If we receive compressed data we will decompress it as it comes in,
in some cases the final byte we receive doesn't actually contribute
to the final output. If this byte is handled by itself then, when
combined with QNetworkReply's signal compression, we ended up not
emitting the readyRead signal for any of the data we received while
compressing the signals.
So, instead of relying on checking difference in bytes downloaded
for each batch of data received we keep track of how much data
we had received the last time we emitted readyRead.
Pick-to: 6.3 6.4 6.4.0
Fixes: QTBUG-106354
Change-Id: I4777be29b46bf0de968667e9de9d975c735ac6e6
Reviewed-by: Konrad Kujawa <konrad.kujawa@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Because it's convenient to have it build automatically.
Pick-to: 6.2 6.3 6.4
Change-Id: Iecd5bd1f9fe518f683b8ee9eb6529e48ce76fb33
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Some tests in corelib/kernel need threading support, but they are not
guarded against compilation if Qt is built without threading.
Such tests have been disabled in this case.
Change-Id: I2f5dc9582f2a59b6af2a9e56638b045dca06193d
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Test was not really good to start with - it was assuming the presence of
particular ciphersuites. Furthermore, it was ignoring the fact that
TLS 1.2 and TLS 1.3 set ciphersuites differently in OpenSSL.
Pick-to: 6.4 6.3 6.2 5.15
Task-number: QTBUG-106018
Task-number: QTBUG-95123
Change-Id: I6c8ba20154cdeb9275878462ab945729d6c82ecc
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The collection of translations available to us need not have anything
to do with whether CLDR has matching data, so preserve the system UI
language list's entries as they are, rather than forcing them through
the QLocale constructor's exercise of likely sub-tag rules.
Instead, simply parse the given locale tags to QLocaleId instances and
use these in the likely-subtag processing to determine what other
entries to add to the list in addition to those supplied by the
operating system. Since going via QLocale did usually supply a
territory, that was included in the BCP 47 name, it's now possible for
the given entry to lack the language_territory name, so be sure to add
that if missing.
This incidentally reduces heap traffic and saves a fair deal of hidden
likely-subtag processing in calls to the constructor and bcp47Name().
Expand testing of QLocale::uiLanguages(), both plain and system. In
the process, cross-link the two closely-related tests, move a comment
on one's _data() to the other's, where it really belongs, and add
reporting of the actual lists on failure. Enable MySystemLocale to
remember the requested locale's ID, before likely sub-tag processing,
so that we can make query() report results for language, script and
territory as requested, to ensure the fake system locale really does
match what was requested. The new german-britain test failed without
it, because there is no de-GB locale in CLDR.
Task-number: QTBUG-99531
Change-Id: Ide041577772c442a4413e3b9a590e11140c48f49
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Have QSystemLocale manage a stack, so that tests can install an
over-ride for the actual system-specific one reliably and restore the
system-specific one when finished. Leave the QSystemLocaleSingleton
out of the stack, all the same. In the process, mark the QDoc comments
for QSystemLocale all as \internal, since this is not public API.
Change-Id: I8faed49780215e42f32be10cf936c32bb46105bf
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
When the system locale is en_DE, macOS seems to think we should use
en_GB as the right translation. While that probably is a sensible
choice in the absence of an en_DE translation, we should definitely
use the en_DE translation if available, especially if en_GB isn't
available (which lead to a fall-back to de_DE, given later entries in
macOS's list). So prepend the system locale's own pcp47Name() if it
(isn't the C locale and) is missing from what we would otherwise have
used for uiLanguages(), after likely sub-tag perturbations.
Add a test simulating (some approximation to) what macOS was doing
that would have caught this case; and add a scope-guard reporter to
the test to report what shows up when lists don't match.
Fixes: QTBUG-104930
Pick-to: 6.4 6.4.0 6.3 6.2 5.15
Change-Id: I116234708067e1717d9157aebc84da76e04a9f38
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Certain masks are not supported outside 32-bit x86, and will assert on
x64.
Pick-to: 6.2 6.3 6.4
Fixes: QTBUG-106000
Change-Id: Ic9f58e5a19c1db3309edeb5ec529e7a78c929665
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
We only want to enable writing BOM if we have _not_ started
writing.
Fixes: QTBUG-106279
Pick-to: 6.2 6.3.2 6.4 6.4.0
Change-Id: Ibcbc101b931615fddb2507f01307bf9619772d7b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Replace the deprecated operator*() calls with various
overloads of QMatrix4x4::map()
Add a separate test for deprecated API and guard it with
QT_DEPRECATED_SINCE checks.
Task-number: QTBUG-104858
Change-Id: Ief2e03198696382dc626f01b209614fe320e70b2
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
qdbusxml2cpp has a -m/--moc option. Change and modify the behavior
such that
-p foo -m includes moc_foo.cpp in the generated .cpp
-p :foo.cpp includes moc_foo.cpp in the generated .cpp
-p foo.h:foo.cpp includes moc_foo.cpp in the generated .cpp
Change the Qt6DbusMacros.cmake file accordingly.
[ChangeLog][qdbusxml2cpp] The -m/--moc option now generates
idiomatic moc file names (moc_base.cpp for headers, base.moc
for implementation files)(was: always base.moc). Build systems
using workarounds for the non-idiomatic naming of moc files
used by qdbusxml2cpp in the past can now drop these workarounds
for Qt versions >= 6.5.
Fixes: QTBUG-103313
Change-Id: I754b1b276f130cb8645166470e1b457a676590f7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
OpenSSL 3 by default disables TLS < 1.2 in the conf files. We're not
going to work against that, so we bump the version of TLS used.
Keep the client-side TLS 1.0 Or Later just to keep testing that this
part works.
Task-number: QTBUG-95123
Task-number: QTBUG-106018
Pick-to: 6.4 6.3 6.2 5.15
Change-Id: Ia6cb10495875de232d69a886832ae74c5f5ac15f
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
When TopMargin is set in TextBlock and the mouse click position is
between Margin (the mouse is not on the textrect), the cursor will
usually jump to the end. So topMargin should be considered when
hitTest() calculates coordinates.
Fixes: QTBUG-91774
Pick-to: 6.4 6.3 6.2
Change-Id: I231377263855b9cd7152684203fc4ed2e9299bb9
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Tang Haixiang <tanghaixiang@uniontech.com>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
On Wayland there is a special case when decorations are enabled,
where we post an initial update request to ensure widgets are
redrawn after the buffer is recreated (see
af7b60ade5c4be81cbc58eb18307c017d5594071 in qtwayland). To
make sure the test behaves the same on Wayland as on other
platforms, we disable client-side decorations when running
these tests.
The flag is primarily required for the requestUpdate() test
at the moment, but it is only queried once per application
run, so it has to be set before any tests are run.
Pick-to: 6.4
Fixes: QTBUG-100889
Change-Id: Ica6d744083ecae4a3722b6d04b956e7615f0dfe5
Reviewed-by: Inho Lee <inho.lee@qt.io>
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
The initial problem was in server not starting encryption, because
its certificate was rejected by OpenSSL (v3) saying 'weak md'.
After the certificate was replaced, we got another problem - due
to the fixed TLS v.1 protocol for both client and server - handshake
is interrupted early with 'tls v1 alert' and error message
saying about not found signature algorithms. Don't fix the protocol
version, use proper QSsl::SecureProtocols (default).
Pick-to: 6.4 6.3 6.2 5.15
Task-number: QTBUG-106018
Task-number: QTBUG-95123
Change-Id: I0f22f5304f4405933ad511195bc61c640518e7d8
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The certificate was using SHA-1 for signing its certificate, which is
considered 'not good enough' by some TLS libraries.
Regenerate it with SHA-512 and a larger RSA key.
Also include the files needed to generate it.
Pick-to: 6.4 6.3 6.2 5.15
Task-number: QTBUG-106018
Task-number: QTBUG-95123
Change-Id: I535e047d540e663c8cec334695196044173949fc
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Batching the tests leads to one of the tests tst_qmetaobject and
tst_qmetaobject_compat not being registered in the batch. Attempts to
batch those together fail as batch test name is defined per-source,
which, in this rare case, is the same across the two targets.
Change-Id: I356931feabc004c39ba0b6863b5f64e06d739a58
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
OpenSSL v3 by default sets SECLEVEL = 2 unless overridden in conf files.
SECLEVEL >= 1 means TLS <= 1.1 is disabled. Remove cases that were
previously expected to work. Arguably, we should not test deprecated
protocols at all, but the cases expecting a failure still work
and logically correct.
Task-number: QTBUG-95123
Task-number: QTBUG-106018
Pick-to: 6.4 6.3 6.2 5.15
Change-Id: Ifa59fa902b433338dbed7abf086cfdef9470384c
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
It was an old test written in a very low-level way, which perhaps is ok
to be independent of testlib in a few tests; OTOH, it was blacklisted
on a couple of platforms. Perhaps doing touch events the standard way
could be more stable.
While we're at it:
- verify that the touch events are accepted, and thus verify the new
bool return value from commit()
- implement paintEvent() to help understand the layout, and touchpoint
locations
- remove repeated QCOMPARE lines
- skip the test if window positioning fails
- try to un-blacklist it, on the assumption that window positioning
failure was the reason
Task-number: QTBUG-87025
Task-number: QTBUG-104656
Pick-to: 6.4
Change-Id: Ie22eb24abf95cd849990a56212be87d06ce8e574
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Doris Verria <doris.verria@qt.io>
[ChangeLog][QtCore][Meta Object] QMetaMethod::invoke(),
QMetaObject::invokeMethod(), and QMetaObject::newInstance() are no
longer limited to 10 arguments.
[ChangeLog][QtCore][Meta Object] The use of the Q_ARG macro is no longer
necessary when using QMetaMethod::invoke(), QMetaObject::invokeMethod(),
and QMetaObject::newInstance(). Types may now be passed
directly. Similarly, Q_RETURN_ARG can be replaced by the free function
qReturnArg().
[ChangeLog][Potentially Source-Incompatible Changes]
QMetaMethod::invoke(), QMetaObject::invokeMethod(), and
QMetaObject::newInstance() no longer support passing forward-declared
types in the argument list (it was possible to pass them by
const-ref). From Qt 6.5 onwards, all types in the argument list must be
fully defined.
[ChangeLog][Potentially Source-Incompatible Changes] Attempting to use
the internal types QArgument, QReturnArgument, QGenericArgument, or
QGenericReturnArgument directly with QMetaMethod::invoke(),
QMetaObject::invokeMethod() or QMetaObject::newInstance() may fail to
compile. Those are internal types that were never meant to be used
directly and will be removed in Qt 7. If really necessary, ensure all
arguments passed to those functions are directly using those classes and
not mixed with Q_ARG and Q_RETURN_ARG. Implementations of bindings to
other languages should contact the Qt development mailing list to
discuss options.
Change-Id: I36b24183fbd041179f2ffffd1701e3e8e47e0fba
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The doc of QMetaEnum::valueToKey() says to use ::valueToKeys() instead
for flag types.
Pick-to: 6.4
Change-Id: I48e5ba47324137f2ce2710f1d876e93e7c562e9f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
This includes QDBusReply, QProperty, and QStringBuilder expressions.
The new constructor subsumes the QStringBuilder case without requiring
jumping though hoops to delay the definition of the ctor the way we
had to for the explicit QStringBuilder constructor, so remove the
explicit QStringBuilder one again.
[ChangeLog][QtCore][QAnyStringView] Can now be constructed from
anything that implicitly converts to either QString or QByteArray.
Fixes: QTBUG-105389
Change-Id: I0e584dd3e20d591381609a3329ef47cec7356ecc
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Since qVersion() might be called also from C code, disable the parts of
qlibraryinfo.h that are relevant only for C++ code if __cplusplus is not
defined.
[ChangeLog][Potentially Source-Incompatible Changes] qVersion() is
moved from qglobal.h to qlibraryinfo.h, '#include <QtCore/QLibraryInfo>'
needs to be added where it's used.
Task-number: QTBUG-99313
Change-Id: I3363ef3fa4073114e5151cb3a2a1e8282ad42a4d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
And include qcore_mac_p.h where needed.
Task-number: QTBUG-99313
Change-Id: Idb1b005f1b5938e8cf329ae06ffaf0d249874db2
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Since we had code to default to zstd as the default algorithm instead of
"Best", we ended up not compressing anything.
[ChangeLog][rcc] Fixed a bug that caused rcc not to compress files with
any compression algorithm if the --no-zstd option was present.
Pick-to: 6.2 6.3 6.4
Fixes: QTBUG-106012
Change-Id: Ic6547f8247454b47baa8fffd170fddae429f82d2
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
With OpenSSL 3.0.5 we see X509_print giving us the text representation
not matching any of previous ones - mostly a question of spaces
and formatting. Let's add a proper version to compare against and
remove useless 0.9.8, 1.0.0, 1.0.1, - we don't support them anymore
and thus QSslCertificate::toText() will never give a match.
Pick-to: 6.4 6.3 6.2 5.15
Task-number: QTBUG-106017
Change-Id: Ida11321c23e74c7313acc682237ab78408cf2803
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
leaf.p12 is using RC2 for encryption and it's disabled by default
in openssl v3.
Pick-to: 6.4 6.3 6.2 5.15
Fixes: QTBUG-106017
Change-Id: I4edd0d29506d1e50b2b618b6a00cceeb4b156204
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The test expects the helper libraries to contain the .dylib suffix
rather than .so.
Replace glob copying of the libraries (which depends on the underlying
shell) with manual copy calls.
This also ensures the libraries don't contain a _debug postfix in the
file name even in a debug build, which would break the tests.
Amends f8c1909320
Amends 1dff26dd95
Change-Id: I20361c33c4a1b9dd4b5273fcdb8cc79c9f266327
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
Because this version does not seem to like MD5 which old cert was using.
Fixes: QTBUG-106019
Pick-to: 6.4 6.3 6.2 5.15
Change-Id: Id354d1e32acb0528685b2788ee9c7326c43397bd
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
[ChangeLog][QtCore][Meta Objects] The QMetaObject::invokeMethod() taking
a method name by string, QMetaObject::newInstance(), and
QMetaMethod::invoke() now support more than 10 arguments.
[ChangeLog][QtCore][Meta Objects] The use of the Q_ARG and Q_RETURN_ARG
macros is now optional with QMetaObject::invokeMethod(),
QMetaObject::newInstance(), and QMetaMethod::invoke(): the type name
will be obtained from the C++ type (the same as QMetaType). The function
qReturnArg() can be used in place of the Q_RETURN_ARG macro. The macros
are still useful in rare conditions where the type was typedef'ed from
its original name.
Change-Id: I36b24183fbd041179f2ffffd17022a2b48c7639b
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
For the case that a newly created and registered accessible
interface gets removed again from the cache before another one
gets registered, the next registered interface was
previously assigned the same "unique ID" again, which e.g. breaks
assistive technology when using caching
with AT-SPI, since that relies on the assumption
that the ID is actually unique for each object.
(But here, the new object was using the same object path
as the old one, so data from the old object would be
used for the new one.)
To prevent that from happening, increment the
counter for the next ID to try at the end of
QAccessibleCache::acquireId, so the next time
the method gets called, it doesn't try again
whether the same ID as used previously is
available again.
For consistency, also rename the variable used
for the counter from lastUsedId to nextId.
This also adds a corresponding test case.
Fixes: QTBUG-105962
Pick-to: 6.4
Change-Id: Iddf4f3b35c57895bcfbb623a5377edf8344ab6c2
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
The paletteChanged() and fontChanged() signals were deprecated
in Qt 6.0, but the test was still using them unconditionally.
This patch guards the usage of the deprecated signals with
the usual QT_DEPRECATED_SINCE(6, 0) check, so that the test
can be built and run with QT_DISABLE_DEPRECATED_UP_TO >= 0x060000
This commit amends 68ea9c0227
Task-number: QTBUG-104858
Change-Id: Idb2da6d91afcdb664f325f23ec625947c9a7fac0
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
The patch includes the following replacements:
* QMetaType::type("name") -> QMetaType::fromType<Type>().id()
* QMetaProperty::type() -> QMetaProperty::typeId()
* QVariant::Type -> QMetaType::Type
* qRegisterMetaType<T>("name") -> qResigeterMetaType<T>()
* The static QMetaType::{load,save} methods are replaced with
non-static versions
* Replace QCOMPARE(property.type(), QVariant::UserType) with
QCOMPARE_GT(property.typeId(), QMetaType::User), because the
deprecated type() method was treating each custom type
(id >= QVariant::UserType) as QVariant::UserType, while the
typeId() method simply returns the actual id.
As a drive-by: remove unneeded QMetaType registration tests
as we have tst_QMetaType to check it.
Task-number: QTBUG-104858
Change-Id: Ia08e002efdf07ff83366a5193164dba96a956f9a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The most common changes are:
* Replace QMetaType::type("name") with QMetaType::fromName("name").id()
or QMetaType::fromType<Type>().id()
* QMetaType::typeName(int) -> QMetaType(int).name()
* QMetaType::typeFlags(int) -> QMetaType(int).flags()
* QMetaType::metaObjectForType(int) -> QMetaType(int).metaObject()
* The static QMetaType::{load,save} methods are replaced with
non-static versions
* The static QMetaType::{create,destroy,construct, destruct} methods
are guarded by QT_DEPRECATED_SINCE calls. The tests are also
extended with non-static calls where they were missing. Fixed
potential memory-leaks in these tests.
Add separate unit-tests for deprecated APIs and guard them with
QT_DEPRECATED_SINCE
As a drive-by: use nullptr instead of 0 in some places
Task-number: QTBUG-104858
Change-Id: I4b0cdd29bc197c186b835002372240aae3098c33
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use consistent defines across both tests in the directory
Change-Id: Idd84603135c8c164d09ae41d5faa7ca3a5fc7c3c
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Since Qt 6.0, QMetaType stores the name obtained from the C++ compiler,
which means we know a type like Qt::Alignment by its proper, full name
of QFlags<Qt::AlignmentFlag>. However, the meta object records only the
bare name of the enumeration, not the full flags.
Pick-to: 6.2 6.3 6.4
Fixes: QTBUG-105932
Fixes: QTBUG-96185
Change-Id: Ic6547f8247454b47baa8fffd170eab977e306377
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Initially, stream receive window's size was quite significant.
At some point when dealing with a server that did not want
our windows' sizes and was closing the connection (they only
accept 64K) we reduced this size, which ended in a regression
with download speed significantly throttled. We return the old
values (or even more, presuming we have 10 multiplexed streams
and not 100). And also making QNAM consistent with its documentation
(it was not updated).
[ChangeLog][QtNetwork] Stream receive window that HTTP/2 protocol in QNAM is using
increased to 214748364 octets (from the previous 64K) not to throttle download
speed. Clients, working with servers, not accepting such parameters,
must set HTTP/2 configuration on their requests accordingly.
Fixes: QTBUG-105043
Pick-to: 6.4 6.3 6.2
Change-Id: I252b6b5eefe92a7304dad15c67928d5a57d9597f
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Unifies our approach to calling CoInitializeEx and CoUninitialize,
removing a lot of boilerplate in the process, and also fixes a few
bugs where we would incorrectly balance our calls to CoInitializeEx
and CoUninitialize.
The optimistic approach of qfilesystemengine_win.cpp of calling
CoCreateInstance without initializing the COM library explicitly
has been removed, as calling CoInitializeEx should be a noop in
the situation where it's already been loaded.
Change-Id: I9e2ec101678c2ebb9946504b5e8034e58f1bb56a
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
It does a check to ensure you aren't comparing outside the container.
Fixes: QTBUG-106001
Pick-to: 6.2 6.3 6.4
Change-Id: Ic6547f8247454b47baa8fffd170eef346b7f4f24
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
The function is used the internal window activation machinery and
should not be called by user code.
Many tests still use this function, and should be ported over to
QWidget::activateWindow(). For now they are using the private
helper in QApplicationPrivate, so that we can progress with the
public API deprecation.
Change-Id: I29f1575acf9efdcbae4c005ee9b2eb1bb0c8e5b5
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
9a369a25dd added a
QAccessibleEvent ctor that takes a QAccessibleInterface*
instead of a QObject*.
Retrieving the QAccessibleInterface* later is done using the
interface's unique ID stored in the m_uniqueId member.
However, the fact that m_uniqueId is a member
of the union alongside with m_child means that setting
a child via QAccessibleEvent::setChild also overwrites
the stored unique ID, which breaks retrieving the accessible
interface later.
Fix this for the case where the QAccessibleInterface has
an associated QObject by assigning m_object in the ctor as well.
This means that a QAccessibleEvent created using either of the two
constructors (the one taking the QObject* and the one taking
the QAccessibleInterface* associated with the object) now behaves
the same.
Fixing the case where there is no associated QObject would require
further changes (e.g. adding a member for the QAccessibleInterface*
or making the m_uniqueId member a separate member instead of having
it in a union with m_child). However, I see no way to do so without
breaking the ABI, so that is left unchanged.
This also adds a corresponding test case.
Fixes: QTBUG-105988
Pick-to: 6.4
Change-Id: I71a548af0277a5034e9e207f066fa3e25c5393f3
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
