there is no particular reason for it being done by qmake.
avoids that the logic is distributed over two source files,
and allows us to generate these headers at pre-build time already,
including not forwarding to a yet unexisting file (which would have a
yet unknown location).
Change-Id: I9c78ab425cf6f01d076c86fd1ee602626f231487
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
It was comparing an absolute path to a relative path.
These could never be the same. Instead of the relative
path, it now gets the canonical absolute path, so the
comparison succeeds when it should.
Task-number: QTBUG-31404
Change-Id: I9c482d4649d493ce6d4f9a522cb61f2c8a5eb21f
Reviewed-by: Topi Reiniö <topi.reinio@digia.com>
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
Since the dx tool has moved in the SDK, we need to update our copy
of it to also search in the new location for dx.jar.
Task-number: QTBUG-31405
Change-Id: If093a9f51f33c5d8666919f516a3b336322a7169
Reviewed-by: Ray Donnelly <mingw.android@gmail.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
This is how it was done in Qt 4.
An issue with the current approach was that it did not consider
modifiers when setting a Qt::Key_* value, which would assign the same
Qt keycode for:
a = a(65)
Alt + a = ā(65) [here it should return a unicode value for 'ā']
This is inconsistent with the other platform plugins.
Also in the combination with a dead keys it was returning nothing in
the output.
Task-number: QTBUG-29005
(cherry picked from commit 6730413fcac1d7eb39af3683b87f965c5823cb6c)
Change-Id: Ic28eb55b3a9798ecb6012cc2e3fb18589b8b0392
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
MSVC2008 compiler fo ARM targets fail to compile qToUnaligned when
using sizeof(T) inside memcpy fynction. The compiler fails at least
when the code is reached through the following macros and templates:
-> tst_QtEndian::toLittleEndian
-> qToLittleEndian(T src, uchar *dest)
-> qToUnaligned(const T src, uchar *dest)
The above sequence produces internal compiler error with
MSVC2008/ARM builds when called from tst_endian.
As a workaround sizeof(T) is called outside memcpy function.
Change-Id: Ib4d382c2cebecb6e54bb99fc8fad72db93825fcd
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Change the hint for
QPlatformIntegration::SynthesizeMouseFromTouchEvents to false
for Windows and suppress the events synthesized by OS.
The synthesized events cause touch events to generate 2 clicks
in Quick2.
Leave code as is for Windows CE.
Task-number: QTBUG-31386
Change-Id: Ia0987342dcdd55c8540810da5e4b90518a501ce6
Reviewed-by: Alan Alpert <aalpert@blackberry.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
These are useful as default implementations of
QPlatformIntegration::drag(), instead of having it return 0 which will
lead to crashes in Qt Quick 2 and widgets applications that use drag
and drop.
Task-number: QTBUG-31288
Change-Id: I70efa139306ced5d879def0f74e3a72d3bcd64f7
Reviewed-by: Jørgen Lind <jorgen.lind@digia.com>
The PlatformToolSet tag belongs into the PropertyGroup with the label
"Configuration". The former location in an anonymous PropertyGroup
tricked Visual Studio into displaying the right PlatformToolSet but
using its default value. If VS 2010 and VS 2012 are freshly installed
on the same machine, the default toolset for VS 2012 is VS 2010.
Task-number: QTBUG-30822
Change-Id: If00a532e92b0812c552b1cac52ff77a1e7039146
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
Precision timers can cause the event loop to lock up
when running MSVC2012 code on pre-Windows 8.
Task-number: QTBUG-27266
Change-Id: Idd73731e82843d0d140859bab825bc1a54eccf1a
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Currently qdoc skips collision nodes (and their children)
when reading index files. This means that cross-linking
between modules does not work for nodes that are defined
under a collision page node. Most notably, the QML global
object 'Qt' cannot be linked to from outside Qml module
as it collides with Qt namespace.
This change fixes the issue by skipping collision nodes
and only processing their children when writing index
files. In addition, we need to adjust the function that
searches for nodes to the possibility that there may be
multiple nodes with the same name but different type.
Task-number: QTBUG-31096
Change-Id: Ic71d714f85539d8537021c73d8f1a527006a6f23
Reviewed-by: Martin Smith <martin.smith@digia.com>
the code makes no sense, and was added with the QNX port without comment.
there is already a detection a few lines up.
Change-Id: I18ec18604c37c7c42f2649a658dd22324d481dd3
Reviewed-by: Andreas Holzammer <andreas.holzammer@kdab.com>
Reviewed-by: Harald Fernengel <harald.fernengel@nokia.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
It was never ported to Qt 5, and 10.7 has simple API for that.
Task-number: QTBUG-31336
Change-Id: Ie00c4ed3af9cd098c9e63eba1c654f1801aa83aa
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
This is what happens in uic generated code. Therefore, we should not
mark an item as text-synced until it's got its text set.
Task-number: QTBUG-31378
Change-Id: I7bb7db8abad922b50546c7669d285369ebf01394
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
This page is generated by a function in qdoc
that was wwrriten especially for this page.
It wasn't checking to see of the group member
pages were marked with \internal. Now it does.
Task-number: QTBUG-31197
Change-Id: If3f0e90f1a3748c47b3975373047b04d011d6748
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
For grouped properties, the property list in
the summary section was not shown owing to a
bug introduced when implementing the abstract
base class concept for QML types. This has
now been fixed.
Task-number: QTBUG-31317
Change-Id: Idc2344539ecf3da53e1be6816f59e01922c5c6fc
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
This warning is pretty annoying and doesn't necessarily imply that an
application is not working properly.
Task-number: QTBUG-28613
Change-Id: Id0a2ebd91f9e4d59dce3e3e29637988d8e6175a9
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
QGLTextureDestroyer will try to make the context current on the GUI
thread, regardless of whether it is owned by another thread. Use
QOpenGLSharedResourceGuard instead which does the right thing and takes
shared contexts into account.
Task-number: QTBUG-31403
Change-Id: I1377f9284995a7ba5af32c85296eef152fc035c8
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Do not touch windows whose geometry (including the unknown frame
size) is likely to be larger than the screen.
Remove fix-up in the Windows plugin.
Task-number: QTBUG-30142
Task-number: QTBUG-31071
Change-Id: I13a8ffb9fb9d8c71d35de75094275388fa427f2c
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Probably a typo since keyCode clearly has no modifiers encoded.
Change-Id: I1c7908b06a759baf7b2c3462861a5d61f8c52b9f
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Due to the way the DEX_CMD is formatted on Windows this would break
every time. Since we actually bundle dx.bat in the repository, there's
no need to check for its existence, so the easy fix is just to move
the existence check into the code path where it's run from the SDK.
Task-number: QTBUG-31405
Change-Id: If1aeb744d3abbd2488153b13aac401436965074e
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Ray Donnelly <mingw.android@gmail.com>
The old implementation was wrong since it did not use the
screen's height (which was already in primary orientation) to
calculate what the new y value of the target rect (which was in
portrait) should be.
Change-Id: Ie5b2241119e244d099e06d85f69953c1d64979aa
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
We're using QWindow::format() to decide opacity or not in
a few places, but this used to resolve to
QPlatformFormat::format() which would in turn return a default
format without alpha set. Instead, return the format requested
by the user.
Masked windows were always broken as converting a 32-bit image
to an Indexed8, doesn't give a grayscale image, but rather
a randomly spreadout set of indices based on the colortable
generated by the converToFormat function.
Task-number: QTBUG-28531
Change-Id: I537288f85c70b1e6194785b9ebcb5ea1f9581cee
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
The 3G versions are iPad2,6 and iPad2,7.
Change-Id: I43a00e84535d494550bca8a533a6d16af4be6722
Reviewed-by: Ian Dean <ian@mediator-software.com>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
updateMatrix will us both redirection matrix and highdpi
scale matrix, so make sure we don't multiply it in twice.
Change-Id: I7394e504746a8de54b4dc79492264deba320538f
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@digia.com>
This implementation will look at the orientation of the main
screen to convert the touch coordinates. This will most
likely change in future work, where we might look at a views view
controller instead to decide orientation etc.
Change-Id: Ic7875c5ecc4f21538f82a4f0467350bdf8ecc0b0
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
If no stretch factors were specified, we used the preferred size as a
stretch factor. Obviously, that didn't work if the preferred size was
actually 0.
This patch works around this by actually setting the stretch factor to
1.0 if this is the case.
This should work fine in most cases, except for the case where there
are also other items with a preferred size close to 0.
In this case, the item with preferred size 0 will just grow
faster than an item with e.g. preferred size 0.1.
Task-number: QTBUG-31217
Change-Id: I966455da0bdd00308591c7f7cfbc4e134d423e57
Reviewed-by: Paul Olav Tvete <paul.tvete@digia.com>
The winapi call used to load the icon does not
support the arguments on windows ce
Change-Id: Ia600eb9b05d5eb40778d8c281e6ce8278bfd7177
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
"technology-apis" doesn't serve a purpose anymore and its product
function is replaced by the new overviews on the landing page.
Change-Id: I1e959981fd163966a54bec0d697bed12007c39e6
Reviewed-by: Geir Vattekar <geir.vattekar@digia.com>
"QTest" is the C++ namespace; "QtTest" is the library name
- Edited the logger output in qplaintestlogger.cpp
- Updated documentation
- Updated expected outputs for self-tests
Change-Id: I43c525c43221a8d4e843a00d6d55b0f06ef55fd7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
this overrides the magic that makes examples only install their sources
in production builds.
packagers may want to force the build of the examples, so they can
package them up for demo purposes.
this is actually why we formerly had the split between demos and
examples ...
Task-number: QTBUG-30788
Change-Id: I5633f69404c5aa6846f5496e8f161a273a7a7da3
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
Reviewed-by: Lisandro Damián Nicanor Pérez Meyer <perezmeyer@gmail.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
This is the easy fix: looking at what is supported by the NDK. If people
have weird setups, then they have to specify -android-ndk-host. We do
actually detect the host architecture later, but using that would be
a much bigger (and riskier) change.
Task-number: QTBUG-31275
Change-Id: I18db878031baa2e1ee2fa4beff364d58d8bd3c7a
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Parts of the code seems to assume that all QWindows are
top-level windows. This will be false when not using
alien, as then, each widget will be wrapped inside a
QWidgetWindow. In that case, we should not tell QPA to
activate the "window".
This bug caused focus handling (and text input) to fall
apart for e.g graphicsview when using a QGLWidget as viewport.
Change-Id: I579db7a84d718973e02e96ed535fe6e25baf4bd5
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
If the liking didn't fail there's no need to print out warnings. The
warnings can still be accessed by the application after calling link()
through the log() function.
This prevents warnings such as these from appearing:
QOpenGLShader:🔗 "No errors."
Task-number: QTBUG-31326
Change-Id: I03c9be5dfada8822c3ab1c3610eac2fc0a91410b
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Reviewed-by: Sean Harmer <sean.harmer@kdab.com>
During some coding style cleanup in 02311c07 ("QNX: normalize braces")
some braces which were actually necessary got removed.
The result was an spurious report of a fatal error when no underlying
error condition had occurred.
This change restores the braces to preserve the prior behavior.
Change-Id: Ic32fbd5961ce59f6c01476fd2cef6fec0bdae93e
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
When linking QtGui and QtOpenGL libraries on linux a multiple defined
symbols linker error causes failure. This change renames the offending
class in the QtOpenGL library. For more information see:
http://qt-project.org/forums/viewthread/27458
Change-Id: I05488bf2c5c612476432c11095b24bc2b29c78f2
Reviewed-by: Samuel Rødal <samuel.rodal@digia.com>
Incorporate some more of the API changes between OpenSSL versions
0.9.8 and 1.0.0.
Task-number: QTBUG-31140
Change-Id: Ie719b34e3ec8751f0fbc07d315e82816c110762c
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Latin1 strings are usually stored as 8 bit data in the json binary
format. But that data structure has a size limitation of 16bit, so
we need to fall back to storing the string as 16 bit data if it is
too long.
Task-number: QTBUG-30946
Change-Id: I0069b1367030b0b2f819fd1f04e34c9e2534a2a3
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Task-number: QTBUG-31297
Change-Id: I96f4652e410334fae54a0012ed917a965becfe5c
Reviewed-by: Shawn Rutledge <shawn.rutledge@digia.com>
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Previously, any foreign window was treated as an overlay, with its
z-order being forcibly set above the native Qt windows.
This change adjusts the strategy to treat foreign windows whose initial
z-order is less than -1 (the default z-order used by mmrender for its
overlay surfaces) as underlays; otherwise the foreign window is treated
as an overlay.
Change-Id: Ifbcfd3a956965fa1a347076e9845afa50a81edfd
Reviewed-by: Thomas McGuire <thomas.mcguire@kdab.com>
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
A few tactical changes were necessary to do this:
* The root window's buffer must be the full size of the window in
order for transparency to work. We now transition this buffer
to full-size upon first request of some QWindow to be translucent.
At the same time the root window is transitioned to being full-
screen, we also set its alpha blending mode to Source-Over in
order to permit lower z-ordered apps to show through.
* Set the root window's buffer position to (0, 0) to avoid a momentary
flicker of unpainted pixels on the far right-hand border as a
translucent window comes on-screen.
* Use Source-Over alpha blending for normal child windows too.
The QtGui core logic for setting the backing surface pixel format to
something with an alpha channel if Qt::WA_TranslucentBackground is
set on the toplevel QWidget already accomplished the remainder of
the work.
Change-Id: I1c2b31aa6323b4555fa194313f0f5e06ada494fc
Reviewed-by: Kevin Krammer <kevin.krammer@kdab.com>
Reviewed-by: Thomas McGuire <thomas.mcguire@kdab.com>
Reviewed-by: Rafael Roquetto <rafael.roquetto@kdab.com>
Reviewed-by: Andreas Holzammer <andreas.holzammer@kdab.com>
For now, we only call 'flushWindowSystemEvents()' for top-level
QWindows.
Change-Id: Ie6f6d3f9b70241820d9c38bcc0caf832d521268b
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
