In rare cases, the Windows event loop can be spinning inside the inner
loop and the message hook is never called. This can be triggered on the
Direct2D platform by opening 32+ window handles.
The issue can be worked around by using the same approach Windows CE uses:
don't rely on the message hook to inform the event loop that the post
message has been delivered. Instead, uninstall the hook and let it be
called directly by the event loop.
Task-number: QTBUG-42428
Change-Id: I10280126dd50729bc260aa5f7029549e2e061c01
Reviewed-by: Louai Al-Khanji <louai.al-khanji@theqtcompany.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
The commit was 9361be58f47ec256bf920c378479a02501219c1f (2008-11-17),
referring to the race condition fix that was applied in commit
d47c05b1889bb4f06203bbc65f4660b8d0128954 (2008-10-08). The fix for the
deadlock reintroduced the race condition and the commit message noted
it.
The workaround is no longer necessary since we've fixed the original race
condition differently now (see the previous two commits).
Task-number: QTBUG-42189
Change-Id: I5a83249597a83c4d4caa2ae57964ad3cc61c1d70
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Reviewed-by: Albert Astals Cid <albert.astals@canonical.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
We don't need two anymore because they now protect the same thing: the
state of the DBusConnection. The difference existed when it was possible
for two threads to access the DBusConnection at the same time: one doing
dispatching and one doing something else. Unfortunately, even though
DBusConnection supports this, QtDBus doesn't.
From d47c05b1889bb4f06203bbc65f4660b8d0128954 (2008-10-08):
Details: if we're removing a timer or a watcher from our list,
there's a race condition: one thread (not the QDBusConnection thread)
could be asking for the removal (which causes an event to be sent),
then deletes the pointer. In the meantime, QDBusConnection will
process the timers and socket notifiers and could end up calling
lidbus-1 with deleted pointers.
That commit fixed the race condition but introduced a deadlock.
Task-number: QTBUG-42189
Change-Id: I034038f763cbad3a67398909defd31a23c27c965
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Reviewed-by: Albert Astals Cid <albert.astals@canonical.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
We lock it before dbus_connection_send_with_reply (the async version) in
QDBusConnectionPrivate::sendWithReplyAsync. We weren't locking it before
send_with_reply_and_block and we apparently should. The locking around
the dbus_connection_send function might not be necessary, but let's do
it to be safe.
The lock now needs to be recursive because we may be inside
QDBusConnectionPrivate::doDispatch.
Task-number: QTBUG-42189
Change-Id: I7b6b350909359817ea8b3f9c693bced042c9779a
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
QString::sprintf does actually support all length modifiers, including
%lld. The format string is also parsed as UTF-8.
What's worthwile to mention, though, is that %lc and %ls is at odds
with the standard, since wchar_t isn't necessarily 16 bits wide.
Change-Id: I30cd22ec5b42035824dd98e3cdcc79d7adcc953a
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Most of the settings there end up overwritten by the clang.conf include
that comes afterwards, except for a few things such as QMAKE_LINK_C,
which remains set to "gcc" and breaks things when one uses
CONFIG=use_c_linker.
QMAKE_LFLAGS_NOUNDEF was coming from g++-unix.conf, though, so we now
manually set it in freebsd-clang's qmake.conf.
Change-Id: Ibd16f59d43eb19e72adf4919da9ce3007100b60f
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
While it does not look like the clang-based mkspecs had any problems so
far with not having QMAKE_LINK_C and QMAKE_LINK_C_SHLIB defined, it
makes sense to set them to $$QMAKE_CC just like the GCC-based ones so
CONFIG=use_c_linker works as expected.
Change-Id: Ib660d12b001dd7a877b6f03e79715db08a272968
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@theqtcompany.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Panels windows are usually outside QScreen::availableGeometry, because
they will usually set extended struts to reserve the screen area for
themselves, but their own screen() must remain the one in which they
are.
This cause one downstream behavior to KDE
https://bugs.kde.org/show_bug.cgi?id=339846
in which a panel got by mistake few pixels on another screen, and
was immediately reassigned to that screen, because its geometry was
intersecting the new screen availableGeometry() but not the geometry
of its own screen, because itself reserved its own geometry away
from availableGeometry()
Change-Id: If6c9defdef62732473687dd336dbcec582bd0ea2
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
The private was not deleted. Adding the dtor in turn
causes a warning about not having a virtual dtor
in the base class, so add that as well.
Change-Id: I24a90caf2cf6192a6f17cf5af96b8f77010d9127
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
This key combination should open the context menu.
Task-number: QTBUG-40085
Change-Id: I7cfc89f766b3734b9fb9d3c9135b4896ffbadb5b
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
Two fromstrerror_helper overloads are defined, to manage the fact that
strerror_r returns an int or a char* depending on the system. The problem
is that then only one overload used (again, depending on the actual
stderror_r return type), leading to one of the two overload to be unused
and thus triggering the unused function warning.
kernel/qsystemerror.cpp:64:27: error: unused function 'fromstrerror_helper' [-Werror,-Wunused-function]
static inline QString fromstrerror_helper(int, const QByteArray &buf)
Change-Id: I6a1c8e1a4b7d14068b682db26002ff68ad36167c
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The result of pressing the key is still a Qt::Key_Return press/release
sequence, which needs to be handled manually.
Change-Id: I72c7b0067bd3ec1bc315ab2c84361800b7be0943
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
Reviewed-by: Marius Bugge Monsen <marius@cutehacks.com>
The technique of sending an action does not always end up at the actual
first responder, but it will end up in a responder in the responder
chain of the first responder, so we continue searching the subviews
recursively until we find the real first-responder.
Change-Id: I6abc9bc18eb127fa4b317cd308783c0ecfcd670a
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Instead of faking it, by returning YES for isFirstResponder, which caused
issues when iOS would try to dismiss the keyboard by resigning the true
first-responder.
Change-Id: I816c4cf9c699d72995ce7968e1f1a4aa9c9c167e
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
In the non-native QFileDialog, a QSplitter separates the two central
views (the list view of "places" and the tree view for navigating
the file system).
Unfortunately, that splitter allows sections to be collapsed,
resulting in a weird status where the user doesn't
understand what has just happened and thinks that (s)he may have
broken something. Worse, that gets actually saved into the
application settings, so the splitter may stay collapsed forever.
Instead, let's simply prevent sections from being collapsible.
Task-number: QTBUG-19467
Change-Id: I11ff7c55a5535680a3edce5f4e70c9338291b94f
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
The old designer generator inserted stray spaces etc.;
save it again with the new designer to fix those.
Change-Id: I3890dd942970d9da71582ccb70b75d59888304bf
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
1. Add comment about the magic TYPE_NUMBER_VARIATION_PASSWORD value.
2. ImhHiddenText, ImhNoPredictiveText and ImhSensitiveData should all
disable suggestions.
3. Only ImhSensitiveData sets TYPE_TEXT_VARIATION_VISIBLE_PASSWORD
4. Don't OR date and time to get DateTime...
Task-number: QTBUG-38080
Change-Id: If456563983130e9af409ffa8cb717ddf259f1d6b
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
The JNI environment should be managed by QJNIEnvironmentPrivate
directly or through QJNIObjectPrivate. There is also a clear
difference between calls coming from or going into Java code.
Calls coming from Java already comes with the 'right' environment and
in most cases no extra considerations or set-up is needed.
Change-Id: I92d935ddfb70332041869185d5a92438930ff9b9
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
This is one of several fixes where the goal is to simplify the jni code
by letting QJNI manage the environment.
Change-Id: Ia714e25fbb3fcd170150392e822b0a3fc3812818
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
Let the QJNI classes manager the jni environment and caching of jni
handles.
Change-Id: I8c238375026adf449d6e6e2b521caa6cd63a0fb4
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
Enables QRunnables to be run on the UI thread. For now this function
is only intended for internal consumption.
Change-Id: I5e2abb06104219a9dd55b3308113056e4da5fa07
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
Gtk library usually takes care of this when the main
Gtk loop ends, but since Gtk's main even loop is not
used in QGtk2Dialog we have to store clipboard's content
with help of gtk_clipboard_store(). This function
sends a SAVE_TARGETS request to X11 clipboard manager
to save clipboards contents as required by ICCCM.
Task-number: QTBUG-34475
Change-Id: If784c425ea4a36ec1c3a8ddc0cdec080f57681a5
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
qglxintegration.h:78:17: error: private field 'm_screen' is not used [-Werror,-Wunused-private-field]
Change-Id: I7ffeda1dd4be962a4bfb43ebd3aa1dbd969a9837
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Note that at one point we need to use Cocoa to render the combo box, but
only if we're dealing with Qt Quick controls.
Also worth noticing, there's currently a bug in Cocoa when rendering
inactive combob boxes. We faithfully reproduce it in Qt for now. We'll
fix it when Apple does.
Finally, we need to start constraininig the combo boxes height. Cocoa
has not supported variable height combo boxes for years, and will even
spit the following warning if we try to do something smart.
This application is trying to draw a very large combo box, 28 points
tall. Vertically resizable combo boxes are not supported, but it
happens that 10.4 and previous drew something that looked kind of sort
of okay. The art in 10.5 does not break up in a way that supports that
drawing. This application should be revised to stop using large combo
boxes. This warning will appear once per app launch.
Task-number: QTBUG-40833
Task-number: QTBUG-42067
Change-Id: I6512a6a581d446a28585db22fe4dbeac09499321
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Defaulting to absolute_library_soname on configure -rpath is no longer
necessary as now we support @rpath install name ids on OS X and iOS.
This also sets QMAKE_SONAME_PREFIX to @rpath for Qt modules when built with
rpath configuration.
This makes Qt libraries relocatable on OS X. Qt SDK is not yet relocatable
though, because plugin location (including cocoa plugin) is still resolved
using absolute path (see QTBUG-14150), also there are several absolute paths
hardcoded in qmake mkspecs pri files.
Task-number: QTBUG-31814
Change-Id: Ie9dffefcd2a946c1580293d433621c1adb7e06c4
Reviewed-by: Jake Petroules <jake.petroules@petroules.com>
This is triggered only when app is using Qt and Qt was built with "rpath"
configuration and project does not specify QMAKE_RPATHDIR explicitly.
Added rpath is made relative to app binary location if target path lies inside
Qt SDK, so all SDK bundled tools and examples will work automatically without
any changes. Tests are an exception here, since they are being run from their
build location by CI, we may not use relative rpath that work only in install
location.
Task-number: QTBUG-31814
Change-Id: I3690f29d2b5396a19c1dbc92ad05e6c028f8515b
Reviewed-by: Jake Petroules <jake.petroules@petroules.com>
This was observed in the weather app, where sometimes we could not find
an items window. This could only be observed in the search results of
the cities. (while VKB was visible).
The old code traversed up to the QQuickListView and then it could not
traversed further up in the parent hierarchy. Because of this it
could also not find the associated window handle.
The reason for this is unknown, but maybe it could be related to the
fact that QQuickListView is a Component.
Regardless of this, invalidate the cache should invalidate everything.
We also traverse through all top level windows, but on iOS there should
not be too many top level windows...
Change-Id: I56a496435bb529a53d5ece8446cd2eeff502af84
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@theqtcompany.com>
I don't know why it was an #if 0. The __has_feature has been there for a
while. But, just to be sure, we check the presence of the header too.
Change-Id: I36e34c9e8fd4ce55c98966d2fad246b77eb16597
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The libc++ header does this:
#if !__has_feature(cxx_atomic)
#error <atomic> is not implemented
So we can't enable the feature until the compiler reports true for that
test.
Change-Id: I96f1c7eea8b93d93bd721fe5a85fa987339d091f
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QTestEventLoop::exitLoop() is used by QSignalSpy to stop event
processing when the connected signal has been received. The design
of QSignalSpy requires QTestEventLoop::exitLoop() to be
thread-safe, which it wasn't. When QSignalSpy is connected
to a signal in a different thread, exitLoop() was called from
the thread which emitted the signal and not the one in which
QTestEventLoop is running. This caused troubles when killing
the internal timer.
This patch adds a check in the beginning of exitLoop(). If
it is called from a different thread, it will post an event
into the message queue in which QTestEventLoop is running
and execute it there.
Change-Id: Icb8c8ff2f5344800ee6c6125b98c677c7a196c32
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Loading a dock window state saved by a Qt 4.2 app is not
something we need to support anymore.
Change-Id: I9ee6e2c742b31114081852e7236cfc8696b9b270
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
Tested with GCC 4.9, Clang from XCode 5.1 and ICC 15 beta.
Clang 3.5 (pre-release) cannot compile qtdeclarative yet with -Werror
due to invalid C++ code there that calls member functions on null
pointers.
Change-Id: Ic2845371a1899716985bc0813dfb820fa418e207
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: If4d9c9c769b598a3194a7cd5bbe5c74e7650694b
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
qtdelcarative's qquickaccessibleattached.cpp contains now some static
instance of QMetaMethod. Marking the constructor as constexpr,
let GCC to remove call to the constructor at load time.
Change-Id: Ic5ab7db0d06caa08f15d65d3bb5f22a34a111fee
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
We know that type id can't be changed, let pass this information to the
compiler.
Change-Id: I105b460417288b84250a954571c247608976f8f7
Reviewed-by: Stephen Kelly <steveire@gmail.com>
Previously, the device was delay-created, which is a problem
if its type is to be used for determinining the pan gesture type.
Task-number: QTBUG-40461
Change-Id: I2dee3d7a3786a0fdf0a9b2b9e174dd121697ab44
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
We shouldn't excluded all volumes under /run since some distos will
mount filesystems there. Instead we should exclude all filesystems with
the type "tmpfs" that /run has, and rpc_pipefs that is mounted below
/run. Tmpfs" is excluded for all UNIX systems since the BSDs have
a similarly named filesystem.
Change-Id: I03fdac515c0bfb1b824b2e3eae1022dd699c0998
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The patch 916dfcb827 while fixing one problem,
introduced another - leaking CGImage.
Change-Id: I08db6ea9fa97ae3489a0bfa1f93e0d18f3671885
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Oswald Buddenhagen
Andrew Knight
Thiago Macieira
Kai Koehne
Simon Hausmann
Raphael Kubo da Costa
Marco Martin
BogDan Vatra
Giuseppe D'Angelo
Friedemann Kleint
Tor Arne Vestbø
Christian Strømme
Nico Vertriest
Maurice Kalinowski
Gatis Paeglis
Gabriel de Dietrich
Adam Strzelecki
J-P Nurmi
Jan Arve Sæther
Roland Winklmeier
Lars Knoll
Olivier Goffart
Jędrzej Nowacki
Allan Sandfeld Jensen
Timur Pocheptsov