| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Symbian loader does not allow DLL import relocations in the data
section. To workaround this, added an accessor function which can be
used to get the class static metadata from the base class even if it is
in another DLL.
Changing the implementation for all platforms would cause a binary
break, so it is only used for Symbian. The Q_NO_DATA_RELOCATION macro
is used to configure this feature instead of Q_OS_SYMBIAN, because
it is possible another platform may need the same fix and Symbian may
eventually fix their DLL loader to support this feature.
Task-number: 258893
Reviewed-by: Kent Hansen
|
|
|
|
| |
Reviewed-by: trustme
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Rectification after the last merge request integration
Reviewed-by: Thierry
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit deprecates the QAIM::reset() method, and adds beginResetModel()
and endResetModel() methods, addressing Qt issue 247023.
http://www.qtsoftware.com/developer/task-tracker/index_html?method=entry&id=247023
If models and proxies use QAIM::reset() alone, then proxies will
emit modelAboutToBeReset after its source model is reset. This means that mapToSource
will not behave as expected (Will always return an invalid index) in a slot connected
to modelAboutToBeReset.
The usecase for this is maintaining viewstate (which items are selected, expanded)
when the model is reset. See BrowserWidget::modelChanged here:
http://websvn.kde.org/trunk/KDE/kdepim/akonadi/akonadiconsole/browserwidget.cpp?view=markup
Task-number: 247023
Reviewed-by: Olivier Goffart <ogoffart@trolltech.com>
Merge-request: 1072
|
|
|
|
|
| |
Task-number: 218125
Reviewed-by: Trust Me
|
| |
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This adds the function beginMoveRows, endMoveRows, beginMoveColumns,
endMoveColumns
Reviewed-by: Olivier Goffart <ogoffart@trolltech.com>
Acknowledged-by: Thierry
Merge-request: 972
|
|/
|
|
|
| |
Task-number: 201882
Reviewed-by: Trust Me
|
|
|
|
|
|
|
| |
This fixes the lifetime of the mutex locker, since before it would
already be destroyed on the next line.
Reviewed-by: Olivier Goffart <olivier.goffart@trolltech.com>
|
|
|
|
|
|
| |
We export one extra private symbol from QtCore instead, and use that.
RevBy: Miikka Heikkinen
|
|\ |
|
| |\ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
After review comments, changed the file server session
from a member of QCoreApplication to a Q_GLOBAL_STATIC, because some
applications will want to access files before constructing the
QApplication.
Reviewed-By: Janne Anttila
|
|\ \ \ |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The service provider (RTimer) will do it for us.
RevBy: Iain
AutoTest: Passed
|
| |\ \ \
| | | |/
| | |/| |
|
| | | | |
|
|/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This fixes click-focus for the Embedded Dialogs demo and the Pad Navigator
demo. In Qt 4.6, the behavior of the spontaneous-bit in events has
changed. This has happened in a number of submits that also modify
Graphics View to have the correct behavior. Somewhere on the line,
the spontaneous bit has been reset, preventing QApplicationPrivate's
ClickFocus handling to kick in. This again causes the Embedded Dialogs
demo to not give focus to widgets that are clicked.
This fix is not likely to be 100% correct. What it does is ensure that
the spont-bit survives a call to qt_sendSpontaneousEvent (as this bit
is for some reason reset at the end of QApplicationPrivate::notify_helper),
and this fixes the most apparent regression in ClickFocus handling.
Pending closer review.
The fix includes an autotest that should be kept if the fix is reverted.
Reviewed-by: bnilsen
|
|/ /
| |
| |
| |
| |
| |
| | |
The compatibility method QMetaObject::connect did not handle the
signal_index = -1 to connect to all signal
Reviewed-by: Thiago
|
|\ \
| | |
| | |
| | |
| | | |
Conflicts:
src/corelib/io/qprocess.h
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
model indexes
Reviewed-By: Olivier Goffart
Task: 231608
|
|\ \ \
| |/ / |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When the variant is invalid the shared is not destroyed.
We even can avoid the creation of the PrivateShared if we know the
variant is invalid
Reviewed-by: Thierry
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Constructing a variant of a pointer type was previously quite slow:
QVariant had no information it was a pointer. It had to create a
QVariant::PrivateShared, ask the QMetaType for a constructor (slow as it
involve locking mutextes) and allocate a pointer.
By detecting a pointer in qVariantFromValue, we can store the pointer
value dirrectly in the Variant union. We then avoid 2 allocations,
and the expensive locking in QMetaType::construct
Reviewed-by: Thierry
|
|\ \ \
| |_|/
|/| | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
operators and 64 bit values in QVariant. This is probably a compiler bug. Workaround is flagged with Q_CC_RVCT specific just in case, but should work for all compilers."
This reverts commit 85a1f7888bd48d5857de2b7c6304a334350fd054.
Conflicts:
src/corelib/kernel/qvariant.cpp
tests/auto/qvariant/qvariant.pro
I cannot reproduce this anymore with the latest RVCT, not even in
S60 3.1.
RevBy: Miikka Heikkinen
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This was done because I initially thought that KRequestPending was a
bit flag, but it turns out it's just an int value, so == is more
correct.
RevBy: Iain
AutoTest: Passed
|
| | |
| | |
| | |
| | |
| | |
| | | |
Private directory is not always in C-drive.
RevBy: Trust me
|
| | |
| | |
| | |
| | | |
RevBy: Trust me
|
| | |
| | |
| | |
| | | |
Reviewed-By: jbarron
|
|\ \ \ |
|
| |\ \ \
| | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | | |
Conflicts:
examples/painting/svgviewer/files/bubbles.svg
src/corelib/kernel/qobject.cpp
src/network/kernel/qhostinfo.cpp
tests/auto/qhostinfo/tst_qhostinfo.cpp
|
| | | |
| | | |
| | | |
| | | |
| | | | |
The code where the infinite recursion could appears should only be
called by code generated by very old version of moc
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Was introduced by commit 919b723
This function is not even supposed to be called anymore by anything
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Make some functions static that are not used anywhere but in the
current file. Others that are used, add the declaration to the _p.h to
ensure we don't forget about them.
Finally, there's no need to enable debugging code if it's not used
anywhere.
Reviewed-by: TrustMe
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-by: Thierry
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
... by not allocating space for slots in the vector.
Before, the vector uses the signal index as index. The problem is that
the slots and signal are mixed in the same index space.
We solve the problem by having a different index space for the signal in
the connectionlists vector. All we need to do is to add the information
about the number of signals in the moc.
Also, we are not connecting to cloned signal but only to the orginial
ones. For example, destroyed(QObject * = 0) would generate two signal,
we now only connect to the first one.
This also improve a little bit the performence while activating signals
since it removed one call to indexOfMethod.
Reviewed-by: Brad
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
be handled by connected views/proxies. Otherwise, the views/proxies may
end up in an invalid state.
Task: 227718
Reviewed-By: Olivier Goffart
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-By: TrustMe
|
| | | |
| | | |
| | | |
| | | | |
Reviewed-By: TrustMe
|
|\ \ \ \
| |/ / / |
|
| | |/
| |/|
| | |
| | | |
Otherwise, the qWarning will be bogus.
|
| | |
| | |
| | |
| | | |
RevBy: Trust me
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Opening and closing sessions has a performance impact on Symbian,
so all users of RFs now share a single session.
If a native file engine is written for Symbian (instead of the posix
one currently in use) then that could own the global data instead of
QCoreApplication.
Task-number: 247617
Reviewed-by: jbarron
|
|\ \
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
doc/src/examples.qdoc
doc/src/plugins-howto.qdoc
doc/src/topics.qdoc
examples/phonon/musicplayer/mainwindow.cpp
src/3rdparty/freetype/src/base/ftobjs.c
src/corelib/global/qglobal.h
src/corelib/tools/qalgorithms.h
src/corelib/tools/qshareddata.cpp
src/corelib/tools/qsharedpointer.cpp
src/corelib/tools/tools.pri
src/corelib/xml/qxmlstream.h
src/gui/painting/painting.pri
src/gui/widgets/qdatetimeedit.cpp
tests/auto/qdesktopservices/qdesktopservices.pro
tests/auto/qgraphicsitem/tst_qgraphicsitem.cpp
tests/auto/qtextcodec/test/test.pro
|
| |
| |
| |
| |
| | |
Makes sure that we can bootstrap QtCore easier on weird platforms
without having to run syncqt first
|