Using Q(Open)GLFunctions concurrently on multiple threads had some issues
due to incorrect synchronization: The m_groups list in the
QOpenGLMultiGroupSharedResource, in which the Q(Open)GLFunctions instance
is stored, became corrupted under certain scenarios, for example in the
tst_qglthreads autotest and any two threads that happen to enter an
initializeOpenGLFunctions() or QOpenGLContext::functions() call concurrently.
Locking in value() has been introduced in 666c25c089
to fix such issues, however using the context group's mutex is not enough: that still
allows two threads using two contexts with a different context group to concurrently
enter insert(). Instead, the MultiGroupSharedResource has to have its own mutex
to protect its own member variables.
Task-number: QTBUG-38771
Change-Id: If01c44c2084b95e487bc9146576ca180ed8044da
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
Call ::ensurePolished from QMenu::exec before the native window gets
created. This ensures that the style handles the menu before its too
late. E.g. a style which wants to create RGBA menus needs to add the
appropriate flag before the native window gets created. Without this
change the style cannot change to RGBA as the native window has already
been created and changing the format used by QWindow is not possible
after QWindow::create was called.
Change-Id: Ic861037a438b4cb74c59a00be0ef2d633db538ed
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Reviewed-by: David Edmundson <davidedmundson@kde.org>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
This involves exposing a new function in the QPlatformNativeInterface
which gets a public function for QGuiApplication
Proof of concept is done through implementing _NET_WM_WINDOW_TYPE
setters for xcb
Change-Id: Ic9544e775fb71cc9b30273595ec41b1cdb1c9d64
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Previously we have rejected this issue as you can work around it
by setting Qt::WA_LayoutUsesWidgetRect per widget. However there
is no reason to apply the negative mac style layout margins on any
custom style so I think we should completely bypass it unless the
native border is used.
Task-number: QTBUG-13050
Change-Id: I8923e07d868c51a13587993c9b2ce79c51beaeee
Reviewed-by: J-P Nurmi <jpnurmi@digia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
That long call chain is screaming "here, I'm dereferencing dangling
pointers! Valgrind me!" but 2006 Thiago didn't see them...
Change-Id: I44de5aea113d05edec2227e9db6be8cec9303be9
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
The main test won't find them if they are.
Change-Id: Iae3ffe4c0289a0c88d46c1bd2e414c20def89ab4
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Previously, this produced wrong results, for example -3:30
became -3:-30.
Change-Id: I10efdfb48e5542b917c86b29cf8a99bfc26f7fe0
Reviewed-by: John Layt <jlayt@kde.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The tests for toString/fromString previously didn't run tests
for timezones with hh:mm where mm != 00.
Change-Id: I74da99c5b6890f46ce06446084a8129b4cbc7a02
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: John Layt <jlayt@kde.org>
This fixes a warning when loading qmake.pro in Qt Creator on OS X.
Change-Id: Iabd70f2b5b9615d0fb1563081f485e6a4c828823
Reviewed-by: Jake Petroules <jake.petroules@petroules.com>
Reviewed-by: Liang Qi <liang.qi@digia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
The list of returned modes is scanned for the first one marked as
preferred, and that is used. If no preferred mode was found then the
builtin default mode of 1024x768 is used. Use that builtin mode only if
no modes were returned at all, pick the first one if any were returned.
Change-Id: Ib355cc92219ced093c605f49dae4e34ff244b639
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
QKmsDevice already requested the drmModeRes and drmModeConnector
information. Simply pass them to the QKmsScreen constructor instead of
requesting and freeing them there again.
Change-Id: I4897d76d7b13d83297c928b6e10e17ccdfdbd242
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Increase the chunk size of the server and remove hard-coded
interval when waiting for the server to listen.
Unmodified, the test takes 170s on Windows.
Change-Id: I65bdc93ff78e1b4fb429fcafd0fdc5e80bb281f9
Reviewed-by: Mitch Curtis <mitch.curtis@digia.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Do as QGLWidget::updateGL does and avoid calling paintGL in updateGL
when we we don't have a valid QSurface to draw on.
We currently end up calling makeCurrent on the context with a null
QSurface in that case, which is the equivalend of doneCurrent, and
causes QOpenGLContext::currentContext to return null later when paintGL
is called on the subclass.
Change-Id: I712ee59274855457b0e4de3757754f56b3528955
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
The Android input method protocol specifies that finishComposingText()
should not move the cursor. Since Qt likes to move the cursor to the
end of the newly committed text, we have to explicitly move the cursor
to where the preedit cursor used to be. Fortunately we already keep
track of that.
Also implement support for the newCursorPosition argument to commitText()
since the function needed to be rewritten anyway. (It was calling
finishComposingText().)
Task-number: QTBUG-38794
Change-Id: Iff2c43bdbd3dda812ccdc71da63f3fa730474eef
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
When applying the five-finger pinch gesture, we get a touchesCancelled
callback with all five touch points, but the pinch gesture ends when
the second to last finger is released from the screen. The last finger
will not emit any more touch events, _but_, will contribute to starting
another pinch gesture. That second pinch gesture will _not_ trigger a
touchesCancelled event when starting, but as each finger is released,
and we may get touchesMoved events for the remaining fingers.
The event property 'allTouches' contains one less touch point than it
should, so this behavior is likely a bug in the iOS system gesture
recognizer, but have to take it into account when maintaining the Qt
touch state.
We do this by assuming that there are no cases where a sub-set of the
active touch events are intentionally cancelled, and always clear the
list of active touches.
Task-number: QTBUG-37304
Change-Id: Icee79978508ecbc6854c0fb55d2da48b99d92f96
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
Use the correct palette and enforce Fusion style to prevent
the Vista style from clobbering the tooltip palette in polish().
Task-number: QTBUG-38183
Change-Id: Id19d548f818d801c4914a343e08207195c343888
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
The intent is to provide compile time validation of signals and to help
detect signal overloading in the future.
Change-Id: I9d5d46ed4b70c5d0cd407deb5928b1e76d37e007
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Since we require a relatively recent system, check the version of the
direct2d dll on disk and report an error to the user if it is too old.
Previously only a cryptic runtime error resulted from a direct2d version
that was too old.
Change-Id: I6c3955e1a98326fca6bcdc871b0a25291391ba88
Reviewed-by: Risto Avila <risto.avila@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
After analysing text drawing performance two things seem to take up most
of the time. The first is font lookup, the second is QVector
initialization.
To address the first point a per paint engine instance font cache is
introduced. At the moment no mechanism exists to clear this cache and
it is unbounded.
To address the second point, we simply switch to using QVarLengthArray
instead of QVector.
In an artificial benchmark that draws text in a tight loop, the first
change raised fps from ~70 to ~100. The second change further raised this
number to ~115 fps.
Change-Id: Iafa25c3e35bc42bd7c1582b0636e721c5193b494
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
When the start point is not on the curve, the end point will be
used as starting point for the QPainterPath, if it is on the curve -
if it is not on the curve, the mid point between start and end point
will be used. The origianl start point will still be used as a control
point for the first bezier curve.
Task-number: QTBUG-38601
Change-Id: I90e07f1141288873263f1f12fa51e6a248d0804f
Reviewed-by: Fabian Bumberger <fbumberger@rim.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
1) toString does not track the string it returns, thus it has to be
deleted by the caller
2) on some platforms vsnprintf crashes if a null string is in the
va_list
Change-Id: Iecf94e93d3a2ddf4186ee20de1f495f2f92dcc60
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Matches the compiler capabilities better and will catch all GCC-like
compilers (including Clang, LLVM and Intel CC on Unix).
Task-number: QTBUG-38544
Change-Id: I102966d307a4e167b6dcf3da08359e656f3af45e
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
GetGlyphOutline fails for ttf fonts with embedded bitmaps. This results in
distancefield rendering not rendering glyphs (for the failed paths). We
need to fall back to texture rendering if this is the case.
Change-Id: Ibdf7dc5c1d34f513c436f88fabbdcc4089bb6fef
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
Determine the visual index by looking up the column of the QModelIndex
in the logicalIndices array instead of looping.
Task-number: QTBUG-37813
Change-Id: I5c3c73c67537877b03cdc2c36a52041d99f7f49d
Reviewed-by: David Faure <david.faure@kdab.com>
Some calls pass len=32, so garbage was shown when maxlen (readBytes) is smaller.
Change-Id: I97e61dcdf5a0db032317c57afb4bfb406437d8d5
Reviewed-by: Markus Goetz <markus@woboq.com>
Reviewed-by: Peter Hartmann <phartmann@blackberry.com>
The collected data doesn't reveal any obvious correlation between sleep
times and test failures but the many warnings are a hassle for anyone
reading the test results.
Change-Id: I71bd0c90c20c730573693f23f4435e538b635d44
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
The image format returned by QOpenGLFramebufferObject::toImage() on
OpenGL ES/2 builds changed in 5.2.0 but was not documented.
Change-Id: I2379c62c7f715803192ffa9bb8ae1171ef4b2314
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
Change-Id: I92fa083665509932b75ff1037904a6f78a950fd6
Reviewed-by: Jeremy Lainé <jeremy.laine@m4x.org>
Reviewed-by: Peter Hartmann <phartmann@blackberry.com>
Change-Id: Ife5b7206fd3d7af57cfca3c0f28f56bb53ede7a7
Reviewed-by: Jeremy Lainé <jeremy.laine@m4x.org>
Reviewed-by: Peter Hartmann <phartmann@blackberry.com>
A pro file that adds files to QMAKE_BUNDLE_DATA using relative
paths will fail building if doing shadow builds. The reason is
that we look for the files inside the build dir.
This change will make sure we resolve files from the source dir
when not using full paths.
Task-number: QTBUG-37054
Change-Id: Ic1067861097b3b6a640ee862472d728d6188576a
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Nothing else seems to report this state and on windows
for example it results in NVDA reading subMenu which makes
little sense.
Task-number: QTBUG-38500
Change-Id: I64820d9f2ea9174034f01da42cb2266a19c19465
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
When moving the mouse over a disabled menu it could appear to be usable
but it should not show any interaction at all.
Task-number: QTBUG-31688
Task-number: QTBUG-37488
Change-Id: I9e6594b40e8b209146d8c3705750dfd75f1258c5
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
This test doesn't fail on Mac any more.
This reverts commit 36493a7a41.
Task-number: QTBUG-23685
Change-Id: Ib7c56494b07de9839b3287758fe228f799bc343c
Reviewed-by: Sergio Ahumada <sahumada@blackberry.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
