... so that QCocoaEventDispatcher can use it to implement timer handling
and benefit from the Qt::TimerType support in QTimerInfoList.
Change-Id: I34b81502465963e2c9d528df463fa2eccd275ad6
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The code for calculating the expected time is only useful for debugging
purposes. Don't compile this into the library unless QTIMERINFO_DEBUG is
defined.
Change-Id: I6530e6a70410a12544410ef286225df98ceddcee
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We're trying to deprecate these, so don't use them anymore.
The inline uses of these have been left intact, for the moment. Inline code will
need to create their own non-inline allocation methods (for future-proofing to
allow alterations in how e.g. individual containers allocate)
Change-Id: I1071a487c25e95b7bb81a3327b20c5481fb5ed22
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This function is never called on systems that are guaranteed to have
a monotonic click (like Mac OS X). Remove the dead code from the
library.
Change-Id: I95852c8dffaa3a9747367f0abe4a4c62e4f86421
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
As stated in the documentation for Qt::TimerType, we allow for up to
5% error for CoarseTimers (the default timer type). PreciseTimers are
not adjusted at all, and VeryCoarseTimers fire with one-second accuracy.
The objective is to make most timers wake up at the same time, thereby
reducing CPU wakeups.
Note that this changes makes it possible for timers to fire early, which
may be unexpected for some applications. Such applications should use
PreciseTimers explicitly.
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: Iaa70314c39a446adbc6dbb6fdfa7bafcd98a7283
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Windows x64 uses 64 bits integer for sockets, to ensure compatibility we
should use ptr sized integers for our socket descriptors.
Task-number: QTBUG-19004
Change-Id: I4b56023874a4f1bad107c66c054fecfedde33d88
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Previously, QObject::startTimer() would warn when called from
QBasicTimer::start() if there was no event dispatcher for the object's
thread. QBasicTimer::start() should output a similar warning when there
is no event dispatcher for the current thread.
Change-Id: I1152f73216c3551c252a7a6995defebc9e1506c8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
3rdparty event dispatchers are impossible to write without using the
internal API QAbstractEventDispatcherPrivate::releaseTimerId(). Fix
this by having each QObject keep track of its own timer ids, and
release them when they are no longer used. As a side effect, this
makes the QObjectData::pendTimer bit unnecessary.
This also removes the QObjectData::inThreadChangeEvent hack that the
event dispatchers used to avoid releasing timer ids when moving timers
to a new thread.
QBasicTimer becomes even more low-level. It cannot use
QObject::startTimer() anymore, since we do not have a way to call
QObject::killTimer() from QBasicTimer::stop(). QBasicTimer uses the
QAbstractEventDispatcher interface directly, and releases the timer id
explicitly as well when stopping the timer.
This change also fixes some rare timer id "leaks" when destroying or
stopping timers after a thread has exited and destroyed its event
dispatcher (the timer ids would never be released when no dispatcher
exists).
Globally destructed QObjects that have running timers may try to release
their timer ids after the timer id freelist has been destroyed. This
commit accomodates such objects by avoiding the null dereference in
QAbstractEventDispatcherPrivate::releaseTimerId().
Change-Id: I2d7cd8221fae441f3cf02b6c0b4bc16063834d00
Reviewed-by: David Faure <faure@kde.org>
Reviewed-by: Denis Dzyubenko <denis.dzyubenko@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
QVariant implementation is based on delegation to a handler. The handler
has rather simple construction, it is a set of function that implements
a switch statement over known types and redirects calls to a right
method of an encapsulated types instance. Unfortunately after qt
modularization project, it is not easy to use types directly from
different modules, as they can be undefined or completely unaccessible.
Which means that each module has to implement own handler to cooperate
correctly with QVariant. We can suspect that list of modules known to
QVariant will grow and it is not limited to GUI, Widgets and Core,
therefore it would be nice to have an unified, from performance and
source code point of view, way of working with handlers.
This patch is an attempt to cleanup handlers. Keynotes:
- Each handler is working only on types defined in the same module
- Core handler implements handling of primitive types too
- Custom types have an own handler
- Each handler is independent which means that dispatch between handlers
is done on QVariant level
- Handlers might be registered / unregistered using same interface
Change-Id: Ib096df65e2c4ce464bc7a684aade5af7d1264c24
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
CallConstructor<T, /* CanUseInternalSpace = */ false> is called when we need to
construct an object that couldn't be fit in qvariantdata, meaning either it is
not a POD type (Q_PRIMITIVE_TYPE), or it is simply too large to fit there.
Change-Id: Ied122b4a6f600e14312a8d515f5b3e91214a94f1
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@nokia.com>
CoarseTimers are worst in their first firing, so we prefer a
PreciseTimer for something that happens only once. If the timeout is
too big, we use a CoarseTimer anyway (current threshold is 2000ms).
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: I30b20acf506e442cd58126abfe3a4d70fc13b075
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The API passes the interval as an int, there's no reason to convert
it to a timeval struct. This also prepares for changing the UNIX
timer code to support the different timer types.
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: Ie3cc1ae8f1be6a9ad3f1766051642cbf3e614418
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
... and deprecate the old registerTimer() functions. The new pure-
virtual registerTimer() breaks source-compatibility. Subclasses cannot
be instantiated anymore, since the pure virtual function signature has
changed.
QAbstractEventDispatcher::TimerInfo is no longer a QPair. It is now a
struct with timerId, interval, and timerType members. This is a source
incompatibility that should only affect subclasses of
QAbstractEventDispatcher, which will need to pass 3 arguments to the
TimerInfo constructor instead of 2. If the subclass used QPair<int,int>
instead of the TimerInfo typedef, the QPair<int,int> declarations will
need to be replaced with TimerInfo.
Call the new registerTimer() function with the type from
QObject::startTimer(). Change all subclasses of QAbstractEventDispatcher
to reimplement the new virtual function. The type argument is unused at
the momemnt, except to ensure that registeredTimers() returns the type
each timer was registered with. Implementations for the various
dispatchers will be done in separate commits.
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: Ia22697e0ab0847810c5d162ef473e0e5a17a904b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The timer type will control the accuracy of the timer. By default, all
timers are CoarseTimers, which allows for +/- 5% interval adjustment.
PreciseTimers will not have any interval adjustments, VeryCoarseTimers
will have intervals adjusted to full second resolution.
Use QTimer::setTimerType() or the QTimer::singleShot() overload to
specify the type.
QObject::startTimer() now takes a Qt::TimerType argument which defaults
to Qt::CoarseTimer. QBasicTimer::startTimer() gets an overload that
takes a Qt::TimerType argument. The argument is unused for now, since
the QAbstractEventDispatcher interface needs to change (done in a
separate commit).
Author: Thiago Macieira <thiago.macieira@nokia.com>
Change-Id: I3100da5aa1fe17ec30b8644897d0fe6ec4a07f52
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove the (-no)-qt3support options from configure, and remove the last
remaining references to Qt3Support, QT3_SUPPORT, and
QEvent::ChildInserted.
The compatibilityChildInsertEvents() tests in tst_QObject and
tst_QWidget have been renamed to childEvents(), which is a more
appropriate name.
Change-Id: Id0b45e9b177efcc8dceee8c9ed8afafedeeace2f
Reviewed-by: Kai Koehne <kai.koehne@nokia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: David Faure <faure@kde.org>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Filter doesn't have to be a parameter because we always use
DefinedTypesFilter.
Change-Id: I19b8eb47a4c50e290cf712f909aaead8fdc94cd9
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
After 8fd64d22ac (Make usage of internal
QVariant space.) change QVariant started to "inherit" movablity from
interned type.
This change fix it by interning only movable type in QVariant and by
using external allocation for not movable ones.
Obviously, this change has negative impact on QVariant it self, but
after it, QVariant will behave a lot nicer with our containers.
Change-Id: Ibffc95833918f65be737f52d694ee81a2036c412
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We need that information to perform some optimizations in QVariant.
Change-Id: Id9a1716e49e4cedd17cd09a32fea4ff003ef61f2
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Use Q_STATIC_ASSERT_X give a better error message.
If C++11 is used, you get the string in the error.
Else, clicking on the QStaticFailure error still shows you the
string in the qobject.h source code)
And report better failure if the return types do not match.
(Without the static assert, you would still have a compilation error,
but in an unrelated place, with no reference to the actual connect()
call. The error was thrown from the virtual call
QSlotObject::call, without saying where it was instantiated)
Previously the error was relying on the existence of a type inside
CheckCompatibleArguments, but the Q_STATIC_ASSERT requires a bool
(hence the introduction of CheckCompatibleArguments::value)
There also was a typo in the return value of
AreArgumentsCompatible::dummy that made that code not work, and that
error not be reported.
(Instead, the error was reported when QObjectSlot::call is instantiated)
Specialization of AreArgumentsCompatible for the void type have been
added because if the return value of a signal or slot is void, the
connection should work.
Change-Id: I5a93ec787ce2a4b94a26630ca31d5001cd294e4d
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This is a regression introduced in Qt 4.8
When QApplication::processEvents is called from a destructor, it is
possible that pending events would still be called on the already
destroyed subclass.
Prevent that by using the same pattern as in QMetaObject::activate
Change-Id: Ida50db07ae089264402dafcde7a41a066479d08b
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
QMetaTypeInterface has to be POD because it is constructed in a static
array. Constructors in POD types are not allowed so we will use a macro
instead.
Change-Id: Iab9ae776dfe4dcd7148558f02d6181c5917aa5c3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Since integrity was pulling in all the same code (except the glib-in-QT_CONFIG
branch), we can just merge this in.
This shouldn't break integrity unless they somehow magically inject glib into
QT_CONFIG without actually having glib, but the removal of redundancy makes this
worthwhile.
Change-Id: I527b5e60bea4452fdca5eedfe729214f16519234
Reviewed-by: Rolland Dudemaine <rolland@ghs.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
This makes it easier to see the guts of the unix event dispatcher, and to
experiment with it.
Change-Id: I715bb68c4de6798e10bc55304a128b88e0249c63
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
This field isn't used at all in qtbase, nor in any of the qt5
submodules.
Change-Id: If57d389935593f797818506a220c6a3cc04b6078
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Commits a6ae75f92a and
6f0f9f6928 were cherry-picked in the wrong
order (despite the dependencies shown in Gerrit), causing the
QObjectData::unused bitfield to be too large.
Change-Id: I65acaa8b507f7f6f2c5735f45bd0ad8343abea54
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
For some strange reason "Object::method ..." was printed, without
the leading Q.
Change-Id: I10b99e8aa8730e4020d15b3e04a01004bade76c3
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QApplication::type used to be static and set by the
QApplicationPrivate constructors.
In QCoreApplication we have the new application_type that should take its place.
QApplication::GuiServer is deprecated (since it doesn't have any functionallity
any more with QWS being removed).
This change prepares QStyle to be called from a QQuickCanvase based application
that does not inherit the QWidget version of QApplication.
Change-Id: Ifbe992e25f1e5821fa047b6eb915f75fa675ab97
Reviewed-by: Paul Olav Tvete <paul.tvete@nokia.com>
QWeakPointer is superior and preferred.
Remove QMetaObject::addGuard(), QMetaObject::changeGuard(),
QMetaObject::removeGuard(), and QObjectPrivate::clearGuards().
Implement QPointer using QWeakPointer<T> instead. This changes the
behavior of QPointer in 2 ways:
- During destruction of a QWidget. Previously, the destructor of QWidget
would reset all QPointers so that they would return zero when destroying
children. Update tst_QPointer to account for this change.
- When constructing a QSharedPointer to take ownership of an object
after a QPointer is already tracking the object. Previously, the shared
pointer construction would not be affected by the QPointer, but now
that QPointer is implemented using QWeakPoiner, constructing the
QSharedPointer will cause an abort(). Fix tst_QSharedPointer by
removing the use of QPointer in the objectCast() test.
These behavior changes are documented in the QPointer class
documentation and in the changes file.
Change-Id: I92d0276219c076ece7bcb60f6e1b9120ce4f5747
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The union in QObjectPrivate of declarativeData and
currentChildBeingDeleted shouldn't use wasDeleted to determining the
meaning of the unioned pointer. Instead, add
QObjectData::isDeletingChildren, set that in
QObjectPrivate::deleteChildren(), and only use the
currentChildBeingDeleted member when the parent's isDeletingChildren is
set.
This solves aborts seen in autotests when widgets are deleting window
children. The abort comes from QWeakPointer on the parent in the child's
close event handler (the abort checks that wasDeleted is not set).
Change-Id: I1a58449159d4a5312aad8ba12e559d05d6c43d93
Reviewed-by: Aaron Kennedy <aaron.kennedy@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Some people like to build their Qt-based projects with -Wshadow
-Werror, and since qobject_impl.h is included from a public Qt
header, their build breaks.
Change-Id: I5d8a5ee1937092149a18c0ff66a723f5bfc9ffc3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This is needed by QtScript, so we need to keep it for now.
Change-Id: Iee6bd7daf7e86a09242523f5aca72ede413a6981
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
After commit 79f675a1e0, the connect and
disconnect callback API effectively becomes useless. Now that these
callbacks cannot/do not, it makes little sense to keep the backdoors
added for Qt Jambi support.
Remove them for now. Should the Qt Jambi team want/need to port to Qt 5,
we can re-add them, possibly designing a better API for doing so as
well.
Change-Id: I6209a1d647d683c979d5294b632b8c12c0f9f91c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This is sent when the object's name changes.
tst_moc must be updated since the method count of QObject has changed.
The test assumed that there were 4 invokable methods in QObject. The
new signal is the 5th, which breaks the test's assumptions. Fix this
as well.
tst_QObject must be updated since the
QObjectPrivate::isSignalConnected() method only supports the first 64
signals. With the addition of this new signal, sig61() in the test
becomes the 65th signal, and will always appear connected now.
Task-number: QTBUG-13999 (related)
Change-Id: Ie87893c71a231fafa7ccf2f16102238a7be8327a
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
By adding anonymous namespace and static linkage we are reducing
visibility of implementation of these macros.
This patch also fixes warning about a declared but unused variable which
was issued by gcc 4.6 for Q_CONSTRUCTOR_FUNCTION.
Change-Id: I2cb70ad4c93f6f77e5518420abcce6fd4cadccfa
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
QWinEventNotifier is an essential class if you're using native Windows
Overlapped IO and need to convert it to Qt signals. However the header
is marked private.
Task-number: QTBUG-68
Change-Id: I22e9a84da97f969ddb82e9ba15e604a01abd80d0
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
These were marked as TODO items for Qt 5. Do them now. (The TODO item
was added when plural support was added back in the early Qt 4 days.)
Change-Id: I3be50bc657582db730401103d691234695784340
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
For compilations without variadic template support
Change-Id: I78af4f6022ad7a0923e5c5788a34eb7d834f50f3
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
For queued connections, the args[0] is set to null as it make no sens to
forward the return value to the signal.
So we need to check in the operator, that the pointer is not null.
(container.data is args[0])
Change-Id: I80bde66f1ec19de0f4962c80e5b2797d2819075c
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
QMetaType should unlock all mutexes guarding custom types data before
calling an user code.
Task-number: QTBUG-22930
Change-Id: I501d011d6cbd467d8f22402d668e84aa848061b8
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
This commit also improves the related documentation a bit.
The test is copied from the test with the old syntax, but all the
connection statement are changed to use the new syntax
Change-Id: Ia5630ca4335b9f8ca6d724ae3c8750d6f0804d8e
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Move from each implementation of QObject::connect to
QObjectPrivate::addConnection the code that adds the
QObjectPrivate::Connection* to the sender's list.
Change-Id: I665af016d5e6673eb0e9c06965e5deed50454b28
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This add an overload to disconnect which is symetrical to the new
syntax of connect.
It is possible to diconnect connection like this:
QObject::connect( sender, &Sender::valueChanged,
receiver, &Receiver::updateValue );
QObject::disconnect( sender, &Sender::valueChanged,
receiver, &Receiver::updateValue );
This overload only work with pointer to member function, and not static
functions or functors.
The test is copied from tst_QObject::disconnect(), just
changed the syntax of the connection and disconnection
Change-Id: Ia8f819100cb12098e32877522b97b732b1e676a8
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
QVariant::Type is redundant, it copies QMetaType::Type enum. In long
term it might be removed completely, but it wouldn't be a source
compatible change.
Change-Id: Ibe79ca0ab43918b4cf767cd7a5040f865abbf03f
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The overloads taking a range were previously used for cloned signals
(ones that had default arguments). Commit
919b723088b8617b202b92d80b8d0983e4fd9500 changed how cloned signals are
handled, making the from,to overloads obsolete.
The 3 argument activate() overload that does not take a QMetaObject
argument was marked obsolete by the same commit, but considering that it
is used by our autotests, I've decided to keep it and not mark it as
obsolete anymore.
Change-Id: I631ce84dce156dec68cf26e10787cb35e3f50e18
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QMetaCallEvent now can handle a pointer to QSlotObjectBase
Change-Id: I94da1e68ce9bb1fd96a9ae013a389552eb625faa
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
In addition to the old connection syntax, you can now connect using function pointers.
connect(sender, &Sender::valueChanged,
receiver, &Receiver::updateValue );
You can connect also to functor or C++11 lambdas
The connections are now type safe (no more problems with namespaces
or typedefs). Implicit type conversion is also supported.
The new syntax forces us to change the meaning of signal form
protected to public, in order to be able to access the signal's
address everywhere
The way it works is by introducing new overload of QObject::connect
that take function pointer as parametter. Those new overload are template
function, that are implemented inline. The actual implementation is
in QObject::connectImpl which take a QObject::QSlotObject* as parametter
for the slot. That slot object contains a virtual function which call
the slot which has to be implemented in the header as it depends on the
template parametter. So the internals of QObjectPrivate::Connection
will store this QObjectSlot* in order to be able to make the call.
You can read a full description here:
http://developer.qt.nokia.com/wiki/New_Signal_Slot_Syntax
History of commits before it was imported on gerrit:
https://qt.gitorious.org/~ogoffart/qt/ogoffarts-qtbase/commits/qobject_connect_ptr
Thread on the mailing list:
http://lists.qt.nokia.com/pipermail/qt5-feedback/2011-August/000796.htmlhttp://lists.qt.nokia.com/pipermail/qt5-feedback/2011-September/001248.html
(The discussions on the mailing list were about trying to find a
solution that do not need making signals public, but no user friendly
solution was found)
Note: support for QueuedConnection, and the symetric QObject::disconnect is
added in another commit.
Qt::UniqueConnection is not supported yet in the new overload.
Change-Id: I67d08436b0720e7f2992be9f7e34770960fa58fa
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
From a bool to a handle to to connection.
Also added a new overload of disconnect that disconnect a handle
This is required because with the new syntax taking lambda or functors,
it is the only way to disconnect a connection (as it is impossible to
compare functors)
The new return value is QMetaObject::Connection, it is a wrapper around
the internal QObjectPrivate::Connection.
QObjectPrivate::Connection is now reference counted.
tst_qglobal.cpp:
This test set up an internal callback, and the callback do not set any
proper connection handle (and tbh, it would be hard for it to do so).
So the returned QMetaObject::Connection is invalid, and ok is false
(Internal callbacks are only used for jambi and should probably be removed)
Change-Id: I111626fb4f47efc4db5e2ea5bff9da15f08fea7b
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This allows for more focussed notification of what part of the
model has changed layout.
The slots in the proxy models can be more optimized later.
Change-Id: I1bd17465b4be6f8efdc107036db897c557fcb519
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
... and use it in QCoreApplication::notifyInterna() instead of
"reinventing" the wheel there. The constructor and destructor for
QScopedLoopLevelCounter are inline, so the class does not need to be
exported.
Change-Id: I7af5a4ef0987f277bcc50c1057f3c74152f3d79d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QMetaTypeGuiHelper was generalized and renamed to QMetaTypeInterface.
From now all types will have common interface which can be used for
basic operations.
Change-Id: I50d67f4a8081fa0f75c9d530a8211593ec37bc55
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
This allows more granular reporting of what has changed.
This change is binary incompatible and source compatible.
Change-Id: I7c5beaee651a24780cc94e41383f7a80210bc603
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Methods of QMetaType; sizeOf, construct and destruct were reimplement
using QMetaTypeSwitcher. The change should have impact only on
maintenance (by removing 3 type switches).
Change-Id: Iac3eb52de75d46f12ac32f229a26aec5b5bc7af2
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
New implementation is slightly faster (~30%) and it is not dependent
on order of definitions in QMetType::Type. Which means that we do not
have to maintain a separate type list.
Change-Id: I0ceb8c919c08afaa55146fbc3f021b78ffc2c063
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Moving these classes to an unnamed namespace reduced chances of
a name conflict and size of executable. GCC 4.6 in release mode produces
a 5kB smaller binary.
Change-Id: Ie1819a88180f9db1cd32cc59ea6fe268ecbd5813
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QAbstractEventDispatcher() does no longer install itself into the current
thread. Instead the new methods QThread::setEventDispatcher() and
QCoreApplication::setEventDispatcher() allow to install a custom event
dispatcher into any QThread as long as there is no default event dispatcher
created yet. That is, before the thread has been started with
QThread::start() or, in case of the main thread, before QCoreApplication has
been instantiated.
Change-Id: I7367e13d8d8aebed5a5651260bb69b8818eb1b90
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
This change adds QMetaObject::IndexOfMethod as a parameter to
the qt_static_metacall function.
It lets the moc generated code return the index of a signal or slot
given its pointer to member function
This is required to support the new connection syntax
Change-Id: I39198c6699b5aa3599d3d282f7ac79b1e3684d33
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Added forward reference to QDataStream instead
In the new connection syntax, I want to include qmetatype.h from qobject.h,
in order to handle metatype for queued connection.
So qdatastream cannot be included here (as it inherit qobject) else it would
lead to circular inclusion.
QDataStream is only used in templated code that is dependent on the
template parametter so it is fine to have it forward declared.
Change-Id: I08ea479f481262a80451a277ee29533c42cb9f89
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@nokia.com>
When normalizing timevals, we need to bound the tv_usec member to
0-999999 inclusive, otherwise select() may return EINVAL.
When rounding timevals to the nearest millisecond in the UNIX event
dispatcher, pass the timeval through normalizeTimeval() when returning.
As discovered on Mac OS X with tst_QFutureWatcher, starting a 10 second
timer would end up calling select() with timeval = { 9l, 1000000 },
resulting in numerous "select: Invalid argument" warnings being printed
to the console.
Change-Id: Ic539e935bf847e0d4c22a73ad203e3a7a81d0690
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QMetaType::isRegistered and QMetaType::typeName are quite expensive.
To compare two custom objects we need to have their type name (for
dereferenced comparison). If the name exist we know for sure that the
type is registered and we do not have to call QMetaType::isRegistered
anymore.
Change-Id: Iba631e012504c8633868a902880fa30d38afb917
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
That was strange that QPolygon was know to QMetaType class but QPolygonF
not.
Change-Id: I67bb99644b8b1d0fe1aec730619d8048cb78219b
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Each QVariant instance has internal storage which may be
used for well-know basic types. This patch changes the
behavior by delegating type dependent operation to QMetaType
class which knows more types than QVariant itself.
The patch significantly reduce amount of code in QVariant
implementation.
There are few side effects of this patch:
- better performance:
* for Core types when using Gui (QGuiVariant is able to
construct Core types)
* for small custom types (QVariant::Private::Data is used for all
types that has size small enough)
- comparing two QVariants can give different result for small custom
types (binary comparison instead of pointer comparison)
Change-Id: Ic17fa500d6a882110bfba896fd456c8e6c7a63a9
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Currently one of the most common coding pattern, when working with
QMetaType types, is to switch over all types ids to convert a given
type id to a real c++ type. The pattern is not perfect, because of:
- code duplication - to convert type id to a real type a user has to
write own switch case
- maintenance - adding new type to QMetaType::Types doesn't propagate
to other parts of code.
Proposed type switcher can solve the issue by switching type id and
delegating found c++ type.
The class is created for internal use only and was created to simplify
common problems in QtDeclarative, QtScript and QVariant.
Change-Id: I0567ef908024b3b05ee18126f98a73a74748fbd2
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
From time to time it is necessary to know from which module a type comes
from. This information couldn't be part of QTypeInfo as id doesn't work
for forward declared types.
QTypeModuleInfo should stay for Qt internal usage only.
Change-Id: Ia53747c88293fc0cdaffef919b94c653412779d3
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QMetaType::Type enum is the main source of type ids. Currently
there are many tasks that can be replaced by a smart macro that
can iterate over all types. The patch introduces series of FOR_EACH_
macros that may be used for code generation.
As the first step the macro was used for Q_DECLARE_BUILTIN_METATYPE
to make sure that no type was forgotten.
Second step was to use created macros in autotest to improve tests
coverage.
Change-Id: I34e9ad7bacf02b44b028bc1aad20b1241aacebd3
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The minimum compile version is 0x050000 now, it cannot be less. Remove
checks for 0x040200 in the widget code, which provided compatibility in
the 4.x series for binaries compiled with 4.0 and 4.1.
Change-Id: Ia01da90de26dca6aeeb79d29c754cbeeecf809af
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
queued_activate() sets the argumentTypes atomic pointer on first use, which
mixes a load, memory initialization, test-and-set-ordered, and
another load. The explicit memory ordering is necessary to ensure
that the memory stores happen in program order.
Change-Id: Id1f8641f9cd081ce81aa8e830692f7af8261e84b
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This function is only used in src/tools/moc/moc.cpp and
src/corelib/kernel/qmetaobject.cpp. We don't need to include the
static, non-inline declaration and definition every time
qmetaobject_p.h is included.
This also silences the related warning from clang:
../../../include/QtCore/5.0.0/QtCore/private/../../../../../src/corelib/kernel/qmetaobject_p.h:171:19: error:
function 'normalizeTypeInternal' is not needed and will not be emitted
[-Werror,-Wunneeded-internal-declaration]
static QByteArray normalizeTypeInternal(const char *t, const char *e, ...
^
Change-Id: I6dfb2cb4d9d82a2ae7795f91169aa62f9a5f2c2f
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
If we have two timers running in an application with the same timeout
and started almost at the same time by the code, they would trigger two
sleeps, the second of which very short (under a millisecond).
This causes us to match the Glib and Windows event loops, which round
all timers to millisecond anyway.
Change-Id: I7eb531e02dadf75925c01192b0f33ef3641ae1ea
Reviewed-by: Olivier Goffart <ogoffart@kde.org>
Clang does not accept this syntax, see
http://llvm.org/bugs/show_bug.cgi?id=8875
The work around is to not use the typedef name. NS() around the
template parameters is necessary to match the real types (and
not the QMetaType::Type enumerators), otherwise we get the following:
kernel/qmetatype.cpp:1647:72: error: template argument for template type
parameter must be a type
...NS(QVariantMap)* >(where)->NS(QVariantMap)::~QMap<QString, ...
^~~~~~~
Change-Id: I3afa0cbbe4ef7ad899cfa9eafb3bcc10bedc20b3
Reviewed-by: Wolf-Michael Bolle <wolf-michael.bolle@nokia.com>
Reviewed-by: Thiago Macieira (Intel) <thiago.macieira@intel.com>
Q_WS_QPA is the only active code path after merging
refactor, other Q_WS-macros are no longer used.
Enable compilation without -qpa.
- Remove Q_OS_MSDOS, Q_OS_OS2
- Remove Q_WS_QWS
- Remove/replace definitions/conditionals of Q_WS_XX
- Remove qpa branches from profiles
- Replace Q_WS_MAC by Q_OS_MAC
- Replace Q_WS_MAC && !Q_WS_QPA by
Q_OS_MAC && !QT_NO_CORESERVICES
- Similarly in profiles: mac:contains(QT_CONFIG, coreservices)
- Replace Q_FS_FAT by Q_OS_WIN
Change-Id: Icce5a6c55b052c8f72b3b979ddf31a4f388ea9c9
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Add a ok return value for whether found or not.
Task-number: QTBUG-21672
Reviewed-by: Olivier Goffart
Change-Id: Ic0ea7455dccf1ac91705bcc1479444eb4091ded3
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Liang Qi <liang.qi@nokia.com>
By making it possible to specify the place in memory where a
type should be constructed, any meta-type can be allocated on
the stack, for example. In the QML/JS QObject binding, this
makes it possible to call slots and access properties from
JavaScript without having to perform any mallocs (e.g. due to
QVariant creation) in the C++ <--> JS value conversion, in
the best case.
In addition to QMetaType::construct() and QMetaType::destruct(),
this change introduces QMetaType::typeSize(), which returns the
size of a type in bytes. This can be used to prepare a suitable
buffer for constructing a type using construct().
Benchmarks indicate that in-place construction is 2-5x faster
than normal construction for core and GUI types on linux-g++.
Note that there is already a QMetaType::construct() function
in Qt 4, which has been renamed to QMetaType::create() in Qt 5.
In order to avoid existing usages of construct() in user code
to call the Qt 5 construct() (when they really meant to call
create()), the third argument ("copy") of construct() is made
mandatory. Hence, calls to QMetaType::construct() written for
Qt 4 will cause a compile error when compiled with Qt 5, and
the user must adapt his code.
Task-number: QTBUG-12574
Change-Id: I836f06f6ee1c1c3edbd199a03424c78c942bdd3e
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Kent Hansen <kent.hansen@nokia.com>
Jędrzej Nowacki
Stephen Kelly
Bradley T. Hughes
Robin Burchell
Jonas M. Gastal
Jason McDonald
Denis Dzyubenko
Alan Alpert
Olivier Goffart
Giuseppe D'Angelo
Frederik Gladhorn
Kent Hansen
Martin Jones
Friedemann Kleint
Debao Zhang
hjk
Toby Tomkins
Sven Anderson
Kevin Simons
Casper van Donderen
Thiago Macieira
David Faure
Liang Qi