summaryrefslogtreecommitdiffstats
path: root/src
Commit message (Collapse)AuthorAgeFilesLines
* Don't assume that raster can do porter duff in dfbAnders Bakken2009-07-171-5/+1
| | | | | | | PorterDuff should only be enabled if the raster engine says it is. E.g. if we're painting on a format with alpha. Reviewed-by: TrustMe
* Fixed bug where line widths were rounded to integers in the GL engine.Kim Motoyoshi Kalland2009-07-161-1/+1
| | | | | | | Regression from Qt 4.4. Task-number: 257990 Reviewed-by: Tom
* Phonon: fixed a big memory leak on WindowsThierry Bastian2009-07-162-2/+4
| | | | | | | If you had deleted a VideoWidget, it could not free the memory taken because we still had a reference on it. Task-number: 258202
* Better caching of file system icon providers.Norwegian Rock Cat2009-07-161-1/+36
| | | | | | | | | | | | It turns out that we weren't doing any caching of icons provided by the file system. We now use the similar trick that's used on Windows which does some caching on the file extension. We do fill up the cache needlessly with extra information (16, 32, 64, and 128) icons. We probably could be better with a iconRef engine that generates these sizes on demand. Still performance is 100% better with this which means using it in itemviews works. Reviewed-by: Jens Bache-Wiig
* Clean up qdirectfbpaintengine.cppAnders Bakken2009-07-151-149/+150
| | | | | | | | - Move implementation of debug functions to the bottom of the file. - Move ImageCache stuff to under QDirectFBPaintEnginePrivate - Move SurfaceCache stuff to under QDirectFBPaintEnginePrivate Reviewed-by: Shane McLaughlin <Shane.McLaughlin@trolltech.com>
* Don't force a lock in QDirectFBPaintEngine::clipAnders Bakken2009-07-152-27/+5
| | | | | | | | | | | Pretty much every paint operation on DirectFB surfaces starts with a clipping operations which until now would always cause a IDirectFBSurface->Lock(). This was to make sure QRasterBuffer::prepare had been called so QClipData::initialize() would allocate a big enough array for its spans. We can safely not make QDirectFBPaintDevice::memory() return 0 until we actually fall back. Reviewed-By: Donald <qt-info@nokia.com>
* Clean up rect/line drawing in dfb paintengineAnders Bakken2009-07-151-82/+63
| | | | | | | | Slight optimization by using the pluralized functions when possible. Also make templatized helper functions to avoid near-duplication of code for the float/int cases. Reviewed-by: Donald <qt-info@nokia.com>
* Fix off by one bug in dfbpe::drawTiledPixmapAnders Bakken2009-07-151-4/+4
| | | | | | | The drawTiledPixmap code iterated while x < right and y < bottom. This should be x <= right and y <= bottom. Reviewed-by: Donald <qt-info@nokia.com>
* Add a "User-Agent" line to our HTTP proxy requests.Thiago Macieira2009-07-151-0/+1
| | | | | | | | | | | | | | | | Apparently some proxy servers can be configured to deny requests based on User-Agent, even for CONNECT connections. See https://bugs.kde.org/show_bug.cgi?id=155707#c155 for an example (packet dump in #c157, analysis in #c158). So send a User-Agent header with value "Mozilla/5.0", hoping that this will be enough to allow the connection. I hope it will, because other tools like libtool send something completely different: User-Agent: curl/7.19.5 (i586-mandriva-linux-gnu) libcurl/7.19.5 GnuTLS/2.8.1 zlib/1.2.3 c-ares/1.6.0 libidn/1.15 libssh2/1.0 Reviewed-by: TrustMe
* QFileDialog static functions doesn't honor the DontUseNativeDialog flag.Alexis Menard2009-07-151-3/+3
| | | | | | | Just add a check before calling hooks. Task-number:258084 Reviewed-by:jbache
* Revert "QFileDialog: When passing an invalid path in static functions the ↵Alexis Menard2009-07-141-18/+3
| | | | | | native" This reverts commit a4c4f994fa51ff216f0d43098824617e14b8a284.
* Fix QDirectFBPixmap::toImageAnders Bakken2009-07-131-0/+5
| | | | | | | | | | Preallocated surfaces can currently be copied to video memory behind our back. This means that we can not use this mechanism for toImage() In later versions of DirectFB this might be possible to toggle with a flag so I'll leave the code in there #if 0'ed Reviewed-by: Donald <qt-info@nokia.com>
* Move ALPHA_PREMULTAnders Bakken2009-07-131-7/+7
| | | | | | | It's only used once and I want to unclutter the top of qdirectfbpaintengine.cpp Reviewed-by: TrustMe
* Remove QDirectFBPaintEnginePrivate::setOpacityAnders Bakken2009-07-131-9/+2
| | | | | | | The function only sets a variable anyway. Replace with: d->opacity = state()->opacity * 255 Reviewed-by: Donald <qt-info@nokia.com>
* Remove unused variables in QDirectFBPaintEngineAnders Bakken2009-07-131-4/+1
| | | | | | fbWidth/fbHeight were never used for anything Reviewed-by: Donald <qt-info@nokia.com>
* Code cleanup in QDirectFBPaintEngineAnders Bakken2009-07-131-13/+13
| | | | | | Take out the QPen member. Reviewed-by: Donald <qt-info@nokia.com>
* Clean up QDirectFBPaintEngine codeAnders Bakken2009-07-131-66/+71
| | | | | | | | | | | | | | | | - Remove unnecessary copy of QTransform - Fold matrixRotShear and scale into transformationFlags. Note that transformationFlags is not a proper bitset of the types of transform but rather set to the most complex operation (from QTransform::type()) and having NegativeScale added if qMin(transform.m11(), transform.m22()) < 0 - Fix a bug whereby setState didn't call setRenderHints - Make everything more readable - Don't initialize state in QDirectFBPaintEngine::begin() QDirectFBPaintEngine::setState will be called from QPainter::begin just before QDirectFBPaintEngine::begin Reviewed-by: Donald <qt-info@nokia.com>
* Fixed a potential memory leak on XPJens Bache-Wiig2009-07-131-2/+4
| | | | | | | | | Calling OpenThemeData directly causes a leak when changing the style as we do not call the corresponding CloseThemeData. Task-number:257916 Reviewed-by:prasanth
* Updated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( ↵Simon Hausmann2009-07-135-6/+34
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | a3e05ad8acdead3b534d0cef772b85f002e80b8d ) Changes in WebKit since the last update: ++ b/LayoutTests/ChangeLog 2009-06-18 Chris Evans <scarybeasts@gmail.com> Reviewed by Adam Barth. Added test for bug 26454 (broken 8-digit hex entities). https://bugs.webkit.org/show_bug.cgi?id=26454 * fast/parser/eightdigithexentity-expected.txt: Added. * fast/parser/eightdigithexentity.html: Added. 2009-06-20 Sam Weinig <sam@webkit.org> Reviewed by Adam Barth. Test for https://bugs.webkit.org/show_bug.cgi?id=26554 Test writing to parent and top. * http/tests/security/cross-frame-access-put-expected.txt: * http/tests/security/cross-frame-access-put.html: * http/tests/security/resources/cross-frame-iframe-for-put-test.html: ++ b/WebCore/ChangeLog 2009-06-18 Chris Evans <scarybeasts@gmail.com> Reviewed by Adam Barth. Fix 8-digit long hex entities. Fixes bug 26454 https://bugs.webkit.org/show_bug.cgi?id=26454 Test: fast/parser/eightdigithexentity.html * html/HTMLTokenizer.cpp: fix off-by-ones. 2009-06-20 Sam Weinig <sam@webkit.org> Reviewed by Adam Barth. Fix for https://bugs.webkit.org/show_bug.cgi?id=26554 Shadowing of top and parent * page/DOMWindow.idl:
* Updated WebKit from /home/shausman/src/webkit/trunk to origin/qtwebkit-4.5 ( ↵Simon Hausmann2009-07-1318-2040/+3375
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | eb4957a561d3f85d4cd5602832375c66f378b521 ) Changes in WebKit since the last update: ++ b/LayoutTests/ChangeLog 2009-06-04 Ariya Hidayat <ariya.hidayat@nokia.com> Rubber-stamped by Tor Arne Vestbø. Added Qt-specific expected result for Canvas pointInPath's test. Qt's QPainterPath::contains(point) returns false if the point is exactly along one of the edges (except the origin). Until we find a workaround for this, compensate it in the expected result. * platform/qt/fast/canvas/pointInPath-expected.txt: Added. 2009-01-11 Simon Fraser <simon.fraser@apple.com> Reviewed by Oliver Hunt https://bugs.webkit.org/show_bug.cgi?id=23242 Add testcase for incremental repaint after use of ctx.transform(), and enhanced isPointInPath testcase to do testing after use of ctx.transform(). * fast/canvas/canvas-incremental-repaint-2.html: Added. * fast/canvas/pointInPath-expected.txt: * fast/canvas/pointInPath.js: * platform/mac/fast/canvas/canvas-incremental-repaint-2-expected.checksum: Added. * platform/mac/fast/canvas/canvas-incremental-repaint-2-expected.png: Added. * platform/mac/fast/canvas/canvas-incremental-repaint-2-expected.txt: Added. ++ b/WebCore/ChangeLog 2008-12-18 Bernhard Rosenkraenzer <bero@arklinux.ch> Reviewed by Darin Adler. https://bugs.webkit.org/show_bug.cgi?id=22205 Fix compatibility with bison 2.4, partially based on older patch by Priit Laes <amd@store20.com> * WebCore/css/CSSGrammar.y: Made compatible with bison 2.4 2009-07-13 Cédric Luthi <cedric.luthi@gmail.com> Reviewed by Tor Arne Vestbø. Fix NPWindow clip rect in PluginViewMac The rect should be in window-coordinates. This bug can be observed with Flash 10 here: http://www.permadi.com/tutorial/cursorTracker/ * plugins/mac/PluginViewMac.cpp: 2009-07-13 Simon Hausmann <hausmann@webkit.org> Reviewed by Ariya Hidayat. Fix Qt implementation of WebCore::directoryName to return the absolute directory name instead of the base file name. * platform/qt/FileSystemQt.cpp: (WebCore::directoryName): 2009-07-13 Simon Hausmann <hausmann@webkit.org> Reviewed by Ariya Hidayat. Fix WebCore::Path::isEmpty() for the Qt port to return true if there is no element in the path. QPainterPath::isEmpty() returns also true if there is one single MoveTo element inside, which makes sense but doesn't patch Webcore's is-empty definition. * platform/graphics/qt/PathQt.cpp: (WebCore::Path::isEmpty): Use elementCount() == 0. 2009-01-11 Simon Fraser <simon.fraser@apple.com> Reviewed by Oliver Hunt https://bugs.webkit.org/show_bug.cgi?id=23242 Fix CanvasRenderingContext2D::transform to do a pre-multiply, rather than a post-multiply into m_transform. This bug did not affect drawing, but did cause m_transform to be incorrect, which impacted willDraw(), and isPointInPath. Test: fast/canvas/canvas-incremental-repaint-2.html * html/CanvasRenderingContext2D.cpp: (WebCore::CanvasRenderingContext2D::transform): 2009-07-04 Sriram Yadavalli <sriram.yadavalli@nokia.com> Reviewed by Simon Hausmann. https://bugs.webkit.org/show_bug.cgi?id=26439 QtWebKit fails in loading www.nytimes.com in Windows/Linux QNetworkReplyHandler is ignoring content associated with 401 error. This causes the XHR response handling to fail. Simon: Added also ProxyAuthenticationRequiredError, to handle the same case when going through proxies, as suggested by Prasanth. * platform/network/qt/QNetworkReplyHandler.cpp: (WebCore::QNetworkReplyHandler::finish): ++ b/WebKit/qt/ChangeLog Fix crash with plugins when the plugin stream is cancelled. Similar to r26667 handle the case where didReceiveResponse on the plugin view results in failure to set up the stream and setMainDocumentError being called instead. This will set the m_pluginView back to 0 and we need check for it before calling didReceiveData. This was triggered by consecutive execution of LayoutTests/plugins/return-error-from-new-stream-callback-in-full-frame-plugin.html followed by LayoutTests/scrollbars/scrollbar-crash-on-refresh.html * WebCoreSupport/FrameLoaderClientQt.cpp: (WebCore::FrameLoaderClientQt::committedLoad): 2009-07-13 Simon Hausmann <hausmann@webkit.org> Reviewed by Ariya Hidayat.
* QNAM: Fix double sending of a HTTP requestMarkus Goetz2009-07-132-0/+9
| | | | | | | | | In some cases, weird timing issues could occur. In those cases, an EOF was sent twice for the upload data, leading to the HTTP code being confused and sending the request headers twice. Task-number: 257662 Reviewed-by: Thiago Macieira
* QFlags::testFlag(): handle the zero case appropriately.Frans Englich2009-07-131-1/+1
| | | | | | | | Brought up by Andy. See perforce change 314809, 17b07e3ab6192b31f77fd2f126705b9ab53b3937. Related to task 221708. Reviewed-By: Andy Shaw (cherry picked from commit cc24c46c117248ecb98200416e7f25375e6bb476)
* Fixes failure when table has null fields to updateBill King2009-07-131-1/+1
| | | | | | | Fixes an issue where too many parameters are bound when updating QSqlTableModel where the stored record has NULLs in it. Reviewed-by: Justin McPherson
* Fixed a crash with input methodsPierre Rossi2009-07-102-6/+3
| | | | | | | | | The inputContext's focusWidget was not reset when disabling input methods. Thanks to Benjamin P. Task-number: 257832 Reviewed-by: Denis
* Fix QDFBPaintEngine::drawTiledPixmap/fillRectAnders Bakken2009-07-101-57/+111
| | | | | | | | | Until this fix very few cases of drawTiledPixmap and fillRect(r, textureBrush) have been handled by DirectFB. This patch makes it possible to accelerate such operations. Reviewed-By: Donald <qt-info@nokia.com>
* Pass the device arg to the keyboard drivers.Anders Bakken2009-07-094-8/+4
| | | | | | | We didn't use to pass the device in to the keyboard handlers rendering them mostly useless. Reviewed-by: TrustMe
* emit dataChanged() signalDerick Hawcroft2009-07-091-0/+1
| | | | Task-number:207874
* Plugged possible memory leak in getPSQLVersion()Stian Sandvik Thomassen2009-07-091-2/+2
| | | | | | | Call PQclear() regardless of the status of the result returned by PQexec(). Reviewed-by: Bill King
* Hopefully fix isnan/isinf for good (for all platforms)Bill King2009-07-081-12/+20
|
* Update QGroupBox on focusBenjamin Poulain2009-07-081-5/+1
| | | | | | | | We cannot assume the position of the decorations when a QGroupBox get the focus. Task-number: 257660 Reviewed-by: Thierry
* Fixes Dericks inf/nan patch for msvcBill King2009-07-081-0/+13
|
* remove debugDerick Hawcroft2009-07-071-1/+0
|
* Handle special number cases (nan,{+-}infinity) in PostgreSQLDerick Hawcroft2009-07-071-0/+17
| | | | | | | | | When contructing the EXECUTE statement, there is a special case that we need to handle whereby we explicitly put quotes around the special float values before submutting the statement for execution Task-number:233829 Reviewed-by: Bill King
* Document limitation in Cocoa cursor handling.Norwegian Rock Cat2009-07-071-0/+5
| | | | | | | | | | It seems there is a bug in AppKit which will automatically reset a cursor even when it is grabbed, but won't reset it when it's brought back into the window. The upshot of this is that doing a setCursor() inside of mouse handling behaves slightly different than on the other platforms (including Carbon). However, we are at the mercy of Cocoa here and I would rather have all the other things AppKit does right and live with this bug which they may fix some day.
* Fix typo in color calculation.Norwegian Rock Cat2009-07-071-3/+3
| | | | | | | | | Argh! It's divide by 256 not 265. The worst part was that I used the same values in Cocoa as well, so they were both "damaged." It should be good now. Task-number: 257499 Reviewed-by: Prasanth Ullattil
* Handle all PostgreSQL notifications sitting in the queueDerick Hawcroft2009-07-071-4/+4
| | | | | Task-number: 257247 Reviewed-by: trustme
* Fix issue where a mainwindow would show two size grips in Cocoa.Norwegian Rock Cat2009-07-061-2/+5
| | | | | | | | | | | | | | | | | | | OK. this is a bit strange. It seems the topdata->resizer value is used to control whether or not we should show a resize handle based on a count (0 no, non-zero yes). Since we somehow decided that this value will never be larger than 15, we made it 4-bits wide. There's a "Qt/Mac" API, QWidgetPrivate::qt_mac_update_sizer(QWidget *, int = 0) which would adjust this value by the int passed in.. We use that in several places, not excluding the QStatusBar where we would pass 1 if we want to show, and -1 if we didn't. Now if you subtract -1 from zero when you are 4 bits wide, well, bad things happen. Therefore protect that (since if it's at zero we have succeeded, we don't want to show the resizer). This seems to work well. The private API is certainly an interesting way of solving the problem, but is easy to abuse (for example, this code will break if resizer = 1 and we are passed -2 in the function. Task-number: 257485 Reviewed-by: Prasanth Ullattil
* Implement QApplication::setOverrideCursor to use pure Cocoa callsNorwegian Rock Cat2009-07-061-1/+11
| | | | | | | | | | | Seems this was a victim of our cursor fixing. Cocoa does a lot for us with setting cursors. This meant that we didn't need to do as much meddling and as a result qt_mac_set_cursor does nothing in Cocoa. Unfortunately, this broke setOverrideCursor. Luckily Cocoa has a stack that works exactly like Qt, so we can just use that. Task-number: 257507 Reviewed-by: Prasanth Ullattil
* Fixed dead code possibly leading to crash.Rohan McGovern2009-07-031-1/+1
| | | | | | | | Looks like this `&&' was meant to be `||'. QNetworkProxy::FtpCachingProxy is 5 so it's clearly impossible for type to be less than 0 and greater than QNetworkProxy::FtpCachingProxy. Reviewed-by: Aaron Kennedy
* Fixes ::record for dialect 3 named tables in interbase/firebird.Bill King2009-07-031-2/+2
| | | | | | The comparison was mistakenly only uppercasing one side, so mixed case table names were reporting back as if they weren't found for both QSqlDatabase::record() and QSqlDatabase::primaryIndex()
* Fixes invalid length for numeric fields in oracle.Bill King2009-07-031-1/+1
| | | | | When the precisionpolicy is high, and the field is numeric, it was getting confused as a string field and pulling the wrong length value.
* QDirectFBPaintEngine return if destRect is nullAnders Bakken2009-07-031-0/+2
| | | | Reviewed-by: Donald <qt-info@nokia.com>
* Better debug output for QDirectFBPaintEngineAnders Bakken2009-07-021-2/+5
| | | | | | Add unsupportedCompositionMode to the output. Reviewed-by: TrustMe
* Remove unused function viewUnderTransparentForMouseView in QCocoaView.Prasanth Ullattil2009-07-022-57/+0
| | | | | | | After we implemented hitTest for QCocoaView, this function is no longer used. Reviewed-by: Norwegian Rock Cat
* Drag and drop events are not delivered correctly in CocoaPrasanth Ullattil2009-07-022-8/+43
| | | | | | | | | | | Drag and drop events should consider the WA_TransparentForMouseEvents attribute like the mouse events. If this attribute is set for a widget, the event has to be passed to right widget under mouse. The widget is identified by calling hitTest. In such cases the leave event has to be delivered to the widget which actually accepted the enter event. Task-number: 252088 Reviewed-by: Norwegian Rock Cat
* Implement hitTestNorwegian Rock Cat2009-07-022-54/+7
| | | | | | | | | | | | | | | Cocoa calls hitTest on our view to determine if the view should get the mouse press. We always said, "yes" and did all the logic ourselves. Turns out that we can say "no" if I'm transparent to mouse events and remove all that code where we do all the work ourselves. Big maintenance win! For the time being I've kept the "transparentViewForEvent" method since it might be useful for others, but no one is using it at the moment and we may just kill it soon. HitTest should handle this situation correctly.
* Fix a regression where dynamic tooltips wouldn't show up in Cocoa.Norwegian Rock Cat2009-07-022-4/+15
| | | | | | | | | | | | | | Tracking of mouse events was only enabled when enableMouseTracking or Hover or a tooltip had been set explictly on the item, but this meant that the dynamic QEvent::Tooltips would never get dispatched. So, in order to help out people that might use this feature, all QCocoaViews must pay the mouse move event tax *sigh*. I added comments in the proper places so that we DO the right thing for a release where we can force the change in behavior. Task-number: 257320 Reviewed-by: Denis
* Fix oops in strcmp in QBuffer.Thiago Macieira2009-07-021-1/+1
| | | | | | Reported via qt-bugs. Reviewed-By: Peter Hartmann
* Add a note about this method not being safe in multithreaded contextsThiago Macieira2009-07-021-0/+7
|
* Tinyint is unsigned, force it to such.Bill King2009-07-021-1/+3
| | | | | Tinyint only supports 0-255, so mark it as unsigned despite sign flag, which have the time is inverted/wrong.
generated by cgit v0.12 at 2025-01-02 21:51:15 (GMT)