Returning NSNotFound from the NSTextInputClient selectedRange
implementation when there is no selection prevents
dictation from activating (for unknown reasons).
Return an empty {0, 0} range instead. Text input
methods such as Pinyin still work after this change.
[ChangeLog][macOS] Speech to text dictation now works
for Qt text input.
Change-Id: Ibf1729bdd271e8ed5ce3c9d2a0373c8ab3613d8e
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
In XCB environment the QClipboard::changed() was not delivered if the
QClipboard::clear() was issued by other Qt app/process.
If the QClipboard::clear() is used, then the owner in
xcb_xfixes_selection_notify_event_t is XCB_NONE, so we need make the
decission to handle this event by the selection_timestamp and our
m_timestamp[mode].
Task-number: QTBUG-56972
Change-Id: If4c486ac02223eac506465cac7ff1a07bd02a187
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This reverts commit 862fa24179,
which attempted to optimize QJsonObject::operator== under the
assumption that the entries it holds are lexicographically
sorted. They should be, because Object::indexOf() finds them
by binary search, but apparently both fromJson(), as well as
construction through op[] leave (some) entries unsorted.
This behavior should be fixed, because other code relies on
sorted entries, too, but until the problem is more fully under-
stood, revert the patch to unbreak equality comparisons.
Task-number: QTBUG-56843
Change-Id: I5b608c16d1bbcb4f01b75ce93bd58b49ff050be2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This strips leading whitespace from asset catalog filenames, which was
causing installation to fail due to incorrectly calculated paths.
Task-number: QTBUG-57090
Change-Id: I80db627262f9d58f4403e2d8ab205bef2113992b
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
The test created a Windows shortcut (.lnk) and checked on its existence.
It was not found in the first test since QFileSystemModel returned
the resolved file name (linktarget.txt). When fixing this by querying
QFileSystemModel::fileInfo()::fileName(), the 2nd test failed since
shortcut files are not considered system files.
Amends change 3b093034b6.
Task-number: QTBUG-53890
Task-number: QTBUG-20968
Task-number: QTBUG-29403
Change-Id: Iec58b52532b44d12759eaa6c8d63a8a4dc8d1bc3
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
sysroot, spec, and xspec have no /src and /get variants.
Change-Id: I8548791f8ea6ba9fd9f10c35f914ed6badbea9d4
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
the autotest was also broken, because it was created by pasting the
bogus message into the result ...
Change-Id: I02b8663b96c7d96cdb3c19639e2213e49fd2bcec
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
qt_x11_set_global_double_buffer() is dead code since Qt 5.
Change-Id: Ie9a33b6f03dc2e39f12bc790292bb0d227f05c44
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Tests are not supposed to write into the build/application directory,
but rather should output to the temp directory.
Change-Id: Idcdf51226a2d547514aea2fbb2054998d8a3437e
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Add better boundary checks and catch (hopefully all)
cases where invalid binary JSON could cause crashes.
Change-Id: I206510b7c5e3ba953802a5f46645878e65704ecc
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Do not move on to glGet since the behavior is undefined.
Task-number: QTBUG-48986
Change-Id: Ifd279635ed1b8441f92697965d15ae3ecb59a7e3
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
Recent Khronos headers decided to break the world by guarding all
function prototypes with GL_GLEXT_PROTOTYPES which has traditionally
been used for extension headers only.
Until this gets corrected - see
https://lists.freedesktop.org/archives/mesa-dev/2016-September/128654.html
- add the define to the config tests and qopengl.h.
While 5.7 already has some of the qopengl.h fixes due to an upgraded
ANGLE shipping with newer headers, this is a cross-platform issue that
will surface everywhere eventually. Therefore we target the full set
of fixes to 5.6.
This time we also make sure the forced define of GL_GLEXT_PROTOTYPES
is removed before including the ext header, thus apps get the ext
protos only if they actually requested them.
Task-number: QTBUG-56764
Change-Id: Ib2c6d2e7b71b8fb8683424f43e6289e64e4ee46c
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
That relation is incorrect. We can have tslib support without evdev.
Task-number: QTBUG-54998
Change-Id: I2c09bad73210fa4a13000077480c70365d482e2e
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
No-break-spaces should not be counted in the space data, but rather
be treated as any other non-breakable character. We were already
taking care of this in the loop we reach if the item starts with
a character which isn't whitespace, but there is a second loop for
items that begin with whitespace characters.
The result of this was that in certain circumstances where you gave
the nbsp its own format and made the line wrap, the previous line
would count an extra trailing space and it would swallow the first
character in its following line.
[ChangeLog][QtGui][Text] Fixed a bug where a no-break space would
sometimes cause the first character of the containing line to not be
displayed.
Task-number: QTBUG-56714
Change-Id: Idd760a389052e6de70f6cc397122b217987fa5f2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Someone cut-and-pasted but forgot one of the changes.
Change-Id: I647dc8117ebfe8ce3d4b26d468b80c15d4e533e8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Parameter in setPageSize is QSizeF. Without a specified unit it is hard to know
what to use as input. Units depend upon the underlying paint device
Change-Id: If001b3e9587d6085cc18017680fa20396e936adb
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Reviewed-by: Nico Vertriest <nico.vertriest@theqtcompany.com>
Reviewed-by: Steve Schilz <sschilz@pasco.com>
QLabel already does that for QPixmap, so just do the same for
QMovie's current pixmap.
Task-number: QTBUG-48157
Change-Id: I7b26460f778e56ff017a5efd433f8929f30e4b41
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
We do not need to check for an invalid timer id, as this can only happen
if the above check is already true. Hence, this was doing the same check
twice.
Task-number: QTBUG-56756
Change-Id: Icca9b26c32ce88eab76dd02c6c10b24af07bfad7
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
The intention has probably been to reset a reference, which is not
required.
Task-number: QTBUG-56756
Change-Id: I1ef44b6c9b8365ac5c8d48234137e518558e9398
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Same change as in QColor (d38f86e50b).
No text and data reduction numbers (ubsan build).
Change-Id: I7280a511e785c9442a3a6a1ed55e10011ce0a84e
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The old code employed a switch statement to filter timer events,
but fell unconditionally through to the default case of calling
QObject::event(). The final return statement following the switch
is thus dead code.
Fix by turning the switch into an if and returning QObject::event()
unconditionally afterwards, which much better describes the intent
of the code, and also fixes the GCC 7 warning about implicit fall-
through in the switch (which wasn't implicit to a human, but GCC's
comment-reading-capabilities are somewhat limited at this point).
Change-Id: I6756a65b3679a446d09fd721dfd0adc24fdf7772
Reviewed-by: Sérgio Martins <sergio.martins@kdab.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
If the QEvent::GraphicsSceneMousePress case falls through, it does so
because d->hasDecoration() == false or the virtual call to
windowFrameEvent() returned false.
It falls through to the case for QEvent::GraphicsSceneMouseMove, etc,
which ensures d->windowData and then checks hasDecoration() again,
with some other conditions on top, and calls the same virtual
function, windowFrameEvent(), with the same arguments again.
Now, it could, theoretically, be possible that that second call would,
due to the presence of a windowData that wasn't there before, return
true when before it did return false. But the only modification to
*this between the calls to windowFrameEvent() is the potential
allocation of d->windowData, which, if actually effected, will have
d->windowData->grabbedSection == Qt::NoSection, hence
windowFrameEvent() won't even be called a second time
It is therefore safe to assume that a break was intended here, so add
it.
Discovered independently be GCC 7 and Coverity.
Coverity-Id: 11149
Change-Id: Id708a1689ed0f0c914622e388c456ea4576fda02
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Use QDataStream to obtain cache key for a palettes that are different
from the default QPalette. This results in unique keys for palettes
created from QStyleSheetStyle's render rules.
Task-number: QTBUG-56743
Change-Id: Icbfe165f705ef3e1c9e88cfc9dca88ff1d1e81e6
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
The existing cursor logic had a couple of issues:
- It made the faulty assumption that we could not use
the NSWindow invalidateCursorRectsForView API for
child NSViews.
- It used NSWindow invalidateCursorRectsForView and
NSView resetCursorRects. This API has been replaced
by the more general NSTrackingArea API.
- It did not implement falling back to the parent
window cursor if the current window has no cursor
set.
Document that QWindow cursors work the same way as
QWidget cursors in that a QWindow with no set cursor
will fall back to the parent window cursor.
Change the cocoa platform code to use NSTrackingArea
exclusively and implement NSView cursorUpdate which
sets the cursor. Handle immediate change on QWindow::
setCursor() manually.
Add QWindow::effectiveWindowCursor() and
applyEffectiveWindowCursor() which finds the correct
window cursor.
Add a manual test for the child window, child widget,
and QWidget::createWindowChild cases.
Task-number: QTBUG-33479
Task-number: QTBUG-52023
Change-Id: I0370e11bbadb2da95e8632e61be6228ec2cd5e9d
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
There is logic for clearing the qgl_curent_fbo pointer
in release(), but it is not always called, causing
the pointer to become stale on QOpenGLFramebufferObject
deletion.
As a last resort, clear the qgl_curent_fbo pointer
on the current context if it’s pointing to the object
that is being deleted.
Change-Id: I36cca511da295412332193524219e32607ef8261
Task-number: QTBUG-56639
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
A pending interrupt of a QEventLoop may interfere with
native runModal calls, resulting in Cocoa's main event
loop to be stopped unexpectedly.
After commit 9ab60b9c processEvents() no longer resets
the event dispatcher interrupt flag.
Add QCocoaEventDispatcher::clearCurrentThreadCocoa
EventDispatcherInterruptFlag(). Use it to clear the
interrupt state before calling runModal and variants.
Work around the inability to use platform API in
the print support code.
Change-Id: I52f26f99a63cbb46969db42f65b09a3c3119ad15
Task-number: QTBUG-56746
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@qt.io>
KDE will set the screen scale factors to 1 by default. Make sure
we don't turn off font hinting in that case.
Task-number: QTBUG-56797
Change-Id: Ieab18a7cfe4c1cb7087caab4d881932a4a991bc8
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
The timerInfo list might get accessed concurrently and cause references
to become dangling. Hence, we need to protect usages with a mutex.
According to tests/benchmark there is no impact on performance.
Task-number: QTBUG-56756
Change-Id: I4bdffccff70d2dca99f4a39defad438afe571ada
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Commented its purpose and the guarded members for
readMutex.
Fixed places where guarded members were accessed without
using the mutex.
Use QMutexLocker instead of manually (un-)locking the
mutex.
Task-number: QTBUG-44357
Change-Id: I3049bb0df30f00659dc284c8e30ad7503c11e7c6
Reviewed-by: David Faure <david.faure@kdab.com>
the CI obtains them from the qt5 super repo nowadays.
Change-Id: I146e6a74763f32bee6651f427dd3664a1236ea0e
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@qt.io>
Make qmake understand the /DEBUG:FASTLINK option in QMAKE_LFLAGS, and
write the corresponding value correctly to VS 2015 project files.
Task-number: QTBUG-55591
Change-Id: I670375ed1523a5ab96bb3cce28635785564edba8
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Check all places where we reallocate our internal data structure
and return a DocumentTooLarge parse error if we can't get enough
memory.
Change-Id: I006d0170d941837220c7dad0508571b68e2cbfd7
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Kati Kankaanpaa <kati.kankaanpaa@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
POSIX does not require that readdir() be reentrant even for operations
on different dirent objects, but all implementations (according to the
glibc documentation) already do that. Moreover, it's not a good idea to
use readdir_r since the buffer space is limited by the caller, so
certain file names may be too long (ENAMETOOLONG) -- we had a workaround
for QNX, but for no other OS. According to the glibc documentation, it
is expected that POSIX will mark readdir_r obsolete and instead require
some form of reentrancy for readdir.
This commit makes everyone use readdir instead. The macros in
qplatformdefs.h are left behind in case someone else is using them.
With glibc 2.24, we started getting:
qplatformdefs.h:150:35: warning: ‘int readdir_r(DIR*, dirent*, dirent**)’ is deprecated [-Wdeprecated-declarations]
qfilesystemiterator_unix.cpp:112:17: note: in expansion of macro ‘QT_READDIR_R’
Task-number: QTBUG-56088
Change-Id: I33dc971f005a4848bb8ffffd14749b4082f62e69
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: James McDonnell <jmcdonnell@blackberry.com>
There are two enums called "Operator"
qdrawhelper_p.h:201:8: warning: type ‘struct Operator’ violates the C++ One Definition Rule [-Wodr]
qopengl.cpp:138:6: note: a different type is defined in another translation unit
Change-Id: I09100678ff4443e6be06fffd1482da1f636614b7
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
The blend_tiled_argb and blend_tiled_rgb565 was not correctly handling
widths larger than the buffer size. This patch adds the same pattern
used in blend_tiled_generic, which worked correctly.
Change-Id: Ie22c2a21d96cb0477cd0990bf01451ab907a4768
Task-number: QTBUG-56364
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
QPluginLoader hasn't unloaded in its destructor since Qt 5.0, but we
missed the equivalent code in QFactoryLoader (which bypasses
QPluginLoader). Besides, QPluginLoader::unload() was still doing
unloading, which it won't anymore.
Not unloading plugins is Qt's policy, as decided during the 5.0
development process and reaffirmed now in 5.6. This is due to static
data in plugins leaking out and remaining in use past the unloading of
the plugin, causing crashes.
This does not affect QLibrary and QLibrary::unload(). Those are meant
for non-Qt loadable modules, so unloading them may be safe.
Task-number: QTBUG-49061
Discussed-on: http://lists.qt-project.org/pipermail/development/2015-November/023681.html
Change-Id: I461e9fc7199748faa187ffff1416070f138df8db
(cherry picked from commit 494376f980)
Discussed-again-on: http://lists.qt-project.org/pipermail/development/2016-October/027476.html
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Dereference after null check (FORWARD_NULL)5.
var_deref_op: Dereferencing null pointer pe.
pe pointer it's being checked at the beginning of the function so it
implies pe might be null.
Coverity-Id: 11106
Change-Id: Ie88b27877a46cdd20a317fb5e21c3fdec1b99dda
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The same way it also mentions QStatusBar.
Change-Id: Ic084466310c989d1a79ba5ba21d6784acaa38e6e
Reviewed-by: Martin Smith <martin.smith@theqtcompany.com>
Update URLs in jennifer.xbel and remove frank.xbel, which
has too many outdated URLs
Change-Id: I8cbc1e3988f45230f849e68202780c76cf4353d9
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Scan lex and yacc sources for dependencies as if they were C source code,
which is close enough to reality.
This will unfortunately result in the generated source files depending on
the generated headers, while it should have been the object files created
from these sources which have that dependency. But qmake cannot do better,
and this is good enough.
Task-number: QTBUG-56507
Change-Id: Ic3e1941bf2e2820bfddf99deba854e1e82f83669
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
This change provides missing documentation for two member
functions of QString.
QString QString::fromUtf16(const char16_t *str, int size)
QString QString::fromUcs4(const char32_t *str, int size)
Change-Id: I94a9437a457062e49e4457f5876e4d7c31fff24c
Reviewed-by: André Hartmann <aha_1980@gmx.de>
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
