| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Enable the remote debugging of QDeclarativeEngines only after
QDeclarativeDebugHelper::enableDebugging()
has been called.
Approved by 4.7 Program Team.
Reviewed-by: Alessandro Portale
Task-number: QTBUG-13762
(cherry picked from commit b2016bbfc9c7389e7b64451417395ceba96af21f)
Conflicts:
src/s60installs/bwins/QtDeclarativeu.def
src/s60installs/eabi/QtDeclarativeu.def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The qml debugging enabler in QtDeclarative made any Qt app crash
which used QDeclarative. Reason was that QtDeclarative.dll tried
to directly access (private) writable static data from QtGui.dll.
This patch adds an accessor function for the data to QtGui, and
the crash is gone.
Done-by: Kai Koehne
Reviewed-by: Kai Koehne
(cherry picked from commit 2362d8b1e9fa86da1d3cc5dbb7d4467ec12311e1)
Conflicts:
src/declarative/debugger/qdeclarativedebugservice.cpp
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Not doing this causes the parent QModelIndexes to be out of sync with
the model.
The fix in fe350ca756df6392b3d0b7840351fccc89f9ef6f is extended to also
use adjusted indexes in the itemMove call.
Merge-request: 849
Reviewed-by: Olivier Goffart <olivier.goffart@nokia.com>
(cherry picked from commit 62dc9c9602987f4381481cc5642e3ea0782bb418)
|
|
|
|
|
|
|
|
|
|
|
| |
This reverts commit 6155050f68cc86c445552da61a5f240c16f5e2cd.
The GL_ARB_ES2_compatibility extension does not mention the lowp,
mediump or highp keywords.
Task-number: QTBUG-14384
Reviewed-by: Samuel
Reviewed-by: Prasanth
(cherry picked from commit 7cd0a90344d340f22b6b2d3afeef092dbaf2cd51)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The decoding from image reader was assuming the image reader do not
keep the image internally. This is not true for the GIF plugins because
the previous image can be used to compose the current image.
This was causing crash on ARM because the 16 bits color depth causes
the image memory to be reduce by half. When the plugin was accessing
the memory, it assumes the images has not changed and is on 32 bits.
This patch disable the in-place conversion if a detach is required.
Regular conversion is the correct solution in this case, and it can
also be made faster by converting while copying.
Reviewed-by: Andreas Kling
(cherry picked from commit 4d974ff0a748b22e668a4cb7ef38101122c85b3b)
|
|
|
|
|
|
|
|
|
|
|
| |
If QDeclarativeWebView was instantiated twice
an assert was triggered by QWebSettings::enablePersistentStorage();
This patch is crtical for tooling (Bauhaus)
Task-number: QTBUG-14278
Reviewed-by: Aaron Kennedy
(cherry picked from commit f532679ca11914b453e22342f2ae5e9f790ce47a)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem was that we always invalidated the cache whenever the item was
rotated. This is however not required for simple rotations such as
90, 180 and 270 degrees. This commit also removes the somewhat arbitrary
logic which takes the desktop size into account. We now use the viewport
size instead.
Auto test included.
Task-number: QT-3779
Reviewed-by: yoann
(cherry picked from commit e84ab1fee7f44a28ee82793f83b0b27d04d28c09)
|
|
|
|
|
|
|
|
| |
Fixes regression caused by optimization added in commit
35a51442ed21f58c06b21293eeb56e843251ee82.
Task-number: QTBUG-14239
Reviewed-by: Martin Jones
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
pixel."
This reverts commit 2acdc634ee895af4a9738c0f6c2496e388afba38.
Originally this fixed QT-3791 but it caused lots of problems in
other sites. So reverting this will fix QT-4001 but reopens
QT-3791. QT-4001 blocks Qt 4.7.1 but QT-3791 doesn't.
Task-number: QT-4001
Reviewed-by: Jason Barron
(cherry picked from commit 25ddfb434f13aa3bbc6ac303aa17dbe9582bf3f0)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In portrait orientation, QDateEdit's calendar popup (QCalendarWidget)
does not fit into screen.
As a fix, the pixel metric PM_FocusFrameHMargin, is reduced from
13 to 3 in qHD screens (640*360). Also, the explicit halving of
the pixel metric for QTableViews is removed.
The pixel metric resulted in 27 pixel text margins (2 * pm + 1)for
calendar widgets, which is clearly a bit too much.
Task-number: QTBUG-14126
Reviewed-by: Miikka Heikkinen
(cherry picked from commit b3f895fd97deda0c7c5bfd704c04450e7afee748)
|
| |
|
|
|
|
|
|
|
|
|
| |
The problem is that with empty layouts we could sometimes divide by 0.
Note: This doesn't fix the whole task...
Task-number: QTBUG-13547
Reviewed-by: ogoffart
(cherry picked from commit c1f9978c9d61bcbdb2f280185a3abdea13d7f532)
|
|
|
|
|
| |
Reviewed-by: Thiago Macieira
(cherry picked from commit b8089f0b7a0fef9318070aea9c8344bfe987bac9)
|
|
|
|
|
|
|
|
|
|
|
| |
the problem was that the item is first painted at its default position,
then moved by the layout and finally repainted.
We now made sure the item is laid out before the first paint event
occurs.
Task-number: QTBUG-13865
Reviewed-by: bnilsen
(cherry picked from commit 33f525e636ef8fa64a15d3e66c56adaea0075bda)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Change c46688b8a88da made us use m_curve_threshold for both
QBezier::shifted and QBezier::toPolygon, and adjusted the threshold
dynamically based on the painter scale. Since the threshold in shifted
was already relative to the pen width, it is independent from the
painter scale. Instead, we need to set a separate threshold for dashing.
Also, in several places we were calling setCurveThresholdForTransform
with the painter matrix even though we were transforming the points into
device coordinate space before stroking.
Task-number: QTBUG-13894
Reviewed-by: Gunnar Sletta
(cherry picked from commit d60dc7cba21794866c9382f83080fab1a129eb08)
|
|
|
|
|
|
| |
Task-number: QTBUG-14070
Reviewed-by: Gunnar
(cherry picked from commit 4cd4160d85dc1e158a545422cac895792b14eda6)
|
|
|
|
|
|
|
|
|
|
|
|
| |
This reverts commit 62b5ef3cc1306e46a4042b14867f2f92d9a110f3.
The implementation of hfw from this patch is unfortunately not robust enough.
It doesn't manage correctly the constraints on the layouts and the cell spans.
It caused bad behaviour or regressions seen in tasks:
QTBUG-13547, QTBUG-13067, QTBUG-13549, and more
Reviewed-By: ogoffart
(cherry picked from commit 5bd6f7eb5c7d87c08539b6c2df416990cc417ec7)
|
|
|
|
|
|
|
|
| |
Always flush sockets after sending data, and make autotests more robust
by using busy wait.
Reviewed-by: Christiaan Janssen
(cherry picked from commit cfe198948f1e4867918176df38b3e0b49757a4b8)
|
|
|
|
|
|
|
|
|
|
| |
When statusChanged() is called during handsake state() was not the same
as the argument passed. Fix this by setting gotHello = true _before_
notifying the clients.
Reviewed-by: Christiaan Janssen
Task-number: QTBUG-14087
(cherry picked from commit a1f050fe4217d3a642ab7f4df8e50c21aa51689c)
|
|
|
|
| |
(cherry picked from commit 9caae83e1f2b9c56ee86b8523391e6a83ea724c5)
|
|
|
|
|
| |
Reviewed-by: Christiaan Janssen
(cherry picked from commit 30959e88498b2c8591145e30c8b497a76c12d8f6)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The protocol so far was client->server only. That is, there was no
sane way for a client to check whether a plugin on the server (service)
was available or not. E.g. calling Client::setEnabled(true) 'succeeded',
without a check whether there is actually a service to talk to.
The new protocol replaces this shortcoming by a service discovery
mechanism: Both client & service announce their available plugins at
handshake time, and later on if there are changes. The status is
reflected in Client::status() and Service::Status() , which are either
NotConnected - no network connection, or not registered properly
Unavailable - TCP/IP connection works, but no plugin with the same
name on the other side
Enabled - You can connect to plugin on other side
The status changes happen automatically (no setEnabled() anymore).
Furthermore a version ID was added to the handshake, so that we can
extend the protocol further in the future :)
(cherry picked from commit fd9771c29d401d88779ab7c5d7715c9ca41dd723)
|
|
|
|
|
|
|
| |
The QmlDesigner(Bauhaus) should not rely on -nokia-developer
Reviewed-by: Marco Bubke
(cherry picked from commit 8ad396891f2a4966ad3cf943f99e208211e956bb)
|
|
|
|
|
|
|
|
| |
Fixes crash in creator. If flag is true notifications are sent. It is important
that positioners work correctly and don't leak.
Reviewed-By: Alexis Menard
(cherry picked from commit 3b43389b78925ce87d660c343c925fa280b91e11)
|
|
|
|
|
|
|
|
|
|
|
| |
This ensures that if the component has already been completed, itemChange()
is called. This is required to modify QDeclarativePositions in the visual
editor, while keeping a legal state. Without this patch notifications are
missing and the Positioner keeps track of already deleted children.
This resulted in a crash.
Reviewed-by: Thomas Hartmann
(cherry picked from commit 1f8bb573f06234a3d13fb57de5eb644824d5024f)
|
|
|
|
|
| |
Bauhaus needs this export to manage layouts.
(cherry picked from commit 858469445ceb745c70dad9cd4f913acc34683e15)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Task-number: QTBUG-13799
Reviewed-by: Martin Jones
(cherry picked from commit 49452ad6b22e080b1dfdfde38c21c48bb910a1ae)
Conflicts:
tests/auto/declarative/qdeclarativeconnection/data/error-object.qml
tests/auto/declarative/qdeclarativeconnection/data/error-property.qml
tests/auto/declarative/qdeclarativeconnection/data/error-property2.qml
tests/auto/declarative/qdeclarativeconnection/data/error-syntax.qml
tests/auto/declarative/qdeclarativelistmodel/tst_qdeclarativelistmodel.cpp
tests/auto/declarative/qdeclarativeqt/data/quit.qml
tests/auto/declarative/qdeclarativeview/data/error1.qml
|
|
|
|
|
|
|
|
| |
Qt 4.7 is still supported, but deprecated.
Reviewed-by: Martin Jones
Task-number: QTBUG-13799
(cherry picked from commit 4fcf055f66cc23c9e60a7add489e394420e71914)
|
|
|
|
|
| |
Task-number: QTBUG-13799
(cherry picked from commit 746954581f06e0bca98f25eb95dca09c663d47bf)
|
|
|
|
|
|
|
|
|
|
| |
The bug was that when querying for the maximum text length, the case
where an invalid QVariant was returned (which is allowed) was not
handled properly. This would lead to input being blocked by the input
context when it shouldn't.
RevBy: Sami Merila
(cherry picked from commit 3bb3af84bef3c0472ca8ed0d5c6bb3c82320956d)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since screen geometry changes before client area geometry is
changed in Eikon, we need to send two resize events, one for screen
area change and one for client area change.
Note that the correct way to detect client area change in applications
is to connect to QApplication::desktop() signal workAreaResized(int)
instead of filtering for resize events meant for QDesktopWidget.
Task-number: QTBUG-14058
Reviewed-by: Jason Barron
(cherry picked from commit 0725abc4607755833a1eb7cb9e083add48d03a88)
|
|
|
|
|
| |
Task-number: QTBUG-13893
(cherry picked from commit 085a121cb1ebba38d62c924500dbc71806b29b3c)
|
|
|
|
|
|
|
|
|
|
|
| |
Starting from 4.7.0, QVariant is recognized as a standard type by the
meta type system. ActiveQt needs to consider this while converting
COM VARIANT types to a QVariant.
Task-number: QTBUG-13845
Reviewed-by: Volker Hilsheimer
Reviewed-by: Olivier Goffart
(cherry picked from commit fc50d7ecaacfbfef1dbefd6ffdc083cb66c5633c)
|
|
|
|
|
|
|
|
|
|
| |
When a native window handle is created for a widget that has override
cursor set, we should reset the cursor on the parent and set the cursor
on the new window handle.
Task-number: QTBUG-6185
Reviewed-by: Olivier Goffart
(cherry picked from commit 9e9a7bc29319d52c3e563bc2c5282cb7e6890eba)
|
|
|
|
|
|
|
|
|
|
|
| |
The assert "Internal error: Plugin imported previously with different uri"
happens on Windows. $QTDIR\imports, which is prefixed to the
search path, is still written with backslashes. Unify this to forward
slashes for all paths.
Task-number:QTBUG-13994
Reviewed-by: mae
(cherry picked from commit 0aa6360b7648ddeb8a905960265eb46d8c12448f)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Change 7c673a4cf64ba04 introduced some autotest failures in the
fpe_steepSlopes test in QPainter.
Since the other rasterizers all deal in a 26.6 fixed point coordinate
space we should snap the line vertices to this to prevent floating point
overflows due to very steep slopes.
This also necessitates keeping track of four different slope / inverse
slope increments for each of the four edges.
This also fixes a previously QEXPECT_FAIL'ed test case.
Task-number: QTBUG-13429
Reviewed-by: Trond
(cherry picked from commit 93199a5b7082fd484b7f21ad4825d71693ecead2)
|
|
|
|
|
|
|
|
|
|
|
|
| |
With a large pen width and a small scale, due to the hacky way we draw
points (stroking a line from (x, y) to (x + tiny_amount, y)), we some
times end up snapping these two points to the same in rasterizeLine().
If we instead apply the SquareCap before we do clipping / snapping we
don't get this problem.
Task-number: QTBUG-13429
Reviewed-by: Trond
(cherry picked from commit 7c673a4cf64ba043bb27f90287517bdcdd7a21db)
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The use of libconninet caused Qt to have a cyclic build dependency.
Which was solved by embedding a copy as a 3rd party library. Maemo has
since donated the libconninet code to Qt because (1) it was a private
Maemo API and (2) the Qt ICD plugin is the only user of the library.
This commit moves the relevent code from src/3rdparty/libconninet to
src/plugins/bearer/icd and deletes the rest.
Task-number: QT-3893
(cherry picked from commit f532d8fcd236be9933e4186a95561e1c264de277)
|
|
|
|
| |
(cherry picked from commit 2ff0dee1b97348a750532223f0a318596c93d412)
|
|
|
|
|
|
|
| |
the refcounting to work in ICD.
Fixes NB#188145 - Network interface doesn't go down after last client disconnects
(cherry picked from commit 4acbb418b1ad093ad848143218a10ee2957db282)
|
|
|
|
|
|
|
|
| |
Includes Jay's fix for the JSC memory consumption on Symbian
https://bugs.webkit.org/show_bug.cgi?id=34349
See also commit 03b7a3cffba609df56b4ccadacae6b6f985e96e9
(cherry picked from commit dbf1d0339a0cfbe60e6b3e45f39e6c02aa3de0a8)
|
| |
|
|
|
|
| |
(cherry picked from commit b750bb085029117497d26a24dcd590fbe34fe281)
|
|
|
|
|
|
|
|
|
|
|
| |
Add a semi-private API to get QScriptEngine for a QDeclarativeEngine. So far the qmljsdebugger lib in QtCreator
accessed the script engine via QDeclarativeEnginePrivate. Replace this by a minimal API that is still in a
private header, where we nevertheless can make some BC checks/guarantees.
Aaron Kennedy agreed with the idea.
Task-number: QTCREATORBUG-2179
(cherry picked from commit c1a7deedf1628e0dd3a5fbf92d97c4151d17fb5a)
|
|
|
|
|
|
|
|
|
|
|
| |
Add a semi-private API to get QScriptEngine for a QDeclarativeEngine. So far the qmljsdebugger lib in QtCreator
accessed the script engine via QDeclarativeEnginePrivate. Replace this by a minimal API that is still in a
private header, where we nevertheless can make some BC checks/guarantees.
Aaron Kennedy agreed with the idea.
Task-number: QTCREATORBUG-2179
(cherry picked from commit 63c4c0449361ced03838e51d18e1113740f27fa9)
|
|
|
|
|
|
|
|
|
|
|
| |
Now returns correct value for Symbian4 platforms with
QSysInfo::symbianVersion(). SYMBIAN_VERSION .pro file variable should
also return correct value even if it doesn't get set in environment.prf.
Task-number: QTBUG-13802
Reviewed-by: Shane Kearns
Reviewed-by: Janne Anttila
(cherry picked from commit 0aad0d2043e061469b7139483c489dc6f823490b)
|
|
|
|
|
|
|
|
| |
The loop looking for posponedLookup in the lookup thread wasn't locking
the mutex before trying to access/modify the list.
Reviewed-by: Andreas Kling
(cherry picked from commit 45c2ac4f1e1218f595b1d21691c8dec4eaa1a021)
|
|
|
|
|
|
|
|
|
|
| |
When configuring with -no-multimedia then the plugins rule should not
be depending on multimedia, otherwise it will not try and build any
plugins.
Task-number: QTBUG-13912
Reviewed-by: Rohan McGovern
(cherry picked from commit c52224f8010bf90732e5b2ad6cef84f9bd248787)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Stroking a path can sometimes result in inconsistent rendering
especially when combined with a clip. For example, if the logical edge
of a clip rect coincides with the logical edge of a path then it can
happen that the edge is not painted correctly because the stroke lies
outside the bounds of the clip rect. To workaround this problem, we add
the 'aliasedCoordinateDelta' such that the rounding will err on the
side of caution. This improves the correctness when using the raster
engine as a reference.
Task-number: QTBUG-13165
Reviewed-by: Samuel
(cherry picked from commit 74f09c794d5fdcb8f4b547b1c55bcd512ef07b95)
|