QWindow uses device independent geometry while QXcb and QPlatform
classes do not.
When QXcbWindow::create is called we have no guarantee that the correct
screen has been set in QWindow so the code
attempts to check if "currentScreen" matches "actualscreen". To perform
that operation though we need to convert
the units from "Device independent" to "native pixels" that we do by
calling QPlatformWindow::windowGeometry which
calls QHighDpiScaling::toNativePixels which requires the correct screen
to be already set.
So basically we have a cyclic dependency, to get the correct screen we
require the correct screen to be already set.
To fix this we can:
1-Remove the dependency (Look for the actual screen using device
independent pixels)
This will imply adding code in QXcb to use
QPlatformScreen::deviceIndependentGeometry to lookup the screen up
2-Make sure the Screen is set before calling QXcbWindow::create
This patch implements the first approach that allows us to keep the
changes within the QXcb backend which seems to be the only one affected.
Task-number: QTBUG-53813
Change-Id: I6dc955d63e17c3b3421f3a1a9e0d841e508b2e5c
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
It was grabbing a QLabel without accounting for the size of
the window in the case where the DPI is larger than 1:
FAIL! : tst_QWidget::translucentWidget() Compared values are not the same
Actual (actual.size()) : QSize(32x32)
Expected (expected.size()): QSize(16x16)
Change-Id: I4873f3c6364ee2696f5612d91e6c97c60b2cd915
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
tst_QSharedPointer can't create a pipe as the OS has too many files
open. Systems like macOS have a lower limit to these simultaneous files
open.
Task-number: QTBUG-60410
Change-Id: I21e89f992ada2a7d09b706522a05b5952f00ec33
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
QSqlQueryModel::roleNames was exposing the default QAbstractItemModel
roles. Only "display" role makes sense because it's a read-only model
and it is not providing any other data information.
[ChangeLog][Important Behavior Changes] The names of the roles
returned by QSqlQueryModel::roleNames now only include a name for the
Qt::DisplayRole. Previously all the roles names of QSqlQueryModel
were returned.
Task-number: QTBUG-60857
Change-Id: Ib18aa0e7083a828648767d700c5af05b6aa84f4f
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
This is a followup to 77a8e90cdd which
didn't handle the case where no columns had been moved.
visualIndices and logicalIndices are empty until initializeIndexMapping()
is called, in which case appending is wrong.
As a result, visualIndex(i) would return -1 for the values over
those added by read(), and an assert would happen at painting time.
The fix is to leave visualIndices and logicalIndices empty if
they are empty already, leaving it to initializeIndexMapping()
to fill them later if necessary (e.g. when moving a column).
Task-number: QTBUG-60837
Change-Id: Ia7e4b9d3122647984acd434dfaa0400df319d065
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Thorbjørn Lund Martsum <tmartsum@gmail.com>
On macOS if an application is no longer active then it will cause any
tool windows to hide until the application is active again. For
applications that did not want this behavior and thus wanted the tool
window to stay visible, the WA_MacAlwaysShowToolWindow flag is
available.
In order to ensure that this flag is respected, the tool window needs
to have its level changed when the application active status changes.
Once it is no longer active the window needs to be seen as a normal
window, and when it is active then it needs to be set to be a window
that is always on top to get the right behavior.
Due to various bugs in AppKit we need to explicitly order windows
in front during this process, which requires us to then iterate the
windows in back-to-front order. For macOS versions < 10.12 there is
no way to get an ordered list of windows, so we fall back to using
the window creation order.
Task-number: QTBUG-57581
Change-Id: If20b4698616707685f83b1378f87593f8169c8c6
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
They have been blacklisted on windows previously and now fail on
macOS 10.12 as well.
Task-number: QTBUG-60993
Change-Id: Ib7a3acfc7f2285c0a587d4abd88a4a218391d623
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
When drawing the branches in a QTreeView, it should use hasFocus() to
determine if it is the active widget or not. This is how it checks when
styling the other aspects of the treeview, so the branches should be
done in line with this.
Change-Id: Id721401424ad95e9fc7ffbdc991e53b5f7feac01
Reviewed-by: Thorbjørn Lund Martsum <tmartsum@gmail.com>
When the mouse is moved over a header section then if there is a status
tip then this should be sent as an event like it would for a typical
QAbstractItemView.
Also adds a test for the StatusTipRole for the QTreeView itself as well as
the header.
Task-number: QTBUG-2066
Change-Id: Iaef8d91f1bd621c2463cde2dff4b2291fb037975
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Limitation is that the signal needs to be parameter-less
[ChangeLog][moc] moc now supports NOTIFY signals of parent classes in Q_PROPERTY
Change-Id: Iad64c96c3ec65d4be8ad9ff1a9f889938ab9bf45
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Brett Stottlemyer <bstottle@ford.com>
[ChangeLog][Third-Party Code] Sqlite was updated to version 3.16.2.
Change-Id: Ib3e9c3a7021760e3cfa6dbd44f67b9da4abbe367
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Move the objects out of the data structure to avoid needless copies.
Change-Id: I1a69fccc431e040b229d6ea9ded0e041c208c861
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Every /Zc:XXX option qmake doesn't know about yields a "WARNING: Could
not parse Compiler option '-Zc:XXX'; added to AdditionalOptions."
Put all /Zc:XXX options we don't handle into AdditionalOptions without
printing a warning. There's no point in making all options known to
qmake and updating them for every MSVC release.
Change-Id: I319e027791a7b0a29d139ee3074ab1aed8ce8a63
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Same as with receiving data, the "newConnection callback" might be
triggered late and cause a crash if it accesses data of a destroyed
object. By moving the handling of this callback to a worker without
much logic we can prevent these late callbacks from doing any harm.
The signals are no longer connected and thus do not trigger any
problems.
Change-Id: Ic61584f12a46506abe12d7b21403d0c0970c0aae
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
MinGW 6.3 ships a gold linker that pops up a crash dialog
while running configure tests.
Change-Id: Ib25e30df664779e1ecced0ae37fa1e73e98f95ea
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
One was setting FramelessWindowHint, the other didn't.
Settle on setting it.
[ChangeLog][QtWidgets][QSplashScreen] All constructors now implicitly
set Qt::FramelessWindowHint, not just the (pixmap, flags) one.
Change-Id: I5e3919acac80bf31c2c61fbade938ff319a6cea9
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
QList::erase(it) is a linear operation. Calling it in a loop
constitutes quadratic behavior.
Fix by using the erase-remove idiom.
Write a generic function object QtFunctionObjects::IsNotValid, instead
of a lambda, because I am building up a collection of such function
objects in my tree, to be submitted any time now™, so this will reduce
the churn once that gets merged. Wrap it in the unnamed namespace to
avoid ODR violations in case this pattern repeats.
Replace an existing erase-remove idiom's isNotValid lambda function
with the new function object.
Change-Id: I4f0af04e7b201eb168e79beeda332eee3c6f33c3
Reviewed-by: Anton Kudryavtsev <antkudr@mail.ru>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Method parse() was passing down its default value argument to various
methods that expected an argument reflecting what's been parsed thus
far. Pass them what they expected ! Most are fairly insensitive to
the details at present, but the zone parsing I intend to add needs to
know the right date to compute its offset.
This makes it necessary to work out a sensible date and time to use,
from the fragments seen so far, using the default value to fill in
gaps but deviating from it wherever it conflicts with the actual
values seen.
Change-Id: I76750fcd92c1bdcad15c7dd4d8002cb90cec94ba
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Save unnecessary premultiply and unpremultiplication when converting
from an unpremultiplied format like ARGB32 to any opaque format like
RGB32 or RGB16.
Change-Id: I73f58200ff5c62fb07910e6f2b1c29b7e254d327
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
a6b34517e introduced this code as a workaround for stale QCocoaWindow
pointers during event delivery, but these days QCocoaWindow is a
QObject and guarded by QPointer. Disassembly of AppKit also shows that
the view is removed for us, so there's no need to do it explicitly,
especially as this causes two distinct event callbacks from AppKit
for what should be one atomic operation.
Task-number: QTBUG-42059
Change-Id: I212c894adf6aee51256ceff03c9821a995c2a63d
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
abde2a59c6 introduced this code on the basis that setContentView
does not remove the view from its superview, but disassembly of
AppKit shows that it does. Since commit 52767b8ee7 we also
ensure that the previous NSWindow has its contentView property
cleared, so this workaround is no longer needed.
Task-number: QTBUG-39628
Change-Id: I21e179263d006f3af1f8a55e9e2f7e8eeab2a632
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
The function never returns nullptr, so return the matrix by value.
Change-Id: I7c1eeb43b9693866049763565b575348ddd35548
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
... as suggested by a code comment.
This avoids the external bookkeeping using a QSet, and will allow, in
Qt 6, when we can port QWizardPage to the NVI idiom, maintaining the
bit as part of the initialize and cleanup functions directly.
Change-Id: I25a4dbf4c42f2c0286aa1a72ab9c59463d44c5df
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Apart from being more efficient to construct and test, for the
expected very small number of entries, the example code itself shows
that a sorted vector is much more useful than an unordered set.
Change-Id: Ic5e38df0176ac4be08eac6a89c2e1cabab2a9020
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
If you keep around QTextLines referring to the previous contents of
the layout, you can experience crashes when using these after the
layout has been cleared. Since QTextLine is a value type, this is
not very obvious in the API, so we should at least give a warning
in the docs.
Task-number: QTBUG-60804
Change-Id: I81b2b640eec5f62b0af1e878aadd9fa23654ec18
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Venugopal Shivashankar <Venugopal.Shivashankar@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Our OpenSSL backend allows missing symbols and in general protected
from possible related failures. Unfortunately, for OpenSSL 1.1 this
means not only missing symbols (removed functions), but new incompatible
opaque data-structures and our 1.0 code trying to use them and probably
accessing some data-members via macros - we end up in UB and crashes.
SSLeay, which returns a version number, was removed in 1.1. A failure
to resolve this symbol we consider as a version mismatch and we make
'supportsSsl' to return false.
Task-number: QTCREATORBUG-18137
Change-Id: I5cd270f9c61a729105149779ee7277816f9467d7
Reviewed-by: Richard J. Moore <rich@kde.org>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Having special code paths defeats the purpose of QT_STRICT_ITERATORS
and is ugly.
Change-Id: Ie0bdbf1a9639a0903bcd020f526629ab2e69883b
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
The events may come from QPA, as spontaneous events, or from other parts
of Qt, as non-spontaneous events. We should keep the state of the original
event.
Introduces QCoreApplication::forwardEvent() as a wrapper around the
opaquely named notifyInternal2. Ideally this would be the behavior of
sendEvent, with an enum argument to override the flag, but that ship
has sailed.
Change-Id: Ib0209f2b99744bd10590c63239ee7a97b60be4fd
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
QNSView instances check the _q_platform_MacDontOverrideCtrlLMB
window property along with the QT_MAC_DONT_OVERRIDE_CTRL_LMB
environment variable during creation.
Change-Id: Id6457fccdce2dff1fa83448dd2bc4d2757a87e9d
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
The test has been observed to be flaky on Windows. Introduce
QTRY_VERIFY_WITH_TIMEOUT for diagnostics.
Change-Id: I72abdd2e5544f8f35199876486ab15151f60e5f2
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Combining libEGL and libGLESv2 into QtANGLE cannot be the default
any longer, as the change was binary incompatible. We just had to
return to previous behavior temporarily to be able to fix
qtlocation.
This reverts commit 97ab65076e30d8cd0cbe6bdbdb51d1dc2c0ff7e7.
Task-number: QTBUG-60795
Change-Id: I3095cb5f9da30e2d873d9a186cfbc5aee3fb10b2
Reviewed-by: Jan Arve Sæther <jan-arve.saether@qt.io>
