YIELD is ARM's equivalent of x86's PAUSE, available since ARMv6k,
which should be old enough that we can use it unconditionally.
The ARM manual[1] defines the __yield() intrinsic for this, however:
- Clang has __builtin_arm_yield, the naming is compatible with what
GCC would use, so we use that if available, to avoid having to
include even more headers.
- GCC (incl. 12) doesn't have __yield() in arm_acle.h or anywhere
else, and it doesn't seem to have the builtin (yet, one can always
hope), so we use asm() there.
- Windows doesn't have an arm_acle.h header, but the docs[2] say it
should have __yield(). Unfortunately, the docs don't say where
the intrinsic is hiding, but we already include <intrin.h>, and
godbolt[3] confirms that's sufficient to use __yield(). We could use
YieldProcessor(), but we'd need to include qt_windows.h, I guess,
which I'd rather not.
- Integrity doesn't have the arm_acle.h header, we use asm(), here, too.
[1] https://developer.arm.com/documentation/dui0472/k/Compiler-specific-Features/--yield-intrinsic
[2] https://docs.microsoft.com/en-us/cpp/intrinsics/arm-intrinsics?view=msvc-140
[3] https://godbolt.org/z/Evfe79xhE
Pick-to: 6.3
Fixes: QTBUG-103011
Change-Id: Ibf496dfe9949f9a3d24c0dcf75a703c5fdbbb4b8
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
For the WM_SETTINGCHANGE message, lParam sometimes may be NULL [1]
and that will lead to crash without extra safe guard.
Amends commit qtbase/1ed449e168af133184633d174fd7339a13d1d595
[1] https://docs.microsoft.com/en-us/windows/win32/winmsg/wm-settingchange
Change-Id: Ibd1e94b4c1d7882db0719c31a66a5fcc9299c3bd
Reviewed-by: Nodir Temirkhodjaev <nodir.temir@gmail.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
* The rect passed to QToolTip::showText() is in view coordinates,
not in global coordinates.
* Determining this rect in the first place doesn't need calling view->visualRect(index)
The view already did this before calling this method, and stored it in
option.rect, so just use that.
* The widget passed to QToolTip::showText() should be the viewport,
that's what option.rect is relative to (thanks Giuseppe!).
Found these issues when implementing my own tooltip handing (for a subrect
of a delegate) by looking at this code.
Pick-to: 6.3 6.2 5.15
Change-Id: I852e5409def28da98137cd0c4c996083e5e45706
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
The previously defined padding for the tab content was
overwritten to 0 by a more specific css selector. In addition,
code snippets should not have any extra padding, hence the :not(.pre)
css selector.
Pick-to: 6.3
Change-Id: I8f331924c5d01c8971660bb7a5b3aad25e3dee8a
Reviewed-by: Venugopal Shivashankar <Venugopal.Shivashankar@qt.io>
Fix the documentation to show the actual header and module
Fixes: QTBUG-99335
Pick-to: 6.3 6.2
Change-Id: I333a5a95a448fdf3e08801bf794c93b87cd1f81f
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
Reviewed-by: Nicholas Bennett <nicholas.bennett@qt.io>
This removes the hack that creates a copy of bootstraplib sources to
avoid rebuilding the world if a bootstraplib source file changed.
Said hack led to confusing behavior:
- when accidentally editing bootstraplib copies instead of the real
sources
- when header files were changed in a way that were incompatible to the
bootstraplib source copy
This reverts commit 80a8ead08d.
This reverts commit 743bb66744.
The official way to avoid the QTBUG-92269 problem is now to set
QT_HOST_PATH=<host-Qt-installation> and QT_FORCE_FIND_TOOLS=ON.
If you want to build the tools as well, set QT_FORCE_BUILD_TOOLS=ON.
Fixes: QTBUG-92269
Change-Id: I226bf5792f9ca8e7e207dc53e01c2903018d82d3
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The install prefix was determined incorrectly for static relocatable
builds. The reason was that for those builds, QLibraryInfo::path()
returns the application directory, which is <prefix>/bin for qmake and
qtpaths.
Fix this by removing the bin directory part from the installation prefix
that QLibraryInfo returns if qmake/qtpaths are used.
Fixes: QTBUG-102877
Change-Id: I2e9ff96ded0ee2a7802b265741a3cfbe2bf0a4ba
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
The current [basic.life] wording seems to cover the existing code, but
IIRC, older versions of [basic.life] were not so relaxed. In
particular, while not completely pertinent, the second placement new
is awfully similar to http://eel.is/c++draft/ptr.launder#example-1
Just make all of this SEP and use std::optional. That way, the code
gets simpler, too, plus we get rid of the last use of C++23-deprecated
std::aligned_storage.
The reset() before the 2nd emplace() isn't necessary, but, in a test,
it doesn't hurt, either, and keeps code readers from guessing whether
the first-emplaced object's dtor is actually properly run (it is).
Pick-to: 6.3 6.2
Fixes: QTBUG-99122
Change-Id: If31a46f8be3a74499f1176133029d097faf7dfe9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's deprecated in C++23. Just use an explicitly-aligned char array
directly, wrapped in a struct to avoid decays to char*.
Task-number: QTBUG-99122
Pick-to: 6.3 6.2
Change-Id: Ic5807f0161cd9f360baa07464988bc48b9679f64
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's deprecated in C++23. Just use an explicitly-aligned char array
directly, wrapped in a struct to avoid decays to char*.
Task-number: QTBUG-99122
Pick-to: 6.3 6.2
Change-Id: I802761c1af62efa6ffc006b07837a7deed1ea4cc
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Our previous approach of creating a union from individual special
integer bitfields leads to undefined values because only one member of a
union can be active at any given time. Compilers have finally caught up
with us on that and have started removing "no-op" writes to members.
The primary user of the special integer bitfield unions is
qv4compileddata_p.h in qtdeclarative. We want our on-disk format of
QML compilation units to be platform agnostic and space efficient.
Pick-to: 5.15 6.2 6.3
Task-number: QTBUG-99545
Change-Id: I24847bda2c364eb8ba75f074cde2a9bec25ced06
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's a stretch of characters in a QString, so it has to be qsizetype.
This enlarges the Value struct further; created QTBUG-103306 to track
ideas to shrink it again.
Pick-to: 6.3 6.2
Fixes: QTBUG-102465
Change-Id: I753cfd801030c834989452eb5422b2cd72d4ef16
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
... before the port of Value::len to qsizetype.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: I7d3823dc5b39d2afade735cdcf61f77ead89738b
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
The prefix is a part of a name, the length of which is bounded to 4k
in fastScanName(), so qint16 suffices.
The length field also shouldn't stay int, but that's a different
patch, because it's just a relative offset to pos, and so isn't as
easily overflown.
This shrinks the Value struct a tiny bit; created QTBUG-103306 to
track ideas to shrink it further.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: I579815e72501a091360f55e750af63cb4dc5a5a7
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
It's an index into a QString, so it has to be qsizetype.
The length and prefix fields also shouldn't stay ints, but that's a
different patch, because they're just relative offsets to pos, and so
aren't as easily overflown.
This enlarges the Value struct; created QTBUG-103306 to track ideas to
shrink it again.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: Ib318e131420c2c535e26cb03cbf450e031626e64
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
... fixing yet another int/qsizetype mismatch.
This is clearly safe, as the loop body trivially doesn't modifiy the
container we're iterating over.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: I7db65455019cf75663432078d39805be5816cf24
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
We have at least one caller which passes QtContainer::size() as the
length, so don't truncate.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: I8a692bd6a7b75e745c08b072f53b6efe901d5436
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Port function-local variables where local static reasoning dictates
they should have been qsizetype, not int.
Pick-to: 6.3 6.2
Task-number: QTBUG-102465
Change-Id: I0d7b07699b6dd379bce4b44f172815fb6335b1a1
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We don't care about the actual type of oldLineNumber here, we just
need it to be of the same type as lineNumber, to avoid truncation, so
we use auto.
This was already broken in 5.15 (lineNumber was already a qint64).
Pick-to: 6.3 6.2 5.15
Change-Id: I68d7e6e91c8122bf426c9c10e8cc91ff55c61a20
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The variable represents an offset into the QString readBuffer data
member. Given a QString with more than 2Gi of characters, the variable
would overflow before we've consumed all of readBuffer's
contents.
Signed integer overflow is UB, so anything could happen.
Task-number: QTBUG-102465
Pick-to: 6.3 6.2
Change-Id: I05d35a5e7f02f05462b318c86b17fdab7f1afec6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Replace BOOTSTRAP option with the single value CORE_LIBRARY argument
in qt_internal_add_tool and qt_internal_add_executable functions.
The introduced argument now may accept 'Bootstap' and 'None' values.
Use 'Bootstap' to link Qt::Boostrap library instead Qt::Core or 'None'
to avoid any core library linking. This is useful for tools that need
to use the CMake deployment routines, but not require the Qt::Core
functionality.
Task-number: QTBUG-87480
Change-Id: I64a8b17f16ac5fe43c6b385252dc21def0c88d2c
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The operators were exported, but not implemented.
Change-Id: I6c89c1f4b76040d2388b3a10afc6eeeb52638e1b
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
Value 220 is already used in the (poorly, but consistently) placed
GraphicsSceneLeave event type.
Change-Id: I2fd90645475a70380e43e906c5f8459b3119f22f
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Window states and their transitions do not currently work
correctly in wasm (eg. showFullScreen -> showNormal).
In order to fix this, we need to conform to Qt's way of
handling windows and their geometry, which involves
deferring much of the actual geometry changes to when
it actually becomes visible, or else the QWidget code
interferes with what we're trying to do.
Change-Id: I8c7da0be76760363e444dc5dc676036e70863f6e
Fixes: QTBUG-102190
Pick-to: 6.3
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
The `inline` keyword does not control inlining, as we all know (use
Q_ALWAYS_INLINE to force inlining, if required). However, it does
control the emission of the out-of-line copy of an inline function, if
the compiler did inline it where it got used: with the keyword, no copy
is required, which is our objective here.
Furthermore, we compile with -fvisibility-inlines-hidden with
GCC-compatible compilers (all but MSVC and Integrity's), which means
those functions will never be in the ABI of QtCore: they'll either have
been inlined with no out-of-line copy, or that out-of-line emission is
hidden.
Change-Id: If2e0f4b2190341ebaa31fffd16e31584561669f2
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
If a QWindow is embedded into a native view it's effectively the
root window, similar to a top level QWindow, and we need to forward
hover events on behalf of itself and all child windows.
Fixes: QTBUG-103084
Pick-to: 6.3
Change-Id: I1e94b9447fb561c1b02db87738de235c5d12b3a4
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: <stefan.wastl@native-instruments.de>
Apart from being consistent with qtpaths (which uses qVersion()),
this also ensures that Qt Creator loads correct debug helpers for types
like QString when debugging qmake.
As a drive by, remove all QT_VERSION_MAJOR, QT_VERSION_MINOR,
QT_VERSION_PATCH defines which are not used anywhere.
Change-Id: Ibc8f2a6af833e1ec6e6cd6e1937ac3c1ab328555
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
This reverts commit 27c25fc909.
Currently Qt only supports core device of keyboard, not every
real keyboard. Create a new xkb state during
xcb_xkb_new_keyboard_notify_event_t, it will lose the correct
state before the event, for example, when using the second
or later layout than the first one. The new xkb state will
use the first layout. It's difficult to sync the xkb states.
Fixes: QTBUG-102493
Fixes: QTBUG-102640
Task-number: QTBUG-95933
Pick-to: 6.3 6.2 5.15
Change-Id: Iaa5369ff9f5495e194577dcbb8f78303158c9a73
Reviewed-by: Lu YaNing <luyaning@uniontech.com>
Reviewed-by: Zhang Hao <543985125@qq.com>
Reviewed-by: Liang Qi <liang.qi@qt.io>
Reviewed-by: Ilya Fedin <fedin-ilja2010@ya.ru>
We can not alter the value of 'android_sdk' and 'android_ndk' options
when qtbase's package_id is being calculated. We have attempted to
convert the option values (paths to file system) to actual version
strings.
While the conversion itself works the usage of Conan lock files in CI
fails: "Locked options do not match computed options".
Exclude 'android_sdk' and 'android_ndk' from package_id. Instead the
'android_sdk_version' and 'android_ndk_version' will be used for
package_id which affects the binary compatibility.
Pick-to: 6.3
Change-Id: Ia99a69d9b91cebb8dea821fd8b9c6c59091f1a41
Reviewed-by: Toni Saario <toni.saario@qt.io>
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
the generated `-D` and `-I` flags contain quite a number of duplicate
entries. since cmake-3.15 it is easy to strip them via a generator
expression
Change-Id: I83915c9074e0e2e62e79cd932cad2fa07e8a12aa
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This prevents false-negatives and false-positives, as e.g. Clang
10.0.0 masks as GCC 4.2, so Q_CC_GNU is 402 on that
compiler. Depending on the test (Q_CC_GNU > NNN or Q_CC_GNU < NNN),
the result of the test is almost random.
Q_CC_<comp>_ONLY makes sure we match only GCC or MSVC, not bycatch
such as Clang or ICC.
Pick-to: 6.3 6.2 5.15
Change-Id: I4c550a11ecf85fc9a2216b330b69bd03d45b47e0
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
It's not async-signal-safe, so don't use it. Instead, let's use writev()
(which should write atomically and we won't need to have a buffer) and
std::to_chars to format numbers where available (where not, we roll out
our own implementation with divisions).
Pick-to: 6.3
Change-Id: I5ff8e16fcdcb4ffd9ab6fffd16eb70a158f44864
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The std variant is faster and more compact.
Use the QtPrivate wrappers to avoid the Integrity mess.
Change-Id: I4a1b1626d29472af059fee55ca26c31d9522b179
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Don't use QSet::toList(), and fix an obvious typo.
Also, core5compat is now needed due to the use of QRegExp.
Change-Id: I766455996c07d354e97a4ed4939f1774fc449331
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Amends cf52d725156811754cd4e094b0984864795d1d58, after which most
indicator styling was no longer taken into account unless some other
item aspects were styled. Calling the baseStyle to draw the entire item
doesn't call back into the style sheet style for the indicator itself.
The QCommonStyle code that breaks the item up into individual sub
elements cannot be called for each element. E.g. turning off the check
indicator feature changes the layout of the item.
So if the indicator is styled, then we have to draw an otherwise empty
item with the style sheet style, and then clip the already painted rect
before calling the base style to draw text and highlighting with the
correct palette.
Add baseline test for QTreeView with different style sheets.
Fixes: QTBUG-102820
Pick-to: 6.3 6.2
Change-Id: I1da5676cc63556230eac525bc550457ebd495a58
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Use the string literals instead.
Change-Id: Ic85d0845c211744a79af7e48b180129797d3d612
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
Reviewed-by: Assam Boudjelthia <assam.boudjelthia@qt.io>
Move the QtConfigureOption and QtOptionParser to qt-conan-common where
existing shared functionality can be further re-used by these parsers.
Pick-to: 6.3
Task-number: QTIFW-2585
Change-Id: Ief56762f3a27bb0d6d474e8d1668234856afd732
Reviewed-by: Toni Saario <toni.saario@qt.io>
The QWasmCompositor destructor deregisters event handlers, which
means it needs to look up the the canvas element, which again means
that this needs to happen before we clear the canvas from
specialHTMLTargets.
Also, calling destroy() is not needed since the QWasmCompositor
destructor makes this call.
Change-Id: I3004b94d0459c28642d3bd8bf9fe794c9b658477
Reviewed-by: Aleksandr Reviakin <aleksandr.reviakin@qt.io>
Reviewed-by: Lorn Potter <lorn.potter@gmail.com>
With the old value, they were placed above the
header bar. Moving them a few pixels below looks
less broken.
Pick-to: 6.2 6.3 5.15
Change-Id: Iddd9d5e0b5c199fe35a9c8b8a7cac9a472901a21
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
When linking Qt statically, the image fomats are not separated into
different DSOs. when linking them into the same target, gcc warns
about an ODR violation between icohandler and bmphandler:
```
/usr/src/debug/qtbase/6.3.0-r0/build/include/QtGui/6.3.0/QtGui/private/../../../../../../git/src/gui/image/qbmphandler_p.h:69:8: warning: type ‘struct BMP_INFOHDR’ violates the C++ One Definition Rule [-Wodr]
/usr/src/debug/qtbase/6.3.0-r0/git/src/plugins/imageformats/ico/qicohandler.cpp:98: note: a different type is defined in another translation unit
/usr/src/debug/qtbase/6.3.0-r0/build/include/QtGui/6.3.0/QtGui/private/../../../../../../git/src/gui/image/qbmphandler_p.h:70:13: note: the first difference of corresponding definitions is field ‘biSize’
/usr/src/debug/qtbase/6.3.0-r0/git/src/plugins/imageformats/ico/qicohandler.cpp:87: note: a field of same name but different type is defined in another translation unit
/usr/src/debug/qtbase/6.3.0-r0/build/include/QtGui/6.3.0/QtGui/private/../../../../../../git/src/gui/image/qbmphandler_p.h:69:8: note: type ‘qint32’ should match type ‘struct quint32_le’
```
we therefore wrap the `BMP_INFOHDR` into an anonymous namespace
Change-Id: I0e6ee66dd09ebda0c3e399738b5c172a57beb8f4
Pick-to: 5.15 6.2 6.3
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
This reverts commit 043529c9dc, that
was introduced to keep compatibility with Windows XP. This is not
necessary anymore, as support for Windows XP got dropped.
The now removed code also had an issue in the format of the fifth
argument of FormatMessage, that is supposed to be a pointer to a
pointer for FORMAT_MESSAGE_ALLOCATE_BUFFER.
Pick-to: 6.2 6.3
Change-Id: Ib75b6a53a778801388d71388701340d3b79dacce
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Ihor Dutchak <ihor.youw@gmail.com>
As the standard compilers on their respective platforms, other
compilers tend to mask as Clang, GCC, or MSVC, leading to complicated
code when you actually mean just one of these compilers:
#if defined(Q_CC_GNU) && !defined(Q_CC_CLANG) && !defined(Q_CC_INTEL)
This is particularly problematic when combined with version checks:
#if defined(Q_CC_GNU) && Q_CC_GNU >= 900
will, e.g., not match Clang 10.0.0, which masks as GCC 4.2. The
correct way (until a new kid on the block starts to mask as GCC, too)
to check for GCC >= 9.0 atm is:
#if defined(Q_CC_CLANG) || defined(Q_CC_INTEL) || !(defined(Q_CC_GNU) && Q_CC_GNU >= 900)
The new macros make such checks intuitive and hard-to-misuse:
#if defined(Q_CC_GNU_ONLY) && Q_CC_GNU >= 900
Use it in qcompilerdetection.h.
Pick-to: 6.3 6.2 5.15
Change-Id: Idcdf973fbc4708f58ed91c800be3d5e599e5b7e6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
This reverts commit 583668005d, and
provides an alternative fix for QTBUG-75106.
Reason: This introduced QTBUG-103009, due to lack of caching. To fix the
original issue, we still reduce the amount of caching we do, by only
considering the first non-dynamic meta-object for QML related objects.
Task-number: QTBUG-75106
Fixes: QTBUG-103009
Pick-to: 6.3 6.2 5.15
Change-Id: Ic76af26a719d1114208be9555286239c6c6df615
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
