| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commit ed375675d4a4f6fd63edeb242e23c87b3de4be6f triggers a behavior in
Glib's mainloop implementation where some event sources are not
"serviced" every iteration of the mainloop context. This breaks an
invariant that many tests relied on, so we need to solve the problem.
The invariant is that a newly added timer that would normally fire on
the next pass of the event loop (liker a zero timer) SHOULD actually
fire. We do this by registering 2 timer event sources with Glib's
mainloop: one normal priority source and one idle priority source. The
idle priority source is the one that will send events most of the
time, with the normal priority one taking over only when
processEvents() is called manually.
Task-number: QT-877
Reviewed-by: jbache
Reviewed-by: thiago
Reviewed-by: denis
(cherry picked from commit d0d0fdb8e46351b4ab8492de31e5363ef6662b57)
|
|
|
|
|
|
|
|
|
|
|
|
| |
If you set the flag itemClipsChildrenToShape to true on a parent, an
optimization was made in 4.5.0 to not add children of this parent in
the index. But when you set the flag back to false all the sub-tree
of the parent should be re-added to the index otherwise the index will
never find all children. This code is not relevant for 4.6 since
the index part of QGraphicsView has been refactored and handle this case
with itemChange in QGraphicsSceneIndex.
Reviewed-by:andreas
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When QGLWidget was used as a viewport for QGraphicsView, it was
still treating the window surface as RGB16. Use the screen's actual
pixel format.
Also ensure that PvrEglWindowSurface::image() returns a non-null
QImage if the drawable hasn't been created yet.
Reviewed-by: trustme
Back port of 53b3a0572242d0a425e74848afba1293f195d29b
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When the editor had been created inside the QtPropertyEditorView (inheriting
QTreeWidget), the subsequent show sent a synthetic mouse move event down to
the QLineEdit, and a new selection was made on the text because the mouse
button was marked as pressed in the event.
QApplicationPrivate::sendSyntheticEnterLeave() now sends a mouse move event
without any button pressed. Auto-test included in tst_QWidget.
Task-number: QTBUG-4055
Task-number: 253159
Task-number: QT-659
Task-number: 245398
Reviewed-by: bnilsen
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes handling selection requests for invalid targets - when
someone asks for a target that is not supported by the clipboard content
we shouldn't do anything (unless it's MULTIPLE).
Fixes copying data when using Synergy which tries to get all targets it
knows about even if they are not listed in TARGETS.
Task-number: QTBUG-4652
Reviewed-by: Bradley T. Hughes
|
|
|
|
|
|
|
|
| |
Send a header that we accept every language. Works around a bug in some
broken websites and is hopefully of no harm for us.
Task-number: QT-952
Reviewed-by: Thiago
|
|
|
|
|
|
|
|
| |
Task: QT-1027
RevBy: Andrew den Exter
Details:
All gst disk sources support setting device name with "device" property,
otherwise the default device is used.
|
|
|
|
|
|
|
| |
RevBy: Andrew den Exter
Details:
Phonon gstreamer backend used cdda:// URL and corresponding
gst source element regardless of disk type.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Revert this change:
----
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
-----
Cherry-picked from qtwebkit.git 4.5 repository/branch with commit
23d13ba45ee59379c04afdef8472acc8611bb36a
|
|
|
|
| |
Task-number: QT-754
|
| |
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
updated
When calling QGraphicsItem::update() on a cached item, the cache is
meant to be invalidated.
In the reported bug, the user had a fixed scene rect
set for his scene, and removing an item caused the entire scene to be
updated (marked as "all needs to be updated"). In this case, calling
update() on the cached item did not cause the item's cache to be
invalidated. The item's new appearance didn't show up until the next
invalidation, which was the same call to update(), but this time without
a preceeding full scene update.
The fix is to always invalidate the cache, regardless. But only
schedule a repaint of the item in some cases (e.g., in this case the
whole scene was marked for update, in which case it's unnessary for this
one item to schedule a repaint of itself).
It's worth noting that in 4.6, removing an item be delete does not cause
the whole scene to be updated, and because of that this error was not
exposed. It's there nevertheless.
Reviewed-by: bnilsen
|
|/
|
|
| |
Reviewed-by: Trust Me
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Regression against Qt 4.4. Children of items with ItemClipsChildrenToShape
would only be discovered if the view's expose region contained the outer
bounding rect of all items, _if_ there was at least one item in the
scene that enabled ItemIgnoresTransformations.
The reason for this bug is that the presence of an untransformable item
causes the item lookups to go through a different path
(QGraphicsViewPrivate::itemsInArea()). This function had the bug that it
didn't correctly discover children of clip-items. Because of this, in
the provided test case you could "work around" the bug by either removing
the clip flag, or the transformation flag.
Task-number: QTBUG-4151
Reviewed-by: Alexis
|
|
|
|
|
|
| |
This fix just fixes up coding bugs here and there
Reviewed-by: Brad
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Removing dubious intermediate detection code that also had a buffer
overflow. The results were inconsistent and not dependable on.
Processing was inefficient and end value to user experience dubious.
Test cases that abused the former behaviour were changed to consider
input in an Intermediate where it was previously considered Invalid.
With this change, user input will mostly be considered in an
intermediate state, until it is effectively validated.
Task-number: 255019
Reviewed-by: Anders Bakken
|
|
|
|
| |
Task-number: QT-1683
|
|
|
|
|
|
|
|
| |
Revert parts of adf322c514a5781dcb9ec304d44229fa47d5e8b3 to
get this to work as in 4.5.2 again. What the original
patch fixed, we don't really know..
Reviewed-by: Simon Hausmann
|
|
|
|
| |
Reviewed-by: Samuel
|
|
|
|
|
|
|
|
|
| |
[NSWindow orderFront:] on a hidden window will make it visible. So
raise_sys() will now check if window is visible before this method is
called.
Task-number: 255428
Reviewed-by: Richard Moe Gustavsen
|
|
|
|
|
|
|
|
| |
This fixes a regression introduced in Qt 4.5.0, when dragging an item
within a QTreeView it would not call clone() when it was dropped even if
a prototype was set on the QStandardItemModel.
Reviewed-by: Marius Bugge Monsen
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This issue affects both tool buttons and buttosn
with menus. The problem is that the Windows metric
we were using to calculate the size changed on
Windows 7. It seems not to be the correct metric so it
seems more reliable to use the dpi-scaled constant 12 as we
were doing in common style already. This should make appearance
consistent between the windows versions.
Task-number: QTBUG-4241
Reviewed-by: denis
|
|
|
|
|
|
|
|
|
| |
Fix QGraphicsItem to clear clip path for items with an ancestor that
clips to shape. Added autotest to demonstrate clipping path problem.
Merge-request: 810
Reviewed-by: Alexis
Reviewed-by: Andreas
|
|
|
|
|
|
|
|
|
|
| |
On Windows CE the plugin path is likely to be the same as the
application directory. That's why we must not compare app_location
with the plugins path.
The check app_libpaths->contains(app_location) should be enough.
Autotest: tst_QApplication::libraryPaths
Reviewed-by: thartman
|
|
|
|
|
|
|
|
|
| |
Cocoa does not allow NSMenu to have multiple supermenu's. If a menu is
added again as submenu, Qt will now disable the menu item or the menu
will not be added at all if it is added again to the menubar.
Task-number: 258822
Reviewed-by: MortenS
|
|
|
|
| |
Reviewed-by: trustme
|
|
|
|
|
|
|
|
| |
The pvr2d.h/wsegl.h headers we ship with Qt (in src/3rdparty/powervr)
are meant for the PowerVR SGX. However, we use an MBX-specific define in
the powervr driver.
Reviewed-by: Jørgen Lind
|
|
|
|
|
|
|
| |
If we define QT_COORD_TYPE, src/testlib doesn't compile for embedded
Linux.
Reviewed-By: joerg
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since the width is multiplied into the dash, it needs to be divided out,
otherwise you can get a dashOffset which is greater than the pattern at
the index, and the dash can become negative. This will in turn lead to
passing a negative width to the rasterizer, which at some point will
get cast to an unsigned int and overflow. Depending on the position of
the line and size of the buffer, this will either crash or produce
garbled output.
Task-number: QT-4444
Reviewed-by: Samuel
|
|
|
|
|
|
|
|
|
| |
Previously you had to set forward only on non-scrollable datasets
explicitly. This queries ODBC, to determine if it's a scrollable
dataset, and sets forwardOnly to false if it isn't.
Task-number: QT-353
Reviewed-by: Justin McPherson
|
|
|
|
|
|
|
|
|
|
|
|
| |
RevBy: Andrew den Exter
Details:
gst_element_query_duration(element,format,duration) doesn't always
return duration in format being asked for (tracks in this case),
it can also return duration in format it can (Time) and modify
format parameter, so check the format is still the same
as requested is necessary. This bug prevented Phonon to emit finished()
signal with some files, since it expected next tracks to exist.
|
|
|
|
|
|
|
|
|
|
|
| |
Change 979d1d3bbc0c68789edbe93f03464d41d7a8469a requires
qt_format_text() to honor the Qt::TextForceLeftToRight flag. Without
this, the text will be laid out RTL twice, and the output will be
broken. Since printing is done through QPicture, this fixes printing
when the UI is reversed.
Task-number: 261033
Reviewed-by: Trond
|
|
|
|
|
|
|
| |
Uses the primary key from the index in the query, not the resulting
location in the modified dataset.
Task-number: 222678
|
|
|
|
| |
Reviewed-by: Justin McPherson
|
|
|
|
|
|
|
| |
max_keycode wasn't retrieved.
Merge-request: 1308
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
|
|
|
|
|
|
|
| |
We have to munmap() what we have mmap()ed
Reviewed-by: Gunnar
Reviewed-by: Jeremy
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For some reason iconv support was broken in Qt 4.5 on FreeBSD, and this
patch was made to fix that.
Original patch by Max Brazhnikov <makc@freebsd.org>.
Original message:
Enable libiconv support.
Didn't get where the real problem is, but using UTF-16LE/BE as default
encoding (see patch) fixes the issue.
Merge-request: 1480
Reviewed-by: Thiago Macieira <thiago.macieira@nokia.com>
|
|
|
|
|
|
|
|
|
|
| |
The deleteLater was beeing created with loopLevel of 1, causing
it to be defferd until QApplication::exec() returned.
Add a QScopedLoopLevelCounter to increase the loopLevel while
triggering the action.
RevBy: Brad
|
|
|
|
|
| |
_HIViewScrollRectWithOptions needs to be declared as a weak-linked
symbol in order to make static linking work.
|
|
|
|
| |
HIViewSetNeedsDisplayInRect was added in 10.4.
|
|
|
|
|
|
|
|
|
|
| |
After restoring a minimized window we only saw the window decoration.
All content was missing. That's because we don't get a WM_SIZE message
for restoring the window. We must react on WM_ACTIVATE in this case.
This fixes the issue for Windows mobile too.
Task-number: 260702
Reviewed-by: thartman
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since 925912ebf552417306a5bd20fd986afda6a582be QNetworkReplyImplPrivate
no longer holds its own pointer to the network cache and relies on the
backend to get it. Since the cache is used in our call to finished, we
must reset the backend only after that.
This fixes a crash I was seeing in Arora...
Reviewed-by: Markus Goetz
Reviewed-by: Peter Hartmann
|
|
|
|
|
|
|
| |
Stops the license checker failing each time the file is re-generated.
Also added "_P" to the generated header guard, as it's a private header.
Reviewed-by: Gunnar Sletta
|
|
|
|
|
|
|
| |
Ported this fix backwards from 4.6 to 4.5
Adds some cleanups (using QVarLengthArray), and reverting to the
initial and correct calculation (when the driver doesn't deem fit to
return SQL_NO_DATA).
|
|
|
|
|
| |
Reviewed-by: Takumi Asaki
Also-applied-as: 3f257dcf8c8ca812d60d4a6a4c2cd2da96bba526
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In the Windows print engine, we try to send a text item as a raw string
of characters to the printer driver if this is possible. This is to
facilitate using PDF-printers as much as possible, allowing them to
save the text in the document so for searching etc. We can only safely
do this if all the characters in the string are ASCII-compatible, i.e.
in the 7 bit range, since this is the only part of the set which is
guaranteed to be compatible across code pages.
Task-number: 180655
Reviewed-by: Trond
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
We need to floor instead of round to prevent rectangles that are on the
edge from being shifted one pixel down / right.
Task-number: 258776
Reviewed-by: Kim
|