QXcbKeyboard only deals with core_device_id. With the reporter's test
case: using xdotool to send a string, the deviceID is not changed,
then we assume xkb_state_new() is more stable(or correct) than
xkb_x11_state_new_from_device() in this case.
See also https://www.x.org/releases/current/doc/man/man3/xcb_xkb_new_keyboard_notify_event_t.3.xhtml .
Fixes: QTBUG-95933
Pick-to: 6.2 5.15
Change-Id: Ic595e1f1424fbc6814871a85ac159907f1aeb12a
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Zhang Hao <zhanghao@uniontech.com>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Prevents QML app using QtActivity never calling
AndroidJniMain::startQt..'s to jam as ANR
Pick-to: 6.2 5.15
Task-number: QTBUG-97115
Change-Id: Ibfe8579dbb701068f4896b6d826ff487094bdf56
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
Mostly a removal of dynamically loaded Win32 APIs.
Since Qt 6's minimum supported platform is Win10 1809
(10.0.17763, code name RS5), all these functions will
be available and no need to resolve them at run-time.
Things not remove:
WinTab functions in "qwindowstabletsupport.cpp".
Not my familiar area, so not touch it.
Pick-to: 6.2
Task-number: QTBUG-84432
Change-Id: I7ad6c3bc8376f6c0e3ac90f34e22f7628efeb694
Reviewed-by: André de la Rocha <andre.rocha@qt.io>
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
To enable a list of windows the app has open in the Dock's menu.
Not to surprise existing applications with a 'Window' menu where
they did not have it before, make the item hidden.
Fixes: QTBUG-59433
Change-Id: I1ac3d3de69f4313f39c4631dc4b68bf6e096532a
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Otherwise the original backingstore image will detach from the
m_xcb_image data on the next backingstore paint or scroll.
Pick-to: 6.2
Change-Id: I73f68d9c2e7c106951541831a5df8b97695f2001
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Some multi-touch devices send touch information for each finger using
different WM_POINTER messages/frames, instead of a single one with
a list of touches, like most devices. This would result in the generation
of multiple touch events, which can cause unexpected behavior in
applications (the QTouchEvent documentation specifies that it should
contain all simultaneous touches). This patch adds a workaround to
ensure all simultaneous touches are included in the events, to comply
with the expected behavior.
Pick-to: 6.2 5.15
Change-Id: I12a2f84b35a6bdd49ee53d25de580c0941a9aea6
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
The m_pendingFlush variable is used to track what is missing in the
server side backingstore. If we're doing a client side scroll the
pending area is still the same.
If we were to always discard the scrolled area from m_pendingFlush
we would get in trouble on the next non-client side scroll, as
we think the content exists server-side.
Pick-to: 6.2
Change-Id: Ie50a99a8e5d8a83d1299c53534a1c83c6bfb47bd
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
The area we're going to fill is the one after applying the scroll delta.
Pick-to: 6.2
Change-Id: I254830a15e5f4c93ba28ed8f0a9b35c40f1d1af2
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
QMacStyle has a single NSSlider that is used to render any QSlider. For
each QStyle API operating on a slider, the style sets the slider up with
respecive properties. On macOS 12, the NSSlider maintains some states
that make QSlider instances influence each other's knob position when
rendering, resulting in uncontrollable jumping of the slider.
This can be fixed by not using startTrackingAt/stopTracking APIs, which
are however the only way we have to make the slider knob get rendered
pressed - there is no property in NSSlider(Cell), and none of the NSCell
attributes have any effect. So we need to use startTrackingAt, and work
around the side effect by reinitializing the NSSlider by calling
initWithFrame.
This fixes the positioning error, but also causes flickering of the knob
when dragging. To fix the flickering, we have to always call
startTrackingAt for a slider that is pressed, even for calls to
setupSlider that are made in QStyle APIs that are not drawing anything.
Also tried with no complete success (either positiong bug or flicker):
* call prepareForReuse on the NSView
* always call stopTracking on the NSSlider
Fixes: QTBUG-98093
Pick-to: 6.2 6.2.2
Change-Id: I3423b9f7cb125a59831c6722509ab3b74742b6ae
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Calling doneCurrent() on a QCocoaGLContext only clears the current
context, but doesns't reset the drawable (view) of the context. In
most cases this is fine, but it has been observed to cause issues
when using the software GL renderer on Big Sur and above.
To be a good citizen we clear the drawable of any of our contexts
that are tied to the NSView this about to be go away.
Pick-to: 6.2 6.2.2
Change-Id: I8c845727c50871f30fbebc2ed62a7d0485651ecf
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
They map to the data exposed by libinput exactly the same way as
touchpad gestures on Wayland. The implementation is functionally the
same and follows the same patterns to preserve similar behavior across
X11 and Wayland.
For example, we use the last known pointer position as the position for
gestures, even though on X11 this data is available as part of events.
The new implementation is only enabled if the used xcb supports the
required APIs.
[ChangeLog][Platform Specific Changes][X11] Touchpads can now detect
multi-finger gestures and send RotateNativeGesture, ZoomNativeGesture
and PanNativeGesture events, since XInput 2.4 and X Server 21.1.
Change-Id: If404dcf8385210deadeb7e7c6d29171e9abc9e50
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
The slider is very touch friendly from macOS 11 on, and the knob is quite
large. Give it some extra pixels, and adjust the test accordingly.
Pick-to: 6.2
Task-number: QTBUG-98093
Change-Id: Iedf6db1081cdd4013ca29ce760aea1e0361b1123
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
There are still more failures, but these are low-hanging fruit.
Pick-to: 6.2
Change-Id: I3d3a1dc9e754bd4865c850cc1549b0b6fd6fa2f8
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The focus object can emit selection updates (e.g from mouse drag), and
accept modifying it through IM when dragging on the handles, even if it
doesn't accept text input and IM in general (and hence return false from
inputMethodAccepted()). This is typically the case for read-only text fields.
So we should listen for selection changes and enable handles also for
this case (unless the IM hints tells us explicitly not to use handles/edit menu).
Fixes: QTBUG-91545
Change-Id: I2855505fc229e954b2c43f5e11374e64bba7eb4e
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
In Qt, Qt::ImEnabled means that the focus object accepts text input from
input method (IM) events. But the IM API also contains API for dealing
with text selections. Text input and text selections are logically two
different operations, but since IM makes use of selections to implement
text input (like selecting a word to suggest a spelling correction), it's
understandable that they are combined into to same API.
So when a focus object reports Qt::ImEnabled to be false, it only means
that it doesn't accept input. E.g a TextArea in QML with "readOnly:true"
will set Qt::ImEnabled to false. At the same time, it can have
"selectByMouse:true", which lets you select text with the mouse.
This behavior is consistent in Qt, for both Quick, Controls 2 and Widgets.
Since we want to support any selections done in controls/widgets on iOS
with selection handles and edit menus, regardless if the focus object
accepts input or not, this patch will set the QIOSResponder (with read-only
actions) as first responder when we detect a focus object with Qt::ImReadOnly.
This means that if a query for Qt::ImReadOnly returns "true", we take that
to mean that it implements the IM API, but without accepting input.
Task-number: QTBUG-91545
Change-Id: I07349909a3bca81f484a2e9af9672428dca62c49
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
QIOSTextInputResponder has two responsibilities; It takes care of
handling text input from UIKit, and to implement first responder
actions related to the edit menu, like copy and paste.
Currently the responder offers both writable (paste) and
readable (select, copy) actions. Because of the former, it means
that it can only be used for focus objects that accepts text input.
Since we also want to be able to show an edit menu for selections
done on a read-only input field, this patch will factor out the
read-only actions we want for that case into a QIOSTextResponder
base class. An instance of this class can be used as first responder
for a focus object that has read-only text, but otherwise doesn't
support text input. This part is implemented in a subsequent patch.
The remaining set of writeable actions, together with input method
handling, will continue to be in the QIOSTextInputResponder subclass.
Task-number: QTBUG-91545
Change-Id: I1c215bb509eb7820c6c60f7ad806f61a5de02ded
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Cargo-cult copied this from somewhere, unnecessarily. This just made the
code bigger for no reason.
Task-number: QTBUG-97950
Pick-to: 6.2
Change-Id: Ice04365c72984d07a64dfffd16b48520d42c7043
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
And bump NTDDI_VERSION to 0x0A00000B (NTDDI_WIN10_CO) at the same time,
to unblock the developers from accessing the latest Windows APIs.
Pick-to: 6.2
Change-Id: Ifbc28c8f8b073866871685c020301f5f20dc9591
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
As a drive-by: fix some improper indentation
Task-number: QTBUG-91024
Change-Id: I29e04aff3638dfb2aab9d40650c55a48baba7222
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The need to finalize the back buffer isn't limited to flushing.
Pick-to: 6.2
Change-Id: I98b04ab49ec27ea536e99462deab8d48a8e40e82
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
When introducing support for scrolling the backingstore it doesn't make
sense to track the painted region explicitly.
Pick-to: 6.2
Change-Id: I370932f02490ac526fb049908f99af678884e807
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
CoreAnimation doesn't immediately mark a surface as in use the moment
we assign it to a layer, but defers it until the surface has bee picked
up by the window server. In theory this would allow us to defer the swap
until the next beginPaint(), which would allow painting to the back buffer
again before Core Animation has time to flush the transaction and persist
the layer changes to the window server, and would also automatically deal
with requests to flush without painting anything.
But, since a client may do several rounds of beginPaint/endPaint before
flushing, we might end up in a situation where we detect that a surface
is in use in the middle of several paint rounds, and end up swapping in
a new back buffer without copying over the previously painted content,
like we do in prepareForFlush. To be on the safe side we swap the back
and front buffer straight away.
We also need to mark the surface in use, to prevent the same problem
from appearing when the window server doesn't pick up the surface in
between two rounds of flushes.
Pick-to: 6.2
Change-Id: Ib418852424773c399df710da33a388e8d2c2e92c
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Doris Verria <doris.verria@qt.io>
We can simply propagate the minimum window size.
Pick-to: 6.2 5.15
Fixes: QTBUG-83503
Change-Id: Ia9f67c92ca37fd0f4e89f08e5153e12c1385f6f8
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Instead, similar to 'Fusion' style use what common style can do,
thus respecting text alignment and not always forcing VCenter
alignment.
Fixes: QTBUG-97698
Change-Id: I89ad01807afc217c3c33650e1f1d255d0b81ed46
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Extract and reuse a lambda for initialization and slot object.
Pass a QDBusObjectPath parameter by const-ref.
Avoid implicit ascii-to-utf16 string conversion.
Change-Id: I591ae6e58e010ccabcf9012db866949492e0191d
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Doing it the opposite way, by associating the new screen first, will
result in the external screen not going back to mirroring the main
display.
Pick-to: 6.2 5.15
Change-Id: I63970380fc4f0902af5032043809a9c1b1f9f95b
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Associating the UIWindow with a different screen will trigger layout
of the child views of the window, including the view that we're in
the process of removing, which doesn't have a platform window anymore.
Instead of protecting every possible code path in the view code
with checks for a platform window we defer the restoring of
mirror mode until after the view has been removed.
Pick-to: 6.2 5.15
Task-number: QTBUG-94530
Change-Id: I8c66106cafa67e06721e621c019b2d10acf02326
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
The window may have been destroyed, but the UIView may still be
be referenced and kept alive by the a11y subsystem.
Pick-to: 6.2 5.15
Task-number: QTBUG-94530
Change-Id: I24f1c9d45e80c1bb4c92536e7f91533a94fd077f
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
It's somewhat difficult to reproduce, but having a mac with a trackpad
and a huge scrollview (meaning I have to scroll fast and a lot), I am
able from time to time to trigger an assert on the line:
Q_ASSERT(pixelDelta.isNull() && angleDelta.isNull())
In all such cases, I can see that deltas are not zero, but equal
to the ones that the next event (momentumPhase == Begin) has.
The code is based on Tor Arne's patch.
Pick-to: 6.2 5.15
Fixes: QTBUG-97945
Change-Id: I874c776b265d3950cc2b6c1d8054363b3d0d1fde
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This allows Qt Quick Dialogs to accurately check how it should show
files and directories.
Change-Id: I0f5102553ff9a0484b3714ba176f7e5e668fd05c
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Remove useless overrides of QAbstractFileEngine methods from the derived
classes. Also remove "This virtual function must be reimplemented by
all subclasses" passages from the QAbstractFileEngine's documentation.
There are pure virtual methods for such use cases. QAbstractFileEngine
already contains useful defaults for classes not supporting all the
functionality.
Change-Id: Ia25965854f3809b15d7502da3749cc2f3414bbc3
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
This reverts commit 816c5de460.
Reason for revert: Causes crash on Nvidia when Qt is configured
with -opengl es2
Fixes: QTBUG-97738
Change-Id: I9fe43146b922fc770890a144fdb1bd1c564635bb
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
This variable was used in Qt5 to specify the install location of the
MySQL library. In Qt6, MYSQL_INCLUDE_DIRS and MySQL_LIBRARIES serve
this purpose, and MYSQL_PATH is unused.
Change-Id: I2df12b79777af50c719a5b6bc5feadc2c59c0c3d
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Change e946e6895a implements proper
support for ICCM 4.1.4 window state handling. One issue it addressed
is that a show() after a hide() needs to be delayed until the window
manager/X-server has processed the previous event.
This was handled with a deferred task list to send the map/unmap
events. According to ICCM, we should wait for the _NET_WM_STATE
notification before processing the deferred events.
But this is only true for top level windows, as child windows are
not handled by the window manager and will never receive any
_NET_WM_STATE notifications. For those, we should use the unmap
notify event, which means that the X-server has processed the unmap
and handle deferred events once that notification has been received.
This fixes an issue in Qt Multimedia, where QVideoWidget would not
show the video anymore after a minimize of the player or when making
the QVideoWidget fullscreen. This is because QVideoWidget uses an
embedded QWindow to render video using HW acceleration.
Fixes: QTBUG-97257
Pick-to: 6.2 5.15
Change-Id: I5c47eba3276a1f243bdafd5346f353c7843403bb
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
“Handle invisible child windows gracefully on dpi changes”
This fix was accidentally removed by commit cd96d870
“Move VM_DPICHANGE handling to QWindowsWindow”.
Fixes: QTBUG-96466
Pick-to: 6.2
Change-Id: I3774f6305631ba47282d43e8480e2acaba517a96
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
It can only differentiate between cellular and not cellular
and then we can determine if it's disconnected or (presumably) using
wifi.
It is also explicitly not supported on macOS, which adds to the
confusion.
Task-number: QTBUG-91023
Change-Id: I1d002ba06dd9acf1a0daabfb2a4193c07871e9b4
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Change 64d62c53c1 started treating an
ACTION_CANCEL as a TouchPointReleased. This leads to unintentional
presses, if e.g. an edge swipe for the android back gesture starts on a
MouseArea.
When Android takes possession of the motion, an ACTION_CANCEL is
delivered, which needs to be handled as such. It should not be treated
as a normal up event that triggers a press. Otherwise, we get the
above-mentioned issue, where an unintentional action is performed.
So let's use QWindowSystemInterface::handleTouchCancelEvent to treat it
as a canceled touch in Qt.
Task-number: QTBUG-72110
Pick-to: 6.2 5.15
Change-Id: I68650a180130501568e5a1e9a7f20ea469461e97
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
The previous implementation would occasionally result
in a crash when a window was closed.
Apparently 'auto' cursor does not work so well,
so we now use 'default' name.
Fixes: QTBUG-96178
Pick-to: 6.2 5.15
Change-Id: I0179d20dbdc01f0e3021d746324e1e39c678a298
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
This reverts commit 0921caf976.
Reason for revert: This change breaks mouse press handling in popups themselves. I'll make another solution.
Change-Id: Ib9fd41f3f078e6abf22e5b0f75724d4acc737ead
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Liang Qi
Rami Potinkara
Yuhang Zhao
Timur Pocheptsov
Tor Arne Vestbø
Morten Johan Sørvig
André de la Rocha
Volker Hilsheimer
Povilas Kanapickas
Mitch Curtis
Richard Moe Gustavsen
Thiago Macieira
David Skoland
Mårten Nordheim
Lorn Potter
Marc Mutz
Ievgenii Meshcheriakov
Paul Olav Tvete
Joerg Bornemann
Lars Knoll
Mike Achtelik
Øystein Heskestad