On Unix systems where the GUI event dispatcher uses a notification
system for socket notifiers that is out of band compared to select(),
it's possible for the QSocketNotifier to activate after the pipe has
been read from. When that happened, the ioctl(2) call with FIONREAD
might return 0 bytes available, which we interpreted to mean EOF.
Instead of doing that, always try to read at least one byte and examine
the returned byte count from read(2). If it returns 0, that's a real
EOF; if it returns -1 EWOULDBLOCK, we simply ignore the situation.
That's the case on OS X: the Cocoa event dispatcher uses CFSocket to get
notifications and those use kevent (and, apparently, an auxiliary
thread) instead of an in-thread select() or poll(). That means the event
loop would activate the QSocketNotifier even though there is nothing to
be read.
Task-number: QTBUG-39488
Change-Id: I1a58b5b1db7a47034fb36a78a005ebff96290efb
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
The previous implementation used [NSApp orderedWindows]
which does not return NSPanel subclasses, which is
used by Qt dialogs and pops.
Use [NSWidow windowNumberAtPoint:belowWindowWithWindowNumber]
instead, which hit-tests on all window types. This
can potentially include windows from other processes
and non-Qt windows which needs to be filtered out.
Add EXPECT_FAIL to tst_MacGui::nonModalOrder. The
correct topLevelAt() implementation now exposes that
this test is failing.
Task-number: QTBUG-39322
Change-Id: I81afa3da964e08fe682802220d8fe81e9284205e
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Prevent crash on platforms that don't support accessibility by skipping
tests.
Change-Id: I42ba44df3200e0abd62797c76a5c538fb1d2757c
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Nothing else seems to report this state and on windows
for example it results in NVDA reading subMenu which makes
little sense.
Task-number: QTBUG-38500
Change-Id: I64820d9f2ea9174034f01da42cb2266a19c19465
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
VoiceOver expects press to open the combobox.
Task-number: QTBUG-37922
Change-Id: Iee7b7974db097e4e2444202c703bd587e1576fe0
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
The big change is that we now keep the id objects representing accessibles
around so that they are persistent for ATs.
This improves performance of Mac accessibility significantly.
This is required for notifications which are now sent so that many things work much better,
for example the VoiceOver focus follows the keyboard focus.
The parent element in QCocoaAccessibleElement was removed, we can
dynamically access it more reliably.
Change-Id: I686d212f40d28b392dcc22f16f3c3430f08bdc98
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
For Mac this makes QTextEdit work nicely with VoiceOver.
Task-number: QTBUG-37204
Change-Id: I1326d24ca6a932ad667ee395f62881b6ec64e892
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
It doesn't depends on windget for quite a long time already.
Change-Id: I251e2e0d52028d17a61c892308d099f344728e79
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
That #ifdef is placed before any include, so it has no effect.
Change-Id: I4bbf967bda720d287b363847669c5c7a22807bc4
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@digia.com>
Actually guarantee cursor doesn't move in this case for both logical
and visual modes (just what the documentation says we already do ;)
Change-Id: Iabdca7aa1d205672386a0095e3487e585611cdb5
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Change b94493e revealed that in the lance script rendering, at the
surface_end command, cleaning up the fbo stuff and enabling the
ordinary painting again was done in the wrong order.
Change-Id: I358dafeffe95b25303fc2a8dc1d61384b2ca64fb
Reviewed-by: Gunnar Sletta <gunnar.sletta@jollamobile.com>
Fixes failures on Windows 8.1.
Task-number: QTBUG-37822
Change-Id: I30460f430e2249555c4e857237e27aa9b03dfa1b
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
This simplifies how we handle QNSView for accessibility purposes.
Instead of trying to half-merge the top level widget
(window->accessibleRoot) into the view, just have the view
always return it as child.
This makes the accessibility implementation for QNSView simpler
and makes applications that show a top level widget such as a button
possible. (We would return accessibility ignored for the button before).
As a side effect finding the active focus and hit-testing should be more reliable as
well.
Task-number: QTBUG-37794
Change-Id: Ib52037f88da8887a0bdc77204b0f3daddfe7709d
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
EditableTextInterface was implemented but not reported to the
accessibility bridges.
Newlines in QTextEdit when using QCursor::selectedText are returned as
unicode paragraphs, replace them by newlines.
[ChangeLog][QtWidgets][Accessibility] Fixed QTextEdit not reporting
newlines to accessibility frameworks and add editable text interface.
Change-Id: Iac21e70f5468a16f8abf242ae148290dbab3f8e4
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
This simplifies the compiler control semantics by always using the
packaged compiler if it is available. With no packaged compiler, the
service is assumed to be running if the directory structure is present.
Use of a packaged library can be overridden by setting the environment
variable QT_D3DCOMPILER_DISABLE_DLL to 1.
When the runtime compiler is used, the source will no longer be
logged, and the compilation output will no longer be cached.
Change-Id: Ib07f517e7043d7785bdfa9da55abd34df518eeed
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Correct the tense of send vs sent in comments and documentation.
Change-Id: I1c5ce9a7b1e49b8b0e8dcfde7d732e4c69acf73a
Reviewed-by: Kurt Pattyn <pattyn.kurt@gmail.com>
Reviewed-by: Laszlo Agocs <laszlo.agocs@digia.com>
That's a bug, but this test causes CI problems until it's fixed.
Task-number: QTBUG-35169
Change-Id: I27b3a61437312d2217743670fa14510d9340d8c1
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Simo Fält <simo.falt@digia.com>
Try to fix frequently failing test:
FAIL! : tst_NetworkSelfTest::ftpProxyServer() Failed to receive data in step 32: timeout
tst_networkselftest.cpp(230) : failure location
on Windows. Introduce timeout constant, add message.
Change-Id: I709f0b34cd1cfe5d3c64cf61ccb7907bd616bc54
Reviewed-by: Peter Hartmann <phartmann@blackberry.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
by default QT contains both core and gui already
Change-Id: I6f5b551104e40a024468e7cb62e302134e9472ec
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
Showing, moving and resizing
Contrarily to what an NSWindow does to its NSViews, child NSWindows need
to be explicitly shown and hidden, and clipped if the parent NSWindow
changes geometry. Also, hiding an NSWindow will not hide its child
windows. This needed to be managed manually, adding 2 additional states
to QCocoaWindow to reflect whether a child window has been clipped out by
any ancestor geometry change, or hidden by any ancestor being hid. Also,
ordering out an NSWindow will remove it fromm its parent's child windows
array, making necessary to maintain a parallel list of child windows in
QCocoaWindow.
Stack order
Although child NSWindows can be ordered relatively to each other, they
need to be added again to be moved lower in the window stack. This also
means the windows above it need to be added on top.
Key (focus) status
One of the remaining issues, is to make sure the top level window keeps
the "key status" while still forwarding key events to the child window.
Keeping same event propagation
This use case is best illustrated with undocking QDockWidgets (if these
are child NSWindows). The main issue is to make sure the QDockArea will
get the mouse events right after undocking a dock widget. We used a similar
workaround as the "key status" problem, and manually forward the mouse
events to the dock area's QWindow.
Manual test, by Morten Johan Sørvig, included.
Task-number: QTBUG-33082
Task-number: QTBUG-22815
Change-Id: I50e34936fb82bff013e99f4bcb3bd0db0704c6ae
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
The compiler service needs to know what options to pass to the compiler,
and these options can affect the outcome of the shader blob. Runtime
compiled shader output must match this file name in the binary directory.
Pre-compiled shader blobs (e.g. those placed in a resource file) are
permitted to drop options from the file name if the implementor can
ensure the blob is compatible with the target.
Defines are not written out as options, but written into the shader
source, and therefore affect the cache key.
Change-Id: I11e4a43fcf6818ddb29aca5eba3d8647ba4021a1
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
the diff -w for this commit is empty.
Started-by: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I77bb84e71c63ce75e0709e5b94bee18e3ce6ab9e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
While there is no implementation for accessibility yet, enabling it allows
the interfaces to be used and an accessibility plugin to be developed
for this platform.
IAccessible2 and MSAA bridge autotests are disabled for this platform.
Change-Id: I2bfd07f6b21ca469b27d88ef11df723ac8ff8202
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Use of mkpath/mkdir is removed in order to make the proxy less
invasive when there is no compiler service running. Creation of the
directory structure is fully the service's responsibility.
Service availability is now only checked once, at the first invocation of
D3DCompile, as it is expected to be started before the application.
Change-Id: Ib8c4f062c418497c2253daf524654e1db30dae47
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
d3dcompiler_qt is a DLL for use with ANGLE which replaces d3dcompiler_XX
at runtime to proxy shader compilation calls. This is useful for:
- Loading the newest D3D compiler DLL found, instead of loading the
version specified when Qt was compiled
- Reporting better debug information when the compiler cannot be loaded
- Caching shader blobs for later use
- Returning cached shader blobs
- Deferring compilation to another mechanism, such as a D3D compilation
service running on a host debugging machine *
The above use cases are especially important for Windows Store apps, as
they are not allowed to ship the d3dcompiler. On Windows Phone, where
there is no runtime compiler, this is essential for handling QtQuick apps
which require runtime shader compilation.
* This requires a separate service which monitors a directory for shader
source files, compiles these files into D3D bytecode, and places
the bytecode in the qtd3dcompiler cache directory. This directory is
monitored by qtd3dcompiler, which is then able to then load the blob.
Change-Id: I9889c8d66d2ddbe5a7a1dc44bfe5d8ad229b0e43
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
It's unexpected that all messages generated by the stream version
of qDebug and friends have a trailing space. It also makes switching
to categorized logging (which only supports the stream version) difficult,
since all autotests checking for debug output would have to be adapted.
Task-number: QTBUG-15256
Change-Id: I8d627a8379dc273d9689f5611184f03607b73823
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
When we unlock the mutex, we need to take in account that the Connection
pointed by 'node' may be destroyed in another thread while it is unlocked
Doing 'node->prev = &node' will make sure that 'node' is actually
updated when it is destroyed.
Setting isSlotObject under the mutex is safer and ensure that no other
thread will attempt to deref the object.
The regression was introduced in 5885b8f775
tst_qobjectrace was updated to catch races arising when we are
connecting with function pointers.
Change-Id: Ia0d11ae8df563dad97eb86993a786b579b28cd03
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@digia.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
CGPostMouseEvent is obsolete and known to have "undocumented special
cases and undesirable side effects." The newer Quatz API doesn't
allow neither multiple mouse button events nor preserving the mouse
cursor location.
Change-Id: I121b02fd01e2990488b05e45431cbdc13589656e
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
QAccessibleImageInterface already has an API to return the
size of the image.
This function ensures that their API's are not overlapping.
Alternatively, we could merge both functions into
QAccessibleImageInterface::imageRect(), but the assumption is that
images change position more often than their size.
Change-Id: I55c25cdff187b9f497828f04cfd5f969cfbc451f
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
This patch merges the two classes that were used to make
dock widgets accessible into one.
The title bar does not need to be represented by its own
accessible object.
In addition the buttons on the toolbar are now labelled.
Task-number: QTBUG-33946
Change-Id: Id90d8c09f15ed683e64dbe3f6ac55bca7a0b300f
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
If create() or makeCurrent() calls fail, the test should fail fast.
Change-Id: I025c541f94c8eee492cf0e1545bfbb8027eff2a7
Reviewed-by: Gunnar Sletta <gunnar.sletta@digia.com>
Unfortunately we returned the column description when the AT client
asked for the row description....
Change-Id: I46bc0edb4fd0f7cc6d98d7d6e0d8ca6f77553a26
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
This avoids the assertion in iaccessible2.cpp(510)
Change-Id: I1a4c007ffcbcda70f0e37ef3cf55a303683b58c1
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
There is no point in having QAccessible2, so merge it with the normal
QAccessible. The header will be removed in a subsequent commit as it is
still needed by declarative at this point in time.
Change-Id: I1fc47d484d482f25387eba827bc5a373536b7a8b
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
There was a disagreement between the a11y plugin and QTreeView
whether the horizontal header should have been exposed or not.
When the header was hidden, this resulted in that we sent an event
with a child id that was wrong, or in worst case higher than
QAI::childrenCount(). This was the reason we got the warning
output as described in the task.
With this commit, we consistently *expose* hidden headers both for
QTreeView and QTableView, but ensure that their state().invisible is
set to true instead.
This makes it consistent with how hidden cells are exposed.
This also fixes a bug in QTableViewPrivate::accessibleTable2Index
where we always added 1 to the index, which was spotted while
writing the test.
Task-number: QTBUG-33247
Change-Id: Ifd1f83d56296dd071424fdb81fce7628bc24fe0a
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
During Qt Contributor Summit 2013 we agreed that we will not support
exception safety anymore.
http://comments.gmane.org/gmane.comp.lib.qt.devel/12004
Task-number: QTBUG-32642
Change-Id: If57917fe8af45e787e215431c94579bc86fc7683
Reviewed-by: Simo Fält <simo.falt@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Change the version number to 1.0, and use the public doctype.
Change-Id: I9b071c80c410c31c38813c4447edd7b186226fab
Reviewed-by: Jake Petroules <jake.petroules@petroules.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
We're rolling back exception safety support
Task-number: QTBUG-32642
Change-Id: I25f20b554a93f25d00cca19b3e308d6cc8fe85e2
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
On Mac it's expected that some elements are
filtered out of the a11y hierarchy.
We do this with the shouldBeIgnored function.
The problem is that we would ignore some objects
and then return them in the child attribute function.
This is inconsistent and leads to voice over not working.
For example having a plain QWidget with other widgets as
children would cut off all of these widgets, since the
plain QWidget would be ignored.
Change-Id: I5f6c26b272e5ca57d59c1ed1ef47e9a2b1181295
Reviewed-by: Gabriel de Dietrich <gabriel.dedietrich@digia.com>
Reviewed-by: Jan Arve Sæther <jan-arve.saether@digia.com>
Previously, accessibleTree->child(0) would return an interface for the
header even if it was hidden.
Also, the assertion was wrong since the index would be 0 if both row
and column were 0. The assertion was actually found while using the
project explorer of Qt Creator (2.7)
Change-Id: I9f3cc2c13b6887569d10c4e062a64552f898231a
Reviewed-by: Frederik Gladhorn <frederik.gladhorn@digia.com>
WEC7 does not have time() function. Let's use Qt APIs to query time and
initialize random number generator, Qt APIs have already been adapted
to work on all supported platforms.
In addition use QByteArray instead of QString to avoid unnecessary type
conversions later on with toLatin1().
Change-Id: I1ae3729397b0f4a1bd138022a4e122021e10f1e9
Reviewed-by: Björn Breitmeyer <bjoern.breitmeyer@kdab.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Andreas Holzammer <andreas.holzammer@kdab.com>
