- Cache dialog geometry when going fullscreen, so we can restore it when
we're not fullscreen anymore.
- Remove temporary var l, we have m_location for that.
The amount of warnings that plasma-shell has, makes it super hard to
make use of tools like journalctl or to grep ~/.xsession-errors.
We need these tools to diagnose possible bugs in the session start or
any other software that redirects stderr or those places.
We can remove this option once all the Warnings are fixed, specially
the one in Qt: https://codereview.qt-project.org/#change,73943
CCMAIL: plasma-devel@kde.org
This is needed to let KLauncher know that Plasma Shell has been
executed.
We are using plasma_shell because there is a bug in KDBusService
that prevents us from using plasma-shell. Fix is already done
but repositoiries at this very moment are frozen.
This is a needless duplication of the property, and it causes bugs.
Removing it, and porting all users to the flags property from QWindow
magically makes layering and resizability bugs vanish.
One problem is left: the dialog dismisses too easily, it should not
dismiss when being resized, moved or when a child dialog or QMenu opens
from it.
Interestingly, this hack yields the correct behaviour in my system. We
quickly set the always on top hint once we're supposed to hide, so the
window stays on top for the duration of the animation -- without
apparent side effects on the positioning and resizability.
CCMAIL:mgraesslin@kde.org
Previously shells, activities, shellcorona and corona all tried to
manage
which screen a containment was on.
This version moves all screen management into ShellCorona so we have a
central place for keeping tracking of containments, activities and
screens.
This allows to control whether to automatically close panel popups when
they lose focus. Applets can set this to false in case they want to keep
the popup open (even temporarily).
This property is meant to bring back the functionality provided by
PopupApplet. If the property is set to true the dialog gets hidden
when it loses focus.
REVIEW: 114378
Recalculate position when screen has changed or when the location changes.
Recalculate position when the thickness changes, otherwise it grows outside
of the window when a bottom panel thickness changes (we change it through
setHeight).
Ues QScreen::geometry instead of QScreen::virtualGeometry. The virtual
geometry is, according to the documentation and empirical experience, the
geometry of all the virtualDesktop.
REVIEW: 114362
They do the same on Unix, but on Windows ConfigLocation is erroneously
pointing to an app-specific directory. Fixed in qtbase 44d48862c0 by
adding GenericConfigLocation for this purpose. Available in Qt 5.2.0 rc1.
Reason for this change: a QIcon might not contain the disabled pixmaps.
Using the icon effect allows to get also the disabled state for such
icons.
REVIEW: 114334
We restart plasma-shell after crashes. When it crashes subsequently on startup,
and more than two times in a row, we give up in order to not endlessly loop.
Once the shell process stays alive for at least 15 seconds, we reset the crash
counter, so a later crash, at runtime rather than on startup will still be
recovered from.
This logic is very similar as to how kwin handles it.
DIGEST:
DesktopView incorrectly filled the geometry of screen()
screen() will be the screen of the parent shell, not the correct
screen.
As we are using QScreen, shellcorona is ported to use that instead
of QDesktopWidget.
REVIEW: 114149
KLocale is now kde4support and we have QLocale QML bindings. This patch
removes the KLocale bindings from plasma-framework (they are not used
anywhere anyway).
If we miss something from KLocale in QLocale, we should upstream it.
REVIEW: 114150
These methods are in the middle of other together-related methods, for
no reason. Looks like corn flakes in the middle of beer. Sure, you can
do that, but why would you. Actually I should try that.
Some CLDR (locale) data used by Qt have standalone month names
starting with lower-case letters. So if we want to provide consistent
look across locales, we need to capitalize it ourselves
See https://bugreports.qt-project.org/browse/QTBUG-35100
Also rename startDay property to firstDayOfWeek property so it's more
clear what it does. Handling the first day of week is changed a bit too
as QML uses 0 for Sunday while QDate uses 7 for Sunday, so the
getter/setter is accomodated to that fact and converts from 0 to 7 on
setting and vice-versa in the getter.
REVIEW: 114099
Stop using classes from kde4support:
- KComponentData is deprecated, it will be using QCoreApplication::
applicationName() and QCoreApplication::applicationVersion(). Of course,
this means that the existing shells will have to be ported. I have no
problem with doing that port myself, if I'm told where to look into.
- Drop usage of KLocale, ported to QLocale
- Drop usage of KStandardDirs, ported to QStandardPaths
- Drop usage of KIcon, ported to QIcon
Furthermore, there's a module in src/declarativeimports/locale that IIUC
exposes KLocale bindings to QML. A specific plan to port it should happen
as well.
REVIEW: 113920
