| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
| |
Apparently Nokia X86 compiler is not able to use templated qCompare
when given arguments have different different type but same base class.
This error should be isolated and reported to Nokia X86 team.
See task: 259508
|
| | |
|
| |\ |
|
| | |\ |
|
| | | | |
|
| | | |\ |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The change mostly removes duplication and does minor touchups.
Reviewed-by: axis
|
| | | | | |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This test hasn't been updated in years, and never worked in the first
place.
RevBy: Trust me
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
causing applications to panic with KERN-EXEC 3. Fixed by providing
accessor for the bool in question instead of exporting it directly.
Note: Unknown why exactly this panic occurs, as exporting bools works
just fine when tested with small test executable + dll. Something
in QtGui causes this to fail; I suspect it might be the massive
code segment size and relocation of the data segment that is
contributing to this problem.
RevBy: Jason Barron
|
| | | | | |
|
| | | |/
| |/|
| | |
| | | |
The cast/constructor is not needed anymore due to commit 5b80fbad.
|
| | |\ \ |
|
| | | | | |
|
| | |/ /
|/| | |
|
| |/ / |
|
| | |
| |
| |
| |
| |
| |
| |
| | |
Qt/Master uses qt-test-server config by default but our S60/master
was configure to use different server by default. The assumption is
that this caused several QtNetwork autotests to fail when testing
QtS60 branch on other configurations in QtSW premises. This because
the old test server was very likely unavailable...
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QDesktopWidgetPrivate::rects etc variables are shared between
QDesktopWidgetPrivate objects and private objects maintains
refcount to data.
We called cleanup for private object directly from QDesktopWidget,
which caused the shared private data to be freed after some instance
of QDesktopWidget was destroyed. Since we bypassed the refcount also
the data for other QDesktopWidgets was invalidated and this cauyse crash
when resizeEvent was received.
There is no need to call QDesktopWidgetPrivate::cleanup directly since
when QDesktopWidget is deleted, it will call destructor for
QDesktopWidgetPrivate which deletes the shared data based on ref count.
|
| |\ \ |
|
| | | |
| | |
| | |
| | |
| | | |
Compiler would for some reason choose operator bool()
instead when comparing subclasses of QScopedPointer.
|
| | | |
| | |
| | |
| | |
| | | |
Stop confusing MSVC by correctly forward declare QScopedPointerDeleter
as struct
|
| | | | |
|
| | |\ \
| | |/ |
|
| | | |\ |
|
| | | | | |
|
| | | |\ \
| | | |/ |
|
| | | |\ \ |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The default widget size is normally calculated from the root window size, but when running on an X server with Xinerama enabled, the root window spans all screens, making the widget far to large.
Use the default screen's size instead to give the widget a size that
should fit on a single screen.
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
compiler mangles "struct" and "class" differently, so make sure that
QTestCharBuffer is always forward declared as "struct".
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
this line was dragged in from the Xml schema branch, but there is no
schematron in Qt.
Reviewed-by: TrustMe
|
| | | | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The code block associated with QBENCHMARK_ONCE macro will only be executed once,
whether the backend returns a valid result or not. The "-iterations" command line
argument is also ignored.
This is useful for benchmarking code that has side effects.
Revby: jasplin
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The documentation implies that start restarts. Which it did, but only if the
timeline was finished _and_ the duration hadn't been changed meanwhile.
So after a setDuration(), start() would do nothing, which was unexpected.
Merge-request: 1145
Reviewed-by: Olivier Goffart <ogoffart@trolltech.com>
|
| | | | |\ \ |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | | |
Reviewed-by: Aaron Kennedy
|
| | | | | |\ \ |
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
isn't available.
Reviewed-by: Michael Goddard
|
| | | | | |/ /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
The docs said that the default axis was (0, 0, 1), but the code
and unit tests were using (0, 0, 0). Modify the code to match the
docs because (0, 0, 0) is not very useful. Also optimize the
calculation of sin/cos values for 90, 180, and 270 degrees.
Reviewed-by: Aaron Kennedy
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Custom OpenVG pixmap filters written by users will need to access
the VGImage form of a QPixmap. The qPixmapToVGImage() function
is provided for this purpose, as a private API.
Reviewed-by: Sarah Smith
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
can not be found change that behavior to warn by default.
Currently the default behavior of include() in a .pro file is not to
warn if the supplied file argument can not be found which can lead to
hard to find build errors. The include() will now,by default, warn if
the specified file can not be found.
If a warning is not required because the included file is optional then
example use in the .pro file:
include(SomePriFile.pri", "", true)
Task-number:259398
Reviewed-by:Marius Storm-Olsen
|
| | | |\ \ \ \
| | | | |/ /
| | | |/| |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Conflicts:
src/corelib/kernel/qobject.cpp
src/corelib/tools/qsharedpointer_impl.h
src/gui/widgets/qdatetimeedit.cpp
src/gui/widgets/qlinecontrol.cpp
src/gui/widgets/qlineedit.cpp
tests/auto/qcssparser/qcssparser.pro
tests/auto/qicoimageformat/tst_qicoimageformat.cpp
tests/auto/qmultiscreen/qmultiscreen.pro
tests/auto/qresourceengine/qresourceengine.pro
tests/auto/qresourceengine/tst_qresourceengine.cpp
tests/auto/qscriptv8testsuite/tst_qscriptv8testsuite.cpp
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
in a tree-like model with one of its grand-children being selected.
Added recursive deselection for the model.
Task-number: 232634
Reviewed-by: thierry
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Before this patch gcc would fail with this message:
internal compiler error: in add_virtual_operand, at tree-ssa-operands.c:1317
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@nokia.com>
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
gcc was complaining about format string not being a string literal.
Also fixes the small chance that error messages would be bogus if
they contain printf control characters.
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | | |
Reviewed-by: gunnar
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Use a static buffer for small strings, and making it oom safe.
We can now see messages up to 512 bytes even if we run out of memory
(important for OOM tests). Also, testlogging (< 512 bytes per line)
should again work without a single allocation.
Reviewed-By: Rohan McGovern <rohan.mcgovern@nokia.com>
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | | |
Reviewed-By: TrustMe
|
