For some reason it seems to be supported to call moveToThread(0).
That call will allocate a new QThreadData for the object. However,
if we then detect that we're calling moveToThread from a thread
which is not the one the QObject has affinity with,
we error out leaking that QThreadData object.
Change-Id: I0fe6625a2a9887535e457f3897b514d2a03d1480
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
An early return in case of errors leaked memory tracked in a variable
in scope.
Change-Id: I68cd77890608caff54df7476d38850e5541ce76e
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
The QClipboard::Mode returned from modeForAtom should be checked
everywhere because values greater than Selection (i.e. FindBuffer)
aren't supported on X and should mean error conditions.
The lack of such a check did an out-of-bounds array access, which
could lead to a crash.
Change-Id: I70f70b5f713ab2f892e258d4df2f7afeb434f0c1
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Gatis Paeglis <gatis.paeglis@digia.com>
An early return caused a leak of a new'd allocated object.
Change-Id: I9fbc37238dd49066d24363a2e8ee8bf35b155301
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
This bug can be reproduced (AFAIK) only on 10.9. To fix it I suggest we create
NSOpenPanel/NSSavePanel
_every_ time it must be displayed. Actually, that's what I've seen in all code samples
I was able to found - nobody tries to retain this panel and re-use it.
If we re-use it, "Media" section magically disappears.
I believe this bug is not Qt's bug, but something weird in Cocoa.
Task-number: QTBUG-40655
Change-Id: Ic0e76e0a9a5444a76f336d511c0ff93f9fd05797
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
When ordering a key window out Cocoa
tries to find a new KEY window. Looks like it prefers the current MAIN
window and since QNSPanel is never a main window, Cocoa is breaking
the stack order. To avoid this - try to change the key
window BEFORE ordering out a window.
The application has a stack of all open windows (visible and hidden),
we iterate through this stack starting from our current key window
and look for the nearest window below, that can become a new key window.
Most probably, it will be our transient parent :)
This code will change (potentially) the key window _only_
if there is a transient parent.
Task-number: QTBUG-39809
Change-Id: I96b630799341875fc7e38dabf1ff6416338e687b
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
A regression was introduced by bdebec4e2e.
The intention of the change was to avoid using Qt's shortcut mechanism
to trigger menu items which were already triggered through the regular
menu API in Cocoa. However, Cocoa has trouble with key equivalents that
do not have any keyboard modifiers. Thus, we have to allow these
particular key sequences to go through the regular system. I've verified
that the original bug is still fixed with this change.
[ChangeLog][OS X] Fixed menu item shortcuts without keyboard
modifiers.
Task-number: QTBUG-41192
Change-Id: I8f5a9cbc7a448b3cb0519baed95be5cbb630205c
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Deleting a QWindow automatically deletes its child windows because
they are in the QObject hierarchy. However, if the user sets both
WA_NativeWindow and WA_DontCreateNativeAncestors, we can't just
delete that widget's QWindow. First because the widget doesn't get
notified (and maybe it should be), and then because we may invalidate
any reference to the QWindow the user may have kept.
Our solution is to reparent the child QWindows into the new parent's
closest QWindow. We must, however, take the precaution of not keeping
any reference to the backing store in the platform window. Reparenting
operations can trigger repaints on the platform window, but the backing
store is not set and flushed until later.
Task-number: QTBUG-38377
Change-Id: I353f5528f227a227b6d10419367cbe1d5d07a94e
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
socket(2) is allowed to return 0, so 0 should not be included
when checking for errors.
Change-Id: I0454ea60347d90078d3ab3046969add8d5c37935
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
device is private, always null and class has no friends, so no need to have it at all
Change-Id: I320d47f1a712a3202c08b494563533e29d185501
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
Some IMEs will perform an insertNewLine when Return/Enter is pressed while
composing text as well as causing the edit to finish. By handling this it
will ensure that the extra enter/return event is sent when the IME
requests it.
Task-id: QTBUG-39125
Change-Id: Ice6eded68e6b1dc51703a38316f76f78099923da
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
qatomic_sparc.h was removed previously as part of general cleanup of
old atomic code for Qt 5.3. Unfortunately SPARC include reference was
not removed resulting in build failure on that platform.
Task-number: QTBUG-41384
Change-Id: Ic6e31b32324b0e5dd3700a6a21515a8eea5668bd
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Qt sets a QCocoaMenuDelegate on every menu it creates in order to emit
the (OSX specific) aboutToShow signal. However, there are a few cases
where OSX will copy a menu without copying the delegate. One of those
cases is the dock: the result of -[NSApplication applicationDockMenu:]
is used to create a new menu, to which a few more items are copied.
This copy is then send back to the dock.
This patch invokes the delegate's -menuWillOpen: method when
-[NSApplication applicationDockMenu:] is called. Note that sub-menus
won't receive the call-back, because the dock doesn't tell the
application what happens after returning from applicationDockMenu:.
Task-number: QTBUG-39604
Change-Id: I0e06df371a3d77342ae4b7148041214e5c4579d7
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
QToolButton::sizeHint() takes into account the presence of a menu.
However, setMenu() doesn't retrigger a size hint recalculation. Hence,
(un)setting a menu on an already sized tool button won't properly
reset the size hint. Since the calculated size hint is cached,
delete the cached value and call updateGeometry to cause a
recalculation.
Task-number: QTBUG-38949
Change-Id: I6e79e5e70e31afdfd129282b3668875eca86f51d
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
The code transformed the coordinates from the view to the window's
content view, and since that content view is flipped in the pure
Qt world (content view == QNSView), it manually flipped the
coordinates to transform from that to window coordinates.
Instead just directly transform the view coordinates to window
coordinates using standard Cocoa methods, which then works with
any kind of content view and NSWindow configuration.
Task-number: QTBUG-40958
Change-Id: Idddd327fe9cff3309606379d0e04ee8b4bd5eece
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
And not a normal pointer. The problem is that in certain scenarios,
if the TLW containing a QSizeGrip changes and the old TLW gets
immediately destroyed, then the mechanism which updates the tracked
TLW is run too late, and ends up accessing a dangling pointer.
Therefore, we need to protect that pointer via a smart pointer.
Task-number: QTBUG-22867
Change-Id: Icfb051132bacde604f660ac7a98bc0a9d1022c68
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
thread-local variables are known to be broken on Windows XP if the
library is loaded dynamically. This has been fixed since Windows Vista,
but to avoid any elaborate runtime checking of the OS version we're just
removing the logic alltogether for MSVC / Windows.
Task-number: QTBUG-41008
Change-Id: I64ee95270d142294c975a5890d6e1b62a833b6ef
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Added the notion that enter and return keys are not handled
if editor is of type QTextEdit or QPlainTextEdit.
Change-Id: I9d5cb529bf154c687fab1949fccf37e9da096c85
Reviewed-by: Topi Reiniö <topi.reinio@digia.com>
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
Right now, we use them for inactive non-editable combo box, check
box, and radio button only on Yosemite. We keep as much as possible
the previous behavior on older versions.
In addition, we add a way for QQuickStyleItem to specify the window
the item is on. This is currently without effect, since we don't
seem to take the inactive window state into account.
Task-number: QTBUG-40833
Change-Id: I2fb2a99e6adf1972f881195b79b07ce85a960273
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
If the font engine for some reason fails to get font metrics
for the font, the application should still not crash.
[ChangeLog][Widgets][QPlainTextEdit] Fixed a possible divide
by zero crash when font metrics were missing for the font.
Task-number: QTBUG-40347
Change-Id: I571bc3eace07cdbee6f9ce9aa649df95412aed71
Reviewed-by: Simon Hausmann <simon.hausmann@digia.com>
Add QMAKE_CFLAGS_XCB to xcb-static,
needed for successfully building with -qt-xcb.
Add QMAKE_CFLAGS because some code gets built with the C compiler.
Also, this fixes a compile error
when xcb include is not in system include path.
Task-number: QTBUG-41129
Change-Id: I842ebe391a7a857e234becc451c33601fce9b07a
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
As fb_types.h says, Firebird requires (S)LONG to be 32 bit, and it defines
SLONG to int. This leads to sqllen being 4, so qsql_ibase.cpp allocates
4 bytes... and was writing 8 bytes into it.
Fixed by checking sqllen, the same way QIBaseResult::gotoNext does.
Change-Id: Ie8680d32f98c354dfc8430b8efbfe95450556956
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
In some cases, e.g. when bad credentials are provided in an ftp URI,
QNetworkAccessAuthenticationManager::cacheCredentials is called with a
null authenticator. This authenticator should not be cached, because
it is useless, and leads to inconsistencies in the use of the cache
Task-number: QTBUG-40622
Change-Id: If2a0a422b915f268648f5eef1d68601446123371
Reviewed-by: Peter Hartmann <phartmann@blackberry.com>
qmake variables using $$system() were incorrectly parsed by the custom
qmake parser in the configure script, when using GNU awk 3.1.8 or
earlier. They are parsed correctly with GNU awk 4 or mawk.
This was occurring with such an assignement (from an extra mkspecs file):
QMAKE_CC = $$system($$CMD QMAKE_CC 2>/dev/null)
The custom qmake parser in the configure script first attempts to
expand $$UPPERCASE variables, before running $$system(), using this:
match(value, /\$\$(\{[_A-Z0-9.]+\}|[_A-Z0-9.]+)/)
But when using non-ASCII locales with GNU awk 3.1.8 or earlier,
$$system was expanded (to an empty string) because these earlier awk
versions match lowercase letters for the [A-Z] regexp, which is
traditionally used to match uppercase letters.
This behavior has been changed in GNU awk 4.0.0, which only matches
uppercase letters for [A-Z] by default. A workaround for earlier GNU
awk versions is to run awk with the C locale.
See GNU awk NEWS "Changes from 3.1.8 to 4.0.0":
25. Gawk now treats ranges of the form [d-h] as if they were in the C
locale, no matter what kind of regexp is being used, and even if
--posix. The latest POSIX standard allows this, and the documentation
has been updated. Maybe this will stop all the questions about
[a-z] matching uppercase letters.
THIS CHANGES BEHAVIOR!!!!
See also gawk.info "A.7 Regexp Ranges and Locales: A Long Sad Story"
Change-Id: Ibb3eb28738c3e77d496c634e1f5c9f630957e730
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Since 5.0 it is possible to use one call to bindValue to bind
values to multiple placeholders with the same name.
Task-number: QTBUG-23360
Change-Id: Ic838150d25dd07bca7bc9e5d91ab3362a73833d6
Reviewed-by: Matt Newell <newellm@blur.com>
Reviewed-by: Mark Brand <mabrand@mabrand.nl>
This statement was left from when toAscii was used, instead of toUtf8.
There is no loss of information when converting to UTF-8.
Change-Id: Iad92977af319b324cbf06f1a24712b31a7836622
Reviewed-by: Louai Al-Khanji <louai.al-khanji@digia.com>
Reviewed-by: Robin Burchell <robin.burchell@viroteck.net>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Widgets and layouts added or inserted to a layout are checked for:
- Not being NULL
- Not being the parent widget of a layout or the layout itself,
respectively
Without this commit, adding a widget to its own layout would result in a
CPU-hogging infinite loop. Now, a warning is written to stderr and the
add or insert function call is ignored.
The checks are implemented as public functions of QLayoutPrivate and
thus accessible in QLayout's descendants to be used in various
"addWidget", "insertWidget", etc functions.
Unlike 'classical' layouts like QGridLayout, QFormLayout does indeed
accept widgets that are NULL. To not break this behavior, any call for
the check functions first tests if the widget or layout, respectively,
to test is NULL or not and calls the check only in the latter case.
Automated tests for QBoxLayout, QGridLayout, and QFormLayout were added.
For an unpatched Qt 5.3, each of those automated tests will freeze as
explained in QTBUG-40609. For a fixed version, warning messages about
invalid parameters to addWidget/addLayout/... calls will be read by
QTest::ignoreMessage, resulting in a passed test.
Change-Id: I1522d5727e643da3f7c025755975aca9f482676d
Task-number: QTBUG-40609
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
ATSFontNotify() allocates over 30MB for no apparent reason for
applications that don't register fonts.
Change-Id: I7bbc97f53b76be1e638de8ca0d1be3a08586cf00
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
This reverts commit f6c36917f1.
Depends on 203ce7c5ce
which can't go into 5.3 because it introduces new enum values.
Change-Id: I13040a7268fae4baeace244a3e75dc982d3a50b8
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Maintaining a black list of devices can't be reasonably done.
The workaround is now enabled by default. It can be disabled with the
QT_ANDROID_DISABLE_GLYPH_CACHE_WORKAROUND environment variable, in
which case the current list of blacklisted drivers will still be taken
into account.
Task-number: QTBUG-40483
Change-Id: I9b90ab45e86569e8e4679333ddb9b835189892c7
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
When 'cursorPos' is out of bounds ([0, lineEnd]), this method crashed.
Change-Id: Ia0540ab3afbffb5c598f7b8515263cce3b3928e4
Task-number: QTBUG-40753
Reviewed-by: Dominik Haumann <dhaumann@kde.org>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
A blocking isSoftwareKeyboardVisible() is a deadlock
waiting to happen: when the android input method performs
blocking metacalls from the android thread to the GUI
thread all the time, we cannot block the GUI thread
waiting for the android thread.
Task-number: QTBUG-40750
Change-Id: I2490897b0f65e0d92214907e239b10b372d949dd
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
For iOS8 and up [UIScreen bounds] changes based on the interface orientation,
so we need to use [UIScreen nativeBounds] instead.
Change-Id: I3fc12cfa417df26ca94c803e970bc2dc18a94378
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
As usual, HITheme is not helping, and renders inactive controls
as disabled. Also, given the design changes in Yosemite, we can't
just desaturate the active pixmap and render it. In this case, we
render the inactive control and enhance it, making very close to
the expected result. The pixel-exact version would be to render
a plain push button, and then add the combo box arrows. However,
these arrows would need to be created (and updated) since there
seems to be no more API to render them.
Task-number: QTBUG-40833
Change-Id: If1bc366c0bc83123972fabebbc8beeb9f071e7a1
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Some applications may use native file dialogs and some may use the
widget implementation, but QFileDialog always saves all settings to
the same place each time it destroyed. So it is important to remember
all settings read in restoreState and put them back in saveState.
This bug was introduced by 37ca2224ec
Task-number: QTBUG-36888
Change-Id: Ied1db83817480312e963a042784d7bdd16f0a2b1
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
Qt expects scroll operations to be executed immediately. They cannot be
postponed since Qt may paint on the scrolled area afterwards.
The new code will only use an accelerated scroll operation (screen_blit)
once before the window is posted, because the blit is from the previous
buffer to the current one. Hence an additional scroll operation could
copy outdated pixel data. Additional scroll operations will be handled
by Qt. Performance issues were not perceived with this approach so far.
Task-number: QTBUG-39958
Change-Id: I6d7c3274c5db6a831a169615518fcdb4d926db70
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
When setting a new screen, the code calls QWindow::destroy(), which
recursively destroys all child windows. It then calls create() on the
top-level window, leaving child windows destroyed. This causes crashes
if you have embedded native widgets.
Task-number: QTBUG-40817
Change-Id: Iaace2589f48bbfd5faaf5ff95357ff43b310504a
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
On iOS 8, the value 0 is returned as a kCFNumberIntType.
The code still works - CFNumberGetValue converts it
to a 0.0 double.
Change-Id: Ic50900b22e4fa19ad1481e8e0e293559bbfd8cd2
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
