Just so we can get this cleaned up as well and remove it from
Qt Core.
Change-Id: I2b5b821b039ce2c024ec3cb7338a1a9becdd2157
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
There's no real dependency to QTextCodec in those files anymore.
Change-Id: Ifaf19ab554fd108fa26095db4e2bd4a3e9ea427f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Document QStringConverter, QStringDecoder and QStringEncoder.
In addition, do some touches to the API, renaming one enum value,
add a flags argument to one constructor and make some members private.
Change-Id: I8f99dc3d98fb8860cf6fa46301e34b7eb400511b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use QStringConverter instead. Also change the default
encoding of QTextStream to utf8.
Change-Id: I30682e75fe0462d1a937539f773640c83a2d82e1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
As a first step add setEncoding/encoding() methods that use the
QStringConverter::Encoding enum, and port all uses of setCodec()/
codec() over to the new API.
Internally QTextStream still uses QTextCodec, this will be ported
over to QStringConverter in a follow-up change.
Change-Id: Icd764cf47b449b57f4ebd010c2dad89e6717d6c0
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This is a replacement for Qt::codecForHtml().
Change-Id: I31f03518fd9c70507cbd210a8bcf405b6a0106b1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add method that tries to determine the encoding of the data
from an initial byte order mark.
Change-Id: I348c51a3d4db9b434af53359b739a7e17acfc760
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add static methods that allow converting between a name for an
encoding and the Encoding enum.
Change-Id: I12bc503cf757ea31d3ca8d5e1f1216efddcb16d4
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add a constructor, that allows constructing a string converter by
name. This is required in some cases and also makes it possible to
(in the future) extend the API to 3rd party encodings.
Also add a name() accessor.
Change-Id: I606d6ce9405ee967f76197b803615e27c5b001cf
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Always encodee INI files as utf-8 in Qt6. This is mostly backwards
compatible, as old ini files would encode all non ascii characters.
[ChangeLog][Important behavioral changes] QSettings will now always
encode INI files as utf-8 (and the iniCodec/setIniCode methods are
removed). This is a change from Qt 5 and earlier, where QSettings would
by default escape all non ascii characters. The behavior is equivalent to
what you got in Qt5 by setting a utf-8 iniCodec on the settings object.
Settings files written in Qt 5 will still be readable in Qt 6 (unless
an iniCodec different from utf-8 was used), but to read Qt6 based ini
files in Qt 5 applications, setting the iniCodec to utf-8 is required.
Change-Id: Ic7dffcca17779bd5e3dae50d42ce633170289f6c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
Feed the data one by one to the encoder or decoder to
verify that the handling of incremental decoding is
correct.
Change-Id: I565e4f1872e00859026334f7662b6778772e159d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
IgnoreHeader was a rather badly defined enum, in addition the
utf8 and utf16 codecs where handling BOMs somewhat different
for stateless decoding.
Fix this by introducing explicit flags for writing a bom when
encoding and not skipping the initial bom when decoding.
Source compatibility for QTextCodec is done with a couple of
static constexpr variables.
Change-Id: I0b2d94f84c937cec1e0494c16ef448c00382691d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove support for setting a codec different from UTF-8
for writing XML files.
All XML readers today can handle UTF-8, and there is no
reason anymore to write a file in a different encoding.
Change-Id: If89fb2d2474a2b55644d9bed7473c11ad91033eb
Reviewed-by: Simon Hausmann <hausmann@gmail.com>
The new QStringEncoder and QStringDecoder classes
(with a common QStringConverter base class) are
there to replace QTextCodec in Qt 6.
It currently uses a trivial wrapper around the utf
encoding functionality.
Added some autotests, mostly copied from the text codec
tests.
Change-Id: Ib6eeee55fba918b9424be244cbda9dfd5096f7eb
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Handle wheel grabbing via wheel_widget in a single place, and
propagate events in the same way for all (spontaneous) events.
Handle ScrollMomentum the same way as ScrollUpdate to allow
partial sequences.
Fix the incorrect ignoring of wheel events by default; like all
other input events, they are now again accepted by default and
ignored in the default event handler implementation of QWidget.
This way, implementing the handle suffices to accept the event.
Note that QWidget::wheelEvent doesn't need to be changed, as the
event is ignored there today (an oversight of the change made in
f253f4c3, perhaps).
This also fixes changing of direction of a wheel event while
the event sequence is grabbed by a widget.
Change-Id: Ia0f03c14dede80322d690ca50d085898a0497dbe
Fixes: QTBUG-67032
Task-number: QTBUG-79102
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
The former messes in bad ways with the overload set (it, fatally,
attracts char16_t, e.g.). The latter was probably added in response to
ambiguities between (char) and (QChar). While it's harmless now,
remove it, since it no longer pulls its weight.
The no-ascii warning is now coming from QChar(char), so the protection
isn't lost.
[ChangeLog][QtCore][QString] The += operators taking char and
QChar::SpecialCharacter have been removed as they cause adding a
char16_t to QString to call the char overload, losing information. The
append() function was not affected.
Change-Id: I57116314bcc71c0d9476159513c0c10048239db3
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Will be useful when DecompressHelper gets taken into use for both.
Task-number: QTBUG-83269
Change-Id: Iaf253219bed193025c2b82d6609f4dcc4de33df8
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Also port its callees. These functions scream to the QStringView-ified...
Change-Id: I13c95d65941eb8d02223306d80efd1437b4bd9b7
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The fromUtf16(ushort*) and fromUcs4(uint*) overloads are going
to be deprecated. Use the newer fromUtf16(char16_t*) and
fromUcs4(char32_t*) overloads.
As a drive-by, use std::end()/std::size() where applicable.
Change-Id: I5a93e38cae4a2e33d49c90d06c5f14f7cb7ce90c
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
We use the Catch2 testing framework to test Qt Testlib, which also opens
up the possibility of using it for other internal testing once it's made
available through the build system.
The test now has a --rebase mode which will write out the actual results
as new expected files. Once we add the required post-processing to the
results to remove timestamps and other testrun-specific data we can
remove the standalone python script generate_expected_output.py that
today has to be kept in sync with the test itself.
No attempt has been made to clean up the comparison-functions, but
these could all benefit from moving their logic from the comparison
to the sanitization step. This will both make the expected files
more generic, and will reduce the diff once a failure occurs, since
we're not seeing all the hunks that the comparison-functions ignored.
Change-Id: I1769d42e7958d56d1ad5da958db0e8fe3a2a3c23
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Update the auto test accordingly, and at the same time remove
all uses of QTextStream (as they aren't required).
Change-Id: I71b7cf6a6b54ea59507f27d5d2d04cc5ae5885fc
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
The fromUcs2() named ctor is designed to replace all the non-char
integral-type constructors of QChar which make it very hard to control
the implicit QChar conversions, which have caused a few bugs in Qt
itself. As a classical named contructor, it simply returns QChar.
The fromUcs4() named "ctor", however, needs to expand surrogate pairs,
and thus can't just return QChar. Instead, it returns a small struct
that contains one or two char16_t's, can be iterated over and be
implicitly converted to QStringView. To avoid bikeshedding the name
(FromUcs4Result, of course :), it's defined inline and thus can't be
named outside the function. This function replaces most uses of
QChar::requiresSurrogates() in QtBase.
[ChangeLog][QtCore][QChar] Added fromUcs2(), fromUcs4().
Change-Id: I803708c14001040f75cb599e33c24a3fb8d2579c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This part of the patch changes all remaining occurrences of the local
versions of Orientation to Qt::Orientation.
Change-Id: Ic9ec19b8f069f614061f319abd30841e10cdd626
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
wchar_t hasn't been QStringView's storage_type for a very long time
(at least since 5.10). And in Qt 6, we require char16_t support not
only in the compiler, but also in the stdlib, so drop the guards and
the alternative code paths.
Change-Id: I99f28b575f61c16a2497840708beaa4b54a80f57
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
We end up with an absurd situation:
QSslConfiguration conf;
//we set CA certs so that on demand loading is disabled:
conf.setCaCertificates({...});
QSslSocket s;
s.setSslConfiguration(conf);
Q_ASSERT(conf == s.sslConfiguration); // this assert fails.
Pick-to: 5.15
Change-Id: I272bed145434082acc6ead95fe3640d222b21131
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
They don't work with std::span, since std::span has neither
const_iterator nor cbegin()/cend().
To fix, perfectly forward the return type with decltype(auto) instead
of mentioning T::const_iterator, and rely on the const-qualfication of
the help::c{r,}{begin,end}() functions' arguments to select the
correct T::{r,}{begin,end}() overload.
Change-Id: I4992d4bd521d2dc0f9ea51ae70cde8286ae543a5
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The message to ignore is given in source code, hence UTF-8; it was
being ingested as local 8-bit, which lead to problems when a debug
message wasn't 7-bit clean and the system's native encoding wasn't
UTF-8. Modified QtTest's selftest to check encoding failure.
Pick-to: 5.15
Change-Id: I898744a450115b6d2ee992f1d3b36d8efaeeff7e
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's documented to be undefined if the number of seconds is outside
the allowed range, but it doesn't hurt for that undefined behavior to
happen to be that the result is invalid. Added a simple test.
Change-Id: I20c3f680c7948b3904f213452272133be77e4d62
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
[ChangeLog][Important Behavior Changes][QDateTime] Removed all
locale-dependence from QDate, QTime and QDateTime, including the
Qt::DateFormat members that select the formats of the default and
system locales and the toString(Qt::DateFormat, QCalendar) overload,
which only used its calendar for these formats. All toString()
methods now use, and all fromString() methods only recognize, the C
locale's names for days and months. Use QLocale's methods if you need
to take locale into account.
Fixes: QTBUG-80441
Change-Id: I3a8968438741afb00f44262f79659c51e9b06c35
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It clearly belonged in tst_QDate::toDateTime(), for which it adds a
few more test-case and in which it inspires some further testing.
The new testing of case-insensitivity doesn't work if the format
contains stray non-format characters, so added a new data column to
take care of that.
Pick-to: 5.15
Change-Id: I73619be02091c97024a84cb963c7029e9fd0569a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
They use the Qt::TextDate format, so do depend on the feature.
Rename one in the process; nothing in its test has anything to do with
de_DE locale.
Pick-to: 5.15
Change-Id: I2adae5c46e6009c13b433993ed2c3c761a500bfb
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The native command buffer handle was not updated, so the subsequent
finish() call attempted to record an invalid VkCommandBuffer. The
problem was not present with offscreen frames, only when finish() is
called with a swapchain-based frame active.
Task-number: QTBUG-84066
Change-Id: I9c4cb701c3dbbc28f237d6ae1cbf65aafd1fa95f
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
We checked against integer overflow, but not against overflowing the
QByteArray size limit. That caused a std::bad_alloc to be thrown, which
is bad when decoding unknown data. QCborStreamReader wasn't affected,
since it doesn't merge chunks.
Change-Id: I99ab0f318b1c43b89888fffd160c36f495fada87
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
It simplifies the API and reduces surprise to have rotation working by default.
On Android, the manifest specifies which orientations the application has
been designed to support; on iOS, it is controlled via the
UISupportedInterfaceOrientations property list key.
In addition, QWindow::contentOrientation() is another way to give
a hint to the window manager, or on iOS to directly control whether
the window's rotation is locked or not.
Task-number: QTBUG-35427
Task-number: QTBUG-38576
Task-number: QTBUG-44569
Task-number: QTBUG-51012
Task-number: QTBUG-83055
Change-Id: Ieed818497f686399db23813269af322bfdd237af
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Test the case where reduce function of the form:
V function(T &result, const U &intermediate)
has T and U types different.
Task-number: QTBUG-83802
Change-Id: Ic89396caba16e7e47ae3ec1527e31b8620f1b08c
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Get rid of code repetition in concurrentfilter test.
Move tests with initial value next to their original version.
Join new lambda tests into a common functions
testing all possible 16 combinations of
functor / function / member / lambda, as they
test in fact the same function. There is no need
to distinguish lambda case over other cases.
This helps in test readability and maintenance.
Add missing tests for lambdas with a combination
of initial value.
Task-number: QTBUG-83802
Change-Id: I45930c1e18a9e4e561909f46a5cbbdf0ad7ba333
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
If any of the temporary directories and files can't be created, skip the
test. Otherwise, the cleanup routine would recursively delete "/".
Change-Id: I51f908a468be8fd2ebd523ff7ce27a7c78d1b4e2
Fixes: QTBUG-83863
Pick-to: 5.15
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The system APIs expect an absolute "display name" of the file path,
so make it absolute.
The test was overly tolerant in accepting failure, as a QStorageInfo
initialized with a file path that doesn't exist is invalid, and thus
always different from the QStorageInfo of the home directory. Fix the
test to compare only valid QStorageInfo objects, and postpone the check
until the file we want to move has been created.
[ChangeLog][QtCore][QFile] moveToTrash supports relative file paths
on Windows
Change-Id: I94c8cd40c60fde469e38f76a98f867f20c6a0b15
Fixes: QTBUG-84015
Pick-to: 5.15
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Also remove dead code that isn't compiled anymore in Qt 6 builds.
Change-Id: I7a7ae35e61fb2ad9cc21180fb7224357ade1505f
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
The replacement is to set, and configure, a QCompleter directly via
setCompleter. With the removal of the separate properties in QComboBox,
the configuration of the completer is not maintained if the line edit
is replaced. A QCompleter is created and set implicitly when the line
edit is set, unless the line edit came with a completer. This is what
the auto test verifies as well.
Change-Id: I9a4c73db5e39a2558aad346c0904be6deb4f1cd2
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
This amends 23b998fa45, and the commits
3e7463411e and
947883141d. This change restores the
pre-5.13.1 behavior of setFocusProxy to not move focus away from a
previously set focus proxy.
With the previous changes, focus would move away from a proxy when a
new proxy is set, if the old proxy had focus. While there are arguments
in favor of this behavior, it is a change of behavior that shouldn't
be introduced to 20+ years old functionality in order to fix the real
bugs addressed by the initial commits.
Instead, move focus only to the new proxy when the focus widget was
the widget that gets a focus proxy.
[ChangeLog][QtWidgets][QWidget] setFocusProxy no longer moves focus
away from a previously set focus proxy, restoring pre-Qt 5.13.1
behavior.
Change-Id: Icf2ad7cba5b860014aeef91cb274c442a2ab9d42
Fixes: QTBUG-83720
Pick-to: 5.15
Reviewed-by: David Faure <david.faure@kdab.com>
We already had this approach for QtWidgets, so to avoid
adding the specific cases every time, we import all the
classes from QtCore and QtGui.
Change-Id: I38dd2eec3bbdfb83f156a48f35a2f4fda44528c7
Pick-to: 5.15
Fixes: PYSIDE-1287
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
This file mustn't have changed at all, aside from the copyright headers,
since we switched to Git in October 2008.
Quick check shows the data from W3C is still online:
https://dev.w3.org/cvsweb/2001/XML-Test-Suite/
I didn't check that the CVS server is actually running because none of
my machines even has the cvs command anymore...
Pick-To: 5.15
Change-Id: I4e559af2a9a1455ab770fffd15f4e740114a2433
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Turns out that crashing on purpose is more difficult than it seems. It
should be easy, given how often we do it accidentally...
Let the null pointer dereferencing be the fall back. Some compilers are
too smart for their own good and remove the fault. Instead, let's rely
on raise(SIGABRT) on Unix and on the UD2 instruction on Windows.
Pick-To: 5.15
Change-Id: Ibdc95e9af7bd456a94ecfffd1603f1c9b73b167d
Reviewed-by: Simon Hausmann <hausmann@gmail.com>
Expected and actual were inverted.
Pick-To: 5.15
Change-Id: Idc3fae4d0f614c389d27fffd15e9fa6a0a8f25e6
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
In that specific wrapping mode, it will first try a normal word wrap. If
it doesn't fit within the specified line width it will discard the
result of that and try WrapAnywhere by calling layout_helper()
recursively. The problem was that at the point it called itself again it
had already adjusted eng->maxWidth:
eng->maxWidth += line.textWidth;
This was not restored, but carried on to the recursive call to
layout_helper(), so the end result was that the maximumWidth would
accumulate text widths from parts of the same line twice.
Due to the same recursive behavior the minimumWidth also had a problem:
It always returned the width of the widest word because it took the
qMax() of the minimum widths of the two passes, (WordWrap and then
WrapAnywhere) effectively making the minimum width always be the width
of the widest word (even though it could wrap at finer granularity).
Pick-to: 5.15
Task-number: QTBUG-77337
Change-Id: Ie7e9c17b157506352c2da38cc7f4a8dfa1283966
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
Should improve performance and is going to be required in
the future anyway.
Change-Id: I89d7c50441d2491da1ab0a4d564dcc91f52ade85
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
Qt Designer let's you add custom slots and signals to the main
form; they should use string-based connection syntax since the
class is not known in setupUI(). Amends
da3cb1deb6.
Task-number: QTBUG-76375
Change-Id: I5a3a5630f77c812d48db1cdb7a8658a4d2718228
Reviewed-by: Liang Qi <liang.qi@qt.io>
A failure in testlib is apt to imply misleading failures elsewhere, so
catch those first. Likewise, broken tools or corelib break
everything. Put the rest of the list in alphabetic order.
Restructure auto.pro to use conditional SUBDIRS += (in the right
order) instead of setting SUBDIRS to a full list and then doing
conditional SUBDIRS -= for most entries. This more closely matches the
way the generated cmake config does things, although it still doesn't
regenerate cleanly.
Change-Id: Idc15326c3534eb4fdce55394269f2dfbc17fcd99
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
As discussed in the mailing list and in the Qt Contributor Summit 2019.
Tested on Linux, macOS and FreeBSD, showing these fallbacks:
OS environment setlocale() call
FreeBSD empty "C.UTF-8"
FreeBSD LC_ALL=C "C.UTF-8"
Linux empty "C.UTF-8"
Linux LC_ALL=C "C.UTF-8"
Linux LANG=en_US "en_US.UTF-8"
Linux LANG=de_DE@euro "de_DE.UTF-8"
Linux LANG=en_GB.iso885915 "en_GB.UTF-8"
Linux LANG=hy_AM.armscii8 "hy_AM.UTF-8"
Linux LANG=ja_JP.sjis "ja_JP.UTF-8"
Linux LANG=ru_RU.koi8r "ru_RU.UTF-8"
macOS empty "UTF-8"
macOS LC_ALL=C "UTF-8"
Versions tested: FreeBSD 12.1, Linux w/ glibc 2.30, macOS 10.14.2.
See
* https://wiki.qt.io/Qt_Contributor_Summit_2019_-_QtCore
* https://lists.qt-project.org/pipermail/development/2019-October/037791.html
Change-Id: Ia2aa807ffa8a4c798425fffd15d97ddb4f35b0ae
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The proper solution is to use qint32 everywhere, but that is left as
a separate exercise.
Change-Id: Id0c06b102b56a1b3b48dd67c6c29c28da7d1f22d
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
Add a enum and formatting for member function pointer based
connections. Now preferably use member function pointer for Qt classes
or parameterless connections.
This should not require qOverload() within Qt classes after the
Signal/Slot disambiguation.
Add command line option to force either syntax for all connections.
Task-number: QTBUG-76375
Change-Id: Icdb4051e1173172a71cd536bdbc7d1ab1edf267d
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
If there is a negative right bearing on the last character of the string,
it will stretch past the tab position (which aligns the advance position,
and not the right edge of the glyph). For certain fonts, this would cause
the actual ideal width to be calculated as 301 instead of 300 (which is
the right alignment edge set in the code).
Pick-to: 5.15
Fixes: QTBUG-46206
Change-Id: I03e8b8fb86e9ebe5337d3ba3384ade73d2ccdd69
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
We don't rely on a latin1 locale anymore for the test,
and the other code was not doing anything.
Change-Id: I08bc08d200c9e037884d8b680dfbb24c129f3d2e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
We must move the functor properly into the binding object, otherwise we
end up with stale pointers as pointed out by ASAN.
Change-Id: Icd84f4c113dd48e1e3e2d744abac0902cdf9339e
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The API reduces the amount of manual plumbing required to offer a
conceptual property through the traditional setter/getter API as well as
through QProperty<T> API. Since the latter would require inlining the
type and thus making it impossible to add new properties without
breaking binary compatibility, this patch introduces a fake API that
behaves similar but does not contain the property by value.
Change-Id: Ib9bccd867f0e4e36a520e5583ba348e728284253
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Recently the moc learned that Q_PROPERTY(int x ...) can mean that "x" is
implemented as QProperty and then allows installing bindings, etc. -
this works by scanning the same class' members.
For our own use of QProperty, we need to place the QProperty member
itself into the d-pointer to be able to maintain the ability to add new
properties without breaking binary compatibility. That however means
that moc can't know that a certain property is backed by QProperty - we
don't scan the members of the private class.
As a workaround, this change enables the syntax where the property type
used in Q_PRIVATE_PROPERTY may be wrapped with QProperty<T>.
In addition this patch fixes the compilation of such declared properties
by ensuring the accessor prefix (t->$accessor) is applied also for the
QProperty related meta call variants.
Change-Id: I8fbdc49319048b57f4eb0b65b56daba0459e9598
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
We hope this shall avoid some flaky failures noticed in quick tests,
e.g. tst_QQuickMenu::Material::subMenuPosition(cascading,flip) was
recently seen failing with 3.88e-11 != 0. This required some revision
to test data in the testlib selftest for floats; the resulting
expected output differs in details but not in which tests pass or
fail. QEMU, naturally, made life difficult, requiring special-case
code in the test-driver.
[ChangeLog][QtTestLib][QCOMPARE] QCOMPARE() now treats its values as
equal when qFuzzyIsNull() is true for both of them.
Change-Id: Icc6ad5164b609937eddbe39cc69120f0abf0f3b4
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
QUrl will reject invalid URLs for us, so we don't get normalization. The
original junk should be retrievable, of course.
Change-Id: Ibdc95e9af7bd456a94ecfffd160610f5b2c8e1a2
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
We rely on QDateTime::fromString being proper, so this is not extensive
testing.
Change-Id: Ibdc95e9af7bd456a94ecfffd160610cdac5d62e1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
By QCborValue design, we store the textual representation in ISO format,
equivalent of CBOR tag 0, which isn't allowed to have negative years or
beyond year 10000.
Change-Id: Ibdc95e9af7bd456a94ecfffd16060ccff359c296
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
QDateTime::fromSecsSinceEpoch() multiplies by 1000 but does not check
for overflow. That means we must do so in QCborValue validation. We
can't use mul_overflow<qint64> on 32-bit platforms, so we do a compare-
and-branch there. For 64-bit platforms, we prefer to do the
multiplication with checked overflow, as the common case is that it will
not overflow and we'll need the multiplication anyway.
Change-Id: Ibdc95e9af7bd456a94ecfffd16060cba6f1c86b8
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The linker doesn't have a -rdynamic option. Otherwise the build
fails with
g++.exe error unrecognized command line option '-rdynamic'
Task-number: QTBUG-75578
Change-Id: Ie89a19fd25e90bef14e64d1d98fd973fa0315997
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Also removed add_subdirectory calls for subdirs which no longer exist.
Change-Id: I759f408ca812e1721dde495b0e23feffdeeb9c60
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Use QAbstractButton::click(). Refactor and polish the tests
to use QTRY_VERIFY() instead of qWait(), speeding them up.
Pick-to: 5.15
Task-number: QTBUG-81845
Change-Id: I119bede8143ec1db5f5250517dee38b576d5a8d2
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
... and do a local code cleanup:
- add override
- port from foreach to C++11 ranged for loop
- remove dead code (`gestureEvent` (née event) was already
dereferenced when we check it for null'ness, so it cannot be
nullptr, so the condition is always true).
Pick-to: 5.15
Change-Id: Ica5e34fbe65c95f6573630f188582b90be15c8f7
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Cleaning up those that are trivial to remove because they have direct
replacements.
Change-Id: I4f5c25884a01474fa2db8b369f0d883bd21edd5b
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Those can be trivially removed as they have direct replacements, or
are completely unused.
The migration path for QCursor::bitmap and QCursor::mask is
QBitmap *pb = c.bitmap(); // up to 5.15, warns in 5.15
QBitmap vb = c.bitmap(Qt::ReturnByValue); // from 5.15, works in 6
QBitmap b = c.bitmap(); // from 6.0 on
Change-Id: I3b3acd1c7f09c4c8414e98b3ce11986f1ecd5eda
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
For kinetic wheel events, Qt tries to make sure that all events in the
stream go to the widget that accepted the first wheel event.
It did so by directing all events from the stream to the widget from
which the spontaneous event was returned as accepted.
However, that widget might have passed the event on to some other
widgets; e.g QScrollArea forwards wheel events from the viewport to the
relevant scroll bar. The event might then have come back accepted only
because parent propagation kicked in (the scrollbar might not accept
the event, so the parents get a chance, and some parent's scrollbar
ultimately accepts the event).
In this scenario, the wheel widget would be the viewport under the
mouse, when it should have been the scrollbar of the parent. The
next events from the stream were then delivered to a widget that didn't
scroll; and parent propagation is not (and should not be) implemented
for the case where Qt has a wheel widget.
Instead, make the first widget that accepts any initial wheel event
the wheel widget, even if the event was not spontaneous. With this
change, all events from the stream are delivered to the widget that
actually handled the event. That has the effect that ie. a viewport
of a scroll area only gets the first event; all following events are
delivered directly to the scrollbar.
The test case added simulates the different scenarios - nesting of
scroll areas, classic wheel events and a stream of kinetic wheel
events.
[ChangeLog][QtWidgets][QApplication] Wheel events from a device that
creates an event stream are correctly delivered to the widget that
accepts the first wheel event in the stream.
Change-Id: I5ebfc7789b5c32ebc8d881686f450fa05ec92cfe
Fixes: QTBUG-79102
Pick-to: 5.15
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
When hiding a popup by clicking outside of its area, a window is closed.
Depending on the platform specific implementation details, this can
result in multiple calls to QWidgetPrivate::setVisible(false). The first
one from the handling of the close event in QWidgetWindow::event; the
second from the destruction of the window in QWindow::event.
Since the first call already sets the Qt::WA_WState_Hidden flag before
calling QWidgetPrivate::hide_helper, we can test if the flag is set
and skip the second call if it is.
The included test does not reproduce the issue, as that issue only
reproduces if the close event is generated by the mouse event handling
in the Cocoa platform plugin (which doesn't call QWidget::close, but
rather sends a native close event to the platform window). However, it
verifies that the fix doesn't introduce any regressions.
Change-Id: Id0eda9326a8adf0cc1f6a3840f9ac0b635ab39a1
Fixes: QTBUG-79134
Pick-to: 5.15
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
In the declaration, leave them in as comments so that the gaps in the
numbering of the values doesn't create confusion.
Change-Id: I92ff299416896c471e7c7d80b988cd4642b6b756
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
[ChangeLog][QtCore][QString] Now supports appending, prepending
and inserting QStringViews.
Change-Id: I7538c050c67590f27d91443eda0b94a4b80b62f2
Reviewed-by: Anton Kudryavtsev <antkudr@mail.ru>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Rather than have a fontChanged() signal which can be connected to for
tracking when the application font has changed, then it is better to
use the event that is sent to all windows and the application itself.
That way it is easy for a window/widget or item that cares about the
change to the application font to catch it in the event() function.
[ChangeLog][QtGui][QGuiApplication] Deprecated fontChanged() signal in
favor of QEvent::ApplicationFontChanged.
Change-Id: Iae8e832238fc85e385a52305bc04f16e597454b0
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Added asserts to insure the invariant declared in the documentation.
Canceled future object is not valid, hence we don't need to handle
this case separately.
Fixes: QTBUG-83389
Change-Id: Ib0653ef40cd3135574a91740e4ce2c6dc4da8a71
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
This follows up on commits 3e7463411e and
947883141d.
The changing of the pointer of QApplicationPrivate does not transfer
focus properly. It updates the pointer, but it doesn't deliver events
or update the widget hierarchy's focus chain. The result is that
multiple line edits might show a blinking cursor.
Instead, use QWidget::setFocus when the focus proxy has changed while
it had focus, and pass OtherFocusReason rather than NoFocusReason.
Add a basic test for QWidget::focusProxy, which exercises this code
path and verifies that pointers are consistent when focus changes as
a side effect of modifying the focusProxy.
Change-Id: I15a4d868bab2b590cfe4a1daa6a3c8cebc9c9ca2
Fixes: QTBUG-83720
Fixes: QTBUG-79707
Pick-to: 5.15
Reviewed-by: David Faure <david.faure@kdab.com>
The internal removeAt(index) method was implemented as taking cbor
indexes directly, in contrast to the other ...At(index) methods.
Fixes: QTBUG-83695
Change-Id: I16597eb6db1cf71e1585c041caa81bf8f7a75303
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Also extend autotesting, both for rendering into a given mip level
and for rendering into a given cubemap face.
Change-Id: Ida94b71150477ceb50a3b5616d8b7be13174558b
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
When a date-time was parsed from a string, the result was equal (as a
date-time) to the correct value, but had (at least in some cases) the
wrong spec, where it should have had a spec reflecting the zone
specifier parsed.
The time-spec imposed for the benefit of QDateTimeEdit is now moved
from QDateTimeParser to QDateTimeEditPrivate, which takes over
responsibility for imposing it. QDateTimeParser assumes Qt::LocalTime
in member functions (where applicable) and uses the time-spec parsed
from the string when constructing the date-time.
QDateTime::fromString() and QLocale::toDateTime() are updated to
use the full QDateTime returned by QDateTimeParser.
Fixes: QTBUG-83075
Done-With: Edward Welbourne <edward.welbourne@qt.io>
Change-Id: I8b79add2c7fc13a200e1252d48dbfa70b36757bf
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Ensure they are handled as enumerations in QMetaType.
This is required for handling QFlag-type properties in Qt Designer
Fixes: QTBUG-83689
Change-Id: Ifbfb5c5b5cd34fce462e299505d063e22e725c2e
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
When converting a text document to HTML, always convert it to utf-8,
as required by the HTML standard. This also means that we remove
the optional encoding parameter.
Change-Id: I0bd2fc9df2d06734e1c5b8053b964fbfbb6881e1
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
After adding new tests, the source code of tst_concurrentmap
grows rapidly with copy/paste/replace pattern.
This intoduce the issue with maintaining the test.
Get rid of separate functions for testing blocking versions
of mapped/mappedReduced and test blocking versions in parallel
with non-blocking flavors. So again, get rid of duplicated code.
Task-number: QTBUG-83258
Change-Id: I4cac6f4e1bfd1b50363dc094610717674aebf2af
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Test the case where reduce function of the form:
V function(T &result, const U &intermediate)
has T and U types different. Make use of
numberSumReduce function and corresponding functor
that reduce Number class object to the result of int type.
Fixes: QTBUG-83258
Change-Id: I194d290988b48e7bca91228c0cd5d39efd1b4712
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Move tests with initial value next to their original version.
Join new lambda tests into a common functions
testing all possible 16 combinations of
functor / function / member / lambda, as they
test in fact the same function. There is no need
to distinguish lambda case over other cases.
This helps in test readability and maintenance.
Add missing tests for lambdas with a combination
of initial value.
Task-number: QTBUG-83258
Change-Id: I7a3d2bf87384d7feecffaaf39281cd626955b0a4
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
The CommonMark spec shows that it's not necessary to have a space
between the code fence and the language string:
https://spec.commonmark.org/0.29/#example-112
This also avoids a needless trailing space after a code fence that
does not include a language string.
Change-Id: I2addd38a196045a7442150760b73269bfe4ffb22
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
The end of a code block nested in a list item is now detected;
and if the text of the list item continues after the code block,
it continues to be indented.
Code blocks should never be word-wrapped.
Fixes: QTBUG-80603
Change-Id: I4427f8b1d4807d819616f5cb971e2d006170d9be
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Various benchmarks were still using the deprecated timing API.
One didn't even *use* the timer it implemented this way.
One was just using start as a short-hand for assigning to currentTime().
Change-Id: If406d0fb606e454fec056f386bcd0aa6726ee96e
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Use the move-only versions of result reporting and getting operations,
if the type of QFuture is not copyable.
Task-number: QTBUG-81941
Change-Id: Ic9fa978380e2c24e190e68d974051a650b0e5571
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Vitaly Fanaskov <vitaly.fanaskov@qt.io>
Cleaning up those that are trivial to remove because they have direct
replacements.
Change-Id: Ie9fecd8c4822ed1a8f378b210cc4c4d9a10f7e36
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Cleaning up those that are trivial to remove because they have direct
replacements.
The QLabel changes to the pixmap/picture getters provide the following
migration path:
QPixmap *ppix = l->pixmap(); // up to 5.15, warns in 5.15
QPixmap pval = l->pixmap(Qt::ReturnByValue); // new in 5.15, works in 6
QPixmap pixmap = l->pixmap(); // from Qt 6 on
The overload with argument can be deprecated after the first LTS or
so.
Change-Id: I8494ceeea55b2aeda0bd340640ad95cb7c91f7d6
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
This is a follow-up to commit ed2b110b6a
to fix indexing errors. Added the test that should have accompanied
that commit, which found some bugs, and refined the Indian number
formatting test (on which it's based).
Make variable i in the loops that insert grouping characters in a
number be consistently a *character* offset - which, when each digit
is a surrogate pair, isn't the same as an index into the
QString. Apply the needed scaling when indexing with it, not when
setting it or decrementing it. Don't assume the separator has the same
width as a digit.
Differences in index no longer give the number of digits between two
points in a string, so actively track how many digits we've seen in a
group when converting a numeric string to the C locale. Partially
cleaned up the code for that in the process (more shall follow when I
sort out digit grouping properly, without special-casing India).
Change-Id: I13d0f24efa26e599dfefb5733e062088fa56d375
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The QT_OPENGL_ES* macros are leftovers from an earlier,
ad hoc configuration system, which has since been
replaced by QT_CONFIG. To clean things up in Qt 6,
we use the new way instead.
Task-number: QTBUG-83467
Change-Id: I578dc7695bff9d5ee303b22e44f60fee22fe0c28
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
And remove a test failure when compiling with asan enabled.
Change-Id: I2b8e676665572adcbbac6a910983d5b209bf6d23
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
Instead of requiring the implementation to do the compare dance, let's
do this in the library. This reduces the amount of duplicated code
slightly and makes it easier to generate binding code from qml files.
Change-Id: Ia3b16cf9769e74d076b669efe4119ab84af3cdf0
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The pre-existing overload passes an int, but this can mean the
descriptor gets truncated in compilations where the descriptor
is 64-bit.
The old overload with int is visible when querying the metaobject system
so string-based connects still work as before, and connecting to it will
produce a deprecation warning in the output.
At the same time the PMF-based connect will, on recompile, pick the
QSocketDescriptor overload. As an added improvement it also comes with
the notification type, removing the need for separate slots where the
code would be mostly shared anyway.
The QSocketDescriptor type can be implicitly converted to and from
qintptr to ensure existing code still compiles. It can also be
constructed from Qt::HANDLE on Windows.
In this same patch I also update the existing string-based connects in
this module, which then includes updating the parameters for some slots
as well.
[ChangeLog][QtCore][QSocketNotifier] Added
QSocketNotifier::activated(QSocketDescriptor, QSocketNotifier::Type).
This replaces the activated(int) signal which in 64-bit environments
could truncate the socket descriptor. If you use "activated" with the
string-based connect() then you need to update the parameter type of the
signal and slot if it had one. If you use it with the pointer to member
function based connect() then all you need to do is update your slot's
parameter type if it has one. If you need to compile your source code
with multiple versions of Qt then connect() to this function using
pointer to member function and update the slot's parameter type if
needed.
Task-number: QTBUG-70441
Change-Id: Ic43d6bc4c5bcb4040867b2ffad8d36fb01eed8af
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
And a few cleanups of out-dated comments and dead code.
Change-Id: I59c6b9129a21b8953626cb63c3ebbf9b6c49a657
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
The context—which lives in QtGui—now knows nothing about versioned functions.
This changes the public API for getting version functions for a context.
[ChangeLog][QtGui][OpenGL] QOpenGLContext::versionFunctions() has been removed.
QOpenGLVersionFunctionsFactory::get() from the QtOpenGL module should be used
instead.
Previously one would call
context->versionFunctions<QOpenGLFunctions_4_0_Core>();
Which now becomes
QOpenGLVersionFunctionsFactory::get<QOpenGLFunctions_4_0_Core>(context);
The rest of the API should be identical.
Since glgen no longer compiles, and the links to its input (gl.spec and gl.tm)
are dead, I've edited the previously generated files manually. If glgen is
fixed, it should be quite easy to make it generate the new way.
Task-number: QTBUG-74409
Change-Id: I800527e0af16a79005b276eeb74417770193c62f
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Use P0608's trick to detect convertibility without narrowing;
and now that we can depend on C++17, use its features.
First, this moves the burden of detecting a narrowing conversion on
the compiler, rather than us maintaining a complicated series
of checks. Of course, this exposes
* bugs in compilers (e.g. GCC < 9 thinks that float->bool is not
narrowing;
* behavior still not (widely) implemented (pointer to bool
conversions are narrowing, P1957);
* interesting compiler choices, e.g. GCC 9 thinks that unscoped
enumerations are non-narrowing convertible to a datatype big
enum to contain all the _enumerators_, even if the underlying
type of the enum (and/or its sizeof()) is wider than the target
datatype.
Second, it allows to detect conversions that have a narrowing
conversion as an intermediate step. Given a type like
struct Bad { operator double() const; };
then an object of type Bad is implictly convertible to a type
like int via a narrowing conversion. Therefore, a connection
is possible between a signal carrying a Bad and a slot accepting
an int. We can now detect and block this.
Tests regarding scoped enumerations have been dropped,
for the simple reason that a scoped enumeration is not
implictly convertible to an integral type, so we don't have
that detection (it would constantly fail). Scoped enumerations
do not take part in narrowing conversions anyhow, cf. [dcl.init.list].
[ChangeLog][QtCore][QObject] The detection of narrowing conversions
when calling QObject::connect() when
QT_NO_NARROWING_CONVERSIONS_IN_CONNECT now takes also
into account user-defined implicit conversions that undergo
through a narrowing conversion.
Change-Id: Ie09d59203fe6283378b36dfbc54de1d58098ef51
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Map setFilterWildcard() and setFilterFixedString() to now use
QRegularExpression.
Change-Id: I2dff2015234decb2badfd306975dcff8553cdd7f
Reviewed-by: Alex Blasche <alexander.blasche@qt.io>
Following the deprecation of the signal overloads, the remaining signals
did not provide equivalent functionality for connecting a slot expecting
an integer. The mapping from QAbstractButton* to the ID is comparatively
cumbersome to do in the connected slot.
Add uniquely named signals that emit the ID of the button directly.
[ChangeLog][QtWidgets][QButtonGroup] Added signals
idClicked/Pressed/Released/Toggled that replace the deprecated signal
overloads.
Change-Id: I77215e4f815c4fb7dd6326e1f431230e6601e8f8
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
* support lambda expressions
* remove the need to specify result_type
* use std::invoke to apply map|filter function
* remove usage of FunctionWrapper* and createFunctionWrapper
Task-number: QTBUG-33735
Task-number: QTBUG-82646
Change-Id: Ibcbe4278f0742c29182bd506081db0abb516f85f
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
The double-swap technique I used was flawed and broke on
self-assignment. What I had meant to use was the move-and-swap
technique. Thanks to Peppe for pointing it out.
This also fixes a compiler bug in the Green Hills compiler. It was
finding the wrong "swap" function in qSwap:
using std::swap;
swap(value1, value2);
It's supposed to find swap(QCborValue &, QCborValue &) due to argument-
dependent lookup. It's instead finding std::swap<QCborValue>, which
recurses.
Fixes: QTBUG-83390
Change-Id: Ibdc95e9af7bd456a94ecfffd1603e1bee90cd107
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Some test cases are sensitive to the exact ordering inside
QHash, and need adjustments when we change QHash or the
hashing functions.
Some rcc tests now also need 32bit specific data, as the hashing
functions for 32 and 64 bit are different now (as we use size_t).
Change-Id: Ieab01cd55b1288336493b13c41d27e42a008bdd9
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The old implementation was either using CRC32 on modern processors
or a trivial, but rather slow implementation.
We can't continue with CRC32, as that implementation can only
give us 32bit hashes, where we now need to support 64bit in Qt 6.
Change the implementation to use MurmurHash, as public domain
implementation that is both very fast and leads to well distributed hashes.
This hash function is about as fast as the SSE optimized CRC32 implementation
but works everywhere and gives us 64 bit hash values.
Here are some numbers (time for 10M hashes):
14 char 16 char
QByteArray QString float
old qHash (non CRC32) 127ms 134ms 48ms
old qHash (using SSE CRC32 instructions 60ms 62ms 46ms
new qHash 52ms 43ms 46ms
Unfortunately MurmurHash is not safe against hash table DoS attacks, as
potential hash collisions are indepenent of the seed. This will get
addressed in followup commit, where we use SipHash or an SSE optimized
AES based hashing algorithm that does not have those issues.
Change-Id: I4fbc0ac299215b6db78c7a0a2a1d7689b0ea848b
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
At the same time use the opportunity to refactor the
insertion code inside the implementation of QHash to
avoid copy and move constructors as much as possible
and always construct nodes in place.
Change-Id: I951b4cf2c77a17f7db825c6a776aae38c2662d23
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
This is required, so that QHash and QSet can hold more
than 2^32 items on 64 bit platforms.
The actual hashing functions for strings are still 32bit, this will
be changed in a follow-up commit.
Change-Id: I4372125252486075ff3a0b45ecfa818359fe103b
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Make use of the new features available in QHash and do a more
performant implementation than the old one.
Change-Id: Ie74b3cdcc9871cd241aca205672093dc395d04a7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
A brand new QHash implementation using a faster and more memory efficient data
structure than the old QHash.
A new implementation for QHash. Instead of a node based approach as the old
QHash, this implementation now uses a two stage lookup table. The total
amount of buckets in the table are divided into spans of 128 entries.
Inside each span, we use an array of chars to index into a storage area
for the span.
The storage area for each span is a simple array, that gets (re-)allocated
with size increments of 16 items. This gives an average memory overhead of
8*sizeof(struct{ Key; Value; }) + 128*sizeof(char) + 16 for each span.
To give good performance and avoid too many collisions, the array keeps its
load factor between .25 and .5 (and grows and rehashes if the load factor goes
above .5).
This design allows us to keep the memory overhead of the Hash very small, while
at the same time giving very good performance. The calculated overhead for a
QHash<int, int> comes to 1.7-3.3 bytes per entry and to 2.2-4.3 bytes for
a QHash<ptr, ptr>.
The new implementation also completely splits the QHash and QMultiHash classes.
One behavioral change to note is that the new QHash implementation will not
provide stable references to nodes in the hash when the table needs to grow.
Benchmarking using https://github.com/Tessil/hash-table-shootout shows
very nice performance compared to many different hash table implementation.
Numbers shown below are for a hash<int64, int64> with 1 million entries. These
numbers scale nicely (mostly in a linear fashion with some variation due to
varying load factors) to smaller and larger tables. All numbers are in seconds,
measured with gcc on Linux:
Hash table random random random random reads full
insertion insertion full full after iteration
(reserved) deletes reads deletes
------------------------------------------------------------------------------
std::unordered_map 0,3842 0,1969 0,4511 0,1300 0,1169 0,0708
google::dense_hash_map 0,1091 0,0846 0,0550 0,0452 0,0754 0,0160
google::sparse_hash_map 0,2888 0,1582 0,0948 0,1020 0,1348 0,0112
tsl::sparse_map 0,1487 0,1013 0,0735 0,0448 0,0505 0,0042
old QHash 0,2886 0,1798 0,5065 0,0840 0,0717 0,1387
new QHash 0,0940 0,0714 0,1494 0,0579 0,0449 0,0146
Numbers for hash<std::string, int64>, with the string having 15 characters:
Hash table random random random random reads
insertion insertion full full after
(reserved) deletes reads deletes
--------------------------------------------------------------------
std::unordered_map 0,4993 0,2563 0,5515 0,2950 0,2153
google::dense_hash_map 0,2691 0,1870 0,1547 0,1125 0,1622
google::sparse_hash_map 0,6979 0,3304 0,1884 0,1822 0,2122
tsl::sparse_map 0,4066 0,2586 0,1929 0,1146 0,1095
old QHash 0,3236 0,2064 0,5986 0,2115 0,1666
new QHash 0,2119 0,1652 0,2390 0,1378 0,0965
Memory usage numbers (in MB for a table with 1M entries) also look very nice:
Hash table Key int64 std::string (15 chars)
Value int64 int64
---------------------------------------------------------
std::unordered_map 44.63 75.35
google::dense_hash_map 32.32 80,60
google::sparse_hash_map 18.08 44.21
tsl::sparse_map 20.44 45,93
old QHash 53.95 69,16
new QHash 23.23 51,32
Fixes: QTBUG-80311
Change-Id: I5679734144bc9bca2102acbe725fcc2fa89f0dff
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add a new QRangeCollection type to store and manage
multiple page ranges. This moves out the parser and validator
logic from the platform dependent (UNIX) dialog and makes it
publicly available from QPrinter.
This improves the usability of QPrinter in those applications
which doesn't use print dialog to configure printer.
(e.g.: QTextDocument, QWebEnginePage)
Change-Id: I0be5a8a64781c411f83b96a24f216605a84958e5
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>