The race goes like this:
1) We destruct QFseventsFileSystemWatcher, which calls FSEventStreamStop
and FSEventStreamInvalidate/FSEventStreamRelease
2) The FSEvent* calls will happen on the same thread as the destructor
is being called on, which will be different to the thread that the
FSEvent* events are popping out on.
3) So, there could be a case where we are in the middle of processing an
event, but the QFseventsFileSystemWatcher has already died.
The fix is to dispatch the stop/invalidate/release on the queue
associated with the stream.
Patch by Matt Galloway!
Fixes: QTBUG-85594
Pick-to: 5.15 5.12
Change-Id: Ie168bbe91e55c5559632b37bc008e11597e4fdaf
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Avoid parsing over the buffer limit, or interpreting non-hex
as hex.
This still leaves parsing of lines longer than 300 chars
unreliable
Change-Id: I1c57a7e530c4380f6f9040b2ec729ccd7dc7a5fb
Pick-to: 5.15 5.12
Reviewed-by: Robert Loehning <robert.loehning@qt.io>
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
When the compile time metatype support for methods was introduced, we
needed to allow incomplete types to avoid breaking a large amount of
code. However, this mistakenly enabled using incomplete types for
properties, too. In contrast to methods, properties lack the fallback
code to retrieve the metatype at runtime. Thus, this commit restores the
completeness requirement for properties again. This is done by always
calling QMetaTypeForType for properties in qTryMetaTypeForType.
Amends fa987d4441
Change-Id: I5f66ff289631c056eecebe40926bf321d283eea7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
As defined by https://llg.cubic.org/docs/junit/
Change-Id: Ic7683f3d49c529674f8467d591528d4a65d3add8
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
The attributes are, like the elements, maintained in reverse
order in the underlying QTestCoreList, so we need to iterate
them backwards when printing out the resulting XML to reflect
the order they were added.
This results in e.g.:
<testcase name="passingBenchmark" result="pass">
Instead of:
<testcase result="pass" name="passingBenchmark">
Change-Id: Ic2eeab8de05ffedd0c41977358d5b40ff77878b1
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Entering Chinese in some fonts in an English environment causes
the cursor drawing coordinates to exceed the updated coordinates,
leaving behind stale pixels at the top.
Keep the refresh and draw area calculation methods the same when
rendering the contents in QLineEdit::paintEvent, and when
calculating the update area in
QLineEditPrivate::adjustedControlRect.
Fixes: QTBUG-85569
Pick-to: 5.15
Change-Id: I978cb56f24f961086b1271e56d07ad1ced16f8ff
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Change added dep in qt_add_cmake_library():
PlarformModuleInternal -> PlatformCommonInternal.
Change-Id: I303a6a520f0dda7260b12cb21673b8f1a23d536b
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
QAbstractSocket::flush stopped being virtual and, instead, QAbstractSocketPrivate::flush
is now virtual. So the code in QSslSocket is essentially a duplicate.
Change-Id: Iebb33c8490268270b90a662ab64b1ae47c4cbac7
Reviewed-by: Alex Trotsenko <alex1973tr@gmail.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
It doesn't understand the arguments we usually pass to testlib.
Change-Id: Iea83d1d8c31a510b2bc442cbfc810eac631322e7
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Deprecated since 5.0. Renamed a function in a manual test that no
longer needs to say it's Qt5-specific.
Change-Id: I6f2159c702f389d378a0e4d86bd4fe633298b100
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Since 5.15 - toTime() with a calendar in the argument list
The calendar was never used.
Change-Id: I37897d47b42e69d1a8312e6389d6349febb889ae
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The openglextensions module is dicontinued and removed in Qt 6.
Change-Id: I0b98f9b6575026828c188786aeb5c1a62de1f8a3
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The shift operator is undefined for negative values.
Fixes: oss-fuzz-23735
Pick-to: 5.15 5.12
Change-Id: Ib4bad0bf2135f9ddc95f6d7de64d935a02b4791e
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Instead of deferring the creation of the test suite until logging
stops, we create it up front, matching the logic of adding test
elements on test function enter.
Change-Id: I78b1ccdfde5493d78ef478d4b3c45d5a49358979
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Simplifies debugging of issues where the watchdog thread is involved.
Change-Id: I4862167bca4a942c7d4319a9374f1f83f292d831
Reviewed-by: Simon Hausmann <hausmann@gmail.com>
We don't want the signal dumper to pick up signals that our own
test machinery produces, such as the ones emitted from the watchdog
thread startup and shutdown. This would otherwise produce:
tst_Signaldumper::initTestCase() Signal: QThread(7fc969e0d870) started ()
At startup, and at shutdown even more confusingly:
tst_Signaldumper::UnknownTestFunc() Signal: QThread(7fc969e0d870) finished ()
Change-Id: I9e81fa168eaa92551d38d5576973bbf95ac23364
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
The logic was changed in eb52d78e90 to only produce a single
result for all the benchmark iterations, but only the plain text
output was rebaselined. This was not caught by the test since we
only run this test with the plain test logger. Regenerating the
results with the python script however runs it with all loggers,
so let's fix the baseline.
Change-Id: I9ebe8a59ac3ab36694a739f98dbfffbae9a08552
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Exercise the touch settings of the native interface.
Task-number: QTBUG-41433
Task-number: QTBUG-48849
Task-number: QTBUG-83252
Change-Id: I5ae95a79c00b55236dbbed9d8549f4fdf5b10b8e
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Change TouchWindowTouchType(s) to be (global) property of
QGuiApplication's native Windows interface since it does not make
sense to set it per window.
It appears the previous code setting the types per Window has never
worked since registerTouchWindow() bailed out due to the checks for
the flags TouchRegistered and IsTouchWindow() (setting in
HCBT_CREATEWND). In addition, registering windows for touch after
plugging in a device would not observe the setting.
Move the checks around to make this work.
Task-number: QTBUG-41433
Task-number: QTBUG-48849
Task-number: QTBUG-83252
Change-Id: I4306fdf13208f6eef22655875f3bd1769270e617
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
All Mac Pros launched since 2019 have AVX512 support (CLX processor) but
the Darwin kernel does not enable the state bits in the XCR0 register
for all processes by default. Instead, it traps the #UD exception and
decodes the instruction to find out if it was an AVX512 instruction. If
so, it adds memory to the thread's context switch space. See [1].
Good solution, but forces us to have OS-specific code to determine if
the OS really supports AVX512. No other OS appears to require this. For
future features (namely, Intel Advanced Matrix Extensions), there's an
official way of implementing this, via the eXtended Feature Disable
register.
Qt has no AVX512 code yet, so this is just being future-proof. Apple has
yet to announce when or if their Mac Pros will switch from Intel Xeon to
ARM.
[1] https://github.com/apple/darwin-xnu/blob/xnu-4903.221.2/osfmk/i386/fpu.c#L174
Change-Id: Ieffc3453b88c4517a1dbfffd162175ead343e622
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This is no longer a problem, and this condition should be
handled by QCocoaGLContext in any case.
Change-Id: Iaac9d1a8962b27bf6f0394b8c1ea3e46dc28e29e
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
QPlatformWindow initializes its view of the geometry based on the
QWindow geometry during construction. If the initial geometry we
then compute is the same, we would end up exiting early from
QCocoaWindow::setGeometry(), because we compared the new geometry
against the QPlatformWindow::geometry(), and the geometry would
never be reflected in the NSView.
Due to other setGeometry calls this was in most cases masked, but
could in theory happen, and is preventing us from cleaning up other
parts of the code.
The call to QWindow::setGeometry() after setting the initial geometry
is also broken, as the initial geometry is available through the
platform window and QWindow::geometry() already, so setting it again
serves nothing except disabling d->positionAutomatic, which is not
correct.
Change-Id: I0db3cfe7d7c3d14070beee6ae3ea3dfd49da9e05
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
To unlock work while waiting for QISP. QESDP is lacking
a few crucial APIs (isDetached() / isShared() / reset(ptr) / deep
const), so we can't go there right away.
Change-Id: I647eb0db4ea800488a323f3b64661e848998168d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Commit a2bdda8e3b added an unconditional hide of child NSViews,
which was found to be too strict, and worked around in a199a87ad
by undoing the hide during setParent(). The unconditional hide
was then changed in 89842b97d7 to use the visibility state of
the window, which means the workaround in a199a87ad is no longer
needed.
Change-Id: If0df2de65693e03c5fb53a906b1399accab3fcc4
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Remove their associated enums, too.
All deprecated since 5.9.
Change-Id: Ieefba1a08c221ea7b62500f219baea8a3ccfb7db
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
A QWindow should only become Active when it's inside an
NSWindow that is Key. If the NSWindow is not key, we need
to wait for it to be so, and handle window activation
from QCocoaWindow::windowDidBecomeKey() instead. Otherwise
Qt will report a QWindow as Active when, in reality, it
is not.
Change-Id: Ib7e63b374f26af527a668c7f7d863c4168a4446d
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
After the QVector->QList rename it made no sense.
Change-Id: I4a422f48b1f5d42c1c4d402ea947a0f4098172b7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The -arch_blank argument to lipo is no longer supported.
Pick-to: 5.15
Pick-to: 5.12
Change-Id: I47efcb3633f23b7a18d66211bc16d5c9dbc067c8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
For those who are providing their own launch images for their iOS
projects then QMAKE_IOS_LAUNCH_SCREEN can be set to point to the
location where the launch image to be used over the default.
[ChangeLog][Platform Specific Changes][iOS] Added support for
specifying a launch image to be used for an iOS project. This can be
achieved by using QMAKE_IOS_LAUNCH_SCREEN.
Change-Id: Ibb236655b282132ab5eee747986a93abb9802200
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
It is unnecessary to create a QList container just for comparison.
Split out helpers for comparing sequence sizes and sequences from
qCompare(QList) and add a template for an array with a non-type template
parameter for the size.
One can then write something like:
const int expected[] = {10, 12,...};
QCOMPARE(QFontDatabase.pointSizes(...), expected)
Unfortunately, any commas in such an array will be misread by macro expansion
as macro argument separators, so any expected array with more than one entry
needs an extra macro expanding __VA_ARGS__.
Change-Id: Ie7c8dc20bf669bbb25f6d7f8562455f8d03968c8
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
On Windows, and possibly other platforms, a touchpad can send a mouse
button press followed by an unexpected mouse move event to the same
coordinates, before sending a mouse button release, which may confuse
applications. Before the enhanced mouse event processing was added, the
code in QGuiApplication was responsible for deducing the mouse event
type and other info, and in the process performed a checking that
discarded events that did not change state. The enhanced mouse
processing code lacked this checking. This change adds an equivalent
checking to the enhanced mouse event processing.
Fixes: QTBUG-85431
Pick-to: 5.15
Change-Id: Ie3e2ae8cbf9870d465dfd2c8808942dd6fc647d2
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
The comment claims it is needed for "the deprecated
QHash::iterator::operator--()", which has been removed already.
Change-Id: I90f5abdcd14e6cac4ed7144e96eb26bf34477391
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Implement the static parts of QWindowsWindowFunctions.
Move some things around and make them static in tablet support
to make this easier.
Task-number: QTBUG-83252
Change-Id: I24c3b0b3c6548eb96c4f3285a4e46385a1e79ef5
Reviewed-by: André de la Rocha <andre.rocha@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
We don't want to print the QString as represented by the debug
operator, but instead want to expand line feeds and other character
codes as normal.
Change-Id: I7261d8f94e7b4382733f06eb22f9a740a5c0488f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
