This helps get better symmetry with the other modules,
and to unify the code paths for both conversion and
retrieving the interface for static types.
Change-Id: Icbd20de2563f36e3de20d826323acd057734abfb
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Always ask QMetaType to convert first before trying the builtin
code in QVariant. That way we can migrate conversions piece by
piece to QMetaType.
Change-Id: I4fd1bad3ef045d37f84f68e748a6357e0ee0c16c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Avoid some atomic operations by using this instead
of type().id().
Change-Id: Ic664794908d3a9fc3686f954b582740b73c052c6
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Remove all the internal members of the union. Instead replace
it with raw storage (uchar[]) aligned to max_align_t.
Place all accesses to the internal members with get<> methods
for consistency.
Change-Id: Icebf46b90c9375aa6ea0b5913b2132608e8c223d
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Replace it with QVariant::Private::get<>().
Change-Id: I801742300d9b15f748bb91fcd6cfa176d2057271
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Adds convenient access to the data from the Private pointer.
data() determines the storage location at run time, get()
at compile time. internalStorage() can be used if we're accessing
one of multiple types, but know that the type is stored internally.
typeId() is an optimization as it allows retrieving the type id
of the metatype without atomic refcounting operations (which
type().id() would be doing).
Change-Id: I39a508c530a1588053248607c8932e501fd474dc
Reviewed-by: Maurice Kalinowski <maurice.kalinowski@qt.io>
Move the type conversions from QVariant::Helper to QMetaType. Only do
this for Qt Gui in a first step.
This makes it possible to completely remove the Handler struct in
QVariant, and now allows QMetaType to also convert Gui types.
Moving the conversion of Core types into QMetaType will require
further work.
Change-Id: I061f789deca1b595d92bb29227eb54b8e71a3ee3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use std::function to register the converter functions
instead of our own handrolled interface.
Change-Id: Ifc1d1d383d21ee8d4239dbc3970c1f31bf0f4037
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
That parameter is duplicating the return value, get rid of it.
Change-Id: I8d6ecee8aca90aecaf08e6d0072d83e9a08ce3d6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Streamline code using if constexpr and remove some unused code paths.
Change-Id: I602acffab4b3e53fab9e2433856f6b7e8210cc60
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This should speed up quite some of the QVariant operations
Change-Id: Ifae2df39e47e2e1ef1d9a947802bdd17fc66aa66
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove the additional indirection through ptr and replace
it with an offset calculation. Get rid of PrivateSharedEx
that was handling certain types differently.
This also fixes the support for overaligned types, by
using the alignment field from QMetaType to determine
the alignment requirements.
Change-Id: Icc6a78eb57f664c7747450578fe58dfee8f2863b
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
There's no point in storing small types with an external
refcount, even if they aren't movable. Simply copying
the type should be faster in pretty much all cases, while
this uses less memory.
Change-Id: I127474f8e3c5fa042f530684f9d5bfccbba134ca
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
There is no need to generate wrapper functions for equals,
lessThan or debugStream for pointer types, as those can
easily be handled by a few lines of code in QMetaType itself.
Change-Id: If79b3bc3a629249c1d17c9e592202f08b59f80ef
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Refactor the methods retrieving data in Q*Iterable so
that we don't return pointers with unclear ownership. Instead,
copy the data into a out pointer provided by the caller.
This also means there is no need for the metatype flags
anymore and we can remove those.
Change-Id: I517de23a8ccfd608585ca00403aca0df2955f14b
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Store a QMetaType, not a meta type id in the classes.
Change-Id: If27a60512a46fa029cc914d65b8cad7f89d7f3b0
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove the old revision of the interface, this is not
required with Qt 6 anymore, as everything is being
recompiled anyway.
Change-Id: I66070c4dc6b5e2a6d22f5a9ebea7688ed38333fe
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
This code is now unused, as streaming operators are
registered automatically.
Change-Id: I0e48944c33a92cf1f2f158fb2dc0ca49256d7938
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Use the builtin support in QMetaType instead.
Change-Id: Ifc0e88719a384aa7fb525652bada22b6f7ee1c45
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
And remove the old manual registration code for those operators.
Add some special handling for long/ulong, as these types could be
streamed as a QVariant so far, but are not directly streamable
through QDataStream.
[ChangeLog][QtCore][QMetaType] The QMetaType::registerStreamOperators()
and QMetaType::registerDebugStreamOperator() methods have been
removed. The streaming operators for a type are now automatically
registered together with the type registration. This implies that the
operators should be visible wherever the type is visible and being used.
[ChangeLog][Behavior Incompatible Changes] Because the QDataStream and
QDebug serialization operators are automatically registered with
QMetaType, the declarations of those functions must be present at any
point where the type is used with QMetaType and QVariant.
Change-Id: I4a0732651b20319af4a8397ff90b848ca4580d99
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
We shouldn't rely on include path containing also /QtCore
Task-number: QTBUG-82615
Change-Id: I0f98f6097865cc5aa2c2fc9a0fabf2c4513b7afe
Reviewed-by: Sona Kurazyan <sona.kurazyan@qt.io>
I'm pretty sure we want to keep supporting statements like
bool ok = connect(~~~);
that is, statements that do use copy initialization and not
direct initialization. As such, QMetaObject::Connection has
to keep using the restricted bool trick and cannot be ported
over to explicit operator bool.
Change-Id: I07a38f7c134686e67b5984aa93fe5cf3201561d7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The debug feature for the CMake build is solely determined by the value of
CMAKE_BUILD_TYPE or CMAKE_CONFIGURATION_TYPES. There is no point in
restricting the autoDetect expression to developer builds or MSVC and
macOS builds.
User-visible result: -DCMAKE_BUILD_TYPE=Debug now actually turns the
'debug' feature on, also for non-developer builds on Linux.
Change-Id: I65c92575c9483bff16beb78b60606c55aeadcece
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Very simple AUTODETECT expressions worked fine, but expressions with
parentheses, for example, were not correctly evaluated.
Change-Id: Ibec4fa55474e149a701f861838fdea41d31beca0
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This configure switch controls the feature 'optimize_size'.
This isn't merely a mapping to CMAKE_BUILD_TYPE=MinSizeRel, because
we potentially want to combine -optimize-size with -force-debug-info,
which maps to CMAKE_BUILD_CONFIG=RelWithDebInfo.
Task-number: QTBUG-85373
Change-Id: I1a9343ebc54816f52e90e9d33ea3df4c99b1ec9f
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Command line options like -debug result in setting the variable
INPUT_debug. INPUT_debug is mapped to the CMake argument
-DCMAKE_BUILD_TYPE=Debug. INPUT_debug also matches the feature
'debug', and the CMake argument -DFEATURE_debug=ON was passed.
Do not pass -DFEATURE_xxx CMake arguments for inputs that are already
handled.
Change-Id: If096dff1c6dd694545c6f671f5f3512a43c3ec50
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
tst_moc uses qt_wrap_cpp, which tries to dispatch to qt6_wrap_cpp
depending on the value of QT_DEFAULT_MAJOR_VERSION. That value is only
set in Qt6CoreConfigExtras.cmake.in, which is not loaded as part of
the qtbase build unless something does find_package(Qt6Core).
Set the QT_DEFAULT_MAJOR_VERSION to 6 before including the
Qt6CoreMacros in the qtbase top-level project.
Change-Id: I7b81d89d965f755e51727007e68771ac3931ac55
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
The -host prefix was added to the 'preliminary' directory rather than
the actual wrapper shell script. This caused it to be overwritten
every time by the cross-compiled qmake binary.
Change-Id: I103811c6aa9181fa701e3e7ecaf828ecdd1e3c90
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
While trying to implement instructions for building examples with
qmake in the CI, an issue has surfaced.
When building examples with CMake with -DBUILD_EXAMPLES=ON in the CI,
the examples are built in-source, aka source dir == build dir.
This means that the header files generated by qdbusxml2cpp will be
placed in the qtbase source dir.
The instructions that try to build examples with qmake build the
examples in a separate build dir after building the examples with CMake.
Unfortunately the qtbase/examples/dbus/remotecontrolledcar/car example
includes the generated DBus adaptor header via a statement like
#include "car_adaptor.h"
and the compiler prefers to pick up the header file from the example
source dir (the one generated by CMake), rather than the one generated
by qmake in the example build dir.
Because CMake's DBus integration uses different flags than qmake's
DBus integration, the generated header file code is not compatible
with the qmake generated cpp file, and the example fails to link when
building with qmake, because it can't find an appropriate constructor
symbol.
In an ideal world, we wouldn't do in-source builds with the CMake
build, but that leads to other issues which I currently don't recall.
To circumvent the issue, adapt the CMake DBus qt6_add_dbus_adaptor
function to allow not passing the problematic '-l' flag by making it
optional. This shouldn't break existing code, but allows us to
generate a compatible header that will be used by qmake and succeed in
linking the example.
Task-number: QTBUG-85986
Change-Id: I06759f79aeb66bb32da7f158f55dd4734c4a9887
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
The code was developed against the narrow-contract version of mid(),
but not updated when mid() became wide-contract.
Change-Id: I038054fb3f5acc0085c48fbf36af13dd14c917b8
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reduce the scopes so that also the result of 1-arg-sliced() can be
called 'sliced'.
Change-Id: Ie156f76838f8650d6926d3c198007aaf12f90734
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QLatin1String::mid() etc were changed from narrow to wide contract,
but the narrow-contract replacements weren't added. This blocks using
the narrow-contract functions in QStringTokenizer.
As a drive-by, Q_REQUIRED_RESULT -> [[nodiscard]] and Q_DECL_CONSTEXPR
-> constexpr. Also centralize most Q_ASSERT()s in a single function,
verify(), in an attempt to reduce the amount of string data generated
from the asserts in assertive builds.
[ChangeLog][QtCore][QLatin1String] Added from(), sliced(), first(n),
last(n) functions.
[ChangeLog][QtCore][QLatin1String] size_type/size() is now qsizetype
(was: int). This makes QLatin1String(ptr, 0) ambiguous now between the
(ptr, ptr) and (ptr, qsizetype) constructors.
Change-Id: Ie195f66ae1974eb0752c058aa9f3b0853ed92477
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Quanifying a parameter passed by value (which is correct for QTime,
there is no need to pass it by reference) as const is pointless.
Change-Id: I86850f65dfac28ffcc4a8633dd6d212259bb33d0
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The serialization code neglected to check against null. Sinze zones
are saved either by IANA ID or in our special OffsetFromUtc format,
representing an invalid zone by a string that cannot possibly be a
valid IANA ID will do.
Fixes: QTBUG-86019
Pick-to: 5.15
Pick-to: 5.12
Change-Id: I6882026403d00f8b254aab34c645f1cf8f9fcc2d
Reviewed-by: Taylor Braun-Jones <taylor@braun-jones.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The read data wasn't encoded into the state correctly.
Change-Id: Ib0a3b50bfeb56968de5c5e8353b28383cb586271
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
