The test expects a very explicit list of events during show, but on
macOS we also get an InputMethodQuery event as a result of the window
becoming active. The test needs to be written significantly to support
these kind of platform differences.
Change-Id: I395c1e9e4e9baf7d9f88f0d067586fc15afb9a16
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
For some reason the firstChild gets the focus when clearing the focus.
This seems to be timing dependent, as removing the 30ms qWait 'fixes'
the issue. So does a processEvent call before minimzing. Both of these
require further investigation.
Change-Id: I62833a5541712f97dc24bc63384fa4c051096537
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
There's a timing issue that affects the position of the vertical
scrollbar when scrolling by pixels.
Change-Id: I29d73574785be539a5870b498a902b1aba887e9c
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
It's not clear why this is failing, but we need to blacklist it so that
we can move over to testing macOS 10.14 and 10.15 in the CI.
Task-number: QTBUG-75786
Change-Id: I208d5af92406c5da8d0210e0188568466b78b2a9
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
The combobox popup can overlap a little with the button, and that's the
expected behavior.
Change-Id: I245bfce85cb5ee661ceb51dbe0d844492878a2bc
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
We were being inconsistent in how we handled this, some tests skipping
while others using QVERIFY. It makes more sense to skip the tests, since
the problem is a missing pre-condition of the test, not the test itself
being bad or exposing real failures in the implementation.
Change-Id: I20eacfe12dbce0b0d926e48cbe2d2772819fa4a5
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
The logic is now mostly handled in QGuiApplication, with QApplication
only dealing with the widget-specific palettes and interaction between
the style and the palette.
The application now picks up changes to the platform theme and will
re-resolve the current application palette appropriately. This also
works even if an explicit application palette has been set, in which
case any missing roles are filled in by the theme.
The palette can now also be reset back to the default application
palette that's fully based on the theme, by passing in the default
constructed palette (or any palette that doesn't have any roles set).
This is also correctly reflected in the Qt::AA_SetPalette attribute.
Conceptually this means QGuiApplication and QApplication follow the
same behavior as QWidget, where the palette falls back to a base or
inherited palette for roles that are not set, in this case the theme.
Behavior-wise this means that the default application palette of the
application does not have any roles set, but clients should not have
relied on this, nor does QWidget rely on that internally.
It also means that setting a palette on the application and then
getting it back again will not produce the same palette as set,
since the palette was resolved against the theme in the meantime.
This is the same behavior as for QWidget, and although it's a
behavior change it's one towards a more sane behavior, so we
accept it.
[ChangeLog] Application palettes are now resolved against the platform's
theme palette, the same way widget palettes are resolved against their
parents, and the application palette. This means the application palette
reflected through QGuiApplication::palette() may not be exactly the same
palette as set via QGuiApplication::setPalette().
Change-Id: I76b99fcd27285e564899548349aa2a5713e5965d
Reviewed-by: Vitaly Fanaskov <vitaly.fanaskov@qt.io>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
The most common mouse wheel movement corresponds to angleDelta().y().
We have previously allowed the user to use either wheel to scroll either
a horizontal or a vertical scrollbar when the mouse is hovering over it;
but 7d29807296 changed it so that the
vertical mouse wheel could no longer scroll a horizontal scrollbar.
The behavior is now restored as it was in
59cc316620.
Task-number: QTBUG-81007
Change-Id: Ieacdce539d5311499a86af645bbe0d5098e16be6
Reviewed-by: Christian Ehrlicher <ch.ehrlicher@gmx.de>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
...but keep support for deserializing for all older versions in order to play
nice with existing .qsb files.
The usage of binary JSON and then CBOR is a historical artifact: relying
on the QJsonDocument (which we generate for purposes unrelated to binary
serialization) was a convenient shortcut. However, writing to and
reading from a QDataStream instead (which QShader already does) is trivial.
In order not to be limited by potential CBOR requirements in the future,
take it all into our own hands.
Extend the qshader autotest accordingly.
Task-number: QTBUG-81298
Change-Id: If0047b659bd6601ca47b5bbbce1b719630cde01e
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
The new float16texture_with_compute manual test demonstrates a case
that was not handled correctly before: multiple dispatch() calls in
a row storing to and then loading from the same subresource(s).
Without the appropriate barriers subtle data corruption issues appear.
For Vulkan this also adds better batching for image/buffer barriers when
using deferred recording.
Also, for OpenGL this fixes the case of updating a buffer or rendering into
a texture and then using it for load/store in a compute pass (previously this
also lacked an appropriate glMemoryBarrier).
Task-number: QTBUG-81217
Change-Id: I7970c445564473f9452662f4b1a20618cb8627a3
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
What's more, demonstrate two types of rendering to a cubemap:
one by one to each face, and by attaching all faces as color
attachments in one go.
Both are used by Qt Quick 3D in connection with shadows, so this
proves that the same is possible to implement with QRhi.
Task-number: QTBUG-81261
Change-Id: I5c7077224d7cae0dd6ea02ac30a9e6f9f1f0c229
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
Uses the two compute shaders from Qt Quick 3D. Demonstrates
and tests both RGBA16F textures and using them (and doing
load/store with mip levels individually) in combination with
compute.
Task-number: QTBUG-81213
Change-Id: I3f0f250d5997a26c857b7c45517684c63b44e58e
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
When not doing fat builds, QMAKE_MAC_SDK_PATH defaults
to iphoneos sysroot, which breaks compilation.
This fixes compilation when user sets also CONFIG-=device
Change-Id: I1d18269946cd2a5608e26d943159c824f31db09a
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Doing only debug build is not supported at the moment.
Add error to configure.
Task-number: QTBUG-71990
Change-Id: If35d00c3f09d733efe63d5ea66d07e4a3ec922cc
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
It has been a QPoint for a long time now, to support both the usual
vertical mouse wheel and also the wheel tilt for horizontal scrolling,
or the actual horizontal wheel if the mouse has one, or the simulation
of a horizontal wheel via touchpad gestures; but the docs continued to
read as if it was just one value.
Task-number: QTBUG-71575
Change-Id: I3efa686ace4f09c7f237f72bf0500fbfbd3213cb
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Based on Glen Johnson's patch, but rearranged to avoid repetition.
Fixes: QTBUG-81248
Change-Id: I9c23ab233ebd5514bc05fd155999597ada7295ef
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The call to qRegisterMetaType() has been removed since the type
is registered as a meta type in Qt 5. Assert on the meta type id instead.
Adapt the documentation accordingly.
Fixes: QTBUG-81254
Change-Id: I0cb459d0dda7a82fc37871605ff634004af0f9f9
Reviewed-by: Paul Wicking <paul.wicking@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Because MSVC warns. Added by me on commit.
c496fee2a5
Fixes: QTBUG-81310
Change-Id: I596aec77785a4e4e84d5fffd15e93a8e367e035e
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
An empty QTextDocument already contains a block; so when the formatting
is fully determined, if the document is still empty, then instead of
inserting a new block, we can set formatting on the cursor, which
affects the pre-existing block, before inserting text. This avoids
leaving a blank line (the default block) above the inserted content.
Fixes: QTBUG-81060
Change-Id: I14e45e300a602493aa59680417d74d4c2b25862d
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
You cannot be sure that property with specified key in
a global object is really a canvas.
Should use `document.getElementById`.
Change-Id: Ife55adaad5517aed64122b0c9bff32489cf19a2f
Reviewed-by: Lorn Potter <lorn.potter@gmail.com>
The standard makes them as big as int_least16_t / 32_t, i.e. big
enough to hold UTF-16 / UTF-32 code units. We're relying on them
to be *exactly* 16/32 bit instead.
Drive by: move the check that a char is 8 bits before saying that
some other type is not its expected sizeof() * 8.
Change-Id: Idbf3d33331667d417702d9dde221905bcfc4d021
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Minimal fix for the missing wakeup issue, in leu of
a new event loop implementation.
So far, emscripten_async_run_in_main_runtime_thread_
looks to be our option for scheduling calls on the
main thread. This function is available on emsdk 1.38.22
and higher.
Requires making from QEventDispatcherUNIX::wakeUp()
non-final. The future event dispatcher implementation
will not inherit QEventDispatcherUNIX, so this is a
temporary change.
Fixes: QTBUG-75793
Task-number: QTBUG-76007
Change-Id: Ie6f6ee6f7674206fc0673a4fe866ac614432ab65
Reviewed-by: Lorn Potter <lorn.potter@gmail.com>
Use QStringIterator rather than indexed loops. This fixes handling of
non-BMP code points (which may be lower or uppercase, see the test).
Change also the semantics of the functions, adopting Unicode §3.13
definitions: a string is lowercase/uppercase if it's equal to its
own toLower/toUpper folding.
As a side effect, empty strings are now correctly reported to be
lowercase AND uppercase.
[ChangeLog][Important Behavior Changes] The semantics of
QString::isLower() and QString::isUpper() have been changed to match the
Unicode specification. Now lowercase (resp. uppercase) strings are
allowed to contain any character; a string is considered lowercase
(resp. uppercase) if it's equal to its own toLower() (resp. toUpper())
folding. Previously, a non-letter character would make the string not
lowercase nor uppercase, and the mere presence of an uppercase (resp.
lowercase) letter would make isLower() (resp. isUpper()) return false,
even if the letter wouldn't change under case folding. As a
consequence, now empty strings are lowercase and uppercase.
[ChangeLog][QtCore][QString] Fixed a number of bugs of
QString::isLower() and QString::isUpper(). Empty strings are now
correctly reported to be lowercase (resp. uppercase), and strings
containing code points outside the BMP are now correctly handled.
Note that the behavior of these functions has also been changed.
Change-Id: Iba1398279a072399a9f21295fe75f6e414f3f813
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Sort alphabetically and add recent relevant options with version
information.
Change-Id: I10c8cc82ce357775ed68cb811a0c906cd38633a5
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
To disambiguate &QSslSocket::sslErrors() expression. Add a new
getter - sslHandshakeErrors().
[ChangeLog][Deprecation Notice] QSslSocket::sslErrors() (the getter) was deprecated and superseded by sslHandshakeErrors()
Task-number: QTBUG-80369
Change-Id: I9dcca3c8499800c122db230753dc19b07654f8a2
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Binary Json is deprecated and we should not use it. We can already
serialize shaders to CBOR. Push the deprecation warning mechanism to the
only user of fromBinaryJson().
Change-Id: I1d56157ab92f74eaab49400be647317e7833e35e
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
QLocalSocket::error is overloaded as a signal and an accessor (for the
error reported by the signal). This means connecting to the signal using
a pointer to member function would require ambiguity resolution. We
deprecate the old accessor (to be removed in Qt 6) and introduce a new one -
'socketError'.
[ChangeLog][Deprecation Notice] QLocalSocket::error() (the getter) is deprecated; superseded by socketError().
Task-number: QTBUG-80369
Change-Id: Iab346f7b4cd1024dee9e5ef71b4b7e09f6d95b12
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
The one that is a getter for the last error found. This is to disambiguate
the expression '&QAbstractSocket::error'. Introduce a new member-function
socketError as a replacement.
[ChangeLog][Deprecation Notice] QAbstractSocket::error() (the getter) is deprecated; superseded by socketError().
Task-number: QTBUG-80369
Change-Id: Ia2e3d108657aaa7929ab0810babe2ede309740ba
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Clang does seem to understand __declspec(nothrow) as
used by the COM macros like STDMETHOD. Suppress the warning,
fixing errors like:
qwindowsfontenginedirectwrite.cpp(105,24): error: 'AddBeziers' is missing exception specification '__attribute__((nothrow))' [-Werror,-Wmicrosoft-exception-spec]
Task-number: QTBUG-63512
Change-Id: If582cb0c12c62a7d12c4ae702747aac1f735db3c
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
The windows filesystemwatcher did not watch for attribute changes for
directories (e.g. hidden flag) so it was not in sync with other
backends. Fix it by adding FILE_NOTIFY_CHANGE_ATTRIBUTES to the watch
flags when watching a directory.
[ChangeLog][QtCore][QFileSystemWatcher] Fixed a bug that caused QFSW not
to watch for attribute changes on Windows. Now it will correctly report
when files and directories become hidden or unhidden, for example.
Fixes: QTBUG-80545
Change-Id: I31767a0da899963e3940b4f5b36d1d581e6aa57c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
A direct logical replacement. Not sure what is going on though or why
an inversion is necessary, but logic is unchanged.
Change-Id: Id9b5531895371f6467018fa82336aff6238ae126
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Eirik Aavitsland <eirik.aavitsland@qt.io>
If the user derived from QProcess, it's likely to override the
setupChildProcess() virtual. Unlike fork(), the system calls for
enhenced functionality (FreeBSD pdfork() and Linux's clone()) won't call
pthread_atfork() callbacks and the environment in the child process
could be inconsistent. Qt's own code is safe, but we can't make the same
statement about users' code.
So we err on the safe side for correctness and run the old, less safe
implementation (thread-unsafe, subject to hijacking of the signal
handler, etc.).
Change-Id: Ia2aa807ffa8a4c798425fffd15d9703bd03f6f09
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The detection code for Linux 5.2 and 5.3 went unused. We had it in the
initial commit so it got recorded for posterity, in case someone wants
to detect them (the road not taken).
Change-Id: Ib5d667bf77a740c28d2efffd15cccdff5aa5a622
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Linux 5.2 added the CLONE_PIDFD flag to the clone(2) system call. Linux
5.3 also added the clone3(2) system call, which allows both CLONE_PIDFD
and CLONE_PARENT_SETTID, which we could use but don't really need.
Unfortunately, 5.2 missed one crucial and one convenient change. Without
the P_PIDFD support in waitid(2), we would need to either remember the
PID of the child process or obtain it from /proc. But without support
for polling on pidfds, we can't find out when the child process
exited. Support for the latter was added in 5.3, which would be
sufficient for us, but it's not easy to detect it at runtime. The
support is also being backported to Android's 4.9 kernel[1], so we can't
do version checks, only functionality.
This commit relies on waitid(2) support, which was the last to be added
and is fortunately the easiest to detect. If it's present, we assume the
whole series is present and use pidfds.
[1] https://android-review.googlesource.com/q/topic:%22pidfd+polling+support+4.9+backport%22
See-Also: https://lkml.org/lkml/2015/3/12/1044 (original idea)
See-Also: https://lkml.org/lkml/2019/4/19/441 (CLONE_PIDFD)
See-Also: https://lkml.org/lkml/2019/4/25/884 (poll(2) support)
See-Also: https://lkml.org/lkml/2019/7/27/334 (P_PIDFD)
Change-Id: Ia0aac2f09e9245339951ffff13c94b3aa13d8b63
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Can be relevant for Qt Quick 3D shadows, where the shadow map is R16F.
Task-number: QTBUG-81268
Change-Id: Ic33e100929e133d1cbe0b062a15697c82536f62a
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
Checking for nullptr is insufficient: just because there
is an empty map present, it does not mean it is valid. The
two cases must be handled identically.
This fixes a regression when using QShaders that do not
have an associated native resource binding map.
Amends 4639660ded
Change-Id: Icb239bf9a9261ed32f2cb7b22c60b608195618fc
Reviewed-by: Paul Olav Tvete <paul.tvete@qt.io>
