We previously only supported compose-over-border when the centre was not
set to tile.
It's possible to use compose over border with tile.. even if it doesn't
make much sense.
Given that creating an opacityMask loads most of the image anyway, we
can make use of the FrameSVG painter path and avoid any additional code
complexity here.
REVIEW: 119406
Revert the changes in contentsRect, it was clearly not the best place to
put the code that contains the code that computes the content size.
Instead move the code in FrameSvgItem, duplicates data and code but works.
Use FrameSVG as 9 tiles instead of uploading a big texture of the finished frame each time.
This also saves the cache being populated with full created frames in different sizes; which end up taking up space in the disk and shared memory cache as well as the GPU memory.
A code path falls back to the original uploading the entire texture if obscure settings are used, i.e overlay.
Benchmarks:
apitrace when resizing a frame goes from an average of 7.6ms per frame of CPU time just for the swizzling and uploading to 1.4ms
GPU time also drops from 40us to 10us
Themes will need to remove stretch-borders (when we gain nothing from stretching; i.e Breeze) to get the most out of it.
REVIEW: 119330
It's unreasonable to use private API, so make everything public API so that
every user of FrameSvg have as much features exposed as possible.
Reviewed by David Edmundson
Fallback to the old code if there's something not (yet?) handled by the
new implementation. This way we get to optimize for the the common use-case
without breaking former, complex, code-paths.
Reviewed by David Edmundson
This patch simplifies the grid a bit more by using Canvas instead of
Rectangles, so actually lowering the number of Items needed to create
the grid from 16 to just 1.
It works pretty neatly with no junctions problem at all.
REVIEW: 119283
Currently the grid itself is composed of 88 rectangles that draw all the
lines in a way that two big rect draws the whole two topmost horizontal
and leftmost vertical border lines and then each day rectangle is
drawing small bottom and right rect.
This patch reduces it to 13 rects only where one rect draws the whole
frame around the grid and then 1px wide/high rects draw the inner lines.
Results in much cleaner & simple code.
Plus there's a small refactor on the id names so it makes more sense.
This does not require any additional changes in the applets.
REVIEW: 119232
Qt Quick accessibility already picks up the text automatically. Instead
set the name (which is a descriptive element) to the placeholder text
which is more descriptive of the element. Otherwise Orca will read the
text twice.
the color properties are remotely useful only when used from QML
and risk to explode in number in the future,
it's not a good thing for a public c++ api.
Also, they will become kinda deprecated as in largely replaced by
the ColorRole api. Too late to remove those properties completely,
but still last days to move them at least out of c++ api.
(and in qml they could be removed in the future in a 2.1 version
of the import while still being present in 2.0 if needed)
This moves the color properties is a theme subclass available only from QML
REVIEW:118972
Included is a plasma theme to make it look the same.
A known regression is that SectionScroller is temporarily removed
but that is currently not used in plasma-desktop
and this fixes a lot more problems than it causes.
REVIEW: 114283
it's an import in core, and advertises itself as an "attached property"
with this we can:
say that all its chidren are of a certain context, like "button" or
"complementary"
then anywhere there will be available an attached property, as ColorScope,
so like:
PlasmaCore.ColorScope {
group: PlasmaCore.Theme.Complementary
PlasmaComponents.Label {
text: "foo"
color: ColorScope.textColor
}
}
- internally, smallSpacing, largeSpacing and gridUnit are entangled, and
all use the font's rendered size. This means we can save a
QFontMetrics construction by merging the updating of these three
props.
- A wanted side-effect is that the updating of spacing and gridunit now
happens at the same time, so they're updated consistently in the UI.
- Also, use compile-time connects for the Plasma::Theme::themeChanged
connection
1/8 is really ridiculously small, it ends up being 2px on almost all
systems, and using it between UI elements makes them look really
cramped. 1/4 works much better, and makes this property actually useful.
Basically splits the Calendar::m_startDate into 'today' and
'displayedDate', where displayedDate is the date that is displayed (it
controls the days model etc) and can be manipulated by the user by eg.
changing months in the plasmoid, and today is the current day, populated
by our dataengine (which means it auto-updates with no need for a
timer).
This allows for greater flexibility and things like "Go back to today"
when eg. the plasmoid is hidden or when the user have browsed too far in
the calendar and just wants to get back to today (the button to do that
pending).
Also this fixes a problem where the time dataengine is being polled
every 30secs for the clock and would reset the calendar view as the
startDate is currently bound to the dataengine and the view resets on
that change.
REVIEW: 118668
CCBUG: 336304
This enables and disabled menuitems with a timer, so we can see whether
dynamically updating these properties are propagated to the QMenu in the
proxy.
CCBUG:336281
* trigger an update after the animation finished as the IconItem needs to
switch to SVGTextureNode again
* Don't connect needlessly to a lambda slot
* FadingMaterialShader had the texture bind swapped
* Fix virtual method hidden warning in FadingMaterialShader
* Use dynamic_cast instead of static_cast to convert to SVGTextureNode
or FadingNode.
REVIEW: 118251
experiment in dynamic repacement of Text and Background
colors, in order to be able to generate icons of different colors
based on where they are (for instance if the normal background
is dark and the button background is light
at the moment supported an option to invert colors, one
to use the "highlight" color (if we want colored icons
on mouse over)
Currently we have 3 applets using the calendar grid and all of them put
the extra heading to the calendar. We also offer ways to change the
displayed month so we should also show /which/ month is the current one
displayed, implicitly.
REVIEW: 118328
BUG: 334147
This introduces new property on MonthView - "today". This property is
updated by dataengine (or can be a timer too, but we already get signals
from dataengine every 30 seconds to update the clock) and is never
updated by the code itself.
Furthermore, the DayDelegate is now bound to it, so when "today"
changes, the selection rectangle in the calendar should also change.
And finally, the selected item in the grid is cleared if MonthView's
date property is cleared (which is second part of this patch, to the
applet itself).
REVIEW: 118277
Removing eliding in Toolbutton, this causes a binding loop as the
sizeHint is dependent on the paintedSize, which will change if you
elide.
In all cases we are trying to make toolbuttons fit the contents, so this
actually got in the way of that.
Otherwise the label can get out of the button boundaries. Tested with
RTL text too, it's elided properly as the elipsis switches to Left on
its own.
Reviewed-by: Marco Martin
BUG: 334867
ToolTipAreas are often the critical path for mouse handling, since
hover events are not filterable from below in the item hierarchy.
When disabling tooltips also stops updating containsMouse code has
no way to adapt, since there's no API to read the pref from within
QML. This changes the implementation of the pref to only prevent
tooltips from being shown, but still continue to handle hover events.
We need to use a GLXFBConfig which matches the depth of the window
pixmap's depth. So far it used the GLXFBConfig of the GL context.
This worked fine for RGBA windows, but failed for RGB windows on e.g.
some NVIDIA drivers.
After this change the FBConfig of the context is completely ignored,
instead it tries to find a good FBConfig for a given depth.
Whenever a FBConfig for a given depth is found it's inserted in a
cache shared between all WindowThumbnails so that we don't have the
X roundtrips all the time.
BUG: 334241
REVIEW: 118110
The WindowThumbnail needs XCB_EVENT_MASK_STRUCTURE_NOTIFY on the window.
For this it needs to change the window attributes. But it needs to keep
the existing event mask to not break other code. Also it cannot remove
the event mask again, as there is no chance to know whether another
component requires STRUCTURE_NOTIFY.
This fixes the not updating items in the tasks applet as the event mask
needed for KWindowSystem got overwritten.
BUG: 331956
REVIEW: 118137
Port IconItem to native QSGTexture including the animation.
This will save constantly uploading a new texture to OpenGL throughout the animation.
REVIEW: 116024
If the window holding the WindowThumbnail item goes away the texture hold
by the node is destroyed but our bound window pixmap is not yet freed.
This results in incorrect state the next time the WindowThumbnail is
shown.
To get back into a clean state discardPixmap() is called if there is no
texture but a bound low level pixmap.
BUG: 333482
REVIEW: 117824
viewTextColor is wrong here, we have to use buttonTextColor. This
follows Plasma 1, and makes texts in fields and areas in white-on-black
themes work.
BUG:333837
At the moment, we could say that plasma is co-installable by chance,
it's only working because KF5 dropped the apps relative directory.
This change introduces a PLASMA_RELATIVE_DATA_INSTALL_DIR that will
be available to know where in share the components are and
PLASMA_DATA_INSTALL_DIR to know where data has to be installed to.
Reviewed by David Edmundson
CCMAIL: plasma-devel@kde.org
The documentation was copy-pasted and was wrong.
The property name started with m_, which is odd and I don't see a reason
for it. Also I didn't find any code that uses it.
REVIEW: 117760
Drop dependency to QtWidgets from this file. We can start assuming that it
might not be that functional in some platforms.
Use QGuiApplication counterparts, based mostly on QScreen, which could make
it more powerful in the future.
REVIEW: 117161
This messes up the fontmetrics needed to size the tab button. Before,
the labels would be elided no matter what.
This is more a band-aid than a real fix, we'll need to investigate how
we can make the sizing work correctly, but at least it "unscrews" the
layout of all of our tabbars.
Introduces a new optional dependency to EGL. If EGL is found the
WindowThumbnail gets compiled with EGL support. The EGL support is
unlike Qt's XCB plugin not mutual exclusive to the GLX backend, thus
it's possible to compile with EGL and GLX at the same time.
By that this implementation is prepared for the case that Qt supports
EGL or GLX through runtime selection.
In practice EGL support is only useful if Qt is compiled with GLESv2.
In that case the corebindingsplugin gets linked against Qt5::Gui_GLESv2
to get the dependency to GLES.
The implementation makes use of the EGL_KHR_image extension
(or EGL_KHR_image_base and EGL_KHR_Image_pixmap) and
the GL_OES_EGL_image extension to bind the X pixmap to an EGLImageKHR.
REVIEW: 116627
Qt's xcb plugin updates _NET_WM_STATE and ignores anything we set,
thus we need to ensure that our additional states are set after Qt
set the state. Luckily KWindowSystem does not remove the state which
is already set.
Reviewed-By: Marco Martin
