When doing a shift-select while moving the mouse then the start point
should be based on the start of the current selection and not the
pressed position. If there is no current selection start index, then
we can safely depend on pressed position as this will be the previous
index pressed on.
This resolves an issue introduced by
e02293a76d when fixing QTBUG-78797
Fixes: QTBUG-81542
Change-Id: Ia66c42b220452fdcbc8cfccc05dbc8a3911c3f5e
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
QByteArray doesn't like it.
Apply the same protection to QString, which we know uses the same
backend but uses elements twice as big. That means it can contain
slightly more than half as many elements, but exact half will suffice
for our needs.
Change-Id: Iaa63461109844e978376fffd15f9d4c7a9137856
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
If the map or array is known to be empty, we don't need to allocate a
QCborContainerPrivate.
Change-Id: Ief61acdfbe4d4b5ba1f0fffd15fe212b6a6e77c3
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
A simple 16k file can produce deep enough recursion in Qt to cause stack
overflow. So prevent that.
I tested 4096 recursions just fine on my Linux system (8 MB stack), but
decided 1024 was sufficient, as this code will also be run on embedded
systems that could have smaller stacks.
[ChangeLog][QtCore][QCborValue] fromCbor() now limits decoding to at
most 1024 nested maps, arrays, and tags to prevent stack overflows. This
should be sufficient for most uses of CBOR. An API to limit further or
to relax the limit will be provided in 5.15. Meanwhile, if decoding more
is required, QCborStreamReader can be used (note that each level of map
and array allocates memory).
Change-Id: Iaa63461109844e978376fffd15fa0fbefbf607a2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The next commit will need to do so from outside QCborContainerPrivate,
where QCborStreamReader::d can't be accessed (private).
Change-Id: Iaa63461109844e978376fffd15fa0f6f04081bf2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Flushing sublayers via QImage copies of the root IOSurface was causing
performance regressions due to the constant allocations of new images
each frame.
We now re-use the QCALayerBackingStore implementation for sublayers,
which gives a dynamic swap-chain.
We're still paying the CPU cost of the copy from the root backingstore
to the layered backingstores, as well as the memory cost, but at least
improves the situation.
We do not try to be smart and paint directly into the sublayers,
as that would leave the root backingstore stale, potentially causing
glitches when views are repositioned. Investigating this is left
for future work.
Fixes: QTBUG-82986
Change-Id: I758a3d8e1e40e2ed4fe6bc590a4a5a988d87a3a7
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
(cherry picked from commit ce2d68ebe1aefeae78ff2fd8ec5ff7e20790ef69)
This allows to pass and receive possible drop actions from other
processes, including GTK applications.
Fixes: QTBUG-75744
Change-Id: I944edc6fa00f8801a25912e70eb104a647a9fc0e
Reviewed-by: JiDe Zhang <zccrs@live.com>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Constantly re-reading the timezone information only to be told the exact
same thing is wildly expensive, which can hurt in operations that cause
a lot of QTimeZone creation, for example, V4's DateObject - which
creates them a lot (in DaylightSavingTA).
This performance problem was identified when I noticed that a
QDateTime binding updated once per frame was causing >100% CPU usage
(on a desktop!) thanks to a QtQuickControls 1 Calendar (which has a
number of bindings to the date's properties like getMonth() and so
on).
The newly added tst_QTimeZone::systemTimeZone benchmark gets a ~90%
decrease in instruction count:
--- before
+++ after
PASS : tst_QTimeZone::systemTimeZone()
RESULT : tst_QTimeZone::systemTimeZone():
- 0.024 msecs per iteration (total: 51, iterations: 2048)
+ 0.0036 msecs per iteration (total: 59, iterations: 16384)
Also impacted (over in QDateTime) is
tst_QDateTime::setMSecsSinceEpochTz(). The results here are - on the
surface - less impressive (~0.17% drop), however, it isn't even
creating QTimeZone on a hot path to begin with, so a large drop would
have been a surprise.
Added several further benchmarks to cover non-system zones and
traverse transitions.
Done-With: Edward Welbourne <edward.welbourne@qt.io>
Task-number: QTBUG-75585
Change-Id: I044a84fc2d3a2dc965f63cd3a3299fc509750bf7
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
No sanitizer is needed, just looking at the code is enough.
It was wrong.
Change-Id: I9df417c137d6b3361c3161865e099a8be40860de
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The paths in the build properties require forward slashes
apparently. On Windows, we would default to native backslashes
and they would be stripped from the path. Converting to forward
slashes fixes the problem.
Issue was introduced by dd04fb639b,
since before that, the NDK path was retrieved from the environment.
Fixes: QTBUG-82944
Change-Id: I6c51113efcf671461a5871991b3225a52b95266c
Reviewed-by: BogDan Vatra <bogdan@kdab.com>
Since the test was refactored and QCursor::setPosition() replaced
with QTest::mouseMove(), the test is completely crippled on macOS,
since it relies on the parts in widget's code, ifdefed with condition
!Q_OS_OSX and commented as "Cocoa tracks popups". Yes it does,
but not for "fake" events generated by QTest. The original test
was introduced when fixing different problems on non-Apple platform(s)
anyway. Let's make QSKIP message saying the truth.
Task-number: QTBUG-63031
Change-Id: If54f195ccc0d4409cc2e7f4ae0b0fbf43989b286
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Due to the changes in 5.14.1 this code now actually seems to be hit here,
throwing NoSuchMethodError exceptions all over the place and breaking
date/time handling quite spectacularly.
Change-Id: I9bee3de39ec98f86d7944b94e89119505f62dc6c
Reviewed-by: BogDan Vatra <bogdan@kdab.com>
This ammends change baed8534bc, which
might have introduced a hard to reproduce segmentation fault when the
screen number is out of bounds, or when the QScreen object doesn't
return a valid pointer for QScreen::handle.
As the issue doesn't reliably reproduce, this is a speculative fix that
adds bounds and nullptr checking.
Change-Id: I0cec0a344e80159ee1723d840f207267a608cef4
Fixes: QTBUG-82807
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
drawTree() does
QPoint hoverPos = d->viewport->mapFromGlobal(QCursor::pos());
d->hoverBranch = d->itemDecorationAt(hoverPos);
and itemDecorationAt does
const QModelIndex index = q->indexAt(pos);
which might very well be an invalid index.
Change-Id: I7db98871543bd7e1c57fcc475d2646757bf2bb42
Reviewed-by: Christian Ehrlicher <ch.ehrlicher@gmx.de>
UWP expects these functions to be opened via the XAML thread, so we
ensure this is done by running those functions on that thread.
Change-Id: I57ae3a7d9b45d0b1a00ac23b103386bd34b65c6d
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
QWindowsWindow::updateFullFrameMargins() which is called
from the screen changed handling does not take native menus
into account.
Since the size of the menu is not known when using
EnableNonClientDpiScaling(), obtaining the correct frame
size requires triggering a WM_NCCALCSIZE message. Extract
the helper forceNcCalcSize() from QWindowsMenu and
use that from updateFullFrameMargins() in case menus are present.
Amends d2fd9b1b98.
Fixes: QTBUG-82580
Change-Id: I306f1faf84e26c88608cb22ffd42eccc848905c3
Reviewed-by: André de la Rocha <andre.rocha@qt.io>
On a high-dpi screen and AA_UseHighDpiPixmaps set, QIcon will ask its
engine for a scaled-up pixmap. When the icon has been created from a
theme, the engine is a QIconLoaderEngine. For a SVG icon, that engine
would recursively use QIcon to load the scaled-up pixmap, leading to
double scale-up. Fix by bypassing the QIcon API in the SVG case,
loading the SVG icon directly from the SVG icon engine.
Done-with: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Fixes: QTBUG-73587
Fixes: QTBUG-75039
Change-Id: I7fba02b6454decb5fcbca9c5a092e75954261dfd
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Keep using the old/deprecated behavior for the Qt 5.14
series.
Task-number: QTBUG-74601
Change-Id: Icee99803f300dfa0116a4de75f9fb26d1010625d
Reviewed-by: Lorn Potter <lorn.potter@gmail.com>
This has been flaky on OpenSuSE; if the stored geom.topLeft() is 0,0
it apparently means the window manager (probably kwin) didn't get around
to decorating and repositioning the dialog by the time
qWaitForWindowExposed() returns. Because we check later to see whether
it moved, we need to be certain of its initial position.
Waiting for the extra "fleece" widget to be shown was based on the
theory that by the time the X server has processed messages related
to that new window, the WM should be done processing the consequences
of the resized dialog window. But there's no such guarantee, so let's
try removing that. On the other hand, removing the delay does open
us up to miss a regression (maybe the dialog gets moved after we have
checked that it didn't move).
Rename because we don't name autotests after bugs.
Amends 26ddb586ac
Task-number: QTBUG-32473
Change-Id: I6bbfe2b4baaee389db0d4112f0fec3b7cb9da554
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Liang Qi <liang.qi@qt.io>
QVLA *does* have iterators and *can* be used with foreach
(... I didn't say it should). Move its description together
with the other containers.
Change-Id: Ib60d1f7b3dc0e8c7004991bd4fdff95b3f23af60
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The changed buffer usage priority that was introduced in our ANGLE
update caused severe performance regressions for Qt applications.
Fixes: QTBUG-73835
Change-Id: I49839bb272cdeec0027264f2751b88bc149665ad
Reviewed-by: André de la Rocha <andre.rocha@qt.io>
Reviewed-by: Miguel Costa <miguel.costa@qt.io>
We are, arguably, not testing QProcess and its ability to start or finish,
we test QUdpSocket. If, for some reason (as we discovered on some
specific machines recently) the process does not start or does not produce
any output (canReadLine), we QSKIP instead of failing. Also, all those
QCOMPARE will bypass the part there we stop processes - so must be
RAII-protected.
Fixes: QTBUG-82717
Change-Id: Idfb0d4a483d753f336b3827875eeaf51c79270e2
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
If XLIB_SKIP_ARGB_VISUALS is set, we might fail to find visuals for
supported configurations.
Task-number: QTBUG-81904
Change-Id: Ib0d26faabe430925881b7f2acfc5361df3af416b
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
The effect of the soft-hyphen needs to be updated once the final the
break point has been found.
This change cleans the logic by using two variables keeping track of
soft-hyphen at current evaluated position and at last confirmed break
point. Also adds tests for supression of soft-hyphens in the tight
WrapAnywhere case.
Fixes: QTBUG-35940
Fixes: QTBUG-44257
Change-Id: I7a89a8ef991b87691879bb7ce40cec4a3605fdd5
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
QSFPM::reset() is deprecated and begin/endResetModel() should be used.
Therefore adjust the documentation to reflect this.
Fixes: QTBUG-82470
Change-Id: I786b3f25e5674d97d0ef6a0c91342973d5e952e9
Reviewed-by: Sze Howe Koh <szehowe.koh@gmail.com>
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
The std::is_pod trait is deprecated in C++20; is_trivial
and is_standard_layout exist since C++11.
Change-Id: I4b901d8edf1a55001764445aee9c338d3dc23b21
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Amends 2f366a63b2.
Task-number: QTBUG-82682
Task-number: QTBUG-81342
Change-Id: I30f465bf432e27828db460f6dbbb59eee0cca8f2
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Reviewed-by: André de la Rocha <andre.rocha@qt.io>
printsupport features are only available if the module is available for
the configuration. Every printer feature check has to be coupled with a
check for the module itself.
Change-Id: Id2ca84e71d5d74463d0ff35e8b18b252a779a883
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
This was taken from abc473fb8fb99900 in SQLite, ref:
https://www.sqlite.org/cgi/src/info/abc473fb8fb99900
Fixes: QTBUG-82533
Change-Id: I9840e29f19a0b861229987f5b59d8585ba2e55dc
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Ammends 4ac872639e. Change return type for getPrefix to eliminate
toLatin1/fromLocal8bit conversion of initial value for relocatable
prefix to preserve not-latin1 characters and return valid prefix path.
Fixes: QTBUG-81462
Change-Id: I15cfa49e9e440e257b04dd31803cd1478f3b07f5
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
The test was fixed and metrics show no flaky failures anymore.
Task-number: QTBUG-58713
Change-Id: I50c0844db099f45bb5b7ca51a510bf0318554c44
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
And make pHnd atomic.
The majority of the variables is updated in QLibraryPrivate::load_sys
and updatePluginState(), which get the mutex protection.
QLibraryPrivate::unload_sys() doesn't need a mutex protection because we
have the refcounting.
[ChangeLog][QtCore][QLibrary & QPluginLoader] Fixed a number of race
conditions caused by having two QLibrary objects pointing to the same
library being operated in different threads.
Fixes: QTBUG-39642
Change-Id: I46bf1f65e8db46afbde5fffd15e1a5b3f5e74ea4
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
resolve() is technically thread-safe if the library has been loadaed. We
don't promise that, but it's there. More importantly, because
QLibraryPrivate is shared among QPluginLoader and QLibrary that point to
the same file, we can't thread-safely set the error string.
[ChangeLog][Important Behavior Changes] QLibrary::resolve() will no
longer set or clear the error string based on the success of finding the
symbol. The error string will reflect the result of loading the library.
Change-Id: I46bf1f65e8db46afbde5fffd15e1a4f4c2713c17
Reviewed-by: David Faure <david.faure@kdab.com>
There was a race condition in accessing the cached instance factory
member, so rework loadPlugin() to return the cached or newly discovered
instance, with proper, atomic caching. Because I had to change that, I
took the opportunity to fix the QFactoryLoader code that calls
loadPlugin().
Note that QLibraryPrivate::loadPlugin() returns non-nullptr now if the
instance is known, which means the last return in QPluginLoader::load()
will convert to true, not false, if the instance got cached between the
earlier check and the call to loadPlugin(). That's probably what was
intended.
Task-number: QTBUG-39642
Change-Id: I46bf1f65e8db46afbde5fffd15e1a42d2b6cbf2c
Reviewed-by: David Faure <david.faure@kdab.com>
It has been reported that 7447e2b337
needs amending to build on MSVC 2019, although it was OK in CI.
Change-Id: Id22c2a3608529abebd66c0e8f401bc6f26f45e18
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
d7cb21ac08 change the way a double is
converted which resulted in less precision.
Fix it by explictily setting the precision in QString::setNum()
Task-number: QTBUG-80068
Change-Id: I1fd9d00837155ceb707e84bfeb9deff03b5ab57e
Reviewed-by: André Hartmann <aha_1980@gmx.de>
Reviewed-by: Konstantin Shegunov <kshegunov@gmail.com>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
When _NET_WM_STATE_HIDDEN is not contains in the _NET_WM_STATE
window property, the window should not be considered to be minimized
According to
https://specifications.freedesktop.org/wm-spec/1.3/ar01s05.html
_NET_WM_STATE_HIDDEN should be set by the Window Manager to indicate
that a window would not be visible on the screen if its desktop/viewport
were active and its coordinates were within the screen bounds. The
canonical example is that minimized windows should be in the
_NET_WM_STATE_HIDDEN state. Pagers and similar applications should use
_NET_WM_STATE_HIDDEN instead of WM_STATE to decide whether to display a
window in miniature representations of the windows on a desktop.
For mutter/GNOME Shell, without _NET_WM_STATE_HIDDEN, window manager
will not reply XCB_ICCCM_WM_STATE_ICONIC settings in WM_CHANGE_STATE
client message.
Task-number: QTBUG-76147
Task-number: QTBUG-76354
Task-number: QTBUG-68864
Done-With: Liang Qi <liang.qi@qt.io>
Change-Id: Ic9d26d963979b7f0ef4d1cf322c54ef8c40fa004
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
When there are no translations on the UI file,
the function was left empty but without a 'pass'
statement, generating a SyntaxError,
this change includes it to avoid problems while
using the generated Python file.
Fixes: PYSIDE-1234
Change-Id: I30482a95c95fb4b4f4456531946a79c960d76318
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Starting with Android API 23+ gradle enables uncompressed native libs by default.
We must set android.bundle.enableUncompressedNativeLibs = false to gradle.properties,
to force it to compress them and extract them on the device.
Fixes: QTBUG-80766
Change-Id: Ia6d8d9179a341bbe7f8dc254a3b31d2ee8d7a5d7
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
