The window is created before the backingstore, so the destruction should
happen in the reverse order. Also, the backingstore is created based on
a pointer to a QWindow, so the dependency is from backingstore to window.
Thirdly, this fixes a crash when using QQuickWidget/QOpenGLWidget.
Task-number: QTBUG-37045
Change-Id: I6e782c553fddfeef877be30ca2610aff75a425a0
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
When a child of a widget is spontaneously revealed due to a call to
the parent 'resize' method, the child will not receive a paint event
if it has the WA_StaticContents and WA_OpaquePaintEvent flags set.
This is caused by the backing store being pre-emptively resized by the
call to setGeometry_sys, which causes QWidgetBackingStore::sync to skip
the block which handles the static contents.
There doesn't appear to be any reason to preemptively resize the backing
store, since it is always resized as-needed during the the 'sync' method.
This change-set removes the code which preemptively resizes the backing
store.
Task-number: QTBUG-35282
Change-Id: Ie9942854ca5322dfe0f98ed8100810161576be80
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Make sure the alphamask in the backingstore is updated whenever the
state of a QQuickWidget/QOpenGLWidget child changes.
Task-number: QTBUG-37062
Change-Id: Iffa80fc8993499a15bc9e773a162a3f98b81ae03
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Enable child widgets (without a native window) that render to an FBO
and are composed with the raster backingstore by the platform plugin.
A preliminary version of QOpenGLWidget is included as private API.
Change-Id: I8f984a4d7db285069ce3d6564707942c823d890d
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
It's not "preliminary" anymore, and it doesn't return a QPlatformWindow,
but a QWindow.
Change-Id: Iaa1938728e689d97011dce8cf0d88ff8cb4a83b9
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Topi Reiniö <topi.reinio@digia.com>
Removed QMoveEvent generating code from QWidgetPrivate::setGeometry_sys
for widgets with native window handles. A move event is already
generated for them by QGuiApplicationPrivate::processGeometryChangeEvent.
Task-number: QTBUG-32590
Change-Id: I73313a012851516047ac017f1e15a21774d8ffe2
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
The native child windows need to inherit the parent's visual in order
to have a translucent background as well.
Surface with type QSurface::OpenGLSurface should not be forced to use
the parent window's visual - the parent visual for instance, might not
even be GL capable.
Changing WA_TranslucentBackground during runtime is not supported, for
two reasons:
1) Other platform plugins seem not to support it
2) It would require recreating X windows.
Task-number: QTBUG-29625
Change-Id: Ic1474dd2de99069027481c7db6bf865f9b8d616d
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Change childWidget->windowHandle() to childWidget->internalWinId() in
q_createNativeChildrenAndSetParent() to determine whether should we call
childWidget->winId().
This is because in some circumstances Qt will crash due to accessing
deleted QWidgetWindow object if we use windowHandle(). Think about the
following scenario:
1) create a widget A without parent and add two child widgets B and C to A
2) create a native widget D as the child of B, note that when we set
Qt::WA_NativeWindow attribute to it, its QWidgetWindow will be created
which means its windowHandle() is not null.
3) create a top level widget E as the child of C and show it. This will
make Qt call createWinId() to A and then
q_createNativeChildrenAndSetParent() will be called to create A's native
children recursively and finally make D's QWidgetWindow object become a
child of A's QWidgetWindow object. Please note here that B will not become
a native widget just because at that moment windowHandle() of D is not
null and Qt will not call winId() to its parent B
4) Set A's parent to another widget which has been shown, setParent_sys()
will be called to A and then Qt will call destroy() to A. in destroy() Qt
will try to call destroy() to its children recursively with a condition that
the child has Qt::WA_NativeWindow been set. But D's parent B is not a native
widget right now so B and D is not destroyed. Qt will then deleted the
QWidgetWindow object of A, since E's QWidgetWindow object is a child of
A's QWidgetWindow object, it will also be deleted. Now E hold a deleted
pointer of QWidgetWindow object. This is the source of crash later.
Task-number: QTBUG-35600
Change-Id: I97a20a68e626ee62b15bb4eae580e26f8948923b
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Check q->isTopLevel() before calling setBackingStore() in
QWidgetPrivate::create_sys() to prevent QBackingStore leaking. This is
because QWidget::setBackingStore() will return if isTopLevel() is true and
the newly created QBackingStore object will be leaked.
Change-Id: I2777acd4c317d5019f5b266feae005042026b8be
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Qt::WA_Mapped maps (sic) to windowHandle()->isExposed(), and we set/update
it in QWidgetWindow::handleExposeEvent(). Setting it directly in show_sys
shortcuts QPA and assumes showing a window is synchronous on all platforms,
resulting in trying to flush the widget backingstore when the window was
not exposed yet (due to discardSyncRequest starting to return false).
This reverts commit 829b1d13b2.
Change-Id: I0bd700d4939bc69ba184d8586435b68ec3dd72fb
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
The dockwidget's toplevel window would be a parent of the container's
window when floating. When plugged back into the mainwindow the
dockwidget's window is destroyed and the container's window along
with it. Added a function toplevelAboutToBeDestroyed to unparent
the containers window before this happens so parentWasChanged will
work correctly.
Change-Id: I06679cfb3a8fa3834c0db0be5973c012b8277275
Reviewed-by: Ulf Hermann <ulf.hermann@digia.com>
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
To be set to false by eglfs and similar platforms where no real window
management is provided.
When this capability is not set, some of the changes done for
QTBUG-26903 will be disabled, allowing dialogs to show up at their
QDialog-chosen place on platforms where no WM is present and so
windows are never repositioned by the system.
Change-Id: If1dac3cd11f2a856913a51277431fe3ec644b719
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
In "very basic platform plugins", the platform window ID can be null,
which conflicts with QWidget's special treatment of 0 as a parented widget.
Such is the case with EGLFS, which can return a null native window handle
for winId on some EGL implementations.
Move the assertion winId!=0 into the widget framework, as this assert is
not relevant for non-widget applications. A large proportion of basic
platform plugin users will see relief from this unnecessary assertion.
Change-Id: I25c9d96550cd747c77c9516d773e9cdebf7440ab
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
QWidget::setWindowIcon() and similar call createTLExtra()
which creates a QWindow without setting the native attributes
on the parent, which can cause crashes when setParent_sys()
decides to delete the window.
Task-number: QTBUG-31672
Change-Id: I4c40ee12741be88b2281df90329ffb698d4009eb
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
We need to make sure that we don't reset() the context when the
QGLWidget is being reparented, as that will lead to the QOpenGLContext
being destroyed and not recreated, leading to a crash in makeCurrent().
Also, don't destroy() the widget if it has not yet been created, as in
that case there's no ParentAboutToChange event sent.
Task-number: QTBUG-31016
Change-Id: I409fff94456802a80bd72b470a6fbaee87505baa
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Andy Nichols <andy.nichols@digia.com>
If the application object is an ancestor of QMenu, dereferencing qApp
in its destructor will cause a crash.
Task-number: QTBUG-30756
Change-Id: I31a33db0fd783bb210a420618911ea8b412e9a0f
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
While the raster platform plugin supports multiple top
level windows, this is not supported on the GL plugin,
so if you use GL or QtQuick2 in your app and use several
top levels, the app would crash with an error message.
A problem is that the top-level SurfaceView is a special
overlay View and does not support being stacked in a
layout. So instead, we let all windows share the same
GL surface and draw on top of each other. This works
fine for simple use cases.
We implement a new platform capability to make sure no
top level windows (even combobox popups and dialogs)
get non-fullscreen geometries. That has never
worked properly with the eglfs plugin.
Task-number: QTBUG-30473
Change-Id: Ia1438019638fc739cc93ffe79b46b81631254df2
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
Previously QPainter computed the devicePixelRatio
based on the physical and logical dpi, and expected
that the ratio between them would be either 1x or
2x.
This was problematic for paint devices like printers
where the physical dpi can be much higher than the
logical dpi, and also for QScreen where the physical
dpi would have to be defined as a multiple of the
logical dpi.
Add QPaintDevice::PdmDevicePixelRatio and QPaintDevice::
devicePixelRatio() getter and implement it for the
QPaintDevice subclasses. Use it when calculating the
highdpi scale transform in qpainter.cpp and when scaling
the clip rect in qwidget.cpp.
Remove physical dpi scaling for QImage, QPixmap and
QOpenGLPaintDevice, reverting to the old behavior.
Change-Id: I6c97510613196d4536ff39d08e9750b8782283d4
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Change-Id: Ifacee152e291face69964471d75e92b7784be4a4
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Remove all trailing whitespace from the following list of files:
*.cpp *.h *.conf *.qdoc *.pro *.pri *.mm *.rc *.pl *.qps *.xpm *.txt *README
excluding 3rdparty, test-data and auto generated code.
Note A): the only non 3rdparty c++-files that still
have trailing whitespace after this change are:
* src/corelib/codecs/cp949codetbl_p.h
* src/corelib/codecs/qjpunicode.cpp
* src/corelib/codecs/qbig5codec.cpp
* src/corelib/xml/qxmlstream_p.h
* src/tools/qdoc/qmlparser/qqmljsgrammar.cpp
* src/tools/uic/ui4.cpp
* tests/auto/other/qtokenautomaton/tokenizers/*
* tests/benchmarks/corelib/tools/qstring/data.cpp
* util/lexgen/tokenizer.cpp
Note B): in about 30 files some overlapping 'leading tab' and
'TAB character in non-leading whitespace' issues have been fixed
to make the sanity bot happy. Plus some general ws-fixes here
and there as asked for during review.
Change-Id: Ia713113c34d82442d6ce4d93d8b1cf545075d11d
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Default-constructed geometry does not mean put the window at 0,0,
and it does not mean center the window on the screen: it means
let the window manager position the window. If the window is
explicitly positioned at 0,0 though, that is a higher priority
than the transient hint; without this change, the transientFor property
had no effect. On X11, transient means use center "gravity" to
make the transient window exactly centered. But the user can still
override the geometry of a transient window, as with any window.
On OSX and Windows, neither transient window functionality nor smart
initial positioning are provided, so a window with no position set
will be centered on the screen, and a transient window will be put
at the center of its transientParent.
Change-Id: I4f5e37480eef5d105e45ffd60362a57f13ec55f5
Task-number: QTBUG-26903
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Currently, there is no concept of not having a cursor set on
a Window. Qt::ArrowCursor is always set instead. This causes
bugs when native child widgets are involved, for example setting
a cursor on the native widget's parent no longer works since
the child's Qt::ArrowCursor applies.
Introduce QWindowPrivate::hasCursor tracking whether a cursor
has been explicitly set and clear in QWindow::unsetCursor().
Handle 0 in QPlatformCursor::changeCursor() to mean
"unsetCursor()":
- Windows: Introduce default constructor for QWindowsWindowCursor
meaning "0". Search for applicable parent cursor in applyCursor.
- XCB: No big changes required, set XCB_CURSOR_NONE for no cursor.
- Other platforms: Assume Qt::ArrowCursor when cursor = 0 is
passed for now.
Task-number: QTBUG-28879
Change-Id: Id82722592f3cd5fe577a5b64dcc600c85cfea484
Reviewed-by: Jonathan Liu <net147@gmail.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
When moving QMdiSubWindows that contains native widgets (e.g QGLWidget)
inside a QMdiArea, garbage is drawn to the screen. The reason is
that since we use native QWindow backed widgets, QPA will assume
that it does not need to update/scroll the backingstore. But since the
backingstore can be shared among several windows, we need to do this to
ensure that exposed areas behind the moved QWindow will be updated.
Change-Id: I35f4d4069a73ab6cd6d610e25c176c752789a927
Task-number: QTBUG-26162
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
It is possible that the widget which is to be shown full screen has a
0x0 geometry if an empty layout with 0 margin is set there. In that case
the widget should not be hidden.
Task-number: QTBUG-27602
Change-Id: I87fd20cc3a5bac89ceccdf639780c4bc721ce860
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
windowTitle, windowModality, windowIcon and so on are named that way
to be similar to the ones in QWidget. However QQuickWindow inherits
all of the declared properties, and we would like to have shorter
property names in QML. If you are working with a Window then it's
obvious the title property is the window title. Unfortunately,
there must be patches in many other modules which depend on this one.
In order to avoid the need to merge them all at the same time,
there is also patch https://codereview.qt-project.org/#change,39001
which temporarily adds backwards-compatible accessors, which can be
removed after the other modules are able to build without them.
We should not rename windowState to state, because in QML, state
usually drives the state machine for animation transitions etc.
(although QWindow is not an Item, a user might get confused about it).
Related patches are
https://codereview.qt-project.org/#change,39001https://codereview.qt-project.org/#change,37764https://codereview.qt-project.org/#change,37765https://codereview.qt-project.org/#change,37766https://codereview.qt-project.org/#change,37762
Change-Id: Ie4424ec15fbdef6b29b137f90a2ae33f173edd21
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
The current implementation requests the platform window to set
the window state if it can, and return the actual window state
back.
The problem with this approach is that the platform window is created
as late as possible, so a call to QWindow::setWindowState would in
many (most?) cases never be forwarded to the platform window (instead,
the platform window is responsible to check the current window state
upon creation). As such, the window state might be left unsynched with
the platform window.
This patch suggests removing the return value from
QPlatformWindow::setWindowState. This will at least be consistent, so
that setting/getting state would produce the same result independent of
delayed window creation. If needed, we can later add new API to
QPlatformIntegration or QPlatformWindow for querying supported/actual
window state.
Change-Id: Ie43f56169656854a765ce88b47a808f8f3d51bb4
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Thomas McGuire <thomas.mcguire@kdab.com>
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
If custom cursor was set before the window was created, it didn't
actually get set, and in some cases even caused a crash.
Fixed by making sure the cursor is correct when showing widget/window.
Task-number: QTBUG-27535
Change-Id: I3bc946a9c406c96af5b86869a3a54893f8980aba
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Gatis Paeglis <gatis.paeglis@digia.com>
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
qwidget_qpa.cpp was setting cursor directly to platform window, so
QWindow didn't actually know what cursor was supposed to be shown.
Since QWindow now supports setting cursor and has override cursor
handling there, set cursor via QWindow instead of platform window.
Also changed QGuiApplication override cursor handling so that
it will query the cursor from QWindow if there is no override cursor.
With these changes, widget's custom cursor will be properly shown
when final override cursor is restored.
Task-number: QTBUG-24674
Change-Id: I2d2bb1027779256f9d6de560b6533f45e205ffe9
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Use the native parent's window if the widget in question does not
have one. This should be in line with Qt 4.8 using effectiveWinId().
Remove redundant code in grabMouse(QCursor).
Change-Id: Id6ab192e739221fe89f865f4d2f7a6d4671a190b
Reviewed-by: Qt Doc Bot <qt_docbot@qt-project.org>
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Change copyrights and license headers from Nokia to Digia
Change-Id: If1cc974286d29fd01ec6c19dd4719a67f4c3f00e
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Sergio Ahumada <sergio.ahumada@digia.com>
QWidget's mapToGlobal() and mapFromGlobal() functions assumed that
if the widget reports it's a window or if it has no parent widget, it
must be a top level window whose coordinates are in global coordinates.
This is not true for child QWindows or embedded native windows
(QAxWidgets).
Changed the logic for mapping coordinates to use equivalent methods
from QWindow if widget has a window handle, and changed QWindow's
methods to map coordinates using native methods if window is embedded.
Also fixed newly failing accessibility autotest. The geometry related
failures there popped up because now the position of the rect returned
by accessible interface is actually correct while widget geometry still
reports position 0,0 before widget has shown up.
Task-number: QTBUG-26436
Change-Id: I658fafd0ce01eb1604ba255efeeba3073ca0189f
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Showing a compo dropdown would cause the parent to
become deactivated (Windows). Introduced by:
8368557b9c for
QTBUG-25852 , QTBUG-23699 .
Task-number: QTBUG-27145
Change-Id: I7b89697a2c55d58cc0ab184b61a0cff8bf2275ec
Reviewed-by: Qt Doc Bot <qt_docbot@qt-project.org>
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Jonathan Liu <net147@gmail.com>
Reviewed-by: Jørgen Lind <jorgen.lind@nokia.com>
Size constraints set on a widget before the creation of
the QWidgetWindow were lost (for example, Qt Creator's
preference page).
Task-number: QTBUG-26745
Change-Id: I7c2f5aed9c8817795603e5ad3c24418d66627bab
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: hjk <qthjk@ovi.com>
For child widgets we must use the windowHandle of the toplevel widget.
Change-Id: If7bad8ed4bba694806a2c8b77bae3d25cc200b8b
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
When mixing native and regular widgets in same QMdiArea, some
subwindows didn't properly get set native. This was because
when a native parentless widget was given a parent, it wouldn't
enforce native window on the new parent and its ancestors.
This happened because window flags were adjusted too late in
relation to createWinId() call in setParent_sys().
Fixed by moving the createWinId() call to its proper place.
Also removed some old Q_WS_* ifdeffing in QWidget::setParent() that
masked some native enforcement code.
Additionally removed few QEXPECT_FAILs from QWidget autotest now
that those cases work correctly.
Task-number: QTBUG-26424
Change-Id: Ib6f9d0531e5c7299e2c307734d49c81f1ffa9713
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Ensure the initial size is valid, since we store it as normalGeometry
below.
Task-number: QTBUG-26226
Change-Id: I3a55c389a48504699942930063089c80657687a0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Thomas McGuire <thomas.mcguire@kdab.com>
This fixes tst_QWidget::movedAndResizedAttributes() for platforms
that don't have support for QPlatformWindow::setWindowState().
Change-Id: Id0f123d11b08a75c0c131080d509e6b23b281600
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com>
Don't call show() inside of QWidget, as that might make the window
fullscreen due to QStyleHints::showIsFullScreen().
This regression was introduced earlier when QWidget::show() gained
support for the hint. No auto test caught the problem, as the
affected platforms (EGL and QNX) are not CI-tested.
Change-Id: I647c2362ac5b53ced562f1d77848552cc122fba0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Removed old defunct platform specific code from
QWidget::isActiveWindow() and added call to
QPlatformWindow::isActive() instead. This is done because
the embedded native windows inside QAxWidgets can have
focus but are not part of the parent application's Qt
window hierarchy, so native methods are required to determine
if they are part of the active window or not.
QWidgetPrivate::setFocus_sys() was implemented to activate
the window of the focused widget if the focus was elsewhere.
This is required because embedded native windows can steal the
focus from the main application window.
Focus event handling in Windows platform adaptation plugin was
fixed to correctly identify the active window in cases where
the are embedded native widgets that can have focus.
Also fixed three test cases that were affected by these changes.
Task-number: QTBUG-25852
Task-number: QTBUG-23699
Change-Id: I817e0ce4317e88955bb49b034eacd630a876ccf0
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
The original commit message follows.
Fixes: Fix the windows PaintOnScreen issue once and for all
Details: To allow both the case where X11 people accidentally set
PaintOnScreen (which should not have any effect on windows)
and where people set it and subclass with paintEngine() {
return 0 } to use GDI / DirectX we do this rather nasty hack.
Original commit in Qt4: 07a2f68bd5869152471e4ffc4a63c683ef141ae8
Autotest: tst_QWidget::paintOnScreenPossible
Change-Id: Ifbb7dc4611959be3ecc362c29a1c3436b0e0fa82
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@nokia.com>
Paul Olav Tvete
Chris Colbert
David Faure
Alex Montgomery
Gatis Paeglis
Jian Liang
Tor Arne Vestbø
Gunnar Sletta
Friedemann Kleint
Laszlo Agocs
Andrew Knight
Samuel Rødal
Gabriel de Dietrich
Eskil Abrahamsen Blomfeldt
Morten Johan Sørvig
Jake Petroules
Axel Waggershauser
Frederik Gladhorn
Shawn Rutledge
Sergio Ahumada
Richard Moe Gustavsen
Oliver Wolff
Miikka Heikkinen
Joerg Bornemann
Friedemann Kleint
Iikka Eklund
Joerg Bornemann
Thomas McGuire