Not just fuzzily equal. The fuzzy comparison fails for infinities, since
the expression p1 - p2 where p1 = p2 = infinity is NaN. And NaN
comparisons are always false.
As a nice side-effect, we don't do the more expensive computation of a
multiplication if the two numbers really are equal.
Task-number: QTBUG-50036
Change-Id: I11f559ef75544c50b3f8ffff1420cec7c7273295
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
If CountedStruct is passed a GetSenderObject object,
it will attempt to call a member on it from within
its own destructor.
That works usually quite well, but in this test case,
which tests for function object leaks when a connection
is torn down because the sender object is destroyed,
the destruction of the CountedStruct happens when all
connections are severed in ~QObject. At that point,
what used to be a GetSenderObject instance no longer
is one and the call into one of its member functions
invokes undefined behavior.
Fix by making QObject::sender() public by a using
declaration instead of a wrapper function.
Found by UBSan:
tests/auto/corelib/kernel/qobject/tst_qobject.cpp:6007:104: runtime error: member call on address 0x7ffc6e7538b0 which does not point to an object of type 'GetSenderObject'
0x7ffc6e7538b0: note: object is of type 'QObject'
Change-Id: Ia973140037b3c1b5a670a8a3949d09b956f40349
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
If QObjectPrivate::declarativeData is set, it is
in various places in Qt expected to point to a
QAbstractDeclarativeDataImpl, from which ownedByQml1
is unconditionally read.
In noDeclarativeParentChangedOnDestruction(), the
declarativeData pointer is, however, set to a local
QAbstractDeclarativeData instance, which, being an
empty class, has size 1 and alignment 1.
Depending on the compiler's idea of bit field order,
this code either read uninitialized data from the
dummy object, or else some random stack memory outside
any (valid) object.
What caught UBSan's attention, though, was the
difference in alignment between the two classes:
src/corelib/kernel/qobject.cpp:917:9: runtime error: member access within misaligned address 0x7fffc9cf706f for type 'struct QAbstractDeclarativeDataImpl', which requires 4 byte alignment
Fix by providing a properly initialized object of the
correct type.
Change-Id: Iae83a949ee5a7bc98df13e35ea614c063085fa13
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Don't pass around meta-type IDs in QMetaType::Type
variables. It leads to reading values from an enum
variable that are invalid.
Fix by passing the IDs around as int.
Found by UBSan:
tests/auto/corelib/kernel/qmetatype/tst_qmetatype.cpp:408:5: runtime error: load of value 4028, which is not a valid value for type 'Type'
Change-Id: Idd106ee3d7960fe3d8fefc0fc5830fc22d38a513
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
TEST_HELPER_INSTALLS cannot be used on platforms with no
QProcess support.
Change-Id: I2a6a283d94ca4487fc628449c53fc37140dd291d
Reviewed-by: Oliver Wolff <oliver.wolff@theqtcompany.com>
Also use this for converting doubles with QVariant. We generally want
exact results there, rather than adding rounding errors whenever we
convert.
[ChangeLog][QtCore][QLocale] Added special value for double conversion
precision to get shortest accurate representation.
Change-Id: I905b8a103f39adf31d24b6ce2c8a283cf271b597
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Use QByteArray/QString addition instead in loops and for
test row names.
Change-Id: Ieffb429efdc14aa5932b3fcdef5a18e13a62d35f
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
Prefer QCOMPARE over QVERIFY for equality and use QLatin1String().
Change-Id: If226a0fc7b25be3e6774c7e36ca1e6f99234e5dd
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
If QTimer::singleShot() is used with a functor callback and a context
object with different thread affinity than the caller, a crash can
occur. If the context object's thread is scheduled before
connecting to QCoreApplication::aboutToQuit(), the timer has a change
to fire and QSingleShotTimer::timerEvent() will delete the
QSingleShotTimer object making the this pointer used in the
connection invalid. This can occur relatively often if an interval
of 0 is used.
Making the moveToThread() call the last thing in the constructor
ensures that the constructor gets to run to completion before the
timer has a chance to fire.
Task-number: QTBUG-48700
Change-Id: Iab73d02933635821b8d1ca1ff3d53e92eca85834
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
[ChangeLog][QtCore][QMetaProperty] write() now resets the property if an
empty QVariant is given, or set a default constructed object if the
property is not resettable
Change-Id: I9f9b57114e740f03ec4db6f223c1e8280a3d5209
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
if t >= QMetaType::User, we would not return false nor call convert.
We would then pass a pointer to whatever is in the QVariant to the
qt_metacall that is expecting a pointer to an object of a different type.
Since we have custom converters, we can call QVarent::convert even for
custom types anyway.
[ChangeLog][QtCore] Fixed crash when setting a QVariant of a different
type to a property of a custom type. Attempt to do a conversion instead.
Task-number: QTBUG-40644
Change-Id: Ib6fbd7e7ddcf25c5ee247ea04177e079f6d7de35
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@theqtcompany.com>
If QT_NO_LIBRARY isn't set we cannot test QCoreApplication's library
path functions and none of the plugin and library related tests are
applicable. Also, examples that rely on dynamic plugin loading for
their core functionality obviously don't work.
Change-Id: I2d381ee1bc8d944e1181557895a7e92a364fd778
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The test assumed that an event loop can be tested for emptiness
multiple times, which is wrong because an event can be delivered any
time.
Change-Id: Ic44245321eeed2091b640ada2c83d205b83a1cc2
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
To handle network events, QEventDispatcherWin32 uses I/O model
based on notifications through the window message queue. Having
successfully posted notification of a particular event to an
application window, no further messages for that network event
will be posted to the application window until the application
makes the function call that implicitly re-enables notification
of that network event. With these semantics, an application need
not read all available data in response to an FD_READ message:
a single recv in response to each FD_READ message is appropriate.
If an application issues multiple recv calls in response to a
single FD_READ, it can receive multiple FD_READ messages
(including spurious).
To solve this issue, this patch always disables the notifier
after getting a notification, and re-enables it only when the
message queue is empty.
Task-number: QTBUG-46552
Change-Id: I05df67032911cd1f5927fa7912f7864bfbf8711e
Reviewed-by: Joerg Bornemann <joerg.bornemann@theqtcompany.com>
Disable input and cursor for QGuiApplication instances used in
autotest to initialize it properly.
Change-Id: I78dc9b776269c082c20f244a51f858289129275d
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
The keyword no longer has a meaning for the new CI.
Change-Id: Ibcea4c7a82fb7f982cf4569fdff19f82066543d1
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
- Replace Q[TRY]_VERIFY(pointer == 0) by Q[TRY]_VERIFY(!pointer).
- Replace Q[TRY]_VERIFY(smartPointer == 0) by
Q[TRY]_VERIFY(smartPointer.isNull()).
- Replace Q[TRY]_VERIFY(a == b) by Q[TRY]_COMPARE(a, b) and
add casts where necessary. The values will then be logged
should a test fail.
Tests from corelib/tools were omitted in this change.
Change-Id: I4c8786d33fcf429d11b2b624c7cd89c28cadb518
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Preparing the replacement of Q[TRY]_VERIFY(a == b) by
Q[TRY]_COMPARE(a, b) for non-boolean types.
Change-Id: Iab6ec2f0a89a3adc79e18304573994965013dab5
Reviewed-by: Mitch Curtis <mitch.curtis@theqtcompany.com>
We force a recreation of the library paths with added information on
construction of QCoreApplication. This way we can find plugins in
the application directory which only becomes known when
QCoreApplication is created. When the user changes the library path
we create a new list of the manually modified library paths and
recalculate it from the delta of original vs. modified paths when
QCoreApplication is created.
The upsides of this approach vs. keeping an explicit delta are:
* We don't need to introduce a separate data structure to hold
the added/removed status for delta items or the information that
the whole list got replaced.
* The lists never get larger than the the real library paths. An
explicit delta would have to record all modifications.
* I don't think the delta replay algorithm we would have to do
anyway could be made much more compact than the one this change
introduces.
Of course, if the user actually changes anything, the list is
duplicated. Considering that this is a rarely used function and
that we would have to save some extra information anyway, I think
we can live with this.
Task-number: QTBUG-38598
Change-Id: I3bfbbd1be62dd5804dcc7ac808b053428a4e3149
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Long-lived threads started by Qt itself can now receive events even if
QCoreApplication hasn't been created. This is required in all threads we
start that will handle events, unless we're sure that the thread will
exit before the global application object begins destruction.
Otherwise, those threads will have race conditions dealing with the
event delivery system trying to call the QCoreApplication::notify()
virtual while the object is being destroyed.
Change-Id: I27eaacb532114dd188c4ffff13d4ad2a4bb443e6
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
On Windows, t->timeout was updated only once, at creation time, so the
timer would always show as "overdue" after the first activation.
The timer is updated to indicate the full remaining time during the slot
activation, which is the behavior of the Unix and Glib dispatchers.
Task-number: QTBUG-46940
Change-Id: I255870833a024a36adf6ffff13ecadb021c4358c
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@theqtcompany.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Since Connection can be copied, one copy could be used for
disconnecting, but the other's d_ptr wouldn't get updated and would
continue to report as still connected.
This patch fixes that by making it check the internal state. That is
only done after d_ptr is already known to be non-null. Unfortunately,
that is the common path:
if (connect(sender, &Sender::signal, [] {}))
will call an out-of-line function. I don't see a way out.
Task-number: QTBUG-46213
Change-Id: I66a35ce5f88941f29aa6ffff13dfb45dca68a350
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Global (application-level) event filters are supposed to be run only in
the main thread, so ensure that it is the case.
Change-Id: I27eaacb532114dd188c4ffff13d5a17d991b8bd2
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
They didn't show up in the "old" CI runs because they usually pass the second
time they are executed - which the testrunner does. The new CI doesn't do that
anymore, instead we now mark those tests explicitly and will track their record
of passing and failing in the new metrics database.
Change-Id: Id34dd6f792f38995b07b6fec88f833df64de2f8b
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This allows for example properties with QMap<Foo, Bar>
[ChangeLog][QtCore] Types in the Q_PROPERTY macro can now contain commas
(for example, QMap<Foo, Bar>)
Change-Id: Ibf5c8c9cf20a7c8b3dfec9e891fb8a9ca1bdba7c
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Do not try to automatically register the meta type for Q_GADGET that
are not default constructible.
This fixes a source incompatibility in the function pointer syntax
of QObject::connect when such types are used as an argument of a signal.
Task-number: QTBUG-45721
Change-Id: I3065f6d57bc1f37e16988d2dee99118de250ca56
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Otherwise the type is registered with the wrong name
Change-Id: I68ec3a05e2528816626e648b46ccc9d70b004866
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
My commit 6c973dee2c broke the case where setApplicationName
is called before the QCoreApplication constructor.
Fixed and added autotest.
Task-number: QTBUG-45283
Change-Id: If7bdb0d82be50b50a95a04027f5f9d7143c1a7ac
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Fixed issue that text/uri-list mimedata got from QMimeData::data()
was corrupted after setting it back via QMimeData::setData()
Change-Id: I2377523a9286519402ab9127ed7f3fa66e39a679
Task-number: QTBUG-45486
Reviewed-by: David Faure <david.faure@kdab.com>
This patch addresses a specific Qml problem, where the meta types list
will grow indefinitely when unloading and reloading Qml components over
and over (in an failed effort to save memory).
The implementation is not specific to Qml though, but will cater to all
use-cases where registered types may not live until the application's
termination.
Change-Id: Ic0224dcd19aeb559715ef088b22a30509be2456b
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
QObject::connect tries to determine if the arguments are registered
metatypes. This used to work even for arguments that were forward
declared. But now, the metatype system tries to call
QtPrivate::IsQEnumHelper<T>::Value to know if it is registered.
That fails on gcc if T is forward declared.
Apparently gcc needs to know the full type of T to pass it in the
ellipsis function, even within a sizeof expression.
So change the ellipsis expression to a template one.
Task-number: QTBUG-44496
Change-Id: I7fa07bd3cde470b134c2ec53b0d581333d16a6f1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
