Fix rounding error in QInputMethod::cursorRectangle().

Use QRectF consistently to avoid the rounding done in the conversion
to QRect.

Task-number: QTBUG-24463
Change-Id: If9ea858ebabf8c449ea058b9d379d4a57cb6c82d
Reviewed-by: Joona Petrell <joona.t.petrell@nokia.com>

QMenuBar: Clean up Q_WS_WIN

 - Replace Q_WS_WIN with Q_OS_WIN
 - Remove useless #ifdef Q_OS_WINCE which located in another #ifdef
   Q_OS_WINCE

Change-Id: I6279b6d74902ab3ca6bdb7292c2936a76e3e6952
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>

Harfbuzz-thai - When calling HB_OpenTypePosition we must set doLogClusters to true for thai

This bug resulted in word wrapping not working correctly for thai.

Reviewed-by: Adrian Yanes <ext-adrian.yanes@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
(cherry picked from commit 7495f2c7c55917a98c903547858be923028b7827)

Change-Id: Iedb26a10078c6cc4e90d090176ab63499293d69d
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>

Harfbuzz-thai - fix to_tis620 function for latin characters

This was an obvious typo - a missing "else" in a group of if statements.

Reviewed-by: Jiang Jiang <jiang.jiang@nokia.com>
Reviewed-by: Adrian Yanes <ext-adrian.yanes@nokia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
(cherry picked from commit a784bdcabe895ab927cbc28118d427c6e932b9fc)

Change-Id: Icbd458034e42cf7f2af33c1ecc9b0e4f1aa8724a
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>

Fix height of some Thai characters

Some Thai characters with two above glyphs were higher than fonts ascent. This
caused them to be clipped in for example qtcomponents buttons.

Added checking for this and calculation of new lower offset between glyphs if
needed.

Task-Number: ou1cimx1#898104
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Sami Merilä <sami.merila@nokia.com>
Reviewed-by: Pasi Pentikäinen <ext-pasi.a.pentikainen@nokia.com>
(cherry picked from commit d7c7bf721c93fe7629f725c181b52ad9ca645a7a)

Change-Id: Ie4ac69de75f50b68b2ba87353d83098846f319f4
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@nokia.com>

Fixed tst_qobject signalbug silently not compiled in -fast builds.

When configuring with -fast on Windows, a directory which contains two
.pro files, one SUBDIRS and one not, will have the SUBDIRS Makefile
silently clobbered by the non-SUBDIRS Makefile.  In practice, this may
cause various subdirectories to be silently excluded from the build.

Rearrange .pro files for this test to avoid triggering this bug.

See also  e9015b3bc8827a8276f58bdd41adcac1bcbf2137.

Task-number: QTBUG-21168

Change-Id: I18fac1ac636fdc6b2aaee1b4cdfee9c4bc2a77ff
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Add src/concurrent/qtconcurrentversion.h to .gitignore

Change-Id: I521d9809b5e7a67db1370dd178d3795beee2a388
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>

Silence coverity warning about sizeof(Node*) vs sizeof(Node) in QList

Coverity was complaining about QList::node_copy using sizeof(Node *)
instead of sizeof(Node).  The complete message from Coverity is:

"Passing argument "from" of type "struct QList<QObject *>::Node *" and
argument "(to - from) * sizeof (struct QList<QObject *>::Node *) /*4*/"
to function "memcpy" is suspicious. Did you intend to use "sizeof(struct
QList<QObject *>::Node)" instead of "sizeof (struct
QList<QObject *>::Node *)" ? In this particular case sizeof(struct
QList<QObject *>::Node *) happens to be equal to sizeof(struct
QList<QObject *>::Node), but this is not a portable assumption."

Task-number: QTBUG-24443
Change-Id: I583887a8b4177a224664065e14f8780a9586c9a3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Report failure to open deprecated header file

If a deprecated header file is not writable, report the error message.
The existing error message statement is not processed due to an
operator precedence issue.

Change-Id: Ifb2d4e234c64e560d05c021c621966bce19ce215
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Use the CMake definition of Windows for CMake macros.

Change-Id: I599ddaaf9176f76f2e144c893e607757957dfef2
Reviewed-by: Clinton Stimpson <clinton@elemtech.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Suppport up to 27 mouse buttons from DirectFB Plugin.

Previous to this update, our plugin accepts only 3 moue buttons
(from DirectFB, to be translated into Qt::MouseButton values.)
This update changes our Plugin, so that it will successfully
translate up to 27 mouse button numbers. It is unclear, to the
author, which configurations of DirectFB on Linux might be
capable of actually sending high-numbered mouse button Events
to Qt. But, if/when a DirectFB configuration does so, this
change makes Qt ready to recognize and work with the button event.

Task-number: QTBUG-24335

Change-Id: I8839ed386eb3d261a187355a244e80d511399ea8
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Remove unneeded code to convert to ASCII/Latin1.

This isn't used, and isn't wanted with the upcoming utf8 switch.

Change-Id: Ibec0fa7f36549df6a1c240353ffcd44beb2976f0
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Don't do macro self-expansion for moc anymore.

Change-Id: Ia34cc244a160c6c4abe6dacd7a2ce29bc4fc7bfb
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

Remove unused Q_CANNOT_DELETE_CONSTANT

Change-Id: I697f3f786d84fa766b3bc48d93fc265deee5fdcf
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Add the QT_LIBINFIX to the CMake config files.

Change-Id: If7bec54eca2d28cba314ce94a1ab1831c024ad31
Reviewed-by: Alexander Neundorf <neundorf@kde.org>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Remove detection of GCC versions which are not supported anywhere.

Change-Id: If254af51880e20be486b14f5237e6e1d4e4d6af3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Remove use of Q_BROKEN_DEBUG_STREAM.

No supported compiler defines it, and it was not used consistently
so it didn't work anyway.

Change-Id: Icc9e911e22daaedaee3d9316c15d19be26cd2e72
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Use QMAKE_SUBSTITUTIONS conditionals

Instead of generating CMake conditionals.

Change-Id: I3d987cc08666270e618222be9292558e73bc961e
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>

Fix breakage in dead code

Change-Id: Ibcddfb711a3f47bf957a4b010330e8a775f1a2e8
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Add further theme hints to QPlatformTheme.

- Add hints for QDialog/QDialogButtonBox.
- Add hint for available popup-menu area.
- Add keyboard scheme hint replacing
  QGuiApplicationPrivate::currentKeyPlatform()

Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@nokia.com>
Task-number: QTBUG-24204
Task-number: QTBUG-24315

Change-Id: I6653786b0dcb49a6fc264d3b9891dbfee502bd3e
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>

QEasingCurve: fix a missing */

Change-Id: I344342e950158c9dc832f1cd181d92899ddb1651
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Fail with #error early if the compiler does not support bool or explicit.

The fail mode for bool is moved from later in qglobal.h, and explicit is
used unguarded throughout Qt, so the macro is already useless.

Change-Id: Iff26892b025ba155e360a1f2dc93a67a6622dbc1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Remove custom text codec for C strings.

This setting is extremely harmful, as code cannot know whether or not to expect
it. It also made the behaviour of QString::fromAscii and ::toAscii unintuitive,
and caused a lot of people to make mistakes with it.

Change-Id: I2f429fa7ef93bd75bb93a7f64c56db15b7283388
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

Updated QApplication / QGuiApplication to match Qt 5 reality.

Removed obsolete Q_WS_X11/WIN/MAC stuff as well.

Change-Id: I8b65a0348a8cb170f3f060023fc6d9d982949b7e
Reviewed-by: Casper van Donderen <casper.vandonderen@nokia.com>
Reviewed-by: Laszlo Agocs <laszlo.p.agocs@nokia.com>

Fix typos

Change-Id: Ic0389d36bad001935376fb29d8c925b7ff908e4f
Reviewed-by: Joona Petrell <joona.t.petrell@nokia.com>

Refactor input context tests

Combined test input context instances and adapted changed
interfaces a bit.

Change-Id: Id5422cac5967d45dcaf2eb71b941d9c45e3b4dee
Reviewed-by: Joona Petrell <joona.t.petrell@nokia.com>

Fix QJson autotest on Windows.

- Fix warning: "Character represented by universal-character-name
  '\uFFFF' cannot be represented in the current code page (1252).
- Fix character conversion
- Change source file to plain ASCII, add defines for special
  characters.

Change-Id: I8557e6ba7488f746247f0d78181f14bfb7d5aaae
Reviewed-by: Miikka Heikkinen <miikka.heikkinen@digia.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

Migrate evdev plugins to the new plugin system

Change-Id: I7eb283646b7529080c8c2f41c267b0135c8f57e4
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

CodeCoverage: Move installCoverageTool call.

Call it from qtest_qParseArgs() instead of qExec(). This will allow to
remove the code coverage calls from quicktest.

Both the installation and the saving of code coverage data are now
gathered in testlib.

Change-Id: If3baedb0cff9d894e6f232b947af7e3158703d9d
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>

Fixed tst_qlogging app silently not compiled in some -fast builds.

When configuring with -fast on Windows, a directory which contains two
.pro files, one SUBDIRS and one not, will have the SUBDIRS Makefile
silently clobbered by the non-SUBDIRS Makefile.  In practice, this may
cause various subdirectories to be silently excluded from the build.

Rearrange .pro files for this test to avoid triggering this bug.

Task-number: QTBUG-21168
Change-Id: Ic51941db497d7b8fb004f3c50f5ea24d90ff3114
Reviewed-by: Toby Tomkins <toby.tomkins@nokia.com>

QtCore: add constexpr to geometric classes

This turns code like QPoint(12, 12) into a compile-time
expression, under a C++11 compiler, and allows to define
arrays of these types that end up in read-only memory,
just like PODs would.

Some constructors and QLine::pointAt() needed to be
adjusted to fit into the empty-body/only-return-expression
requirement for constexpr constructors/functions.

Change-Id: Id11ee2752c948930c3e40a91d1f6d7c97db7a373
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Flag QGraphicsView test as unstable on Mac.

Task-number: QTBUG-24296

Change-Id: I9f748c368fbc0cc2b272be9400da95d774d51bde
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Drop file-engine abstraction from public API

This abstraction imposed serious performance penalties and is being
dropped from the public API.

In particular, by allowing file names to be arbitrarily hijacked by
different file engines, and requiring engines to be instantiated in
order to decide, it imposed unnecessary overhead on all file operations.

Another flaw in the design with direct impact on performance is how
engines have no way to provide (or retain) additional information
obtained when querying the filesystem. In many places this has meant
repeated operations on the file system, where useful information is
immediately discarded to be queried again subsequently.

For Qt 4.8 a major refactoring of the code base took place to allow
bypassing the file-engine abstraction in select places, with
considerable performance gains observed. In Qt 5 it is expected we'll be
able to take this further, reaping even more benefits, but the
abstraction has to go.

[Dropping this now does not preclude that virtual file systems make an
appearance in Qt at a later point in Qt 5's lifecycle. Hopefully with a
new and improved abstraction.]

Forward declarations for QFileExtension(Result) were dropped, as the
classes were never used or defined.

Tests using "internalized" classes will only fully run on developer
builds. QFSFileEngine was removed altogether from exception safety test,
as it isn't its intent to test internal API.

Change-Id: Ie910e6c2628be202ea9e05366b091d6d529b246b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

Correct pkg-config usage to respect sysroot

Direct use of pkg-config variables disregards the sysroot offset pkg-config
factors into consideration with the dedicated variables:

--cflags-only-I
--libs-only-L

these parameters exist exactly for this reason.

Change-Id: Ieecf31ebe0640f64b272b84fba22701aacf8f966
Reviewed-by: Johannes Zellner <johannes.zellner@nokia.com>
Reviewed-by: Donald Carr <donald.carr@nokia.com>
Reviewed-by: Girish Ramakrishnan <girish.1.ramakrishnan@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>

Fix potential race condition in data structure of adopted thread watcher

Fix potential race condition in data structure of adopted thread watcher
in Windows. Since QVector is not thread safe, it is not safe to read and
append data to qt_adopted_thread_handles or qt_adopted_qthread
simultaneously in qthread_win.cpp. This patch fix this race condition.

Change-Id: I2d0c7a4cdde5390d38d15c98343f0fc6ddd24aba
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>

Merge overloads

Fixes ### Qt 5

Change-Id: I2846593616635440e17c8a9ce518780561992b84
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Merge overloads.

Fixes a Qt 5 TODO

Change-Id: Ide3159a5b979bd8e8a1edefff7fccb98dbe9a78e
Reviewed-by: Pierre Rossi <pierre.rossi@nokia.com>

QElapsedTimer/Win: Fix 64-bit integer overflow

The ticksToNanoseconds function in qelapsedtimer_win.cpp multiplies
ticks from performance counter with 1000000000 which can sometimes
result in 64-bit integer overflow. This can cause the elapsed time to
reset or jump around.

Task-number: QTBUG-23150
Change-Id: I464503e03cbe64e13906e773beafbf88e7dc256a
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
(cherry picked from commit f30a91ba9d98de1a0ebee5608ba289ac35871a8c)

Fix the qlocalsocket test

This test is broken in a couple of ways. A few one line fixes
combined into a single patch.

1. Linux is the only OS that does
abstract unix domain sockets by prepending a null as the first
character.  Don't test this on non-Linux platforms and expect
it to pass.

2. Change QVERIFY2 to QCOMPARE so we can see why this
fails in CI but no on the local system.  Use QCOMPARE
where possible.

Change-Id: Ic3d2cf9696730dc4d6589539fdfe8a48ccf28de5
Reviewed-by: Alex <alex.blasche@nokia.com>

Enable qlocalsocket auto test

Due to recent changes in the test it should now compile
and run properly.  This re-enables the test.

Change-Id: I6c647d99fa1f1b1c53e006fef2865d6be08ec16c
Reviewed-by: Alex <alex.blasche@nokia.com>

Fix error in addrlen field

addrlen must be initialize to the amount of space available
in the buffer.

Change-Id: I52945d780bba9d22aeaa7ac5a35a0e54dbea60dc
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Give the compile flags an EXECUTABLE_ prefix.

The fPIE flag should only be used with executables.

Change-Id: If799ae4a7fe2492af3aac67651659a52d365024a
Reviewed-by: Alexander Neundorf <neundorf@kde.org>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Add GL_BGR to the list of defined GL constants, right next to GL_BGRA

We use this one in WebKit's OpenGL texture mapper and it would be
convenient to have it here if it's not defined by the platform.

Change-Id: Idae33e49773e21c6290b164bccd4f209cf9820a2
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

make resolution of default spec less inefficient

resolve only once, in particular on unix.

Change-Id: I090698fc6029322a3a16d179d461af3e8336f6ad
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>

revamp reading of bootstrap variables

make the "evaluator" able to process simple variable expansions.
cache the processed spec, so it is not re-read for every variable.

Change-Id: I20e69ec7b65faa7d571e68dbfea6c21c79a62641
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>

regenerate unicode tables after rittk's patches

Change-Id: I60b416fc2dc2f0ccbcf13288a9ba2a42547269ec
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

optimize QString::toLower()/toUpper() for special cases, step 2

from now, QUnicodeTables::specialCaseMap[] starts with a placeholder; so,
if somethingCaseSpecial is true, then somethingCaseDiff is always greater than 0

Change-Id: Ibb1870512836eee71b1521564c0745096c05b2f9
Merge-request: 70
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Olivier
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

optimize QString::toLower()/toUpper() for special cases, step 1

reorganize QUnicodeTables::specialCaseMap as follows:
specialCaseMap contains sequence entries in form { length, a, b, .. }

Change-Id: Iea1f80bc2f4dc1f505428dad981cde26daaa52c7
Merge-request: 70
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Olivier
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

optimize QString::toCaseFolded()

use the codepath similar to QString::toLower()

Change-Id: Ica1948c5e9c82199307d9f823e07d42b50d59480
Merge-request: 70
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Olivier
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

optimize handling of surrogate pairs in toLower()/toUpper()

high surrogate part never changes on upper/lower casing
(which tested explicitly in the qunicodetables generator),
thus we can copy the high surrogate part in the loop

Change-Id: I24642b35eea7596c6ad494f2a7bc670f10b759a8
Merge-request: 70
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
Reviewed-by: Olivier
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

replace 'const QChar &' with 'QChar ' where appropriate

as QChar is actually an ushort and there is no point in taking its
address.

Merge-request: 69
Change-Id: Idcc9d621e5627514ade006aa12a789a88929d48b
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

Move QSysInfo out of qglobal.h and into a separate header

qsysinfo.h is still included by qglobal.h, but it may be possible
to remove the include from qglobal.h and instead include qsysinfo.h
only where needed (e.g. qendian.h, qdatastream.h).

Change-Id: Ifa2c72e0dae206d88eaa192e15a906297673c048
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

QEasingCurve: implement copy assignment operator inline

Implement the copy-assignment operator inline, using the
copy-swap idiom. This makes assignment strongly exception
safe, but also simplifies maintainence, because op= is
formulated in terms of the copy constructor now.

Change-Id: I803c9100a520d659b685992237cb76fd114222c8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Print out the try_compile output for tests expected to fail.

This way, ctest -V shows that it fails for the expected reason.

Change-Id: I97589d4e90d889ea0b10dbd6192526712dbfdd8e
Reviewed-by: Alexander Neundorf <neundorf@kde.org>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

remove bizarre late feeding of user_configs into base_vars

there is no obvious reason why this should happen. if base_vars is used
again, the user configs will be parsed again, too.

Change-Id: Ib56e01a468cdb5e81d610bcaf0163bf730cbae05
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>

qmimetype: Fix test failure when LC_ALL is set.

This code sets LANG=en_US so that the method comment(), which returns a translated
name, can be compared with an expected result in English.
(QMimeType::comment uses QLocale::system().name() and QLocale::system().uiLanguages())

But LANG= has no effect if LC_ALL is set, so LC_ALL needs to be cleared (or set
to en_US) for the test to work.

Change-Id: Icb031057769be9bc8c0fcab65daa45e7bf1d5b18
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Test for QT_NO_SSL instead of QT_NO_OPENSSL

Change the ifdefs in our own code (except openssl backend) to use the
new configure flag.

Change-Id: I8774734771c66b22164b5fae8fdb27814ac3df7b
Reviewed-by: Richard J. Moore <rich@kde.org>

Make qget manual test compile without ssl support

Change-Id: I35f92328b79df1cfcae52fa6ff8290f260183aea
Reviewed-by: Richard J. Moore <rich@kde.org>

Make https requests fail fast when configured without ssl support

Restored the Qt4 behaviour where attempting a https request using
QNetworkAccessManager fails with ProtocolUnsupportedError instead
of timing out or hanging.

Covered by existing autotests.

Task-number: QTBUG-17189
Change-Id: Iceb1ba6558c7d2b1af8ddf8d4ea9315a5b44d970
Reviewed-by: Richard J. Moore <rich@kde.org>

QEasingCurve: simplify and fix copy constructor

The copy constructor used the default Private constructor,
followed by the application of the compiler-generated
copy assignment operator, and finally replaced the config
member with a copy of itself.

This is needlessly inefficient.

Worse: it's incorrect: if config->copy() throws, then
*d_ptr is leaked.

Solution: implement the copy constructor for Private,
and use it in the copy constructor of the public class.

Effect: everything that can throw now prevents the new
Private class from being created, and the compiler ends
up cleaning up after us.

Change-Id: I09ed18bb39ee7cd81aaa8ba01676fc202502a8e3
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

QHash: more robust use of qDebug()

I'm sure 'line' can't contain anything resembling a format string,
but better safe than sorry.

Change-Id: I0aa0f4a5a03d97b18d6a866c27fa721fcbe0f77b
Reviewed-by: Kai Koehne <kai.koehne@nokia.com>

QEasingCurve: implement move constructor

The move constructor sets other.d_ptr to zero. This is safe, because
after being moved from, the object is left in a state in which it
can be safely destroyed (delete nullptr is a no-op).

It cannot meaningfully be used anymore (most members will crash with
a nullptr dereference), but in most cases, the moved-from object
cannot be accessed anyway (not a named object), and if a named object
is moved from, it must have been through explicit std::move(), as in
the test case.

The STL makes better guarantees (moved-from containers are .empty()),
but I don't think it's worth introducing a null state into
QEasingCurve just for supporting a use-case that should be
considered a bug anyway.

Change-Id: I4115b7386cdea6960507da6843a0d0196d8e4139
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

QEasingCurve: implement move-assignment operator

Implemented as in QPen etc.

Change-Id: I65b43c6ec7308ca4b44f614594c15c41ab2f89f9
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

QEasingCurve: add member-swap

Implementated as in QPen etc.

Change-Id: Ia08551bf7902b60e115d1b1d2353030597e34841
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

QEasingCurve: mark as Q_MOVABLE_TYPE

The private data doesn't contain any back-links to the public
class, afaics. So QEasingCurve qualifies for movable.

Change-Id: Id8ab186fc0a25740458341256b53425dc40bce37
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Use requested window format to match rest of the patch in 38257651

Otherwise we end up with an error at runtime:

  QEGLPlatformContext::makeCurrent: eglError: 3009, this: 0x4b5c58

Change-Id: I1e794f427243e40f0a8f572a936b9fb756e89f8a
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Port badxml autotest to QMetaObjectBuilder

The meta-object format is going to change for Qt5. Use QMOB to
insulate the badxml test from such changes.

(It just so happens that the QFAIL("a failure") statement is still on
line 109 after the refactoring, so the expected_badxml.* files'
location tags did not have to be changed.)

Change-Id: I04421d13c4df71c8004fa71cafc4823a59079a41
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>

Rename Q_MUTEX_LINUX to QT_LINUX_FUTEX

As requested by Thiago

Change-Id: Ie5b00cf4e530e19d360d0bc588f0f051e04b338f
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>

moc: Fix loading of plugins in release mode.

Generate the plugin meta data with "debug" set to false/true
in two code sections #ifdefed QT_NO_DEBUG.

Do not use the value of QT_NO_DEBUG set at moc compile
time which does not work in release mode/Windows.

Change-Id: I0252795ed063bebb2c3b3784f880e64845b5b7e7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>

Streamline QPalette initialization code.

Share the brushes to reduce memory usage by QBrushData.

Change-Id: I04044b4c54b1f6a35a73d8bb4d299661ec0136a5
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Use new plugin system for Input Context plugins.

Change-Id: Ifd1f63499afba546a745815762263a58ebf66460
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

QHeaderView - create a union to save some space.

Beside saving some space it also ensures a bit faster vector operations
(remove and insert memcpy) - and lower the risk for a cache-miss.
(Many operations seems to be about 5-10% faster in the benchmark).

Change-Id: If8109e2146c25f642015906375dfbb449706ce8e
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

QHeaderview - many sections removed performance boost.

When the QHeaderView model rowCount is decreased then the headerview
needs to remove some sections.

If e.g swapSections had occured then qheaderview before this patch did
a very slow removal by removing one section at a time and for each section
it restored visual and logical indexes.

By temporarily storing the logical index on the section we
can reduce the execution time of this a lot. The old code is only
faster when we remove one index (and is kept for that situation)

There is a complexity in difference (and many factors in our
qheaderview benchmark (the trunc part))

Change-Id: If45eab17fbca3364d458b004980570b5d013e4bd
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Remove qMacVersion()

Add QSysInfo::macVersion() instead, to match the windowsVersion()
function.

Change-Id: I783e59583ca21653d25586156cbb0cb1f301868b
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>

Fix qlogging test.

Build app sub-process first.

Change-Id: I87a11f7fd5d8a82584e496722f79e236191b0fb3
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>

Improve tst_QPrinterInfo::testForDefaultPrinter().

Be explicit about skipping the test when no default printer is present.

Change-Id: If69b275eb0f490411471ec42798d8aefcc57fd83
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Allow printer names to contain hyphens in QPrinterInfo test.

Change-Id: I473627413d2f1cd5637704f2948eb07a264be49e
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Fix warning about unregistered metatype in QWindow autotest.

Change-Id: If9d8d7e1cd52815ef7231294e4890dfafcd28ec8
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Flag two tests (Win+A, Simon+G) in parseString as expect fail on OSX.

Remove previously defined insignificant test flag as the number of
tests failing has been reduced.

Task-number: QTBUG-24406
Task-number: QTBUG-23058

Change-Id: I01b41f30469cf7a440e21195e105cb30a8db76e2
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Make the QApplication autotest work for shadow builds.

Use QFINDTESTDATA to locate the helper applications.

Change-Id: I604d10e37c9367f2e95225864edf5bf705f1d961
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

tst_qapplication: mark expected failure on mac

The tab key behavior in Qt5 no longer respects the "Text boxes and lists
only" tab navigation option in OSX, which is the default.  This is a
regression since Qt4.

Task-number: QTBUG-24372
Change-Id: I54c1663f8fb259dd847083432102a0bfad7dd69c
Reviewed-by: Toby Tomkins <toby.tomkins@nokia.com>

Changed qnetworksession unittest to work from installation dir

- Made test depend on subprogram
- added install of subprogram

Change-Id: Ib263e9e75ed3c900b52fb1c9b6d319e71d19bdbb
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>

Stabilize tst_qdbusabstractinterface and make it XFAIL

This test was written incorrectly in a way which happened to allow it to
pass most of the time (but not all the time).

Reset the state of test objects between each test function, and mark the
broken functions with QEXPECT_FAIL and a link to a task.

Replace the unusual WaitForPinger construct with the usual
QTRY_VERIFY/QTRY_COMPARE method of verifying asynchronous operations.

Task-number: QTBUG-24262
Change-Id: I82d09002307c0b500bf60cd5b583674321b37609
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Remove many unneeded virtuals.

Cleanup a ### Qt 5 TODO.

Change-Id: I384e868ecc3ca4d80e4e71bb54d402f4ec1337a8
Reviewed-by: David Faure <faure@kde.org>

convert length functions in QDom from uint to int.

Fixes a ### Qt 5

Change-Id: I7a385f2b704d38ae626094f27b06a918d4a9bc48
Reviewed-by: David Faure <faure@kde.org>

Merge overloads.

Fix a ### Qt 5 by merging the overloads.

Change-Id: If33e7592191c81b32caa6d68c73dbf2282437886
Reviewed-by: David Faure <faure@kde.org>

Include qt_windows.h for Windows CE.

The include is needed for OutputDebugString,
as it is defined in winbase.h. To include
use the centralized header.

Change-Id: I486da20d7b054cee352be085e65a0ede1394653d
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>

Silence warnings/debug output from qobject tests

They create noise in the test results.

Change-Id: I40e7239ba7cd41bec577fe8220c86476553a6502
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>

Merges several overloaded functions in QCoreApplication.

The source compatibility is kept.

Change-Id: If66053b271d65062b3c0ce6ec66c8394a37b4e3e
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

don't attempt to build host tools which are gone from qtbase

Change-Id: If9a3fb9792fc5d90496ff834b2c3adc177e1311e
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Marius Storm-Olsen <marius.storm-olsen@nokia.com>

Windows - fix connecting to a socket using IPv4 mapped IPv6

Connecting to an IPv4 mapped IPv6 address (e.g. ::FFFF:127.0.0.1)
requires the IPV6_V6ONLY socket option to be cleared.
This was causing tst_qtcpserver::ipv6ServerMapped autotest to fail.
The same change is not required on MacOS X - the test passes there.

Task-number: QTBUG-24351
Change-Id: I6c08b19f0daa12765da2d44792ffb17299322695
Reviewed-by: Markus Goetz <markus@woboq.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Remove QPlatformIntegration/QPlatformTheme from QGuiAppPrivate.

Forward-declare instead.

Change-Id: I3851994e8bc05b389e94e948478339ba33d521c1
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

Fix warning about converting false to pointer value.

Change-Id: I70c324e6fcfd2bba3ab44837c5ce2c007de8896f
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>

Make sure that we pick up xcb from pkg-config

and run the configure checks with the values

Change-Id: Ie8e0072c686c6a7dce1d02e25a9c1abce4679d34
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>

Remove Native Client special casing from configure.

Change-Id: I20f044b267b6c18e5a3050de22710c45e7573cc3
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>

Remove QAccessibleInterface::relationTo().

QAccessibleInterface::relations() replaces both relatedTo() and
some of the RelationFlags that navigate() handled before.

Change-Id: I4b9d2c28ba3d753efe7cc9bfa0ad68bed3ee02fd
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>

Updated for accessibility

Change-Id: I13b5c2f293e9c1c238c7b7c6aae9d26f5c150f24
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@nokia.com>

Port QDBusAdaptorConnector meta-object to revision 6

We want to drop support for old revisions in Qt 5.

This commit brings the QDBusAdaptorConnector meta-object in sync with
current moc output.

The QtDBus implementation was assuming that the relaySlot() slot
would be created at index methodOffset() in the meta-object, but
since revision 4 that's no longer the case (signals always come
first). Made the code more robust by actually querying the
meta-object what the index is.

Change-Id: Ie0791680cc2e9e5fb1472c4462c391f92ea22ea6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

Remove unused QtDBus meta-object extensions

The inputSignature and outputSignature fields were not being used
anymore.

All tests still pass.

Change-Id: Icbc8fdcd5179a2b1b4843d58b90af925f6bef133
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

QPoint/QPointF: inline manhattanLength(); QLineF: inline isNull()

There's no reason for them not to be, and it's a prerequisite
for making these functions constexpr.

Change-Id: I03c9965147b51014c7af60a4c2d7f25a3f6e21a7
Reviewed-by: David Faure <faure@kde.org>
Reviewed-by: Gunnar Sletta <gunnar.sletta@nokia.com>