In cases where the keyboard layout doesn't have a mapping for a given
event and modifier combination the result will be an empty string.
Fixes: QTBUG-90683
Pick-to: 6.0
Change-Id: Ice06241f0ae71a19cde041410818decc312bc630
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Key events have a wider range of possible values than the unicode range,
as they also include all the special keys as defined in Qt::Keys.
Instead of turning the argument to keyMapForKey into an integer,
we can remove the argument completely, as it was only used as a
fallback in the cases where UCKeyTranslate (or in the future,
charactersByApplyingModifiers:), failed to map the virtual key
and modifiers to a character. But in those cases we should not
fall back to the Qt key from the key event, as that doesn't
match what the keyboard layout defines. Most keyboard layouts
explicitly define the base key as the key for these "undefined"
mappings, but if the keyboard layout does not, we should not
produce any input in the application, to match what AppKit does
in this case.
Fixes: QTBUG-90315
Pick-to: 6.0
Change-Id: Ib9ffd9521049ee8e4b103597c1d34cbe3d23dbdf
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
The former API is the recommended way to present drawables -- avoiding
WindowServer submission throttling and potential issues with lock-ups
when window are not serviced, are being moved to other screens, or users
mistakenly call waitUntilCompleted.
The scheduled block captures the drawable and retains it until the block
is done executing.
Change-Id: I99dca2db3bcfbea67c686dd5d97ba5011598997e
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
The drawable returned from nextDrawable is autoreleased, which means
unless we explicitly retain it we have no guarantee that it will stay
alive long enough to be used again in endFrame(). For example, if the
user had an autorelease-pool in their custom render code, where the
call to beginPass() happens, the drawable would be released when that
pool went out of scope. The only reason we didn't hit this yet is due
to the default autorelease-pool being at the level of the runloop, so
the drawable happened to outlive the rendering of a single frame.
An advantage of manually managing the lifetime of the drawable is
also that we can release it immediately after being done with it,
instead of waiting until the autorelease-pool drains.
Change-Id: Ie6fd271adbe1121eb947bb7075142f1a6c81175d
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
We can avoid the overhead of the out-of-line function calls if the id is
already set.
As a drive-by, use relaxed loads in QMetaType::id (was: acquire).
Change-Id: I1d53dbf7e58eae80776ff36334ebf0642ba7842f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QSI(M) allow users to override their data() functions. This
means that we can't have their data() implemented in terms of
multiData(), or otherwise a subclass that overrides data() will
fall out of sync with its multiData() implementation. We must
keep multiData() implemented in terms of data() instead.
While at it, document QSI::multiData().
Pick-to: 6.0 6.1
Task-number: QTBUG-89423
Change-Id: If30a980d8449135b516745fec73170d53367b4b7
Reviewed-by: Christian Ehrlicher <ch.ehrlicher@gmx.de>
As described in 3279c8e7d7, we can't
depend on the widths of decoration to always be the same for
drawText() and equivalent drawStaticText/drawGlyphRun.
This is typically visible as an off-by-one when using the default
fonts on OpenSUSE.
Since this test was actually made to test positions of RTL glyphs,
we can just disable the decoration and simplify the text.
Fixes: QTBUG-89086
Change-Id: I139fe3e1c5e98d8b1d7e0e7c19645fd4717d1d95
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The various spatial-vector, line, point, region and margin types have
all long taken for granted that their co-ordinates are finite and, in
particular, not NaN. Make this expectation explicit in the
documentation. Added assertions where (chiefly) noexcept and (where
the assertion would be a simple qIsFinite() call) constexpr didn't
preclude doing so. Also assert against zero divisors that would lead
to non-finite results.
Make minor clean-ups to docs in the process. QMarginsF had several
methods whose declaration, definition and docs weren't consistent in
their parameter names.
Task-number: QTBUG-89010
Change-Id: I601a830959d13a73dcb17ce31a1202a1486c8f7f
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Use qHypot() instead of casting to and from double so much,
use qFuzzyCompare(_, 1.0f) rather than qFuzzyIsNull(_ - 1.0f).
Change-Id: I70f38fe2d9aefe1ceb15e1a370f181a7856911b7
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Use qHypot() to implement length(), avoid duplicating its code and use
its result more carefully, saving the need for casting to and from
double.
Subtracting a double from 1.0f still got a double, so the
qFuzzyIsNull() checks were using double's tolerance, where the use of
1.0f indicates the float tolerance would have been more apt. Also, use
qFuzzyCompare(_, 1.0f) instead of qFuzzyIsNull(_ - 1.0f).
In getEulerAngles(), scale co-ordinates by length before multiplying
(to ensure O(1) quantities) rather than scaling the products by the
squared length (possibly after {ov,und}erflowing).
Change-Id: Id8792d6eb047ee9567a9bbb246657b0217b0849f
Reviewed-by: Andreas Buhr <andreas.buhr@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Be more careful about reporting a new selection to Qt. The code for
handling IM selection events in QQuickTextArea is quite complex
and need to take pre-edit text into account. The latter means
that when the pre-edit text changes, as a result of the user composing
a word, the width of the pre-edit text will also change (and as
such, the cursor rectangle). But the cursor position itself stays
the same. And for this reason, it emits cursorRectChanged more often
than strictly needed. But rather than trying to clean that up, we
do some extra checking before we send the IM event from QPA in
the first place.
Pick-to: 6.0 6.1 5.15
Fixes: QTBUG-63018
Change-Id: I689d989c3fe5d61ef2b1dbee7a70418b7790bce9
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
Previously it only returned checked or unchecked for a tri-state
checkbox.
Fixes: QTBUG-84616
Pick-to: 5.15 6.0 6.1
Change-Id: Ife72098e35f8295fd389bda232de5478ffa7e87f
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
On iOS, if you select some text, the edit menu will show on
top of it. And if you tap on the screen (or inside the menu) it
will hide. But if you type on the input panel, it will stay
open. This is wrong.
This patch will keep better track of whether or not the edit
menu was opened by us, and if it was, ensure we close it also
if the cursor moves by typing on the input panel.
Pick-to: 6.0 6.1 5.15
Fixes: QTBUG-90860
Change-Id: I0a51382030560182e7925c8b694b42e50943514e
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
This simplifies the Android platform plugin code, since
it no longer has to make QHighDpi scaling calls.
QPlatformInputContext returns geometry in the native
window coordinates. These can be used directly, or in
some cases be mapped further to native global (screen)
coordinates. Note that this mapping now happens after
High-dpi scaling, so we must use QPlatformWindow::mapToGlobal()
instead of QWindow::mapToGlobal()
The inputItemTransform() is applied by QPlatformInputContext
when needed, these calls can also be removed.
Change-Id: I179d44d148dded43683df5e97ad506940894c423
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
The benchmark used to crash because QMetaType::typeName would return an
empty string, which is not a legal value for newRow.
Change-Id: I9e6c6c1cf153943bfa21181cd2cca596a7943ea0
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Amend ac970d48fd and use
QEXPECT_FAILURE for systemTimeZone test on 32bit systems.
Task-number: QTBUG-89889
Change-Id: I0eed35df871c69a20bcd7c544fc0e9a48dd8db7b
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Clang warning: 'isSequential' overrides a member function but is not
marked 'override' [-Winconsistent-missing-override]
Change-Id: I1a7c5516d2656469eab556e7f9d310192510b99b
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Pick-to: 6.1 6.0 5.15
Change-Id: I9d3f7ca2022b17559dd5bdd5c6bb8c85e89c3f64
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
And remove direct conversion to take advantage of both the SIMD, but
also threading on all platforms.
Change-Id: I782ae7c80e82b694e23abee28d69cecd0d28c32f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Being able to iterate h = h.add(next) over a sequence of numbers makes
it possible to implement arbitrarily-many parameter functions that
need hypot, e.g. the implementation of V4's Math.hypot().
This amends commit b08368d99f.
Change-Id: I76bb65bd5f318e915d42fa4c935129850e258132
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This simplifies the include path addition we need when compiling and
eliminates the need to rewrite all the #include lines on update.
Change-Id: I14e301d277f1eb49286721b012559f0ec84ea520
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
...in the hope that they may work. If not, that's it, but at least we
tried.
Task-number: QTBUG-76970
Change-Id: I134c5cc4cfb5ad1e6f9edbfcf506df20022e127a
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
When QWidgetWindow handles a QTouchEvent, it will call
updateTouchPointsForWidget() for each widget it tries
to deliver the touch event to. This will make sure that
the point's position() is updated to be local to the
widget being processed.
The problem is that we never reset this overwriting of
local positions in the event after we're done. So if we
later try to synthesize a mouse event from it, the local
position in the fake mouse event will be based on the local
position inside a random widget, and not the original local
position sent from QPA.
Rather than trying to store all the original local
positions inside the event before going through this delivery
logic, and reset it afterwards, we base the local position of
the synthesized mouse event on the (unmodified) global
position instead.
Fixes: QTBUG-90033
Change-Id: I1588351482de7cce9c06d102db3686ea8dd0c118
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
QPointingDevice requires using the constructor to set its settings
in Qt 6.
Replace + with | operator and return QKeyCombination instead of int.
Change-Id: Id3da469cc13b34ec7b55afa751dbc04601880df6
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
- Replace vague symbols with descriptive words.
- Use qScopeGuard() to ensure that removeShortcut prints the number of
removed shortcuts for all execution paths.
Change-Id: I4d36366445756b41b1dc40bbb1e987dce669a723
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
This was added by 9ff76c27b9 on
the basis that it signifies a shaping error and would later assert
or crash.
But the line is easily reachable by user code. If Harfbuzz returns
0 glyphs, it just means it is unable to shape the string, for instance
if the input string only contains default ignorables (like a ZWJ)
and does not have any appropriate glyph to use for replacement.
Qt expects there to always be at least one glyph in the output
(num_glyphs == 0 is used to indicate shaping is not yet done), so
to avoid asserts later on, we simply populate the output with a
single 0 token, which is a required entry in the font that is
reserved for representing unrepresentable characters.
This also adds a test and therefore a zero-width joiner to the test
font to reproduce the issue.
[ChangeLog][QtGui][Text] Fixed a possible crash with certain fonts
when shaping strings consisting only of control characters.
Fixes: QTBUG-89155
Pick-to: 5.15
Pick-to: 6.0
Change-Id: Ia0dd6a04844c9be90dcab6c464bebe339a3dab11
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Setters of QObjectCompatProperty need to remove the binding, but
only if it is not in the wrapper. This patch adds this method to
the public interface.
Change-Id: Ibc96418fe15b0aeeb2059411ea910ba1f733c255
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
A static build of Qt with QT_BUILD_TESTS=ON would fail with errors like
this:
FAILED: tests/auto/corelib/io/qprocess/testProcessEcho2/testProcessEcho2
undefined reference to `qRegisterResourceData(...)'
We create object libraries for resources. In this case, for the
mimetypes of QtCore.
For static builds we added the object files to INTERFACE_LINK_LIBRARIES
of the Qt module the resource belongs to:
target_link_libraries(Core INTERFACE
"$<TARGET_OBJECTS:$<TARGET_NAME:${resource_target}>>")
That is, however, a problem for projects that link only against QtCore,
like testProcessEcho2 above. On the linker command line Qt6Core.a is
before qrc_mimetypes.cpp.o and that doesn't work, because the latter
needs symbols of the former.
To have the resource object files before any other libraries on the
linker's command line, we now treat them as source files.
Pick-to: 6.0
Fixes: QTBUG-90582
Change-Id: Ia2786beb5f7c2eba194e8f00ae8785572df1cf6e
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
And remove the direct conversion so we can get both the SIMD
optimization and threading applied.
Change-Id: Id032ea91cc40c1cbf1c8a1da0386de35aa36cfb5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It can do unaligned access directly as well.
Change-Id: I4046cf69e2ad8eb628b8b53bfa521800bb35dc6d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The initialization was duplicated across the different modes. We now do
the setup once, in a shared showPanel:withParent function.
This also simplifies and removes the need to store the return code.
Change-Id: I3c4da48cfef92bcc59c76cffa15b40150de1a9e1
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Relying on Objective-C's no-op behavior when sending messages to nil was
nifty, but a bit confusing when trying to track the ownership model of
the class.
It's now explicit at the call sites what's going on (a cast).
The canSelectHiddenExtension property is valid both save and open
panels, but AppKit will only show it for save panels.
Change-Id: I8e12d629639e2179d155b2ecda1bb2dab2a5757d
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
For Windows. Based on the code I wrote for QNetworkStatusMonitor.
It also renames the netlistmgr feature, avoiding the abbreviation.
Locally my MinGW fails the networklistmanager feature test so it may
not be supported on MinGW, likely leaving it without a backend at all.
Change-Id: I13bbe4127edc2a9c0bb91602c95f1cb206a85a69
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Change the setSelectionOnFocusObject() to take geometry
arguments in native pixels, in order to be consistent with
other QPlatformInputContext API.
Calling code which pass non-native geometry must be updated
(in practice only the Android platform plugin).
Change-Id: I1c61b8cf583afe607567d6056ca7ff23cc3de3f3
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This field has been defined since API level 11, so no reason why we
shouldn't just use that instead.
Pick-to: 6.0
Change-Id: I8e3ffde3680e41f26d776e1f0474f632186a6b27
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
If ImhHiddenText is set then that should take precedence over the
other input method hints. Also, certain combinations aren't really
possible. E.g., ImhEmailCharactersOnly and ImhHiddenText doesn't have
its own specific variation on Android.
ImhSensitiveData and ImhNoPredictiveText are also likely to not work
as TYPE_TEXT_FLAG_NO_SUGGESTIONS is normally ignored by Android
keyboards. A common workaround is to use the visible password variation
but since this will force the layout to use latin characters it's not
something we can use as an universal workaround, so users will need to
manually enable it through the environment variable
QT_ANDROID_ENABLE_WORKAROUND_TO_DISABLE_PREDICTIVE_TEXT
Fixes: QTBUG-85787
Pick-to: 6.0 5.15
Change-Id: I0ff591ec9acf8dd556c7987e6d997cff3ddfe38e
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
registerNativeMethods() should be using the local jclass from
findClass() instead of the global reference.
Task-number: QTBUG-89633
Change-Id: I469a9a1ecff95ab9948421baa5c88735ba7ad776
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
Always try segments from 64k to 128k pixels, this matches
qimagescale logic, and gives a one percent speedup locally.
Change-Id: I3ef468eac9dca4b84f04850e970f3d15a4f16255
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
This enables support for QT_SCALE_FACTOR on iOS.
Fixes: QTBUG-74978
Change-Id: Ibcf0741c178e44802065e472e096a5f4c7d6f3cf
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Passing nullptr to QHighDpi::toNativePixels() is seldom
correct, and will work only on the main screen.
Instead, call frameMargins() on the platform window
(instead of on the QWindow) and get the margins directly
in native pixels.
Pick-to: 5.15
Fixes: QTBUG-90716
Change-Id: Id7b31ec3246010c367781b64ed832c589bbaeb0d
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Also remove tests/tests.pro that would be empty without the benchmarks.
Change-Id: Iaf92a729d1286b3e0c03bf9f877b59e1d83708e6
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Some of the listed source files are already included and must not be
compiled.
Pick-to: 6.0
Fixes: QTBUG-90777
Change-Id: I80b4ca6a50b882445aca513d800ef8375070d786
Reviewed-by: Kai Koehne <kai.koehne@qt.io>