And also, reduce the utter sloppiness, weirdness of the test and
make it more a test and not a joke. Since the test itself depends
on !QT_NO_SSL, why bother building and running its main, to create
a useless tst_QSslError and do nothing then? Exclude test from
no-ssl build.
Pick-to: 5.15
Pick-to: 6.0
Change-Id: I67879b0de036cbc8c2f75a18f4cf94e6c43c5af0
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Unlike the QTextBrowser backend, litehtml does not render elements
inside <iframe> correctly. This prevented external links to
YouTube from working in offline documentation.
Move the <iframe> to a macro override specific to online doc builds.
Pick-to: 6.0 6.0.0
Fixes: QTBUG-88975
Change-Id: Iff7828ddeed353620eaa9ac669a3e0c03749daa2
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
This enables documentation building in add-ons and other modules that
need to build docs.
Task-number: QTQAINFRA-3972
Change-Id: Ic2763d6d36d26e5a1267b312727c3c268ad6d114
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
(cherry picked from commit e84e85dd4c59f2c79f81ba4d8d7f04397eb0003c)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
Use universal references instead of rvalue references for passing
callables in the implementations of QFuture's continuations.
Change-Id: I1288c78f78f84f30c6607e505e7f9807a9272071
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Jarek Kobus <jaroslaw.kobus@qt.io>
(cherry picked from commit b002722dabef794da0e80010b115b2c6cd6dc6b8)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
Constructing a QMetaProperty can be unnecessarily expensive due to the
QMetaEnum resolution. Add a private method to construct only
QMetaProperty::Data. Make us of it in indexOfProperty.
Task-number: QTBUG-82931
Change-Id: If954538106bcfaa7d088db26591f6bd6eeaf3731
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The logic in notify() was doing quite a bit more work than
it needed to. By inserting a dummy node after the current one instead of
replacing it, we can avoid half of the data shuffling that has been
happening and also don't need a back pointer when executing the
notification.
Also avoid calling a semi expensive destructor of QPropertyObserver.
Reduces the overhead of notify() by ~30%.
Change-Id: I7ce16bcf9cd9c4368c18bf875fc959223452fd4f
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
(cherry picked from commit e8ef871e3522f340b4efe32382af7e35ef908665)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Save a function call in the common case where we don't have a binding
This makes a rather large performance difference for setters that do
not have a binding.
Change-Id: I140f29790f6fe868721a33b9fad37205e547b8e9
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit c63901c5f3195596eb81e5f5ae5483ca5a0b6d35)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
No need to do function calls for the case where we return immediately
after checking a boolean.
Change-Id: I3e449850a10fcf82acb843cce6da6dfd98de32ad
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit e165f416a752398079590161a18255f9a0058a3e)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
They are not needed and removing it can simplify the code in some places
and avoid a couple of masking operations.
Change-Id: I0e4241a2784026aa89deed35f408b094e89a11a0
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit b1be6e6e6f355bfcb0c3814516f6009c91d2de89)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
In many cases, it only derefs and does nothing else. Inline the fast
code path.
Change-Id: Ib605c385c1683f7833f7189c84d6cf4eb5b0e59e
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit eda4c29eb26dab32e22040bdda0b9b9109b1408b)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Don't execute instructions that will never do anything.
Directly add the tag to the pointer in the constructor to avoid
additional masking operations, and avoid a masking op that is in
practice a no-op in setTag().
Do the same optimization in QTagPreservingPointerToPointer.
Change-Id: Ia364f89cbe6ccc876ec9bda0c239fc4f57c10501
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit 2721728c9056b442c0281f20792f19eb6a491aa0)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
And inline the fast checks inside the methods in QBindingStorage.
This allows QObjectBindableProperty and friends to inline all the
fast checks and almost completely eliminates the overhead for property
accesses when no bindings are being used.
Read and write times of QObject based properties when no bindings
are being used:
Read Write
Old style property: 3.8ns 7.3ns
QObjectBindableProperty (no notification): 4.5ns 4.3ns
QObjectBindableProperty (with signal): 4.5ns 7.6ns
QObjectBindableProperty (inline accessors): 3.2ns 3.4ns
Numbers without this patch:
Old style property: 3.8ns 7.9ns
QObjectBindableProperty (no notification): 7.2ns 7.7ns
QObjectBindableProperty (with signal): 7.2ns 16.0ns
QObjectBindableProperty (inline accessors): 6.3ns 6.7ns
Change-Id: Ifd1fa3a489c3be8b1468c0b88af547aac397f412
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
(cherry picked from commit 98c82fb445acf45cc4c4bc86a5adda43358127bf)
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
Those entries were always empty. The INTERFACE_QT_MODULE_USES property
was never set.
Map each public dependency to its qmake lib name and place this value
into the module's QT.<module-name>.uses variable.
Take into account the "_nolink" target modifier and translate it to
qmake's "/nolink".
Pick-to: 6.0
Fixes: QTBUG-88951
Change-Id: Ib6ef65b842a1fe1da3ade55867583343b4ee76ee
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Rename QT_QMAKE_LIB_TARGETS_foo to QT_TARGETS_OF_QMAKE_LIB_foo, because
we want to introduce the counterpart QT_QMAKE_LIB_OF_TARGET_bar in a
subsequent commit.
Pick-to: 6.0
Task-number: QTBUG-88951
Change-Id: I33f00f4fe65c5977da6e74c632ebeab3b891c89a
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Consider a qmake lib 'foo' that doesn't have entries in
QMAKE_LIBS_FOO. Then we must not discard this entry in the generated
module pri file, because otherwise any attempt to QMAKE_USE 'foo' will
fail with the error message "Library 'foo' is not defined.".
Pick-to: 6.0
Task-number: QTBUG-88951
Change-Id: Ibed283857f5f66b1b79459fe7b7cf06c7ce0691d
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The correct entry for this lib is 'openssl/nolink', not
'openssl_headers'.
In configurejson2cmake's helper.py we already have
this information encoded such that the no_link_so_name field of
'openssl' points to 'openssl_headers'. Extend configurejson2cmake to
take this into account, and re-generate src/network/configure.cmake.
Pick-to: 6.0
Change-Id: I9817ea7722503a373d309b7e6fa201448d403e8d
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Similar to what has been done in 5138a970f3, we need to adjust the
remotecontrolledcar/controller dbus example to be built with qmake.
In the CI we first build the examples with CMake in the source
directory. That creates some generated source files in the dbus example
dir.
Then, the examples are (will be) built ouf of source with qmake, and
those generated source files will be placed in the build directory.
When building with MSVC, there's a peculiarity with nmake/jom's
inference rules that lead to picking up the generated car_interface.cpp
from the source directory, instead of the build directory.
See QTBUG-13496 for details.
car_interface.cpp, built with CMake is generated to include
car_interface.moc.
car_interface.cpp, built with qmake does not have this include, but the
build system calls moc on car_interface.h.
All this leads to duplication of the meta object for
OrgExampleExamplesCarInterfaceInterface, preventing successful linking.
Work around this issue by turning off the generation of nmake/jom
inference rules.
Task-number: QTBUG-85986
Change-Id: I0a6ca9f880b63f05a4e9f2032e44d3b391d8a756
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
write_library_section traverses the parent/child hierarchy of scopes to
determine whether the scope belongs to a public Qt module. This doesn't
work for scopes that stem from included .pri files, because each
included file has its own parent/child hierarchy.
We already have an include scope hierarchy in the form of
Scope._included_children, but lack a way to get to the including
scope.
Add Scope._including_scope and adjust the is_public_module calculation
to take that into account after hitting the top of the parent/child
hierarchy.
Pick-to: 6.0
Change-Id: I8fee1cfbf048e7afc6783b0a52eaca75be17072f
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Add '_qt_internal_dump_expression_values' function that dumps all
values evaluated by 'qt_evaluate_config_expression'.
'_qt_internal_dump_expression_values' doesn't evaluate undefined
features, only collect actual values.
Fixes: QTBUG-88476
Pick-to: 6.0
Change-Id: I9d09ffbd9f9fa91bc4f36536c58e7f118b06f9b9
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
While it worked as expected with beta, with final release this thing
looks wrong again, especially in the dark mode.
Pick-to: 5.15
Pick-to: 5.12
Change-Id: I6857344de897e0f39d9d8e66431635dbe2f4d077
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
QSslCertificate::verify() has an undocumented and not very desirable property -
on some platorms it updates the default configuration, which can be surprising.
For example, we deprecated QSslSocket::setDefaultCaCertificates() and recommend
using QSslConfiguration::defaultConfiguration(), QSslConfiguration::setDefaultConfiguration(),
and QSslConfiguration::setCaCertificates(). If an application does this to select
CA roots it trusts explicitly, and then for some reason is calling verify, the
application can have its QSslSockets successfully connecting to a host, whose
root was not trusted by the application. Also, on Windows, defaultCaCertificates()
include system roots already, no need to have them twice.
[ChangeLog][QtCore][QtNetwork] QSslSocket::verify - do not change the default configuration
Pick-to: 5.15
Pick-to: 6.0
Pick-to: 6.0.0
Fixes: QTBUG-88639
Change-Id: I1cd40b259d0a6dcd15c78d1e7c027ff10859595c
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
And extend an auto-test for this. When a cookie verification mechanism is
enabled, and verifier, indeed, verifies that some datagram is a 'Client Hello'
message with a proper cookie attached, we start a real DTLS handshake creating a
QDtls object and calling 'doHandshake'. In case cookie verification
was enabled, we need parameters from the verifier (it's a crypto-strong
'number' and hash algorithm) to 'lock and load' the TLS state machine in
a freshly created TLS session object. This code path previously was only
tested manually and was found by LCOV as untested.
Pick-to: 5.15
Pick-to: 6.0
Change-Id: Ieacb8c989997999ea10e15bda6ae106a0338b698
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
QXcbConnection::getTimestamp uses dummy events to get timestamp from
X server. However, in some cases, X server shuts down while client tries
to get timestamp. In this case, QXcbConnection::getTimestamp keeps
getting null event and thus falls into indefinite loop.
This fix checks if xcb connection is still valid and use a special
xcb_timestamp_t value, CurrentTime (0L), as returned value.
CurrentTime should not be generated by X server and if getTimestamp
returns this value, it means an "exception" case is triggered.
This fix is introduced because in kwin_x11 (KDE project), X server can
exit on logout. kwin_x11 should handle disconnection from X server.
But the indefinite loop prevents kwin_x11 to process disconnection
event and therefore kwin_x11 cannot quit properly.
Fixes: QTBUG-88435
Pick-to: 5.12 5.15 6.0
Change-Id: Iaf7ef3f8a35fa8389d22a608e3c49041bf90e1b9
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This reverts commit 39b1dea2ec.
39b1dea2 is wrong, the bug is for static build of qt with static
xcb. It adds the explicit dependency of xcb-util 0.3.9, which is
not available on Debian yet.
Task-number: QTBUG-86287
Fixes: QTBUG-88688
Pick-to: 5.15 6.0
Change-Id: Iffc821f49bdfcad3f2556951d3677c35a7682266
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
And do not run the test with QSslSocket::supportsSsl() returns false -
this may mean unresolved symbols and thus missing functionality,
like i2d_X509 etc. This also makes cases more like other, that already
had those checks.
Fixes: QTBUG-87386
Change-Id: If4e9a650ca325b6f70956f532891a4c1d50465c0
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The container overview documentation did not mention
Q_DECLARE_TYPEINFO and the related optimizations. This patch
adds a short paragraph about it.
Task-number: QTBUG-86584
Pick-to: 6.0
Change-Id: I5b0b8ce92a47da5f0398cc413fbf3e07b0921e59
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Some code-paths were never executed by auto-test, thus giving us LCOV's
diagnostic. Extend existing tests and add new ones.
Pick-to: 5.15
Change-Id: I648747547f0525a482216b1e1972fcc698c73f65
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
- QList iterators are now nested classes inside QList.
- Drop reference to Qt OpenGL Widgets landing page, there is no such
page.
- Fix typos and linking issues.
Fixes: QTBUG-86295
Pick-to: 6.0
Change-Id: I964843deb81aa55ff8ddb9a1c2b004cb72e68de9
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
The \fn commands for friend functions need to include the class
scope for QDoc to find the declarations.
Pick-to: 6.0 6.0.0
Fixes: QTBUG-88932
Change-Id: Icbb45c12d7d657d95243231148483c00baf07c36
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
The example won't show up in Qt Creator's Welcome mode without either
an \image or a generic thumbnail. Add the latter for now.
Pick-to: 6.0
Fixes: QTBUG-88970
Change-Id: Id5ec236ef4760e678921bc2a90608d66ac35394d
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
QMetaObject::indexfOfMethod returns the method corresponding to a
specific signature. In QML, we however only want any of the methods with
a given name (and do overload resolution at a later point).
For this usecase this patch introduces the internal
QMetaObject::firstMethod function.
Change-Id: Ie3820354edffb273c4cbe1399201a955ebe79344
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
std::optional<int> is the wrong datatype to use for compare.
First and foremost, it can't be used in the idiomatic form of
auto r = a.compare(b);
if (r < 0) ~~~ // a is less than b
if (r > 0) ~~~ // a is greater than b
which we *already* feature in Qt (QString, QByteArray).
Also, std::optional<int> (explicitly) converts to bool, which is
a trap, because the result of the comparison can be accidentally
tested as a bool:
if (a.compare(b)) ~~~ // oops! does NOT mean a<b
Not to mention extending this to algorithms:
auto lessThan = [](QVariant a, QVariant b) { return a.compare(b); }; // oops!
std::ranges::sort(vectorOfVariants, lessThan);
which thankfully doesn't compile as is -- std::optional has
an *explicit* operator bool, and the Compare concept requires an
implicit conversion. However, the error the user is going to face
will be "cannot convert to bool because the operator is explicit",
which is deceiving because the fix is NOT supposed to be:
auto lessThan = [](QVariant a, QVariant b) { return (bool)a.compare(b); }; // big oops!
Instead: backport to Qt the required subset of C++20's <compare>
API, and use that. This commits just adds the necessary parts
for compare() (i.e. partial ordering), the rest of <compare>
(classes, functions, conversions) can be added to 6.1.
Change-Id: I2b5522da47854da39f79993e1207fad033786f00
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
(cherry picked from commit 3e59c97c3453926fc66479d9ceca03901df55f90)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
In addition (and as a fallback) from requiring qHash, add support
for std::hash specializations. This catches two birds with one stone:
1) users of Qt can simply specialize std::hash for their datatypes,
and use them in both QHash and stdlib unordered associative containers;
2) we get QHash support for any (stdlib) datatype that is hashable
without having to overload qHash for them.
[ChangeLog][QtCore][QHash] QHash, QMultiHash and QSet now support
for key types anything that can be hashed via std::hash, instead of
always requiring a qHash() overload.
Change-Id: Ib5ecba86e4b376d318389500bd24883ac6534c5f
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
To support qHash overloads without a seed we have a qHash(T, size_t)
catch-all that calls qHash(T) and XORs the seed. The problem is
that this catch-all is not SFINAE friendly. For a type Foo which
does not have any qHash overload, we can't ask if qHash(Foo, size_t)
is callable because it would instantiate the catch-all and fail
to compile.
Add a suitable trait and enable_if.
Pick-to: 6.0
Change-Id: Idffd48a537eebaf77cee7030b8d91a302643ffde
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
If one clicks on a QHeaderView's section, the header view will
sort the view by the respective column/row. By clicking multiple
times, one is able to toggle the sorting between ascending
and descending. Something that is NOT possible to do however is to
un-sort the view -- that is, to restore the model's original
sorting. This must be done via code, by asking the header or the
view to sort by section -1.
This commit adds new property to QHeaderView to make it possible
to unsort models. Basically, the sort indicator becomes a tri-state:
sort ascending, sort descending, unsort (sort by column -1).
[ChangeLog][QtWidgets][QHeaderView] Added the sortIndicatorClearable
property. Setting this property allows the user to clear the sort
indicator on a section, resetting the model to its default ordering.
Change-Id: Ibf4e280b2086b75ccd64d619ea4d70816dc3529f
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Reviewed-by: David Faure <david.faure@kdab.com>
- Add tests for QCollatorSortKey
- Add test for QCollator copy assignment and copy construct
Currently QCollatorSortKey tests are working properly only
with QT_CONFIG(icu)
Task-number: QTBUG-88546
Change-Id: Ic35dfd33038cc736245904b78fe4383a5a11b580
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Q_MOVABLE_TYPE was conceived before C++ had move semantics. Now, with
move semantics, its name is misleading. Q_RELOCATABLE_TYPE was
introduced as a synonym to Q_MOVABLE_TYPE. Usage of Q_MOVABLE_TYPE
is discouraged now. This patch replaces all usages of Q_MOVABLE_TYPE
by Q_RELOCATABLE_TYPE in QtBase. As the two are synonymous, this
patch should have no impact on users.
Pick-to: 6.0
Change-Id: Ie653984363198c1aeb1f70f8e0fa189aae38eb5c
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
This patch removes two unused variables and marks one unused, fixing
three warnings.
Change-Id: I71f59839452590b82ffb5459a968f06bd434fb9a
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Despite being installed in the the binary directory, this
CMake snippet cannot be executed.
Pick-to: 6.0
Change-Id: Ibfd31e51184f0688d19ac063804cc200c95555a7
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Before this change, the prl files always assumed that resource object
files are installed into $qt_prefix/lib when doing a prefix build.
That was true for qt_internal_add_resource calls, but not for
qt6_add_qml_module and qt6_target_qml_files.
Change qt_internal_record_rcc_object_files to take a new required
INSTALL_LOCATION argument. The argument takes a path relative
to CMAKE_INSTALL_PREFIX.
Modify __qt_propagate_generated_resource to save the relative path of
the generated resource source file, which will be used in the
computation of the final resource object file location.
This is needed because the Qml resource functions place the source
files in a different directory layout, e.g. .rcc vs .qmlcache
Modify qt_generate_prl_file to prepend $$[QT_INSTALL_PREFIX]/
instead of $$[QT_INSTALL_LIBS]/ for the resource install paths.
A follow up patch is done in qtdeclarative to pass the new
INSTALL_LOCATION argument from the Qml CMake functions.
Amends f9dcade5e7
Task-number: QTBUG-87702
Task-number: QTBUG-88425
Change-Id: Id17bb517b4cb5d00911bfd10a728ba4e0d44871b
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
(cherry picked from commit 54d0ca93bca78f8fd31b6761f078e7a96283f183)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>
Conform to the XDG Base Directory Specification:
"If, when attempting to write a file, the destination directory
is non-existent an attempt should be made to create it with
permission 0700. If the destination directory exists already
the permissions should not be changed."
At the same time the spec states about XDG_RUNTIME_DIR that
"its Unix access mode MUST be 0700", so don't consider the
directory with wrong permissions correct and use a fallback.
Task-number: QTBUG-68338
Pick-to: 5.15 5.12
Change-Id: I03c6b35b3f7d5ceb8e6326695bfc8207da92ea67
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: David Faure <david.faure@kdab.com>
Except for types marked as unresolved, we're doing it anyway - the only
difference is that now we skip looking up the metatype by typeid.
[ChangeLog][QMetaProperty][Important Behavior Change]
QMetaProperty::typeName returns now always the same name as name() of the
corresponding metatype. This can cause a change for enum properties
which were not fully-qualified.
Change-Id: I1f57743948b7262ac06095d3bbc838d620f6e481
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
If a testcase failed before we reached free, the QMetaObject would have
leaked.
Moreover, defer freeing until all tests have run: dynamic QMetaObjects
might create dynamic metatypes. If those were shared, we would run into
issues.
Change-Id: I2edfcb2605e58eaba454bfe0446a6b187a5210df
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
