Linux kernel version 5.8 added support for the stx_mnt_id field in the
struct statx because "Systemd is hacking around to get it and it's
trivial to add to statx, so...". This allows us to much more neatly
match the lines in /proc/self/mountinfo.
The same kernel version added STATX_ATTR_MOUNT_ROOT so we can tell if a
given path is the mount point of a filesystem. We don't have a need for
that information for now.
We need to retain fallback code for two reasons: first, the user may be
running with an old Linux kernel, in which case we won't get the
STATX_MNT_ID bit set in stx_mask. Second, we may have failed to open()
the path in question, because the user may not have the necessary
permissions.
There's still a race condition because the mount IDs can be reused
immediately after something is unmounted. There's a 64-bit unique mount
ID (available since v6.8) but it's not reported in /proc/self/mountinfo,
so we couldn't us it right now. We can with 6.8's statmount().
Pick-to: 6.7
Task-number: QTBUG-125721
Change-Id: If3345151ddf84c43a4f1fffd17d3f7dbce4ff16b
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
Test this gesture with both mouse and touch.
Change-Id: I0285e7a17302361893c4b7ed0faf97c4f342814e
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
The RHI support and compositor in QPlatformBackingStore were
tied to the surface format of the top level window owning
the backing store.
This meant that inserting an RHI-enabled widget (QRhiWidget,
QOpenGLWidget, QQuickWidget, QWebEngineView) into the widget
hierarchy required recreating the top level window with a
matching surface format that could support the RHI composition.
It also meant that we could not have two RHI enabled widgets
with different surface format requirements (Metal and OpenGL
for example) in the same top level widget hierarchy.
The recreation of the window had various visual side effects,
such as temporarily switching out of full screen state, or the
widget rendering a frame of black, as well as more serious
problems such as not correctly restoring the window geometry.
In addition, if client code had pulled out the winId() of the
window, and did not invalidate these references on window
destruction via QEvent::WinIdChange or QEvent::PlatformSurface,
the client would reference stale window handles. Although
this is a programming error (QWidget::winId() specifically
mentions this requirement), we should avoid recreation if
we can.
We were already supporting flushing the backingstore to
individual native child widgets, but always did so via a
single RHI managed by the platform backingstore. By
expanding QPlatformBackingStore to keep one set of RHI
support and compositor per surface format, we can refine
the logic in QWidget and QWidgetRepaintManager to not
require recreating the top level. Native child widgets
are then flushed independently, including any RHI textures
and raster content that overlaps with the widget.
We still assume that a single RHI support and compositor
can be be used for multiple windows, as long as those
windows have the same surface format. In the future, if
needed, we can refine this to use one set per surface
format e.g.
Fixes: QTBUG-119221
Fixes: QTBUG-121181
Fixes: QTBUG-120096
Task-number: QTBUG-115652
Task-number: QTBUG-108344
Task-number: QTBUG-113557
Task-number: QTBUG-119309
Change-Id: I2635ed3d20c2fb76eab3b8130007dd656a0b93e5
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
We need to be able to have true popup windows in Qt Quick and Controls,
including handling the press-drag-release sequence to select one entry
from a menu or combobox. After the mouse press, a new window is created.
On some platforms (such as xcb), the new window gets window system grabs
of both keyboard and mouse (QApplicationPrivate::openPopup() calls
grabForPopup() and it actually works); while on others, the pre-existing
window continues to get the whole sequence of mouse events until the
release. In the latter case, Qt needs to forward events from the
original window to the popup. Until now, the list of popups was
QApplicationPrivate::popupWidgets.
Now we track the open popups as a list of QWindows rather than QWidgets,
in QGuiApplicationPrivate::popup_list, and add a set of static functions
to manage that list. Functions such as QApplication::activePopupWidget()
QApplicationPrivate::openPopup() and closePopup() are rewritten to make
requests to QGuiApplicationPrivate.
276943c8b791ba5897dcdb1ecfda780ac33a090b made
QGuiApplicationPrivate::closeAllPopups() virtual. That is now reverted,
because we're putting QGuiApplication in charge of popup management
and trying to minimize widget-specific behavior. So far,
QApplicationPrivate::closePopup() is still overridden to take care
of focus changes.
So far, QtGui does not take care of closing popups when the user
clicks outside: the active popup window gets those events, and needs
to close itself if the click occurs outside. An attempt to move this
logic raised some issues with legacy widget test cases.
Using a touchscreen to press on QMenuBar and open a QMenu, drag to
a menu item and release, is temporarily broken for now. The plan is
to fix it in a subsequent patch.
Task-number: QTBUG-68080
Task-number: QTBUG-69777
Change-Id: I02b5034987b5ee8909917d305f414c8b0db9c7f5
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
For backwards compatibility reasons, font files have multiple different
ways to specify vertical metrics (ascent, descent, etc.).
For OpenType, the main two are the usWin* and sTypo* metrics in the OS/2
font table. The usWin* metrics are typically used as the clipping bounds
of the font (so no character will ever draw outside these bounds). The
sTypo* metrics thus make it possible to specify a different set of
metrics for use in text layouts which is smaller than the clipping
bounds (or bigger), so that you can have fonts where some characters
overlap with preceding or subsequent lines.
However, GDI (and thus many applications) use usWin* also for line
spacing, which lead to the sTypo* metrics being untrustworthy in some
fonts and later to the introduction of the USE_TYPO_METRICS in the OS/2
table version 4. The idea of this flag is to tell the font system that
the sTypo* metrics can be trusted and should be preferred over the usWin*
metrics.
But the OpenType specification states that sTypo* metrics should *always*
be preferred and modern font systems such as FreeType and DirectWrite
will respect this. This in turn has lead to fonts where the
USE_TYPO_METRICS flag is untrustworthy instead, i.e. the sTypo* metrics
are preferable, but the USE_TYPO_METRICS has accidentally not been set.
Qt trusts the USE_TYPO_METRICS flag and uses the usWin* metrics whenever
this is unset. Since QFontMetricsF::height() (ascent+descent) in this
case includes the line gap metric, a lot of components have been written
to use it for size without adding any margins over the text. So changing
the default now would break a large amount of components, including the
ones in our own Windows style.
Most fonts should work correctly, by setting the USE_TYPO_METRICS flag
if the typo metrics are intended to be used. For those that do not, we
introduce a PreferTypoLineMetrics style strategy.
[ChangeLog][QtGui][Fonts] Added QFont::PreferTypoLineMetrics for using
the recommended line spacing metrics of the font, even if the font has
not explicitly set its USE_TYPO_METRICS flag.
Fixes: QTBUG-125585
Change-Id: Ib2f7df404fe719186d78733bda26da712f1ab85a
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
... to be in sync with QObject naming.
This amends b4c63b89dfe136d0579bf1b6422c4d878cdd74ab.
Change-Id: I25301f65aa880205d8c0cfd6f4bfa9fdba34a01c
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Also write gray, CMYK and mAB RGB color space profiles.
Fixes: QTBUG-125302
Change-Id: Id3b3b64537b9c08f1d40b8243c228ad111d08289
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
The test data is provided as float, so we must obey [conv.fpint]/1 when
converting from float to integer types. There are values in the rows
that are outside of the range of short and shorter integers, so just
check the range.
Fixes: QTBUG-125889
Pick-to: 6.7
Change-Id: If3345151ddf84c43a4f1fffd17d405ee0cd00d44
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
The test passes locally, and if individual functions should fail, then
skip those.
Change-Id: Ib9123bacaff2a83c2bc378b37201fd1d75dfdb45
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
We don't need the bodyName member, because it's the same as
originalBodyName, just encoded, and we can delay the encoding to
build() time.
This is not worse than the old code, since we anyway toString() the
QAnyStringView unconditionally.
So we don't need to visit the QASV and implement RFC2232 encoding for
all three view types, we can just use the QString version, after
toString().
This not only has the advantage of less code and not storing duplicate
data, but we now also encode u8"ä.txt" the same as "ä.txt"_L1 and
u"ä.txt", ie. using latin1, as required by Postel's Law, and not as
UTF-8, as the old code did.
Change-Id: If82a33a1cd09b859b3a4450a60083b1d3aedf7bc
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Constructing and composing a QHttpMultipart contains some aspects that
are possible candidates for automating, such as setting the headers
manually for each included part. As a reference, when issuing a default
multipart with CURL, one does not need to manually set the headers.
Add the class QFormDataPartBuilder to simplify the construction of
QHttpPart objects.
Add the class QFormDataBuilder to simplify the construction of
QHttpMultiPart objects.
[ChangeLog][QtNetwork][QFormDataBuilder] New class to help constructing
multipart/form-data QHttpMultiParts.
Fixes: QTBUG-114647
Change-Id: Ie035dabc01a9818d65a67c239807b50001fd984a
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Provide the new comparesEqual() helper function as an implementation of
(in)equality operators.
Use QT_DECLARE_EQUALITY_OPERATORS_HELPER macro instead of
Q_DECLARE_EQUALITY_COMPARABLE for the comparison between QPointer<T>
and QPointer<X> to avoid creating the reversed version of the
operators in C++17 mode.
Use new \compares command in the documentation to describe the
comparison operators provided by QPointer.
Task-number: QTBUG-120306
Change-Id: Iff24d3ef5c790de7f06ab825f853e06186fa1471
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
This makes it possible for the stdCompatibility() test to cope with
the backend (based on the Windows Registry, which disagrees with MS's
ICU-based std::chrono::tzdb) successfully constructing a zone, by
finding a known alias, instead of failing because it doesn't know the
name. It may also be useful to client code when dealing with legacy
names.
Amended some existing tests to only expect what they now should and
added some tests specific to aliasMatches(). Also added some
explanative comments where it isn't needed.
Task-number: QTBUG-115158
Change-Id: I095bdbead78df339e29b29518d5010ef905fa8b2
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Follow-up to commit 69555b364d5f398dec2e66c1f969cf3bfdc44057, using
QTimeZone::isTimeZoneIdAvailable() instead of trying to construct an
ad-hoc test for known unknonwn zone IDs.
Reopens: QTBUG-102187
Fixes: QTBUG-102187
Change-Id: I36ff4f1dfbc4035d73df5b8fd68376c1bc9641e1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Also rename the subdirs in the hiddenDirs_hiddenFiles test dir,
normalDirectory/normalDirectory/ is a bit confusing.
Make the test more deterministic by comparing lists of files/dirs
instead of just counts.
Change-Id: I12fdb5428bbef8382d4ee591792d167abcd216cf
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
So far, the keys of icon engine plugins were only interpreted as the
suffix of icon files, loaded via QIcon(filename). However, an icon
engine could provide a lot more flexibility if it could implement an
entire theme.
Match the list of keys a plugin can register itself with also against
the current theme name. If a matching plugin is found, use that plugin
to create the icon engine. Store the factory from the plugin to avoid
costly lookups for each icon.
Extend the QIcon test case by adding a custom plugin that supports two
themes. Since the plugin and icon engine creation infrastructure
doesn't communicate which theme the plugin was created for, use
QIcon::themeName to record the current theme when the engine gets
created.
[ChangeLog][QtGui][QIconEnginePlugin] The keys registered by an
QIconEnginePlugin implementation are now also matched against the
current theme (system or user theme), allowing engine providers
to implement entire themes through a plugin.
Change-Id: I8a5e30ff8b5bb7c78b5204e82760e4328671e4c1
Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
This exercise found an inconsistency involving char16_t. Filed
QTBUG-125588 to keep track of this and decide how to deal with it.
Task-number: QTBUG-125588
Pick-to: 6.7 6.5
Change-Id: I259fecef34539e8841149570d7411d959b48d876
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Extending the fromCharacter() tests to check for the construction to
be noexcept and implicit turned up a few bugs:
The constructor from char32_t was simply not marked as noexcept, even
though all its operations were. So just mark it noexcept.
The constructor from QChar, otoh, was never called, (certainly not for
rvalues), because QString happens to be constructible from QChar (and
QLatin1Char and QChar::SpecialCharactor), so due to perfect
forwarding, the if_convertible_to<QString> ctor won, allocating a
QString.
This is a regression of Qt 6.5 compared to Qt 6.4.
To fix this, exclude arguments that convert to QChar from matching the
if_convertible_to<QString/QByteArray> ctors, taking care to not match
those arguments that are already compatible_char<>s.
This, in turn, creates a problem for implicit QASV construction from
QLatin1Char and QChar::SpecialCharacter, because now that we've
excluded them from the if_convertible_to<QString> ctor, calling the
existing QChar or QString non-template constructors for these types
would require two user-defined conversions (from said type to
QChar/QString and from QChar/QString to QAnyStringView). That works
for explicit construction, but we need implicit convertability.
So bring out the big guns once more and add a perfectly-forwarding
ctor for anything convertible to QChar. QChar itself is actually
already handled by compatible_char<>, so the old QChar non-template
ctor can go. The extra copy of the QChar argument will be optimized
away by the compiler.
[ChangeLog][QtCore][QAnyStringView] Fixed a regression where
constructing a QAnyStringView from a char-like type (QChar,
QLatin1Char, ...) would first construct a QString and only then
convert that to QAnyStringView.
Amends 812a0d3125cb89e340c59aa92cdc946862fb009d.
This change is forward and backward BC and SC, but not
behavior-compatible: certain operations detectably change
noexcept'ness, and some arguments may now cause the resulting
QAnyStringView to have a different encoding (though I really tried
hard to avoid that).
Since it's a regression, I proposed to pick this to the affected
branches, 6.7 and 6.5 (6.6 is already closed at this point).
Pick-to: 6.7 6.5
Fixes: QTBUG-125735
Change-Id: I86f37df5d80bee36db27e529c017cb73995a6831
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Until now, QCryptographicHash had a big drawback over the underlying
C APIs: at least one extra memory allocation: Either you created a
QCryptographicHash object (allocates a Private), and enjoy noexcept
addData() and resultView(), or you used the static hash() function
(which creates a Private on the stack, so doesn't need to allocate
one), but then we needed to return in an owning container.
Enter QSpan and hashInto(), which allow the user to provide a buffer
into which to write, and therefore can be completely noexcept and
(apart from a missing optimization which, however, doesn't affect the
API) deliver near C-API efficiency.
We don't have QMutableByteArrayView, which would deal with the
different types of equivalent char types for us, so we overload for
QSpan<std::bytes>, QSpan<char> and QSpan<uchar>, which makes the
overload set accept roughly what QByteArrayView would accept, too.
Return by QByteArrayView because that has a richer API than QSpan,
which lacks (on purpose) string-ish API.
[ChangeLog][QtCore][QCryptographicHash] Added hashInto().
Task-number: QTBUG-125431
Change-Id: I80ecc7151d3418a36c4d5db6d22d0b82c869b19f
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
These were not covered by the existing tests.
Pick-to: 6.7 6.5 6.2
Change-Id: I909ea3aa5b676904dc72ecf8ce32b73cca1b6af7
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
Create a temporary file instead of trying to use a file from source
directory because that's not available when test was installed via
packaging.
Task-number: QTBUG-118680
Change-Id: Ic152d68c79ac467c0d149cab04ab224c3b64099f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QTZ::displayName() returns an empty string if it doesn't know how to
localize a zone name. In this case, we can't expect it to match the
expected name, which depends on having relevant locale data available.
Task-number: QTBUG-115158
Change-Id: I1cd8c1469399502764c354bf24423298f106f23e
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
We were only handling the case where uri-list was a file.
This fixes a regression.
Also - add dragging url into the clipboard manual test
Pick-to: 6.7 6.7.1
Change-Id: Ifbd087ffd157463b6b903199e3ff22c2de1c4942
Reviewed-by: Piotr Wierciński <piotr.wiercinski@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
This is meant to allow the buildsystem to add additional paths (for e.g.
QML imports) while still allowing the application to pick up the paths
where e.g. Qt's plugins are located – without having to query qtpaths or
the build system to locate them first, and to then also add them to
qt.conf.
QLibraryInfo::paths will return a list containing any values provided in
qt.conf, followed by the value that would have been provided if there
were no qt.conf file.
If no value is provided at all for a given key in qt.conf,
QLibraryInfo::path will behave as if no qt.conf file was provided,
instead of using LocationInfo::defaultValue.
For now, this is considered to be internal only, and won't be
documented.
Change-Id: Ic74c66b5cbde605b43336cd88c3f909999381f46
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Add missing checks for integral overloads uint, short and ushort and
character overloads char and char16_t (the latter being a reproducer
for QTBUG-125588).
Pick-to: 6.7 6.5 6.2
Task-number: QTBUG-125588
Change-Id: I479e4c8c3cd86c3f85ab3c9d44b79158b0a48476
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
setFloating() called on the 2nd last dock widget in a floating tab
didn't cause reparenting to the main window. The dock widget remained
parented to a QDockWidgetGroupWindow, while no longer being part of
floating tabs. The QDockWidgetGroupWindow would therefore have only
one tab, which is an inconsistent state and causes crashes.
Factor out the implementation of setFloating() to a new
QDockWidgetPrivate::setFloating(). Reparent to the main window, if
the dock widget is part of floating tabs.
Add test functionality in tst_QDockWidget::setFloating().
Turn createFloatingTabs() into a macro, in order to QSKIP() in the test
function's scope.
Fixes: QTBUG-122394
Pick-to: 6.7 6.5
Change-Id: I3719785f310b131446cdea908f09b7195c992b3e
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
This tests assumes existence of `PATH` environment variable. This
variable is not a special variable on VxWorks (it's neither set nor
evaluated by command line), causing this test to fail.
Skip this test as it's not applicable on VxWorks.
Pick-to: 6.7
Task-number: QTBUG-115777
Change-Id: I431b3f0a0ff9efefffc356bb5c7e11fdfac78690
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The codepaths for image scaling are a bit convoluted, so some "surgery"
is needed. QImage::scaled() delegates to transformed(), building a
suitable scaling matrix.
transformed() checks if the matrix is a scaling matrix, and then
has several dispatches.
If smooth scaling was requested:
* if the image format is supported by smoothScaled() without needing a
conversion, delegate to that;
* otherwise, if the transform is "non paintable" or the source image is
big enough, then again call smoothScale. "non paintable" here means
that we're scaling more than 2x down, and QPainter wouldn't do a good
job.
Otherwise, images in color formats (>= RGB32) are converted by applying
the needed transformation on a QPainter and draw the source image with
that transformation.
Otherwise, if the matrix is invertible (a scaling matrix with non-zero
scaling always is, it's a diagonal matrix), then dispatch to
qt_xForm_helper.
--
Amend this reasoning to support CMYK images:
* Make smoothScaled support CMYK without conversions. To do so,
make qSmoothScaleImage scale CMYK as if it was a ARGB image.
* Make transformed() call smoothScaled() for CMYK images
* In transformed(), consider CMYK as nonpaintable, because we can't
paint over a CMYK image.
* In the non-smooth codepath, also check that we don't try to paint over
CMYK, and always go through qt_xForm_helper instead.
Note that we still don't support any other transformation for CMYK.
Add a test, adapting the exiting one for RGB.
Change-Id: Ic72d78923a17fb3963aa22c57265904c716792b0
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
QLibraryInfo::path only allows a single value, however, at least for
QmlImportsPath we would benefit from having more than one to ease
deployment (note that this is unrelated to Qml2ImportsPath which was
necessary to support QML 1 and 2 in a single Qt install).
Users inside Qt will ported in follow-up patches where it makes sense.
[ChangeLog][QLibraryInfo] qt.conf now allows providing mutliple paths,
and QLibraryInfo has a new paths method to fetch all of them.
Task-number: QTBUG-124009
Change-Id: I78b3bcfa94c8d975dbd789f6826cda4fc5e60403
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This will build a multi-abi Android AAR target and verifies
its content using a customer target "verify_aar" by running
zipinfo on the AAR, and comparing its content with a subset
of files that are expected to be present in the package.
Task-number: QTBUG-125483
Change-Id: I19ad132bbad1a75c40b00de173ca09ad56500076
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Implement partial support for SQL_INT128 datatype which is used for
DECIMAL/NUMERIC columns with a precision > 18. This support is only
available when QT_SUPPORTS_INT128 is defined and for MSVC.
Binding values to columns which need SQL_INT128 is supported but
numbers given as QString will be converted to doubles even if
QSql::HighPrecision is set.
[ChangeLog][SQL][IBASE] Added support for SQL_INT128 datatype.
Task-number: QTBUG-124575
Change-Id: If3fb1eb0f19dc60f000d163f3bf45da7acb08c87
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Andreas Bacher <andi.bacher@outlook.com>
[ChangeLog][QtNetwork][QNetworkAccessManager] QNetworkAccessManager now
supports using full local server name, as in, named pipes on Windows or path to
socket objects on Unix.
Task-number: QTBUG-102855
Change-Id: Ifc743f5025b3d8d0b558ecffff437881897915d9
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Decided to keep the strong_ordering semantics, as user code is very
much guaranteed to rely on it.
Update the docs.
[ChangeLog][Potentially Source-Incompatible Changes] Made several
operators (relational, QDebug streaming, etc) hidden friends. This may
break code that calls these operators a) with types that require
implicit conversion on the arguments, or b) using member-function
syntax (lhs.operator<(rhs)). The backwards-compatible fix for (a) is
to make at least one of the argument's implicit conversions
explicit. For (b), use infix notation (lhs < rhs).
[ChangeLog][EDITORIAL] Replace all "Potentially Source-Incompatible
Changes" items that mention making operators hidden friends with the
catch-all one from the QPersistentModelIndex change.
Fixes: QTBUG-117660
Change-Id: I8175af5e2cb3c77e7486e972630b9410b3a0896c
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Use Milian's asTuple() trick and compareThreeWayMulti().
The pointer variable presents a challenge, because, like op<, op<=>
cannot order unrelated pointer values. Trying to is UB; one is
supposed to use std::less and std::compare_three_way instead, but, you
guessed it, less<tuple<T*>> uses op< to compare the tuples, which, in
turn uses op< on the T* member, causing UB.
[ChangeLog][QtCore][QtCompare] Added a Qt::totally_ordered_wrapper
class whose relational operators use the std ordering types. Use
it to wrap the pointers when doing the comparison to avoid UB.
This allows a smooth migration once we depend on C++20 and class
authors start using lhs.asTuple() <=> rhs.asTuple() or even start
to =default op<=>. The wrapper type is robust against all these
transformations.
Update QModelIndex comparison documentation part.
Task-number: QTBUG-117660
Change-Id: I126b2ebff458800134ed6e774a389d85d76a395f
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
In Qt 5, we had QVector<QRect> QRegion::rects(), but it was
deprecated, because just iterating over the QRegion as a container of
QRects was more efficient (QRegion has a SSO for the case of one
rectangle). With QSpan, we can now bring it back with the same
efficiency as iteration, supporting Qt 5 code that never made the move
away from rects() and new code that wishes to make the conversion into
rectangles more explicit. Re-add the Qt 5 tests, which show that the
function is nearly a drop-in replacement for the Qt 5 rects() (QSpan,
at the time of this commit, doesn't have relational operators, yet).
Also add a QSpan overload of setRects(). The old (ptr, n) function
(now obsoleted, but not deprecated) allowed nullptr + n != 0, which
QSpan doesn't accept, so print a warning in that case. Also, QSpan can
hold more than INT_MAX elements on 64-bit platforms, but QRegion's API
was never ported from int to qsizetype, so we need to catch oversized
spans used as inputs, too.
[ChangeLog][QtGui][QRegion] Added QSpan overload of setRects();
re-added Qt5's rects(), but returning QSpan instead of QVector now.
Fixes: QTBUG-124712
Change-Id: I24570c886cbf77abd8d1f4a3f42ae53c892cd9ff
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
They don't work unless #! are the first two bytes of the file.
This is a fix-up for commit d6bc7613ac91e663f573486135946039eeccf8d7
Change-Id: I83c976a538a67eef3d40b50674a255d6e2f3205a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lucie Gerard <lucie.gerard@qt.io>
For std::span, these are free functions in namespace std and therefore
work for anything implicitly convertible to std::span,
incl. QSpan. But they're C++20, QSpan needs something for C++17
builds.
By adding them as hidden friends for QSpan, we allow unqualifid calls
to transparently resolve to the respective overload, ensuring
source-compatibility between std:: and QSpan, and, eventually, a
transition back to std::as_*_bytes.
I considered the alternative to add these functions in the q20
namespace, but q20::as_bytes() would have to take QSpan, and QSpan is
convertible from more types than std::span, so we wouldn't be able to
guarantee that std::as_bytes(t) works for all T t for which
q20::as_bytes(t) works, the fundamental guarantee for namespace qNN.
[ChangeLog][QtCore][QSpan] Added std::span-style as_bytes() and
as_writable_bytes() functions.
Fixes: QTBUG-125489
Change-Id: Ia9a7560c7843e182892608178433be7349c825ba
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Move the constexpr fromArray() check from fromArray() to constExpr()
and there add the protection necessary to work around
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71962, copied from
tst_QByteArrayView.
As a drive-by, add the test for constexpr construction from a char16_t
array, which was missing here (probably because of the GCC bug) while
the corresponding test was already in tst_QByteArrayView.
Amends 107ff4c1d6b5da2cb11c65b2bd9106817f7fdb02(!).
(iow: 6.0, and no, I don't know why it became a problem only now)
Pick-to: 6.7 6.5 6.2
Change-Id: Id9d2a08175709f8bf85b3b192e7aa49783b9e715
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Move the constexpr fromArray() check from fromArray() to constExpr()
where there is already the protection necessary to work around
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71962.
Amends bbe7570ddcc6fcce707a355b76d2d0024c44ea38(!).
(iow: 6.0, and no, I don't know why it became a problem only now)
Pick-to: 6.7 6.5 6.2
Change-Id: Id9d2a08175709f8bf85b3b192e7aa49783b9e714
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Because of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71962, and
because marking constExpr() as __attribute__((no_sanitize_undefined))
doesn't fix a thing, we opted to disable the triggering parts of
constExpr() from all GCC builds.
This is, of course, unfortunate, since it meaningfully reduces
compile-time coverage in the general case for a rather obscure build
type most won't ever use.
While GCC doesn't give us a predefined macro to check for in the .cpp
file, the cmake build system knows whether we use ubsan, so just
define a macro of our own.
As a drive-by, simplify GCC detection by using Q_CC_GCC_ONLY.
Amends de6a004bc5a5b9cd3ecfbb14818bb42fb1ecfd68.
Pick-to: 6.7 6.5 6.2
Task-number: QTBUG-101307
Change-Id: I4be5bd103b9d2386b2ac9fd22e0c34f9c63fee04
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Replace public friend operators operator==(), operator!=(),
operator<(), etc of QUuid and GUID with friend methods
comparesEqual() / compareThreeWay().
Use Q_DECLARE_EQUALITY_COMPARABLE_LITERAL_TYPE, because
the (in)equality operators are constexpr.
And then we use helper macros, because the other relational
operators are not constexpr. Cannot make relational operators
constexpr, because it requires to make variant() and isNull() methods
constexpr and QT_CORE_INLINE_SINCE. But the experiments
show that it does not work with adding constexpr to
QT_CORE_INLINE_SINCE.
Put relational operators under !QT_CORE_REMOVED_SINCE(6, 8) to prevent
an ambiguity. On Windows the metatype for QUuid is created
in removed_api.cpp. That leads to an ambiguity, and as a result
the compiler fails to create the equals methods of QMetaTypeInterface.
This, in turn, leads to the failed comparisons.
The usage of !QT_CORE_REMOVED_SINCE(6, 8) solves the problem.
Task-number: QTBUG-120304
Change-Id: I640bdeb8f1f7306ba06b9e4193d008cf2bb6dbfb
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Focus abstraction in 58d5d4b7c2faaeaa2c2ccdabb3da6d37f9db880a was
supposed to be behavior-neutral. QWidgetPrivate::reparentFocusChildren
used QObject::findChildren() to find children inside and outside the
current focus chain. If the re-parented widget had only direct children
and the focus chain was equal to creation-order, the result was
identical to previous behavior.
When the re-parented widget had grandchildren, the behavior differred.
While not being detected by existing tests, it caused a regression.
Re-implement the previous logic: Iterate through the focus chain,
instead of relying on QObject::findChildren() returning a complete and
well-ordered list.
Modify tst_QWidget::focusChainOnReparent() in order to cover the
regression.
This amends 58d5d4b7c2faaeaa2c2ccdabb3da6d37f9db880a.
Fixes: QTBUG-125257
Change-Id: Iff4f1d0d9b6117c50c8980dfb6eebfc6f6d4a710
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Overload set to be used in QUuid soon.
Pick-to: 6.7 6.5
Change-Id: I6f2c180bb7bb884d40f1691409e816405c3e5d47
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
