qt:qt-iphone-clone.git
8 years agorespect QAbstractSpinBox::NoButtons in windowsce / windowsmobile styles
Joerg Bornemann [Mon, 10 Aug 2009 13:12:08 +0000 (15:12 +0200)]
respect QAbstractSpinBox::NoButtons in windowsce / windowsmobile styles

Code is adapted from QCommonStyle which handles this case for other
styles.

Reviewed-by: thartman
8 years agofix QTextFormat::doubleProperty where qreal is float
Joerg Bornemann [Tue, 11 Aug 2009 09:53:41 +0000 (11:53 +0200)]
fix QTextFormat::doubleProperty where qreal is float

This function was too strict. It returned 0 if the property wasn't of
type QVariant::Double. Now it tests for QMetaType::Float too.

Reviewed-by: kh1
Reviewed-by: mauricek
8 years agofix decoration of DontShowOnScreen widgets on Windows CE
Joerg Bornemann [Wed, 12 Aug 2009 09:17:04 +0000 (11:17 +0200)]
fix decoration of DontShowOnScreen widgets on Windows CE

Widgets with the WA_DontShowOnScreen attribute must not have a window
decoration.

Autotest: tst_QWidget::initialPosForDontShowOnScreenWidgets
Reviewed-by: thartman
8 years agoTest naming convention fixes.
Rohan McGovern [Wed, 12 Aug 2009 05:03:27 +0000 (15:03 +1000)]
Test naming convention fixes.

Make sure the testcase and directory name are the same
(excluding `tst_').

8 years agoUpdate embedded copyright headers.
Jason McDonald [Wed, 12 Aug 2009 02:40:14 +0000 (12:40 +1000)]
Update embedded copyright headers.

Reviewed-by: Trust Me
8 years agoRevert "Frame property for QDateTimeEdit can be set independently of"
Matthew Cattell [Tue, 11 Aug 2009 14:32:32 +0000 (16:32 +0200)]
Revert "Frame property for QDateTimeEdit can be set independently of"

This reverts commit 6844dea0cb583a86bc72e7f008720ab76deef040.

added to wrong branch. should be in 4.6 but this was added to 4.5

8 years agoFrame property for QDateTimeEdit can be set independently of
Matthew Cattell [Tue, 11 Aug 2009 13:58:43 +0000 (15:58 +0200)]
Frame property for QDateTimeEdit can be set independently of
calendar popup

A frame was always drawn around the QDateTimeEdit editor if a popup
calendar had been set. QStyleOptionsComboBox options are being set in
paintEvent and initialised from the properties of QStyleOptionsSpinBox but
were missing the frame bool property. Now, if the user sets a frame on the
QDateTimeEdit, this property will be consistent with setFrame() property
of the QDateTimeEdit widget.

8 years agoPhonon: improve locking to make it safer to load a source
Thierry Bastian [Tue, 11 Aug 2009 11:16:51 +0000 (13:16 +0200)]
Phonon: improve locking to make it safer to load a source

Task-number: 259482

8 years agoUpdate URL for Qt FTP site.
Jason McDonald [Tue, 11 Aug 2009 10:50:53 +0000 (20:50 +1000)]
Update URL for Qt FTP site.

Reviewed-by: Trust Me
8 years agoUpdate URL for Qt Labs.
Jason McDonald [Tue, 11 Aug 2009 10:47:48 +0000 (20:47 +1000)]
Update URL for Qt Labs.

Reviewed-by: Trust Me
8 years agoUpdate URL for mailing list server.
Jason McDonald [Tue, 11 Aug 2009 10:44:24 +0000 (20:44 +1000)]
Update URL for mailing list server.

Reviewed-by: Trust Me
8 years agoUpdate URL for Qt web site.
Jason McDonald [Tue, 11 Aug 2009 10:38:02 +0000 (20:38 +1000)]
Update URL for Qt web site.

Reviewed-by: Trust Me
8 years agoUpdate URL of online documentation.
Jason McDonald [Tue, 11 Aug 2009 10:29:21 +0000 (20:29 +1000)]
Update URL of online documentation.

Reviewed-by: Trust Me
8 years agoUpdate old email address.
Jason McDonald [Tue, 11 Aug 2009 09:44:38 +0000 (19:44 +1000)]
Update old email address.

Reviewed-by: Trust Me
8 years agoEliminate some mentions of Trolltech.
Jason McDonald [Tue, 11 Aug 2009 09:43:41 +0000 (19:43 +1000)]
Eliminate some mentions of Trolltech.

Reviewed-by: Trust Me
8 years agoQt's domain name is now qt.nokia.com.
Jason McDonald [Tue, 11 Aug 2009 07:40:03 +0000 (17:40 +1000)]
Qt's domain name is now qt.nokia.com.

Reviewed-by: Trust Me
8 years agoUpdate license headers.
Jason McDonald [Tue, 11 Aug 2009 05:32:51 +0000 (15:32 +1000)]
Update license headers.

Reviewed-by: Trust Me
8 years agoFixes formatting of date strings in psql driver
Bill King [Mon, 10 Aug 2009 23:47:57 +0000 (09:47 +1000)]
Fixes formatting of date strings in psql driver

After the coverity fix, the proper code path executed, which failed to
enquote the date field properly, so this fix fixes that issue.

8 years agoMissing break when formatting from a time value
Bill King [Mon, 10 Aug 2009 23:16:33 +0000 (09:16 +1000)]
Missing break when formatting from a time value

Found by coverity.

8 years agoDoc: Our webdomain has changed.
Volker Hilsheimer [Mon, 10 Aug 2009 15:22:13 +0000 (17:22 +0200)]
Doc: Our webdomain has changed.

8 years agoDoc: Fix a typo
Benjamin Poulain [Mon, 10 Aug 2009 14:43:09 +0000 (16:43 +0200)]
Doc: Fix a typo
Genarated->generated

8 years agoFixed an assert that could happen when the mediaSource is deleted
Thierry Bastian [Mon, 10 Aug 2009 11:45:24 +0000 (13:45 +0200)]
Fixed an assert that could happen when the mediaSource is deleted

When using streaming, it could happen that the last reference to the
MediaSource is in another thread. So the objects are destroyed from
another thread. In which case we would delete QObject (ioDevice) in
another thread. That is fixed by calling deleteLater which will ensure
that they are deleted in their own thread.
Note: there was a nother assert that could happen due to a race
condition in the worker thread. That is also fixed with this patch.

Reviewed-by: jbache
8 years agoFix autotest for qtwidgets::snapshot
Jens Bache-Wiig [Mon, 10 Aug 2009 07:55:37 +0000 (09:55 +0200)]
Fix autotest for qtwidgets::snapshot

Kramer is no longer accessible from the pulse test system

Reviewed-by: jesperht
8 years agoPrevent a crash when there is no QTest::logger when calling warn()
Andy Shaw [Mon, 10 Aug 2009 06:13:55 +0000 (08:13 +0200)]
Prevent a crash when there is no QTest::logger when calling warn()

Also, the info() function was modified to use an assert rather than an
pointer check with an if to make it consistent with the rest of the file.

Reviewed-by: jesperht
8 years agoremove useles QT += network
Oswald Buddenhagen [Fri, 7 Aug 2009 14:05:33 +0000 (16:05 +0200)]
remove useles QT += network

8 years agoFix compilation with xlC 7: operands to ?: must match.
Thiago Macieira [Fri, 31 Jul 2009 08:30:44 +0000 (10:30 +0200)]
Fix compilation with xlC 7: operands to ?: must match.

See 3ae2cab9c8bd1790a00da2755ac036143a3a35f4 for another similar fix.

Reviewed-by: Trust Me
(cherry picked from commit 18fbfdf0f774198e2e1277e064cc3a8eb9dbb29d)

8 years agoFix compilation with xlC 7: operands to ?: must match.
Thiago Macieira [Thu, 30 Jul 2009 14:40:06 +0000 (16:40 +0200)]
Fix compilation with xlC 7: operands to ?: must match.

"../shared/qm.cpp", line 556.45: 1540-0207 (S) No common type found for operands with type "const char [7]" and "QByteArray".

Reviewed-by: Trust Me
(cherry picked from commit 3ae2cab9c8bd1790a00da2755ac036143a3a35f4)

8 years agoxlC 7 cannot compile QtConcurrent with these templates here
Thiago Macieira [Thu, 16 Jul 2009 13:38:27 +0000 (15:38 +0200)]
xlC 7 cannot compile QtConcurrent with these templates here
(cherry picked from commit cb64ac587249f5dc6563a035e2ef5a3ad2bc5d13)

8 years agoAdd functionality tests for XSync.
Thiago Macieira [Wed, 15 Jul 2009 15:22:11 +0000 (17:22 +0200)]
Add functionality tests for XSync.

It was reported to be auto-detected, but wasn't. Apparently, AIX 6's
X11 doesn't have this.

Reviewed-By: Denis Dzyubenko <denis.dzyubenko@nokia.com>
(cherry picked from commit 0a63875d787e1b035ace2c76fa1d0de6329127d7)

8 years agoAutotest: ensure we don't mishandle SSL certificates with NULs
Thiago Macieira [Fri, 7 Aug 2009 11:17:05 +0000 (13:17 +0200)]
Autotest: ensure we don't mishandle SSL certificates with NULs

This is a vulnerability in some implementations. Qt isn't affected
because... well, we never implemented the decoding of escape sequences
:-)

8 years agoAvoid compiling imageformats if you configure with -no-lib*
Marius Storm-Olsen [Fri, 7 Aug 2009 06:32:35 +0000 (08:32 +0200)]
Avoid compiling imageformats if you configure with -no-lib*

Task-number: 239108
Reviewed-by: Andy Shaw
8 years agoUse QFile instead of QTemporaryFile in compilerwarning testcase
Marius Storm-Olsen [Thu, 6 Aug 2009 12:45:51 +0000 (14:45 +0200)]
Use QFile instead of QTemporaryFile in compilerwarning testcase

QTemporaryFile on Windows doesn't open the file as a sharable, and
doens't close the file when you call .close(). So the testcase
fails on Windows with a Sharing Violation when the compiler tries
to compile the file.

By switching to QFile we can at least close the file before letting
the compiler chew on it, and remove it at the end when the testcase
is done. Open the file with Truncate, in case the testcase fails to
remove the file.

Reviewed-by: trustme
8 years agoFix crash in QX11Data::xdndHandleEnter when XGetWindowProperty fails
Octavian Voicu [Thu, 6 Aug 2009 12:09:47 +0000 (14:09 +0200)]
Fix crash in QX11Data::xdndHandleEnter when XGetWindowProperty fails

Task-number: 259143

Merge-request: 1119
Reviewed-by: Denis Dzyubenko <denis.dzyubenko@nokia.com>
8 years agoTests: Make uic/uic3 use QLibraryInfo paths as do the linguist tests.
Friedemann Kleint [Thu, 6 Aug 2009 11:44:12 +0000 (13:44 +0200)]
Tests: Make uic/uic3 use QLibraryInfo paths as do the linguist tests.

8 years agoMac: fix autotest
Richard Moe Gustavsen [Thu, 6 Aug 2009 11:35:28 +0000 (13:35 +0200)]
Mac: fix autotest

On mac, we prefer testing with mp3 rather than ogg, Because in the latter
case, you must install perian to make it work. And the test machines
dont.

Reviewed-by: brad
8 years agoTests: Make tools tests (rcc/moc/uic/uic3) run without X11-connection.
Friedemann Kleint [Thu, 6 Aug 2009 09:10:00 +0000 (11:10 +0200)]
Tests: Make tools tests (rcc/moc/uic/uic3) run without X11-connection.

8 years agoMake the test pass by setting a path where you have dirs.
Alexis Menard [Thu, 6 Aug 2009 07:52:15 +0000 (09:52 +0200)]
Make the test pass by setting a path where you have dirs.

The test was failing because the goal of the test expect to get only
directories displayed. But if you don't have some in the current dir
then it fails. home dir should be enough.

Reviewed-by:TrustMe

8 years agoCleaned up test naming and platform-specific tests.
Rohan McGovern [Thu, 6 Aug 2009 01:04:23 +0000 (11:04 +1000)]
Cleaned up test naming and platform-specific tests.

When an autotest fails to compile, there's no way to determine which
testcase(s) have been prevented from running.

Our results parsing tools have been guessing, under the assumption that
a directory called `qdogwalker' always contains a testcase named
`tst_qdogwalker'.  That wasn't true for all our tests, so let's make it
true.

Also changed the platform-specific tests so that qmake will simply skip
those tests on unsupported platforms, instead of wasting time compiling
a useless QTEST_NOOP_MAIN test.

8 years agoFix a bug when activating HideNameFilterDetails on Windows filedialog.
Alexis Menard [Wed, 5 Aug 2009 11:19:29 +0000 (13:19 +0200)]
Fix a bug when activating HideNameFilterDetails on Windows filedialog.

The HideNameFilterDetails was not taken in account for windows
file dialog. This patch fix that. For Windows Vista and later we use
now the new COMMON_ITEM_DIALOG API (in master) where it add always
the filter extension like *.txt so no point to take that flag in account.

Task-number: None, during testing
Reviewed-by: prasanth
8 years agoFix compilation on AIX: -I/usr/include cannot be used.
Thiago Macieira [Thu, 30 Jul 2009 10:58:06 +0000 (12:58 +0200)]
Fix compilation on AIX: -I/usr/include cannot be used.

The reason is that the C++ compiler needs to add -I/usr/vacpp/include
before -I/usr/include, so our adding of -I/usr/include changes the order.
This causes a compilation error in the C++ header xlocinfo.h.

In any case, these checks done in qmake are unnecessary. If the OpenSSL
headers are in one of those include paths, they will be found on their
own already.

Reviewed-By: Bradley T. Hughes
(cherry picked from commit 04f7834fcd8db4ecb6a4c53de42ac83f571b37c8)

8 years agoFix compilation when QT_NO_HTTP is defined
Markus Goetz [Wed, 5 Aug 2009 08:57:13 +0000 (10:57 +0200)]
Fix compilation when QT_NO_HTTP is defined

Note however you still need to -nomake examples -nomake demos
when configuring Qt.

Reviewed-by: Thiago
Task: 259179

8 years agoFixes hangup when pressing key in QListWidget
Olivier Goffart [Wed, 5 Aug 2009 08:42:46 +0000 (10:42 +0200)]
Fixes  hangup when pressing key in QListWidget

If there is severals matching items but all of them are disabled, it
would have produced an infinite loop.
Block the loop once we reach an already matched item.

Task-number: 258949
Reviewed-by: Thierry
8 years agoCocoa: Menus show old selected values.
Richard Moe Gustavsen [Tue, 4 Aug 2009 13:59:25 +0000 (15:59 +0200)]
Cocoa: Menus show old selected values.

We never told Cocoa that it needed to redraw the window view
when a window was shown. This is implicit if the window is
shown for the first time, but needs to be done explicit
if you hide and show it again.

Task-number: 254672
Reviewed-by: bnilsen
8 years agoGet more database tests working/passing.
Bill King [Wed, 5 Aug 2009 03:38:44 +0000 (13:38 +1000)]
Get more database tests working/passing.

8 years agoMake file extension case insensitive with the Gtk+ file dialog
Jens Bache-Wiig [Tue, 4 Aug 2009 14:54:07 +0000 (16:54 +0200)]
Make file extension case insensitive with the Gtk+ file dialog

The Qt dialog treats extension filters as case insensitive where
the default behavior of the Gtk+ file dialog is to be case
sensitive. The GtkFileFilter allows only simple glob syntax to
be used so we basically rewrite the extension string from .xyz
to .[xX][yY][zZ].

Task-number: 259155
Reviewed-by: joao
8 years agoFix: Gtk native dialogs do not repaint when zero-timers are used
Jens Bache-Wiig [Tue, 21 Jul 2009 11:37:43 +0000 (13:37 +0200)]
Fix: Gtk native dialogs do not repaint when zero-timers are used

The problem was that our glib timers were given too high priority,
effectively starving the Gtk event loop. By assigning our timer source,
G_PRIORITY_DEFAULT_IDLE we ensure that all gtk+ events are handled first,
thus allowing the native dialogs to respond and repaint properly.

Task-number: 258433
Reviewed-by: bhughes
8 years agoFix indentation.
Alexis Menard [Tue, 4 Aug 2009 12:13:57 +0000 (14:13 +0200)]
Fix indentation.

Reviewed-by:TrustMe

8 years agoFix ancestor flags that are not correctly update when reparenting.
Alexis Menard [Tue, 4 Aug 2009 11:05:05 +0000 (13:05 +0200)]
Fix ancestor flags that are not correctly update when reparenting.

updateAncestorFlags was not reseting the flags if you change the parent
that have for instance itemsClipChildrenToShape to a new one that
doesn't have that flag.

Task-number:258956
Reviewed-by:bnilsen

8 years agoBuild on snow leopard.
Morten Sorvig [Tue, 4 Aug 2009 07:31:56 +0000 (09:31 +0200)]
Build on snow leopard.

Don't error out when building qmake, just let it build a 64-bit binary (even for carbon)

RebBy: Richard Moe Gustavsen

8 years agoMerge branch '4.5' of git@scm.dev.nokia.troll.no:qt/qt into 4.5
Jason McDonald [Tue, 4 Aug 2009 03:39:50 +0000 (13:39 +1000)]
Merge branch '4.5' of git@scm.dev.nokia.troll.no:qt/qt into 4.5

8 years agoFix obsolete license headers
Jason McDonald [Tue, 4 Aug 2009 03:38:51 +0000 (13:38 +1000)]
Fix obsolete license headers

Reviewed-by: Trust Me
8 years agoPreserved fill rule for perspective mapped strokes in raster engine.
Samuel Rødal [Mon, 3 Aug 2009 12:31:10 +0000 (14:31 +0200)]
Preserved fill rule for perspective mapped strokes in raster engine.

Task-number: 254407
Reviewed-by: Gunnar
8 years agoAutotest: disable the pointer-tracking tests in 4.5
Thiago Macieira [Mon, 3 Aug 2009 11:18:36 +0000 (13:18 +0200)]
Autotest: disable the pointer-tracking tests in 4.5

The functionality these tests tested was broken in 4.5 and has been
disabled. Therefore, these tests simply cannot pass.

It's fixed in 4.6.

8 years agoFix compilation of webkit with mingw
Thierry Bastian [Wed, 15 Jul 2009 07:07:37 +0000 (09:07 +0200)]
Fix compilation of webkit with mingw

8 years agoFix more database autotests.
Bill King [Mon, 3 Aug 2009 04:21:06 +0000 (14:21 +1000)]
Fix more database autotests.

8 years agoCocoa with namespace breaks the build
Richard Moe Gustavsen [Fri, 31 Jul 2009 07:58:21 +0000 (09:58 +0200)]
Cocoa with namespace breaks the build

Added the needed macros around the classnames the way it
should be done.

Reviewed-by: Prasanth
8 years agoCoverty warnings --. This switch was missing a break.
Alexis Menard [Thu, 30 Jul 2009 15:37:30 +0000 (17:37 +0200)]
Coverty warnings --. This switch was missing a break.

Reviewed-by:TrustMe

8 years agoAllow maximize button on the titlebar for a fixed size window
Denis Dzyubenko [Wed, 29 Jul 2009 13:34:31 +0000 (15:34 +0200)]
Allow maximize button on the titlebar for a fixed size window

On Windows we will add maximize button to the titlebar even if the
window has a fixed size if the user explicitely asked for it by
setting Qt::CustomizeWindowHint | Qt::WindowMaximizeButtonHint.

Task-number: 250188
Reviewed-by: Leonardo Sobral Cunha
8 years agoAutotest: add a newline after the user program headers.
Thiago Macieira [Thu, 30 Jul 2009 12:35:07 +0000 (14:35 +0200)]
Autotest: add a newline after the user program headers.

If the user forgot to end their headers with a newline, the
compilation would fail because the next line is #include
<QtCore/QtCore>.

Reviewed-by: Jesper Thomschütz
8 years agoRemove mem leak / warning in the cocoaport
Richard Moe Gustavsen [Thu, 30 Jul 2009 12:11:07 +0000 (14:11 +0200)]
Remove mem leak / warning in the cocoaport

Remove mem leak / warning in the cocoaport

Reviewed-by: msorvig
8 years agoMac: Fix crash in QCoreTextFontEngineMulti when disabling font kerning.
Morten Sørvig [Thu, 30 Jul 2009 11:41:34 +0000 (13:41 +0200)]
Mac: Fix crash in QCoreTextFontEngineMulti when disabling font kerning.

CFDictionaryAddValue expects a CFNumberRef, not a pointer-to-CFNumberRef.

Reviewed-by: TrustMe
8 years agoUnable to change focus between two line edits on mac
Richard Moe Gustavsen [Wed, 29 Jul 2009 13:48:46 +0000 (15:48 +0200)]
Unable to change focus between two line edits on mac

This is because we try to decide whether the window cocoa tells
us to be active should be active, and if we desagree, we do nothing.
The result is that Qt and Cocoa ends up in different states.
I decided to remove a lot of the logic that went on in this case, and
the resons is:
    1. By checking the callplaces to
onApplicationWindowChangedActivation, we know that we always have a
valid widget pointer, and we know that the widget always is a window
(otherwise Cocoa would never tell us that the widget got active).
    2. We can never end up doing nothing in this response. The best
we can do is to follow what Cocoa tells us. If this turns out to
break something, it would probably be better to check why we get an
activation call in the first place for a window that should not be
activated (e.g. is canBecomeKeyWindow set correctly?)

Task: 253610
RevBy: msorvig

8 years agoFixes hack around show queries not allowed to be prepared.
Bill King [Thu, 30 Jul 2009 01:18:59 +0000 (11:18 +1000)]
Fixes hack around show queries not allowed to be prepared.

Mysql queries other than select can't be prepared, otherwise they fail
to return the necessary meta-information to enable them to be seen as
returning data under certain versions of mysql. This fixes the hack
to work correctly until we stop preparing queries automagically.

8 years agoDoc - Renamed the cursor's images to illustrate splitV and splitH properly
Kavindra Devi Palaraja [Wed, 29 Jul 2009 15:17:12 +0000 (17:17 +0200)]
Doc - Renamed the cursor's images to illustrate splitV and splitH properly

Task: 258895

Reviewed-By: Jens Bache-Wiig
8 years agoComment one of my changes from yesterday
Richard Moe Gustavsen [Wed, 29 Jul 2009 06:18:35 +0000 (08:18 +0200)]
Comment one of my changes from yesterday

Task-number: 258445

8 years agoDisable benchlibtickcounter test by default.
Rohan McGovern [Wed, 29 Jul 2009 03:59:41 +0000 (13:59 +1000)]
Disable benchlibtickcounter test by default.

This test depends on some low-level features of the CPU and the OS
scheduler which seem impractical to automatically verify
independently of the test.  So turn it off to stop its intermittent
failures in our test farm.

8 years agoFixed failure of tst_qmake::simple_app_shadowbuild.
Rohan McGovern [Wed, 29 Jul 2009 03:52:12 +0000 (13:52 +1000)]
Fixed failure of tst_qmake::simple_app_shadowbuild.

Test needs a placeholder file to make sure its build directory exists.

8 years agoFixed failure of tst_moc.
Rohan McGovern [Wed, 29 Jul 2009 02:00:53 +0000 (12:00 +1000)]
Fixed failure of tst_moc.

os9-newlines.h's lack of newline at end of file is significant.

8 years agoSkip uic3 test when qt3support is disabled.
Rohan McGovern [Wed, 29 Jul 2009 01:23:37 +0000 (11:23 +1000)]
Skip uic3 test when qt3support is disabled.

8 years agoFixed compile of tst_qitemeditorfactory.
Rohan McGovern [Tue, 28 Jul 2009 23:49:26 +0000 (09:49 +1000)]
Fixed compile of tst_qitemeditorfactory.

8 years agoFixed failure of tst_q3urloperator.
Rohan McGovern [Tue, 28 Jul 2009 23:44:27 +0000 (09:44 +1000)]
Fixed failure of tst_q3urloperator.

Make test use network-settings.h correctly.

8 years agoFixed failure of tst_q3socket.
Rohan McGovern [Tue, 28 Jul 2009 23:35:03 +0000 (09:35 +1000)]
Fixed failure of tst_q3socket.

Make the test use network-settings.h correctly.

8 years agoQtify void* func => void *func
Anders Bakken [Tue, 28 Jul 2009 21:45:55 +0000 (14:45 -0700)]
Qtify void* func => void *func

Reviewed-by: TrustMe
8 years agoRefactor QDirectFBPaintDevice
Anders Bakken [Tue, 28 Jul 2009 21:30:27 +0000 (14:30 -0700)]
Refactor QDirectFBPaintDevice

Make the engine a member in QDirectFBPaintDevice instead of one in
QDirectFBPixmapData and one in QDirectFBWindowSurface.

Don't use the default screen for the dfbpaintdevices. Pass one
explicitly.

Reviewed-by: Noam Rosenthal <noam.rosenthal@nokia.com>
8 years agoMake autorepeat work for DirectFB
Anders Bakken [Tue, 28 Jul 2009 20:49:03 +0000 (13:49 -0700)]
Make autorepeat work for DirectFB

When holding down a key DirectFB gives us only keypresses.
Qt wants these kind of events:

press not autorepeat
release autorepeat
press autorepeat
release autorepeat
press autorepeat
release not autorepeat

Reviewed-by: Donald <qt-info@nokia.com>
8 years agoFix crash in when resizing directfb windows
Anders Bakken [Tue, 28 Jul 2009 20:13:26 +0000 (13:13 -0700)]
Fix crash in when resizing directfb windows

Since QRasterPaintEngine::baseClip was initialized when the
QDirectFBPaintEngine was created and the paintengine persists for the
window surface we run the risk that the baseClip is of the wrong size.
This could lead to crashes and erroneous clipping.

Reviewed-by: Donald <qt-info@nokia.com>
8 years agoCompile with DirectFB < 1.0
Anders Bakken [Tue, 28 Jul 2009 19:43:12 +0000 (12:43 -0700)]
Compile with DirectFB < 1.0

These porter duff operations were added for 1.0.

Reviewed-by: Donald <qt-info@nokia.com>
8 years agoRegression: QColorDialog crash on Mac when accepting with keyboard
Richard Moe Gustavsen [Tue, 28 Jul 2009 13:54:20 +0000 (15:54 +0200)]
Regression: QColorDialog crash on Mac when accepting with keyboard

Basically, the crash happends because we release the dialog when
the ok button is clicked. But when the button is clicked with the
keyboard (actually cocoa emulates a click when pressing enter), it
also tries to do some focus handling on the button afterwards. But
at that point, the button is already freed up. With a crash as the
result. This patch goes through a lot of the code, and rewrites it
so we doesnt free up the cocoa objects before the destructor of
the QColorDialog is called.

Task-number: 258445
Reviewed-by: msorvig
8 years agoFixed drawImage() so that it doesn't sample outside the source image.
Kim Motoyoshi Kalland [Tue, 28 Jul 2009 07:55:20 +0000 (09:55 +0200)]
Fixed drawImage() so that it doesn't sample outside the source image.

In qt_scale_image_16bit() and qt_scale_image_32bit(), when a sample
point was located on the border between two pixels in the source image,
the sample point was rounded up instead of down. If a sample point was
exactly on the bottom or right edge of the source image, the function
would therefore sample a pixel outside the image. Because of how the
target rectangle is rounded, a sample point will never be exactly on
the top or left edge of the source image, so we will not get a similar
problem there.

I extended the lance test pixmap_scaling.qps.

Task-number: 258533
Reviewed-by: Samuel
8 years agoDon't use layered windows for non-translucent windows
Gunnar Sletta [Tue, 28 Jul 2009 07:41:00 +0000 (09:41 +0200)]
Don't use layered windows for non-translucent windows

Reviewed-by: Samuel
8 years agoRemove ignore system clip stuff in dfbpaintengine
Anders Bakken [Mon, 27 Jul 2009 19:53:44 +0000 (12:53 -0700)]
Remove ignore system clip stuff in dfbpaintengine

Reviewed-by: Donald <qt-info@nokia.com>
8 years agoFix off-by-one bug in QDirectFBScreen::flipSurface
Anders Bakken [Tue, 28 Jul 2009 01:41:39 +0000 (18:41 -0700)]
Fix off-by-one bug in QDirectFBScreen::flipSurface

DFBRegions are exactly like QRects it turns out. x2/y2 ==
QRect::right()/QRect::bottom()

This means I actually can reinterpret_cast QRect to DFBRegions but it's
not really worth it.

Reviewed-by: Noam Rosenthal <noam.rosenthal@nokia.com>
8 years agoMajor clipping patch in QDirectFBPaintEngine
Anders Bakken [Mon, 27 Jul 2009 23:51:51 +0000 (16:51 -0700)]
Major clipping patch in QDirectFBPaintEngine

Handle region clips by iterating over the rectangles and updating clips
accordingly.

Reviewed-by: Noam Rosenthal <noam.rosenthal@nokia.com>
8 years agoMake porterduff operations work in DirectFB
Anders Bakken [Mon, 27 Jul 2009 22:36:26 +0000 (15:36 -0700)]
Make porterduff operations work in DirectFB

This patch makes us not fall back to the raster engine for most of the
composition mode operations when doing blits. Primitives are still
handled by Qt when composition != SourceOver

Reviewed-by: Noam Rosenthal <noam.rosenthal@nokia.com>
8 years agoDisable the pointer tracking feature in Qt 4.5.
Thiago Macieira [Mon, 27 Jul 2009 19:16:34 +0000 (21:16 +0200)]
Disable the pointer tracking feature in Qt 4.5.

The functionality is broken, since pointers can be released by a
QSharedPointer tracking a pointer of different type, which would leave
behind pointers in the hash.

The fix requires Qt 4.6 because of a new symbol being added.

8 years agoRoll back the fancy updating of translucent windows.
Gunnar Sletta [Mon, 27 Jul 2009 14:00:47 +0000 (16:00 +0200)]
Roll back the fancy updating of translucent windows.

The problem with the fix, though it produces less flicker when
resizing, is that it delays telling windows that the window has moved
until after the window has been completely repainted. Problem with
this is that functions that rely on windows to be up to date will fail
until the backbuffer is flushed. This was the case for mapTo/FromGlobal, and potentially other functions too.

Reviewed-By: Eskil
8 years agoFix missing title quoting.
kh [Mon, 27 Jul 2009 13:08:00 +0000 (15:08 +0200)]
Fix missing title quoting.

Reviewed-by: kh
8 years agoAdd docs for Windows Server 2008 R2 to QSysInfo::WinVersion
Bradley T. Hughes [Mon, 27 Jul 2009 11:45:20 +0000 (13:45 +0200)]
Add docs for Windows Server 2008 R2 to QSysInfo::WinVersion

Windows Server 2008 R2 is based on kernel 6.1, the same as Windows 7.

Reviewed-by: TrustMe
8 years agoFix crash when printing to PDF on Qt/Mac
Morten Sørvig [Mon, 27 Jul 2009 10:27:09 +0000 (12:27 +0200)]
Fix crash when printing to PDF on Qt/Mac

Priting to PDF would crash for applications built on Leopard when running
on Tiger.

Add the standard runtime version check to QFontEngineMac::faceId().

Task-number: 251791
Reviewed-by: Richard Moe Gustavsen
8 years agoFixed failure of tst_qlocale if run when LC_ALL is set.
Rohan McGovern [Mon, 27 Jul 2009 04:50:10 +0000 (14:50 +1000)]
Fixed failure of tst_qlocale if run when LC_ALL is set.

This test tried to be unaffected by the locale environment variables
set by the user by explicitly overwriting them and spawning a new
process.  However this was only done for LANG and not the other
variables which can affect the locale.  Do it for all the locale
variables.

Also, the previous approach (store the current value with qgetenv, then
put it back with qputenv later) doesn't really work because that
effectively takes variables which weren't set and sets them to an empty
string, which is not the same thing.  Use QProcess::setEnvironment
instead.

Reviewed-by: Andy Shaw
8 years agoDoc - Mention Google trademarks with the two examples using Google services.
Kavindra Palaraja [Sun, 26 Jul 2009 18:18:02 +0000 (20:18 +0200)]
Doc - Mention Google trademarks with the two examples using Google services.

Reviewed-by: Simon Hausmann <simon.hausmann@nokia.com>
8 years agoFix a running external tests: user program headers must come first.
Thiago Macieira [Sat, 25 Jul 2009 10:05:28 +0000 (12:05 +0200)]
Fix a running external tests: user program headers must come first.

No wonder QT_SHAREDPOINTER_TRACK_POINTERS was having no effect:
there was an #include <QtCore/QtCore> before it.
(cherry picked from commit 4c12010fac555bce0a6c8d69a267a56f4c15087f)

8 years agoMusicplayer example crashes when trying to open unvalid file
Richard Moe Gustavsen [Fri, 24 Jul 2009 13:07:51 +0000 (15:07 +0200)]
Musicplayer example crashes when trying to open unvalid file

Reason: Example tries to get media from QList with an index out of
bounds. Nothing to do with Phonon.

Task-number: 253307

8 years agoPrematurely creating a dialog as a sheet causes problems
Richard Moe Gustavsen [Fri, 24 Jul 2009 10:10:21 +0000 (12:10 +0200)]
Prematurely creating a dialog as a sheet causes problems

Prematurely creating a dialog as a sheet and then calling
exec() on it will show a window w/o decorations. The problem is
that first telling a window to be a sheet, and then tell it to
exec, is unambigious. Because doing the latter implies application
modality (when modality is not set), which again implies not
using a sheet. Calling exec (and setting modality) will win over
window flags, so in this case, we now recreate the window as a
normal app-modal dialog.

Task: 254524
Reviewed-by: Trenton Schulz
8 years agoQNAM: Clarify docs about when a QNetworkReply can be deleted
Markus Goetz [Fri, 24 Jul 2009 08:42:15 +0000 (10:42 +0200)]
QNAM: Clarify docs about when a QNetworkReply can be deleted

Task-number: 258644
Reviewed-by: Thiago
8 years agoUpdated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( eb3afcbf...
Simon Hausmann [Fri, 24 Jul 2009 08:22:11 +0000 (10:22 +0200)]
Updated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( eb3afcbfb4006de4015047555cb256fcde93b954 )

Changes in WebKit since the last update:

++ b/WebCore/ChangeLog
2009-05-27  John Sullivan  <sullivan@apple.com>

        fixed <rdar://problem/6925482> repro crash in WebCore::DragController::dragExited dropping
        bookmarks (at least) over Top Sites (at least)

        Reviewed by Kevin Decker

        * page/DragController.cpp:
        (WebCore::DragController::dragExited):
        nil check m_documentUnderMouse and take the "local file" case if it's nil

8 years agoFixed crash when vectorpath was polygonal only in raster::stroke()
Gunnar Sletta [Fri, 24 Jul 2009 07:35:41 +0000 (09:35 +0200)]
Fixed crash when vectorpath was polygonal only in raster::stroke()

Polygonal vector paths may have types==null, in which case this
would have crashed.

Reviewed-by: Eskil
8 years agoUpdated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( 1535d41a...
Simon Hausmann [Thu, 23 Jul 2009 14:14:01 +0000 (16:14 +0200)]
Updated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( 1535d41a668e5f74f44ff3aa1313a84d5718d2d7 )

Changes in WebKit since the last update:

++ b/WebCore/ChangeLog
2009-07-23  Simon Hausmann  <simon.hausmann@nokia.com>

        Reviewed by Holger Freyther.

        Fix crashes with the QObject bindings after garbage collection.

        There is one QtInstance per wrapped QObject, and that QtInstance keeps
        references to cached JSObjects for slots. When those objects get
        deleted due to GC, then they becoming dangling pointers.

        When a cached member dies, it is now removed from the QtInstance's
        cache.

        As we cannot track the lifetime of the children, we have to remove
        them from QtInstance alltogether. They are not cached and were
        only used for mark(), but we _want_ them to be subject to gc.

        * bridge/qt/qt_instance.cpp:
        (JSC::Bindings::QtInstance::~QtInstance): Minor coding style cleanup,
        use qDeleteAll().
        (JSC::Bindings::QtInstance::removeCachedMethod): New function, to
        clean m_methods and m_defaultMethod.
        (JSC::Bindings::QtInstance::mark): Avoid marking already marked objects.
        (JSC::Bindings::QtField::valueFromInstance): Don't save children for
        marking.
        * bridge/qt/qt_instance.h: Declare removeCachedMethod.
        * bridge/qt/qt_runtime.cpp:
        (JSC::Bindings::QtRuntimeMethod::~QtRuntimeMethod): Call removeCachedMethod
        with this on the instance.

2009-05-04  Jakub Wieczorek  <faw217@gmail.com>

        Reviewed by Simon Hausmann.

        As Qtish implementation of MIMETypeRegistry::getMIMETypeForExtension()
        returns the application/octet-stream mimetype when it can't associate
        extension with any mimetype, it can happen that the application/octet-stream
        mimetype will hit the list of supported image formats. For instance,
        it is possible when QImageReader or QImageWriter support an extension
        that is not in the extensions map.

        Make sure that this mimetype is not treated as displayable image type.

        * platform/MIMETypeRegistry.cpp:
        (WebCore::initializeSupportedImageMIMETypes):
        (WebCore::initializeSupportedImageMIMETypesForEncoding):

++ b/WebKit/qt/ChangeLog
2009-07-23  Simon Hausmann  <simon.hausmann@nokia.com>

        Reviewed by Holger Freyther.

        Added a testcase to verify that cached methods in the QOBject bindings
        remain alife even after garbage collection.

        * tests/qwebpage/tst_qwebpage.cpp:
        (tst_QWebPage::protectBindingsRuntimeObjectsFromCollector):

8 years agoQPainter::stroke() on raster engine would draw moveto's as lines
Gunnar Sletta [Thu, 23 Jul 2009 13:26:11 +0000 (15:26 +0200)]
QPainter::stroke() on raster engine would draw moveto's as lines

The reason being that there was an assumption that any non-curved path
was a continous polyline. For paths with multiple subpaths in it
we need to split this up into multiple strokePolygonCosmetic calls.

Task-number: 257621
Reviewed-by: Kim Motoyoshi Kalland
8 years agoDiagonal dashes are moving when touching the clip boundary.
Gunnar Sletta [Thu, 23 Jul 2009 12:30:57 +0000 (14:30 +0200)]
Diagonal dashes are moving when touching the clip boundary.

We normally pad the clip rect with the size of the pen and miterlimit
to avoid this, but this didn't handle the case where there was a long
diagonal dash. We also need to multiply the padding with the longest
dash.

Reviewed-By: Tom Cooksey