QXcbScreen installs its own event mask on the screen's root window.
This overwrites any existing event mask already set and by that
breaks applications when a new screen is added.
By first fetching the existing event mask and adding it to the newly
installed event mask, Qt does no longer break applications also
installing an event mask on the root window.
Task-number: QTBUG-39648
Change-Id: I8686dd6ae49d0e807c6fe1ea4a231ff728bfcf25
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Gatis Paeglis <gatis.paeglis@digia.com>
If the widget exists in the style rules cache before it polishes for the
first time then it should be removed from styleSheetCache too so that the
latest set stylesheet is used for the polishing.
Task-number: QTBUG-39427
Change-Id: Ic1e7988afe530f16ea9996bae56543ed554d6be9
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
uClibc doesn't have a nsmap member in __res_state. Since it also doesn't have
res_nquery() which is mandatory for QDnsLookup, we can simply disable the code
to fix the build.
Change-Id: Ia872f535519aca3a2de763548c6dd0e3e0ee20d4
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Reviewed-by: Mandeep Sandhu <mandeepsandhu.chd@gmail.com>
Introduce a global-static struct storing atom and class name for
the message window to be shared between threads. This prevents
RegisterWindow()/UnregisterWindow() of different threads (using exec())
from interfering and silently failing.
Task-number: QTBUG-39471
Change-Id: I9bc1106a41f64749c55825a96973921bb831458f
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
IndexSetter can't be passed to QObject::connect() since C++98 templates
require global types (they can't be function-local).
Change-Id: I099322f835661d4679140b7810b50dbeb0e4b9e0
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
We require the intrinsics from immintrin.h, so include it
unconditioanlly with that compiler.
Change-Id: I4a17676631f9d89e2d22e486f40c9b177ca06c1e
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
After 717d39ac08, if HB_Face was instantiated by QFontEngineFT A
and then used by QFontEngineFT B, whilst A already destroyed,
a crash occurs in hb_getSFntTable() due to accessing a stale pointer.
Task-number: QTBUG-39278
Change-Id: I3428669a311f49cdda1725b778f45219cbcf470d
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
This fixes an assert in OS X 10.10 Yosemite, where the pre-
allocated buffer would be too small to hold 3 successive calls
to ALLOCATE_ARRAY.
Task-number: QTBUG-39504
Change-Id: I5a0ae36170636eb97ab21c5903b96674e2a99547
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
Passing in a sample count of -1 should be treated as 0. This is common
when setting up framebuffer formats from a QSurfaceFormat where the
default, unset value is indicated by a value of -1.
This broke QQuickWidget which was unaware of this limitation of
QOpenGLFramebufferObject and was passing format.samples() as the sample
count without making sure it is 0 or higher.
Task-number: QTBUG-39699
Change-Id: I324b8b006eaa992c15ae932f9df305500fefeb65
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Make sure qRegisterGuiGetInterpolator is called even for static builds.
Task-number: QTBUG-37341
Change-Id: I65735a558d5bbfaa02fa4ec47d55ddf33ca1991a
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Added some checks to QRasterPlatformPixmap::createPixmapForImage() to avoid
crashes when QImage::convertToFormat() returns a null image.
Change-Id: I573505a1aff7931d9a2fb452d0a83ae93d8de7db
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Commit cf092abdfc introduces a virtual
"void initialize()" in QPlatformIntegration class.
"void initialize()" was already implemented in QDirectFbIntegration
since commit 6534898cc6 allowing
initialization steps to be overridden by QDirectFbIntegrationEGL.
Therefore the QScopePointer "m_input" handling a QThread is reset twice.
The QThread firstly created is forcibly terminated. The application
displays the error message "QThread: Destroyed while thread is still
running" and sometimes crash with a SIGSEGV.
This commit rename QDirectFbIntegration::initialize() into
QDirectFbIntegration::connectToDirectFb() to fix this issue.
Task-number: QTBUG-38710
Change-Id: I3ca07c373af7c47abf08da2b45bbcf7a6cf573ad
Reviewed-by: Holger Freyther <holger+qt@freyther.de>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
the no_dll switch has questionable semantics: it pro-actively breaks
non-dll builds. therefore its usage needs to be limited to dll build.
Task-number: QTBUG-39594
Change-Id: I98328e502693df835af565b5ec25ada2c1c168ad
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
The non-threaded QXcbEventReader invokes processXcbEvents when the
EventDispatcher is about to block. This method ensures that the xcb
connection is going to flush. Applications can use low level xcb code
in that case without having to ensure to flush the connection before
going to block again.
With the threaded QXcbEventReader this didn't work and applications
which for example changed a window property and waited for the matching
property notify event were stalled.
This change ensures that also in the threaded case the connection gets
flushed when the EventDispatcher is going to block.
Change-Id: If1dc5eb96e2f1bde10b7a40af550b0608c62f70c
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
The trick of creating a copy is not thread-safe. I'd known this since
the moment I wrote that code, but thought "what could go wrong?".
Task-number: QTBUG-39285
Change-Id: If521d4a649c06e6a34926687e85623aa25cb4c35
Reviewed-by: David Faure <david.faure@kdab.com>
Test if the window has a handle before using it.
Change-Id: I728a129722f8ecd021998d483530a8d1687e5fe3
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
The fixed width should not be set on QFontComboBox itself, but on the
popup.
This regressed with commit bfb25c0352,
and caused assistants preferences dialog to become wider than the
desktop.
Change-Id: I2059794fc12d34837cdb7dfded80df57d102a6f0
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
We can't get the actual screen geometry on Android, but in Qt 5.3.0
we would always return the screen geometry minus the size of the
status bar. After the available geometry was initialized to 0x0
instead of this arbitrary value, some applications that depended
on this as a constant value would break if they collected the
information before the window surface had been initialized and they
forgot to listen to QScreen::geometryChanged().
To reduce the risk of regressions, this patch makes sure we return
the same thing as before for the screen geometry and that this is
not linked directly to the available screen geometry.
Task-number: QTBUG-39464
Change-Id: Ie63337b3b10d2eb5130e4fece6c5b144e8230164
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
Reviewed-by: Christian Stromme <christian.stromme@digia.com>
This reverts commit a1c5198387.
The idea of detecting non-spontaneous events by comparing
against the widget's crect has problems when sequences
of programmatic resizes occur. In addition, QWindowSystemInterface's
queueing of events is problematic for this.
Task-number: QTBUG-39611
Task-number: QTBUG-32590
Change-Id: I4674d8d5d5d432d938f7226b5790543335665c1f
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
This reverts commit c3e416296a.
The idea of detecting non-spontaneous events by comparing
against the widget's crect has problems when sequences
of programmatic resizes occur. In addition, QWindowSystemInterface's
queueing of events is problematic for this.
Task-number: QTBUG-39611
Task-number: QTBUG-32590
Change-Id: I9ff8049add147be23e064a513e8645ae04577c6c
Reviewed-by: Andy Shaw <andy.shaw@digia.com>
On Unix systems where the GUI event dispatcher uses a notification
system for socket notifiers that is out of band compared to select(),
it's possible for the QSocketNotifier to activate after the pipe has
been read from. When that happened, the ioctl(2) call with FIONREAD
might return 0 bytes available, which we interpreted to mean EOF.
Instead of doing that, always try to read at least one byte and examine
the returned byte count from read(2). If it returns 0, that's a real
EOF; if it returns -1 EWOULDBLOCK, we simply ignore the situation.
That's the case on OS X: the Cocoa event dispatcher uses CFSocket to get
notifications and those use kevent (and, apparently, an auxiliary
thread) instead of an in-thread select() or poll(). That means the event
loop would activate the QSocketNotifier even though there is nothing to
be read.
Task-number: QTBUG-39488
Change-Id: I1a58b5b1db7a47034fb36a78a005ebff96290efb
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
strace reveals that we do ioctl(-1, FIONREAD) and get EBADF. The only
case where this could happen is inside _q_processDied, which calls the
read functions without checking if the pipe is still open.
Change-Id: I67637fc4267be73fc03d40c444fdfea89e1ef715
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
The QProcessPrivate::Channel object contains some structures that may
survive the closing of the pipe, so calling this function "destroy" is
incorrect. Let it be just the closing.
For symmetry, the createChannel() function is renamed to openChannel().
Change-Id: I2899214c6e4c25835390b10ccf3931315a91589e
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Similar to the previous commit, this simplifies the code.
Change-Id: Ia02b9b5174b4bc6fd04ec2534231b7db5fc914fa
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Since we only scan for XInput2 devices on application start, we will
currently miss any devices plugged in while the application is running.
This patch makes QXcbConnection listen for XInput2 hierachyChanged
events and use them to trigger a rescan of XInput2 devices.
This fixes a regression in Qt 5.3, where the scroll wheel on hot-
plugged mice does not work until the Qt application is restarted.
Change-Id: I2cdc7ca24d3ab00716cedc4b22355b6e4935b184
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
- Open files passed on the command line.
- Point the file dialog to the pictures location
and use a filter string for the supported types.
- Set the window title according to file name.
Task-number: QTBUG-37203
Task-number: QTBUG-39287
Change-Id: I4e5e43875c3a7544c862c054181e75942939c1d5
Reviewed-by: David Faure <david.faure@kdab.com>
In ICU the strength parameter decides whether a comparison is
case sensitive or not.
Fix mac comparison code. It can't have worked before.
Added some basic automated testing for QCollator.
Change-Id: I2646c464fd22ccd3a93c461fa3dba4bd1d4c7b4b
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
It is NOT always the same as isEnabled().
Added a unittest to prove it.
Change-Id: I7717126835923e8c091249bfcdf81767c44fb5f7
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
They would either disappear or be positioned at bogus coordinates.
The MINMAXINFO structure works with coords from the primary screen
then uses an "interesting" algorithm to adjust to secondary screen:
Say you have a primary screen with width=1000 and secondary screen
with width=2000, here's what you get when you set ptMaxSize to:
ptMaxSize.x | Size window gets in second screen
--------------------------------------------------
500 | 500
1000 | 2000
1001 | 2001
1100 | 2100
So basically you can't get any value between 1000 and 1999
How many people use the taskbar on a second display and maximimize
a frameless window anyway ?
Task-number: QTBUG-39537
Change-Id: Ic9b3120e7fb5a9a5d97828a2e44be02ae587b92e
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Namely, the Adreno 205. We used to enable this
workaround for Huawei Honor (Adreno 205).
Task-number: QTBUG-33951
Change-Id: Ic92a6913664f2f0954271c700d9ef83d27c238a7
Reviewed-by: Yoann Lopes <yoann.lopes@digia.com>
According to [1] "%L" should expand to a full path for the
default (based on system's locale) Compose file.
[1] http://www.x.org/archive/current/doc/man/man5/Compose.5.xhtml
Task-number: QTBUG-35943
Change-Id: Ie803a89742d9c0aa3b2d759bea28ed403dc68c9c
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
