qt:kde-qt.git
8 years agoFix for qfileinfo autotest: LocalDiskFile set also for non-existing file v4.6.0-tp1
aavit [Mon, 7 Sep 2009 09:14:12 +0000 (11:14 +0200)]
Fix for qfileinfo autotest: LocalDiskFile set also for non-existing file

Change 53576b4d3c3e7325d01efba6c4da80299492f2db introduced the
behaviour that QFSFileEngine sets LocalDiskFlag regardless of whether
the file exists or not, but it just did it for Windows. This change
makes fsengine for unix/mac behave likewise.

Reviewed-by: trustme
8 years agoFixed qsubmdiarea setFont test.
Gabriel de Dietrich [Mon, 7 Sep 2009 09:30:36 +0000 (11:30 +0200)]
Fixed qsubmdiarea setFont test.

The compared fonts didn't have the same styleHint because it's resolved
setting the font.

Reviewed-by: Olivier
8 years agoFix tst_QItemDelegate::task257859_finalizeEdit on Mac
Olivier Goffart [Mon, 7 Sep 2009 09:29:24 +0000 (11:29 +0200)]
Fix tst_QItemDelegate::task257859_finalizeEdit on Mac

qFindChildren<QWidget*> would also find the QFocusFrame in addition to
the line edit

8 years agoFixed resolving colors of the form "rgb(r,g,b)" in SVGs.
Kim Motoyoshi Kalland [Mon, 7 Sep 2009 09:16:39 +0000 (11:16 +0200)]
Fixed resolving colors of the form "rgb(r,g,b)" in SVGs.

The bug was introduced by 13bcc92274d52fa6df2d636c78cf6ea457d670aa.
Instead of comparing only the beginning of a string with "rgb(", a full
string compare was used. I also added some error handling to avoid
crashing on noncompliant SVG files.

Reviewed-by: Trond
8 years agoSupport setting font size in shorthand font properties
Eskil Abrahamsen Blomfeldt [Mon, 7 Sep 2009 08:54:35 +0000 (10:54 +0200)]
Support setting font size in shorthand font properties

The fontSizeAdjustment defaulted to 0 which means "medium" in internal
semantics. This will override any font size you set in the short-hand.
In other locations, fontSizeAdjustment defaults to -255 which has no
meaning attached. To allow setting the font size in short-hand (as in
"font: 20px Arial"), we can't default to a specific adjustment.

Two tests are added: The first verifies the case that already worked,
where you specify the font size using the "font-size" property. The
other verifies the short-hand case and would previously fail.

Task-number: 207189
Reviewed-by: Simon Hausmann
8 years agoStabilize QComboBox test.
Olivier Goffart [Mon, 7 Sep 2009 08:44:58 +0000 (10:44 +0200)]
Stabilize QComboBox test.

On Mac, there is long fading animation before the popups hide

8 years agoStabilize QFontMetrics test
Olivier Goffart [Mon, 7 Sep 2009 08:37:01 +0000 (10:37 +0200)]
Stabilize QFontMetrics test

On Mac, the text may be elided if we give the exact length as the size.
Giving one pixel bigger makes sure the text is not elided

8 years agomake JavaScriptCore compile on HPUX
Kent Hansen [Fri, 4 Sep 2009 16:01:43 +0000 (18:01 +0200)]
make JavaScriptCore compile on HPUX

Get rid of circular dependency (don't include StructureChain.h from Structure.h).
Disable some time stuff that isn't available on HPUX.

Reviewed-by: Simon Hausmann
8 years agoStabilize QTreeView test
Olivier Goffart [Mon, 7 Sep 2009 07:58:01 +0000 (09:58 +0200)]
Stabilize QTreeView test

8 years agoFixed compilation error for qmdiarea and qmdisubwindow test.
Gabriel de Dietrich [Mon, 7 Sep 2009 08:05:14 +0000 (10:05 +0200)]
Fixed compilation error for qmdiarea and qmdisubwindow test.

Reviewed-by: Olivier
8 years agoNew function added to QScriptEngineAgent autotest
Jedrzej Nowacki [Thu, 3 Sep 2009 15:21:27 +0000 (17:21 +0200)]
New function added to QScriptEngineAgent autotest

Function convert numeric ScriptEngineEvent::Type to QString. It is not
used in test but really useful in debugging

Reviewed-by: Kent Hansen
8 years agoFew expected fails were fixed in QScriptEngineAgent
Jedrzej Nowacki [Thu, 3 Sep 2009 15:12:15 +0000 (17:12 +0200)]
Few expected fails were fixed in QScriptEngineAgent

functionExit event was partially fixed. The time point in JS execution
with JIT enabled works now but still there is no returning value in few
cases.
Autotest was corrected.

Reviewed-by: Kent Hansen
8 years agoFix tiled blit in 16-bit mode
Paul Olav Tvete [Mon, 7 Sep 2009 07:17:55 +0000 (09:17 +0200)]
Fix tiled blit in 16-bit mode

Optimizations in change 8e447e8a did not handle the case when the target
width is less than the width of a tile.

Task-number: 260759
Reviewed-by: Samuel
8 years agoMake dynamic cube maps work again in boxes
Rhys Weatherley [Mon, 7 Sep 2009 07:18:12 +0000 (17:18 +1000)]
Make dynamic cube maps work again in boxes

Reviewed-by: trustme
8 years agoMerge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6
Rhys Weatherley [Mon, 7 Sep 2009 04:25:33 +0000 (14:25 +1000)]
Merge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6

8 years agoqdoc: improve the QGLFormat documentation
Rhys Weatherley [Mon, 7 Sep 2009 04:21:59 +0000 (14:21 +1000)]
qdoc: improve the QGLFormat documentation

Reviewed-by: trustme
8 years agoMark these as fix later.
Bill King [Mon, 7 Sep 2009 03:52:13 +0000 (13:52 +1000)]
Mark these as fix later.

The unicode logic is beyond me, leave these for fixing by someone
else with more unicode knowledge.

8 years agoDB2 - Don't return an error when the table is just empty.
Bill King [Mon, 7 Sep 2009 03:49:36 +0000 (13:49 +1000)]
DB2 - Don't return an error when the table is just empty.

Makes the behaviour consistent with the ODBC driver behaviour

Reviewed-by: Justin McPherson
8 years agoSample buffers are on by default for OpenGL/ES 2.0, off for others.
Rhys Weatherley [Mon, 7 Sep 2009 02:42:03 +0000 (12:42 +1000)]
Sample buffers are on by default for OpenGL/ES 2.0, off for others.

Reviewed-by: trustme
8 years agoUnit tests for QGLFormat option constructor
Rhys Weatherley [Mon, 7 Sep 2009 01:56:58 +0000 (11:56 +1000)]
Unit tests for QGLFormat option constructor

Reviewed-by: trustme
8 years agoUnit tests for QGLFormat::defaultFormat and ::defaultOverlayFormat()
Rhys Weatherley [Mon, 7 Sep 2009 01:42:26 +0000 (11:42 +1000)]
Unit tests for QGLFormat::defaultFormat and ::defaultOverlayFormat()

Reviewed-by: trustme
8 years agoMerge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6
Rhys Weatherley [Mon, 7 Sep 2009 00:54:23 +0000 (10:54 +1000)]
Merge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6

8 years agoUnit tests for QGLFramebufferObjectFormat
Rhys Weatherley [Mon, 7 Sep 2009 00:53:32 +0000 (10:53 +1000)]
Unit tests for QGLFramebufferObjectFormat

Reviewed-by: trustme
8 years agoMerge branch '4.5' into 4.6
Rohan McGovern [Mon, 7 Sep 2009 00:44:44 +0000 (10:44 +1000)]
Merge branch '4.5' into 4.6

8 years agoFixed compile on Solaris.
Rohan McGovern [Mon, 7 Sep 2009 00:37:42 +0000 (10:37 +1000)]
Fixed compile on Solaris.

The `test' builtin in Solaris' /bin/sh does not understand `-e', use
`-f' instead.

Fixes: configure: test: argument expected

Note, using the `-nokia-developer' configure option hides this
breakage.

Reviewed-by: Rhys Weatherley
8 years agoPerformance: Convert QGLFormat to use implicit sharing
Rhys Weatherley [Mon, 7 Sep 2009 00:18:46 +0000 (10:18 +1000)]
Performance: Convert QGLFormat to use implicit sharing

QGLFormat was being deep-copied many times per frame because of
code like this:

    if (context()->format().doubleBuffer()) { ...

This change modifies QGLFormat to use implicit sharing to reduce
the overhead of the above type of checks.

Reviewed-by: Sarah Smith
8 years agoUnit tests for QGLFormat copy constructor and operator=
Rhys Weatherley [Sun, 6 Sep 2009 23:30:20 +0000 (09:30 +1000)]
Unit tests for QGLFormat copy constructor and operator=

Reviewed-by: trustme
8 years agoUpdate QGLFormat::operator== to include all fields.
Rhys Weatherley [Sun, 6 Sep 2009 23:17:08 +0000 (09:17 +1000)]
Update QGLFormat::operator== to include all fields.

The documentation says "Returns true if all the options of the
two QGLFormats are equal", but that's not what it was doing.

Reviewed-by: Sarah Smith
8 years agoCode cleanup: remove friend declarations for non-existent functions
Rhys Weatherley [Sun, 6 Sep 2009 22:42:10 +0000 (08:42 +1000)]
Code cleanup: remove friend declarations for non-existent functions

Reviewed-by: Alex
8 years agoMake QGraphicsShaderEffect private API for 4.6
Rhys Weatherley [Sun, 6 Sep 2009 22:25:01 +0000 (08:25 +1000)]
Make QGraphicsShaderEffect private API for 4.6

The custom shader code in the OpenGL2 paint engine needs time to
mature before we make this official public API.

Reviewed-by: trustme
8 years agoFixed "multiple definition of `QTest::mouseActionNames'" error in
Rohan McGovern [Sat, 5 Sep 2009 21:46:20 +0000 (07:46 +1000)]
Fixed "multiple definition of `QTest::mouseActionNames'" error in
projects where two separate compilation units include `qtestmouse.h'.

8 years agoFixed compile of code which defines QT_NO_CAST_FROM_ASCII and includes
Rohan McGovern [Sat, 5 Sep 2009 11:36:19 +0000 (21:36 +1000)]
Fixed compile of code which defines QT_NO_CAST_FROM_ASCII and includes
qtestmouse.h.

8 years agoMerge branch '4.5' into 4.6
Rohan McGovern [Sat, 5 Sep 2009 01:26:46 +0000 (11:26 +1000)]
Merge branch '4.5' into 4.6

Conflicts:
src/gui/itemviews/qlistview.cpp

8 years agoRemove license header, which cannot be parsed by Sun assembler.
Rohan McGovern [Sat, 5 Sep 2009 01:17:32 +0000 (11:17 +1000)]
Remove license header, which cannot be parsed by Sun assembler.

Commit 2e0d78836becf24c7f27c982316cf1b4492f27aa fixed this for i386
but omitted the fix for x86_64.

8 years agoFixed compile for S60.
Rohan McGovern [Sat, 5 Sep 2009 00:12:54 +0000 (10:12 +1000)]
Fixed compile for S60.

Q_DECLARE_METATYPE must appear before qRegisterMetaType.

8 years agoUse the right type in DFBPaintEngine
Anders Bakken [Fri, 4 Sep 2009 23:11:59 +0000 (16:11 -0700)]
Use the right type in DFBPaintEngine

Reviewed-by: TrustMe
8 years agoOnly export internal raster classes in QWS
Anders Bakken [Fri, 4 Sep 2009 17:05:58 +0000 (10:05 -0700)]
Only export internal raster classes in QWS

QRasterizer, QClipData and QRasterBuffer only need to exported when
QRasterPaintEnginePrivate is subclassed which only is the case for
DirectFB.

Reviewed-by: Noam Rosenthal <noam.rosenthal@nokia.com>
8 years agoDoc: Added info on Q_FLAGS() by providing an example with more context.
David Boddie [Fri, 4 Sep 2009 18:46:10 +0000 (20:46 +0200)]
Doc: Added info on Q_FLAGS() by providing an example with more context.

Reviewed-by: Trust Me
As-seen-on: qt-interest

8 years agoDoc: Clarified that certain KDE widgets are KDE 3 widgets.
David Boddie [Fri, 4 Sep 2009 18:43:44 +0000 (20:43 +0200)]
Doc: Clarified that certain KDE widgets are KDE 3 widgets.

Task-number: 258991
Reviewed-by: Trust Me
8 years agoRemove unnecessary call to exposeRegion in DFB
Anders Bakken [Fri, 4 Sep 2009 16:42:14 +0000 (09:42 -0700)]
Remove unnecessary call to exposeRegion in DFB

If we make sure to erase the background to background color in connect()
we don't have to do it in setGeometry.

Also clean up the code in QDBWindowSurface::flush

Reviewed-by: Donald Carr <donald.carr@nokia.com>
8 years agoQDFBScreen bgcolor is only use in non_wm mode
Anders Bakken [Fri, 4 Sep 2009 16:35:07 +0000 (09:35 -0700)]
QDFBScreen bgcolor is only use in non_wm mode

No need to carry those extra bytes around when QT_DIRECTFB_WM is
defined.

Reviewed-by: Donald Carr <donald.carr@nokia.com>
8 years agoRemove unused function in QDFBScreen
Anders Bakken [Fri, 4 Sep 2009 16:32:48 +0000 (09:32 -0700)]
Remove unused function in QDFBScreen

Reviewed-by: Donald Carr <donald.carr@nokia.com>
8 years agoHandle proxywidgets better in DFB
Anders Bakken [Thu, 3 Sep 2009 21:10:25 +0000 (14:10 -0700)]
Handle proxywidgets better in DFB

This code is roughly the same as in other screen drivers. We shouldn't
flush widgets that are in a graphicsscene as a proxy widget. They will
be handled by the graphicsview.

Reviewed-by: Donald Carr <donald.carr@nokia.com>
8 years agoMerge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into qt46
Espen Riskedal [Fri, 4 Sep 2009 15:42:14 +0000 (17:42 +0200)]
Merge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into qt46

8 years agoAdded Symbian to our list of Tier1 platforms
Espen Riskedal [Fri, 4 Sep 2009 15:12:34 +0000 (17:12 +0200)]
Added Symbian to our list of Tier1 platforms

At the moment we only support S60 3.1, 3.2 and 5.0. Once Symbian
Platforms actually gets publicly released we'll add those too.

Reviewed-by: Volker Hilsheimer
Reviewed-by: Jason Barron
8 years agoFixed object rotations in the boxes demo.
Kim Motoyoshi Kalland [Fri, 4 Sep 2009 15:00:06 +0000 (17:00 +0200)]
Fixed object rotations in the boxes demo.

When Math3D was introduced, the boxes demo was changed to use the
classes in the new module, but the change caused some bugs with the
object rotations. This commit should fix them.

Reviewed-by: Samuel
8 years agoMake the example also work on a desktop with a dark theme.
Thomas Zander [Wed, 2 Sep 2009 20:04:22 +0000 (22:04 +0200)]
Make the example also work on a desktop with a dark theme.

8 years agoUpdating line-numbers for linguist auto tests
João Abecasis [Fri, 4 Sep 2009 11:03:26 +0000 (13:03 +0200)]
Updating line-numbers for linguist auto tests

8 years agoMore fixes for qatomic_windows.h
João Abecasis [Thu, 3 Sep 2009 12:40:10 +0000 (14:40 +0200)]
More fixes for qatomic_windows.h

If platform headers have been included the Interlocked names may be
macros and we end up with double underscores in our function names. By
prefixing "Interlocked" and then the actual function names we work
around that.

Let's hope nobody redefines Interlocked or Increment and friends...  :-)

Reviewed-by: Prasanth Ullattil
8 years agoFixed tst_QFiledialog::filesSelectedSignal where the selected directory to run
Gabriel de Dietrich [Fri, 4 Sep 2009 14:01:41 +0000 (16:01 +0200)]
Fixed tst_QFiledialog::filesSelectedSignal where the selected directory to run
the test may not have any file on it.

Reviewed-by: Olivier
8 years agoImplement symbol hiding for JSC's JIT functions.
Thiago Macieira [Fri, 4 Sep 2009 13:38:35 +0000 (15:38 +0200)]
Implement symbol hiding for JSC's JIT functions.

These functions are implemented directly in assembly, so they need the
proper directives to enable/disable visibility. On ELF systems, it's
.hidden, whereas on Mach-O systems (Mac) it's .private_extern. On
Windows, it's not necessary since you have to explicitly export. I
also implemented the AIX idiom, though it's unlikely anyone will
implement AIX/POWER JIT.

That leaves only HP-UX on PA-RISC unimplemented, from the platforms
that Qt supports. It's also unlikely that we'll imlpement JIT for it.

Reviewed-by: Kent Hansen
8 years agoAdded multiple frames to QPaintBuffer.
Samuel Rødal [Fri, 4 Sep 2009 13:08:44 +0000 (15:08 +0200)]
Added multiple frames to QPaintBuffer.

This lets us stream a single QPaintBuffer instead of one QPaintBuffer
per frame in the trace graphicssystem, which leads to not streaming
pixmaps / images once per frame. Performance when doing a trace is also
a lot better for painting heavy applications.

Reviewed-by: Trond
8 years agoAvoided streaming multiple copies of images/pixmaps in QPaintBuffer.
Samuel Rødal [Fri, 4 Sep 2009 11:20:04 +0000 (13:20 +0200)]
Avoided streaming multiple copies of images/pixmaps in QPaintBuffer.

This reduces the size of the trace files significantly in certain cases.

Reviewed-by: Trond
8 years agoFix tst_QTimer::moveToThread() on Windows
Bradley T. Hughes [Thu, 3 Sep 2009 06:17:05 +0000 (08:17 +0200)]
Fix tst_QTimer::moveToThread() on Windows

We shouldn't fully unregister timers when the event dispatcher is stopped
when a thread exits, since this releases the timerId back to the pool.
Instead, only free the OS resources.

Auto-test included.

Reviewed-by: ogoffart
8 years agoOptimize qt_format_text test operations: try not to detach
Olivier Goffart [Fri, 4 Sep 2009 12:00:01 +0000 (14:00 +0200)]
Optimize qt_format_text test operations: try not to detach

Second try.
This is a fixed version of the one reverted in commit 25014061b8de

The difference is that it does
text[offset] = QChar::LineSeparator;
instead of
chr = QChar::LineSeparator;

Reviewed-by: Eskil
8 years agoFix crashes when moving descendents during destruction.
Andreas Aardal Hanssen [Fri, 4 Sep 2009 09:56:43 +0000 (11:56 +0200)]
Fix crashes when moving descendents during destruction.

This fix prevents crashes in
QGraphicsItem caused by calls to pure virtual functions as an item
tries to access its dying parent. This happens when an item is moved
by an ancestor during this ancestor's destructor.

Autotests included.

Reviewed-by: bnilsen
8 years agoDon't call pure virtuals on deleting QGraphicsItem
Aaron Kennedy [Fri, 4 Sep 2009 03:35:25 +0000 (13:35 +1000)]
Don't call pure virtuals on deleting QGraphicsItem

This also needs to be reviewed by andreas.

Reviewed-by: Michael Brasser
8 years agoMake the view in the flowlayout example big enough.
Jan-Arve Sæther [Fri, 4 Sep 2009 13:19:02 +0000 (15:19 +0200)]
Make the view in the flowlayout example big enough.

8 years agoMake sure that top-level windows respect hfw/wfh constraint.
Jan-Arve Sæther [Fri, 4 Sep 2009 10:37:20 +0000 (12:37 +0200)]
Make sure that top-level windows respect hfw/wfh constraint.

Note that the constraint is only enforced when resizing the window
interactively. Calling setGeometry() will not try to enforce the
constraint.

See the graphicsview/flowlayout example for a "manual test".

Task-number:  257455

8 years agoQGraphicsWidget::sizeHint didn't pass the correct constraint to layout.
Jan-Arve Sæther [Fri, 4 Sep 2009 10:30:02 +0000 (12:30 +0200)]
QGraphicsWidget::sizeHint didn't pass the correct constraint to layout.

Adjust the constraint with the widget margin before passing it on to
d->layout->effectiveSizeHint().

8 years agoMinor fixes to the graphicsview/flowlayout example
Jan-Arve Sæther [Fri, 4 Sep 2009 09:42:52 +0000 (11:42 +0200)]
Minor fixes to the graphicsview/flowlayout example

8 years agoMerge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6
David Boddie [Fri, 4 Sep 2009 13:12:40 +0000 (15:12 +0200)]
Merge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6

8 years agoDoc: Fixed a code snippet demonstrating setStandardOutputProcess().
David Boddie [Fri, 4 Sep 2009 13:11:34 +0000 (15:11 +0200)]
Doc: Fixed a code snippet demonstrating setStandardOutputProcess().

Task-number: 260654
Reviewed-by: Trust Me
8 years agoRemove implicit margin on bulleted lists when exporting to HTML
Eskil Abrahamsen Blomfeldt [Fri, 4 Sep 2009 13:05:46 +0000 (15:05 +0200)]
Remove implicit margin on bulleted lists when exporting to HTML

When we export a bulleted list to HTML, we add a meta-block around the
list items to set the bullet style of the list (<ul> or <ol>.) This
block will implicitly have a margin, even if the list itself does not
in the internal document. The result is that exporting a list to HTML
and then reloading it will give a different apperance, because the 0
margin of the original was not saved. The fix is to always set the
meta-block's margin to 0, since it is only there to set the bullet
style and should not affect the layout at all.

Task-number: 201228
Reviewed-by: Simon Hausmann
8 years agoMake QGLFramebufferObject::isBound() check it is the bound fbo
Tom Cooksey [Thu, 3 Sep 2009 14:24:17 +0000 (16:24 +0200)]
Make QGLFramebufferObject::isBound() check it is the bound fbo

Previously, it was possible that painting using QPainter to a different
GL target (E.g. QGLWidget) would unbind the FBO. In such cases, isBound
returned true, even though the FBO wasn't bound to any context.

Reviewed-by: Samuel
8 years agoFix pixel snow on the combo box popup on Snow Leopard.
Morten Sorvig [Fri, 4 Sep 2009 12:22:17 +0000 (14:22 +0200)]
Fix pixel snow on the combo box popup on Snow Leopard.

On 10.5 and below the pixmap data bytes happened to be initialized by
malloc, on 10.6 this is no longer the case.

Revby: Gunnar Sletta <gunnar@trolltech.com>

Similar to 28f94e1ef94f.

8 years agoFix pixel snow on the combo box popup on Snow Leopard.
Morten Sorvig [Fri, 4 Sep 2009 12:22:17 +0000 (14:22 +0200)]
Fix pixel snow on the combo box popup on Snow Leopard.

On 10.5 and below the pixmap data bytes happened to be initialized by
malloc, on 10.6 this is no longer the case.

Revby: Gunnar Sletta <gunnar@trolltech.com>

Similar to 28f94e1ef94f.

8 years agoWizard background images incorrect in Snow Leopard.
Prasanth Ullattil [Mon, 3 Aug 2009 13:46:04 +0000 (15:46 +0200)]
Wizard background images incorrect in Snow Leopard.

We need to clear the QPixmap before the image is drawn using the
CGContextDrawImage().

Reviewed-by: Norwegian Rock Cat
8 years ago%SystemDrive% folder created in current directory on Windows 7.
Prasanth Ullattil [Fri, 4 Sep 2009 11:25:08 +0000 (13:25 +0200)]
%SystemDrive% folder created in current directory on Windows 7.

This happens only when the standarddialogs demo is run from qtdemo. On
Windows 7, calling the GetOpenFileName() caches something in
%SystemDrive%\ProgramData\Microsoft\Windows\Caches. qtdemo was removing
all environment variables other than PATH for the child processes. This
caused the example to dump the cache in current directory instead.

Task-number: 260898
Reviewed-by: Jens Bache-Wiig
8 years agoMerge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6
Gunnar Sletta [Fri, 4 Sep 2009 11:35:17 +0000 (13:35 +0200)]
Merge branch '4.6' of git@scm.dev.nokia.troll.no:qt/qt into 4.6

8 years agoMade eliding of arabic text work without chaning the look of the glyphs
Gunnar Sletta [Fri, 4 Sep 2009 11:34:01 +0000 (13:34 +0200)]
Made eliding of arabic text work without chaning the look of the glyphs
Done by Lars

8 years agoSlightly improved QGLFramebufferObjectFormat API.
Samuel Rødal [Fri, 4 Sep 2009 11:20:52 +0000 (13:20 +0200)]
Slightly improved QGLFramebufferObjectFormat API.

Renaming setInternalFormat() to setInternalTextureFormat() (and
similarly for the accessor) makes the API a bit more explicit.

Reviewed-by: Trond
8 years agoAdded support for quering a printer for the actual number of copies.
Gunnar Sletta [Fri, 4 Sep 2009 10:13:12 +0000 (12:13 +0200)]
Added support for quering a printer for the actual number of copies.

The logic of numCopies() returning 1 to indicate how many times you
have to run your printing code is not ideal and prevents the
"property" to be used as a placeholder for the value, which was
required by KDE. Added QPrinter::actualNumCopies() to ask
for the value set through setNumCopies() or through the print dialog.

Reviewed-by: Trond
8 years agoQAbstractSocket: Missing initializations
Markus Goetz [Thu, 3 Sep 2009 13:53:15 +0000 (15:53 +0200)]
QAbstractSocket: Missing initializations

8 years agoQHostInfoPrivate: Uninitialized int
Markus Goetz [Thu, 3 Sep 2009 13:45:53 +0000 (15:45 +0200)]
QHostInfoPrivate: Uninitialized int

8 years agoQNetworkAccessManager: Missing break in proxyAuthenticationKey
Markus Goetz [Thu, 3 Sep 2009 13:40:20 +0000 (15:40 +0200)]
QNetworkAccessManager: Missing break in proxyAuthenticationKey

This probably made the proxy authentication saving for ftp proxies
not working.

Reviewed-by: Thiago
8 years agoStylesheet: Do not crop the text in tabs
Olivier Goffart [Fri, 4 Sep 2009 08:44:30 +0000 (10:44 +0200)]
Stylesheet:  Do not crop the text in tabs

Because we did not reiimplement SE_TabBarTabText to force the
parentstyle to call the virtual function back, we got the normal style
margins in addition to the one by the stylesheet.

Also added test for !hasNativeBorder to suport cases where we just
specify a border

Reviewed-by: jbache
Task-number: 196326

8 years agoFix the QStyleSheet test with pulse.
Olivier Goffart [Fri, 4 Sep 2009 07:09:54 +0000 (09:09 +0200)]
Fix the QStyleSheet test with pulse.

The windows manager was moving the window behind our back,  making the
test fails sometimes

Reviewed-by: Thierry
8 years agoFix ./tst_qdatetimeedit::task118867
Olivier Goffart [Thu, 3 Sep 2009 13:16:21 +0000 (15:16 +0200)]
Fix ./tst_qdatetimeedit::task118867

Before the QLineEdit change, the QLineEdit::displayText used to return the
currently entered text when we call validate()  on the validator.

Now, the displayText is set _after_ the validation.

Hence, the QDateTimeEdit::validate did not had access to the real text.

Reviewed-by: Thierry
8 years agodefine MAP_ANON as MAP_ANONYMOUS on HPUX
Kent Hansen [Fri, 4 Sep 2009 09:45:51 +0000 (11:45 +0200)]
define MAP_ANON as MAP_ANONYMOUS on HPUX

Reviewed-by: Simon Hausmann
8 years agoProspective SunCC compile fix
Simon Hausmann [Fri, 4 Sep 2009 06:57:37 +0000 (08:57 +0200)]
Prospective SunCC compile fix

Call the QScriptValue constructor without class scope.

Reviewed-by: Trust me
8 years agoAdd a way of getting the stack base on HP-UX.
Thiago Macieira [Wed, 12 Aug 2009 18:28:22 +0000 (20:28 +0200)]
Add a way of getting the stack base on HP-UX.

Unfortunately, HP-UX's pthread doesn't allow us to get the stack
address of a running thread. We have to suspend the thread. And
obviously we can't suspend ourselves.

The solution is to start another thread, which suspends us, gets the
stack address, and resumes.

8 years agoAdd PLATFORM(HPUX) and COMPILER(ACC) defines, for HP-UX support.
Thiago Macieira [Thu, 3 Sep 2009 12:27:31 +0000 (14:27 +0200)]
Add PLATFORM(HPUX) and COMPILER(ACC) defines, for HP-UX support.

Also add PLATFORM(IA64) and PLATFORM(HPPA), though we don't use them
and probably don't need. Just in case.

Conflicts:

src/3rdparty/webkit/JavaScriptCore/wtf/Platform.h

8 years agoFix compilation on AIX: struct tm has no timezone information.
Simon Hausmann [Thu, 3 Sep 2009 12:24:51 +0000 (14:24 +0200)]
Fix compilation on AIX: struct tm has no timezone information.

"/pulse/qt/src/3rdparty/webkit/JavaScriptCore/wtf/DateMath.h", line 113.42: 1540-0217 (S) "tm_gmtoff" is not a member of "struct tm".
"/pulse/qt/src/3rdparty/webkit/JavaScriptCore/wtf/DateMath.h", line 115.37: 1540-0217 (S) "tm_zone" is not member of "struct tm".

Conflicts:

src/3rdparty/webkit/JavaScriptCore/wtf/DateMath.cpp
src/3rdparty/webkit/JavaScriptCore/wtf/DateMath.h

8 years agoImplement support for getting the stack base on AIX.
Thiago Macieira [Wed, 29 Jul 2009 15:23:44 +0000 (17:23 +0200)]
Implement support for getting the stack base on AIX.

8 years agoFix compilation with xlC 7: operator-> must return a pointer-like type.
Thiago Macieira [Sun, 19 Jul 2009 14:40:01 +0000 (16:40 +0200)]
Fix compilation with xlC 7: operator-> must return a pointer-like type.

"webkit/JavaScriptCore/runtime/Protect.h", line 101.17: 1540-1196 (S) The return type cannot be "JSC::JSValue" because "class JSValue" does not have an "operator->" function.

8 years agoWork around an apparent xlC 7 bug: partial specialisation on HashMap.
Thiago Macieira [Sun, 19 Jul 2009 14:34:50 +0000 (16:34 +0200)]
Work around an apparent xlC 7 bug: partial specialisation on HashMap.

This problem was appearing only in the functions in the HashMap
partial specialisation for Ref<T> that returned std::pair<iterator,
bool>.

"/pulse/qt/src/3rdparty/webkit/JavaScriptCore/wtf/RefPtrHashMap.h", line 205.5: 1540-1174 (S) The member "template std::pair<HashMap<RefPtr<T>,U,V,W,X>::iterator,bool> inlineAdd(const KeyType &, const MappedType &)" is not declared as a template in its containing class definition.

8 years agoFix compilation with xlC 7: apparently problem with nested templates.
Thiago Macieira [Sun, 19 Jul 2009 13:55:54 +0000 (15:55 +0200)]
Fix compilation with xlC 7: apparently problem with nested templates.

1540-0062 (S) The incomplete class "IsConvertibleToDouble<0, char>" must not be used as a qualifier.

8 years agoFix compilation with Sun CC 5.9: std::pair does not call types' constructors.
Thiago Macieira [Sun, 19 Jul 2009 13:08:01 +0000 (15:08 +0200)]
Fix compilation with Sun CC 5.9: std::pair does not call types' constructors.

Error: Could not find a match for WTF::HashMap<std::pair<WTF::RefPtr<JSC::UString::Rep>, unsigned>, JSC::Structure*, JSC::StructureTransitionTableHash, JSC::StructureTransitionTableHashTraits, WTF::HashTraits<JSC::Structure*>>::remove(std::pair<JSC::UString::Rep*, unsigned>) needed in JSC::Structure::~Structure().

8 years agoFix compilation with Sun CC 5.9: the default constructor for HashTableIteratorAdapter...
Thiago Macieira [Sun, 19 Jul 2009 12:59:48 +0000 (14:59 +0200)]
Fix compilation with Sun CC 5.9: the default constructor for HashTableIteratorAdapter isn't called.

"../JavaScriptCore/wtf/RefPtrHashMap.h", line 208: Error: Cannot use std::pair<WTF::HashTableIterator<WTF::RefPtr<JSC::UString::Rep>, std::pair<WTF::RefPtr<JSC::UString::Rep>, StaticValueEntry*>, WTF::PairFirstExtractor<std::pair<WTF::RefPtr<JSC::UString::Rep>, StaticValueEntry*>>, WTF::StrHash<WTF::RefPtr<JSC::UString::Rep>>, WTF::PairHashTraits<WTF::HashTraits<WTF::RefPtr<JSC::UString::Rep>>, WTF::HashTraits<StaticValueEntry*>>, WTF::HashTraits<WTF::RefPtr<JSC::UString::Rep>>>, bool> to initialize std::pair<WTF::HashTableIteratorAdapter<WTF::HashTable<WTF::RefPtr<JSC::UString::Rep>, std::pair<WTF::RefPtr<JSC::UString::Rep>, StaticValueEntry*>, WTF::PairFirstExtractor<std::pair<WTF::RefPtr<JSC::UString::Rep>, StaticValueEntry*>>, WTF::StrHash<WTF::RefPtr<JSC::UString::Rep>>, WTF::PairHashTraits<WTF::HashTraits<WTF::RefPtr<JSC::UString::Rep>>, WTF::HashTraits<StaticValueEntry*>>, WTF::HashTraits<WTF::RefPtr<JSC::UString::Rep>>>, std::pair<WTF::RefPtr<JSC::UString::Rep>, StaticValueEntry*>>, bool>.

8 years agoFix linking with Sun CC 5.9: function pointers for extern "C" are treated differently
Thiago Macieira [Thu, 23 Jul 2009 09:26:22 +0000 (11:26 +0200)]
Fix linking with Sun CC 5.9: function pointers for extern "C" are treated differently

The Sun CC compiler treats C functions and C++ functions differently,
as if they had a different calling sequence (they don't, but they
could). So if you declare a function in C++ having a function pointer
as a parameter, it's understood to be C++ even if it had previously
been declared as extern "C".

This could be a compiler error, though. In any case, the end result is
that WebKit fails to link because of an undefined reference to
NPN_PluginThreadAsyncCall.

"plugins/npapi.cpp", line 177: Warning (Anachronism): Formal argument 2 of type void(*)(void*) in call to WebCore::PluginMainThreadScheduler::scheduleCall(_NPP*, void(*)(void*), void*) is being passed extern "C" void(*)(void*).

There are more of these errors left in WebKit, but they are not
causing problems right now.

8 years agoFix compilation with Sun CC 5.9: ambiguity in ?:
Thiago Macieira [Thu, 3 Sep 2009 12:07:49 +0000 (14:07 +0200)]
Fix compilation with Sun CC 5.9: ambiguity in ?:

Error: Ambiguous "?:" expression, second operand of type "WTF::PassRefPtr<WebCore::DocumentFragment>" and third operand of type "int" can be converted to one another.
Error: Ambiguous "?:" expression, second operand of type "WTF::PassRefPtr<JSC::Label>" and third operand of type "int" can be converted to one another.
[and others similar]

Conflicts:

src/3rdparty/webkit/WebCore/workers/WorkerContext.cpp

8 years agoRemove comma at end of enum. Some compilers are more picky than others.
Thiago Macieira [Thu, 3 Sep 2009 12:06:41 +0000 (14:06 +0200)]
Remove comma at end of enum. Some compilers are more picky than others.

Conflicts:

src/3rdparty/webkit/WebCore/generated/CSSPropertyNames.h

8 years agoFix compilation with Sun CC 5.9: moving elements in a vector requires source not...
Thiago Macieira [Tue, 21 Jul 2009 12:18:35 +0000 (14:18 +0200)]
Fix compilation with Sun CC 5.9: moving elements in a vector requires source not to be const

I don't know why the compiler couldn't call src->~T() on a const T *src,
but fact is it couldn't.

In any case, since move is copying the source and deleting it, formally
the argument shouldn't be const anyway.

8 years agoFix compilation error on Solaris: mmap/munmap take/return a char*, not void*.
Thiago Macieira [Sun, 19 Jul 2009 13:14:06 +0000 (15:14 +0200)]
Fix compilation error on Solaris: mmap/munmap take/return a char*, not void*.

"../JavaScriptCore/interpreter/RegisterFile.h", line 128: Error: Using static_cast to convert from char* to JSC::Register* not allowed.

Error: Formal argument 1 of type char* in call to munmap(char*, unsigned) is being passed JSC::Register*.

8 years agoFix linking with SunCC 5.9: de-inline the operator new and delete in ParserArenaDelet...
Thiago Macieira [Thu, 23 Jul 2009 14:55:01 +0000 (16:55 +0200)]
Fix linking with SunCC 5.9: de-inline the operator new and delete in ParserArenaDeletable.

If you mark functions as "inline", the compiler doesn't have to emit
out-of-line copies. What happens is that Nodes.h declares these
functions, but the inline bodies are in NodeConstructors.h.

ParserArena.cpp used these functions, but didn't include
NodeConstructor.h. I could have added the missing #include, but this
is error-prone, since you have to remember to do that.

Moving the bodies into Nodes.h was also not possible, because it
requires JSC::Parser to be defined and Parser.h needs to #include
"Nodes.h".

So the solution is to de-inline.

8 years agoFix oversize-buffer support for aligning.
Thiago Macieira [Thu, 29 Jan 2009 15:07:27 +0000 (16:07 +0100)]
Fix oversize-buffer support for aligning.

Since Vector initialises VectorBase with the value of inlineBuffer(), it does so before the m_inlineBuffer member has had a chance to initialise. This lead to dereferencing of uninitialised pointers and, as was expected, crashes.

8 years agoDoc: Review of docs for QGraphicsItem::ItemUsesExtendedStyleOption.
Geir Vattekar [Fri, 4 Sep 2009 10:17:34 +0000 (12:17 +0200)]
Doc: Review of docs for QGraphicsItem::ItemUsesExtendedStyleOption.

Task-number: 253733
Reviewed-by: Bjørn Erik Nilsen
8 years agoFixed once-in-a-while failing tst_QGraphicsItem::selected test.
Gabriel de Dietrich [Fri, 4 Sep 2009 09:44:03 +0000 (11:44 +0200)]
Fixed once-in-a-while failing tst_QGraphicsItem::selected test.

Added event information in QTest::mouse* warning message.

Reviewed-by: Olivier
8 years agofix text eliding for arabic and syriac
Lars Knoll [Fri, 4 Sep 2009 10:05:41 +0000 (12:05 +0200)]
fix text eliding for arabic and syriac

Arabic and Syriac are connected scripts where the letter shape changes
depending on the context. Text eliding should not affect that letter
shape if the truncation happens in the middle of a word.

The patch ensures that by adding a Unicode ZWJ character between the
text and the eliding in case the character would connect in the full
string.

Reviewed-by: Simon Hausmann