"Shared objects still referenced" dlerror should actually be treated as
"for your information" only, not as an actual error.
Change-Id: Ie02bd1db0dd2dc93bb759f8b6c7e825070e17bb9
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
offsetHi needs to be casted first before shifting.
Change-Id: I29c773dd13d5b16042629604015bbf5645fab861
Reviewed-by: Andrew Knight <andrew.knight@digia.com>
Enabled for QToolBar to allow it to overlap OpenGL widgets when
expanding.
Task-number: QTBUG-33082
Change-Id: I76dc8da52bc04eedc6d6779c48753da100ed1c9f
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
This is not Qt style, not needed, and clutters the code, so remove
QPageSize:: qualifications where they're not needed.
Change-Id: Ia93ac9523ef43a40cf4dab3bcb383a54af666c96
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
The created page format must always be released,
even if one of the subsequent PM* functions fails.
Change-Id: If42aaeccd6bdb51ba53444f491ca2878783d0678
Reviewed-by: John Layt <jlayt@kde.org>
Fix reference counting error in QCocoaPrintDevice::
createPageSize(). "key" is accessed with a "Get"
function and should not be released. Switch from
using QCFString to a plain CFStringsRef with manual
ref counting.
Task-number: QTBUG-38023
Change-Id: I04d661bffeb5b3122b0c3c8eaaffdd1af51842fd
Reviewed-by: John Layt <jlayt@kde.org>
The big change is that we now keep the id objects representing accessibles
around so that they are persistent for ATs.
This improves performance of Mac accessibility significantly.
This is required for notifications which are now sent so that many things work much better,
for example the VoiceOver focus follows the keyboard focus.
The parent element in QCocoaAccessibleElement was removed, we can
dynamically access it more reliably.
Change-Id: I686d212f40d28b392dcc22f16f3c3430f08bdc98
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
The example is incorrect: no context parameter is needed.
There was also no mentioning of the fact the the context must be
current at the time of calling initializeOpenGLFunctions(). This
is corrected too.
Change-Id: If8695140096e4b4f84927579c099b0af80750703
Reviewed-by: Sean Harmer <sean.harmer@kdab.com>
This way we can customize QPlatformWindow by using
dynamic "_q_platform_" properties in corresponding
QWindow.
Change-Id: I987b7a17d252541fe766af6aa37a6cffb67f1807
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
Commit 0d4918950e added a new place
(QXcbWindow::setTransparentForMouseEvents()) where requests from the
XFixes extension are generated. However, this wasn't checking if the
extension is actually supported before using it.
Fix this by turning QXcbWindow::setTransparentForMouseEvents() into a
no-op if the XFixes extension isn't available.
This means that the window in question won't be transparent for mouse
events, but we cannot do much about that if the X server doesn't support
the required extension.
Task-number: QTBUG-38109
Change-Id: I2931481eb71bab990f2dcf0ec600b9e62d3d799c
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Recognize that the d-pointer can never be nullptr and drop the
nullptr checks.
The d-pointer can never be nullptr as there's no move ctor and all other
ctors create a QPageSizePrivate.
Change-Id: I6c4e165949ed55510aefbc2d933f20fc8e624333
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The old code always compared the Private classes, unless one of them
was nullptr (which it never was, as there is no move ctor and all
other ctors create a QPageLayoutPrivate.
The new code compares the dpointers, and only if they differ, the
Private classes. It also drops the nullptr checks, as they cannot
trigger.
Change-Id: I523c3503e2edb520f98f9b4e2e3bdaf28a9a355d
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Better style, since it treats the left-hand and right-hand size symmetrically,
e.g. for implicit conversions.
Change-Id: Ib5d39b1ebffffcb664ee2f72c756702469e32d3b
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Many setters check for an actual change before making changes to
data members, but because of QSharedDataPointer, the detach had
already happened by the time the comparison returned false.
Change-Id: I320806e74de4a64fa3a340831621c1f5120ebb0f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: John Layt <jlayt@kde.org>
This is not Qt style, not needed, and clutters the code, so remove
QPageLayout:: qualifications where they're not needed.
Change-Id: I62f90c29bcb9f3c137d319051ac79e081fe5fb69
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: John Layt <jlayt@kde.org>
Better style, since it treats the left-hand and right-hand size symmetrically,
e.g. for implicit conversions.
Change-Id: If1ffa9dc2a018e402c884f60bfbc82e799daeb92
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
A QSize or QSizeF is not an accurate representation of a QPageSize, so
the corresponding constructor should be explicit.
Change-Id: I6b1808e5f93e3caef948c0e5300bd3c20e3c4210
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: John Layt <jlayt@kde.org>
This could happen in the following situation:
* a custom animation driver with fixed delta
* a triple-buffering scheme (rendering ahead a frame)
* a second animation timer starting while a first was active
This would cause QUnifiedTimer::startTimers() to trigger
QUnifiedTimer::updateAnimationTimers(-1), and use the current time from
the QElapsedTimer rather than the animation driver. This time could be
less than the last reported time from the animation driver.
Change-Id: Ibf1796fcb99f288d4946b30e5e7225695aa61781
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
They were of the form "warning: using the result of an assignment
as a condition without parentheses [-Wparentheses]"
Change-Id: I049bf0f67073bf41310ca5ee73f17e5e69de569f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: J-P Nurmi <jpnurmi@digia.com>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
The image format of the backing store should use the byte order of the
machine where the application is running, not the native format of the
X server. Then, if the byte order of the backing store image differs
from the X server's native format, it needs to be converted before
being sent across the network.
Task-number: QTBUG-29898
Change-Id: Ic91c8ffb814c6beeb9f1d9195174a47d8bd94a90
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
QNX version of screen doesn't reallocate window buffers dynamically.
The buffers have to be destroyed and recreated when a window is resized.
As the overhead is minimal this will be done on BlackBerry, as well.
Change-Id: I488942879822c64a6ab1871ebf5d6da9aec144d6
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
The CoreText engine will remove the PDF token from the end of
the string (instead of producing a zero-width glyph for it),
thus the output will be different from the OpenType backend
and Qt will get confused. To fix this, we emulate the expected
behavior by molding the output in a special case.
This is a port of e45c4387ae16627d61e30a58ae901d888d375aa7 from
Qt 4.
Task-number: QTBUG-38113
Change-Id: Ia0a078e3a60317981d4d5a4ee7e575a1714a1d75
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
QDateTime::addDays() is a const function and returns a new QDateTime with
the given days added, thus the current statement had no effect.
Found by applying Q_REQUIRED_RESULT in dev branch.
Change-Id: I1b061619d45d7806feaa2bf9fb6d9f0b43d63def
Reviewed-by: Richard J. Moore <rich@kde.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QDateTime::addSecs() is a const function and returns a new QDateTime with
the given seconds added, thus the current statement had no effect.
Found by applying Q_REQUIRED_RESULT in dev branch.
Change-Id: Id712334f91e0adb40bafc23470bf46479334c81a
Reviewed-by: Richard J. Moore <rich@kde.org>
The feature was removed from the C++14 draft standard and moved instead
to a Technical Specification. Since we don't know how to enable TS
features in GCC 4.9 yet, remove it from the definition.
The Clang definition is probably safe, since it is behind an #if
__has_extension.
Change-Id: Ibc32b35657b046680078b39a7678bd8e1e5395d2
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Redefining macros is always a bad idea. On top of that, MSVC doesn't
like when we #define dynamic_cast, even though the C++ standard
explicitly allows it.
Task-number: QTBUG-29093
Change-Id: I6e33d609ce213cf6a9085faa3f991a873d825dc6
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
instead of rolling our own qt_clamp().
Change-Id: I2729ae3ff98e8c29c66f0f5c792b1bc7bf586f06
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
... and not only one. This was a problem e.g. when there were several
requests to the same host and the host was not reachable; only one
reply would get an error signal in case we suppressed other errors in
"happy eyeballs" host lookup style.
Task-number: QTBUG-36890
Change-Id: I1b5757498bd644b0d773cf6c43e4950620949c5c
Reviewed-by: Richard J. Moore <rich@kde.org>
If the style hints it, QComboBox::showPopup() will try to
show a native popup instead. This is currently undocumented
and an opt-in feature (typically by using a proxy style).
Works only on Mac so far.
Task-number: QTBUG-32731
Change-Id: I4447e884cbd6b490f3039c7a95168698c0bed16e
Reviewed-by: Jens Bache-Wiig <jens.bache-wiig@digia.com>
This remains an opt-in solution bound to the usage
of SH_ComboBox_UseNativePopup in a proxy style. The
midterm goal is to make this option on by default,
possibly in 5.4. This solution is and will remain a hint
in the sense that some exotic use cases of QComboBox
(e.g., when setting its view) are inherently incompatible
with the native popup idea.
Task-number: QTBUG-32731
Change-Id: I2a3d780795c22f9989e44325fcaf314538b1de49
Reviewed-by: Jens Bache-Wiig <jens.bache-wiig@digia.com>
Since Qt expects a mouse press/release event to not change
mouse position in the same event, QGuiApplicationPrivate::processMouseEvent
will detect if a QWindowSystemInterfacePrivate::MouseEvent
tries to do this and convert it to two events; one move
and one press/release.
The problem is that the extra mouse event gets posted. So if
delivering the first event causes a flush in the event queue
(which can easily happen if e.g calling processEvents), the
second event will be processed before the first returns.
On iOS we see a crash with DnD as result of this, since drag
data gets deleted on mouse release, and returning back to a mouse
move after that will cause dangling pointers.
This patch will instead of posting the event, call the
event handler recursively with the faked event as argument.
That way a flush will not cause the "pending" event to be
delivered.
Change-Id: Id9d88053b4859083fedd666584815016d67ac51b
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
The Windows version for QS_TOUCH and QS_POINTER was slightly off.
Change-Id: Idb8a8219e09c6aa1a1b24e45b9da640c8d4b7161
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
After c4aabeb2b8 we now try to access
q_ptr in QWidgetPrivate::setRenderToTexture. QOpenGLWidget needs to
delay this call after the QObject constructor is done to make sure
that setTextureChildSeen doesn't dereference q_ptr before it has
been initialized.
Change-Id: Icaee82c8b806f42bc7614b0ac6fe4e6026331750
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Now menu items and key shortcuts for Cut, Copy, Paste and Select All
work in the standard ways in dialogs such as the file dialog, provided
that the corresponding QActions have been created and added to the menu.
This depends on new roles to identify each menu item which is so
broadly applicable that it should work even when a native widget has
focus; but the role will be auto-detected, just as we were already
doing for application menu items such as Quit, About and Preferences.
When the QFileDialog is opened, it will call
redirectKnownMenuItemsToFirstResponder() which will make only those
"special" menu items have the standard actions and nil targets. When
the dialog is dismissed, those NSMenuItems must be reverted by calling
resetKnownMenuItemsToQt(), because to invoke a QAction, the NSMenuItem's
action should be itemFired and the target should be the
QCocoaMenuDelegate.
Task-number: QTBUG-17291
Change-Id: I501375ca6fa13fac75d4b4fdcede993ec2329cc7
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
For Mac this makes QTextEdit work nicely with VoiceOver.
Task-number: QTBUG-37204
Change-Id: I1326d24ca6a932ad667ee395f62881b6ec64e892
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
The new QPageLayout methods weren't originally added to QPagePaintDevice
as no new virtuals can be added, instead static polymorphism was used to
add the methods directly in the derived classes QPdfWriter and QPrinter.
This however means that classes like QTextDocument with print() methods
that take a QPagedPaintDevice are unable to access the QPageLayout
methods. To fix this, instead make the QPagedPaintDevicePrivate a
virtual class and have QPdfWriter and QPrinter implement derived
private classes that are called by the non-virtual QPagedPaintDevice
base methods.
Change-Id: Ieb6e513b1fa05f5ae76ea1f9156b0b1a053089eb
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This allows us to get the current value of the QMutex / QBasicMutex
after the testAndSet operation failed. It saves an extra load from
memory.
Change-Id: I4922a8b3df15e342b177b13f56cf4f1184314520
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
