Previously the blacklisting file format only worked with slot:data or
plain slot names for the items to blacklist. However, tests with
global data report themselves with the global data-row tag in the same
way as function-specific ones do; and tests which have both join the
two as slot(global:data) in the test output name, so the reader is apt
to mistake global:data for a data tag. In any case, it is potentially
desirable to be able to blacklist a function with either or both of
global and local data-row tags specified. Add support for that and
remove a blacklisting that was only needed due to the lack of this
support.
For now, make the new parameter to checkBlackLists() optional, so
that qtdeclarative's qmltest framework can adapt to this change.
Fixes: QTBUG-100870
Change-Id: I9125811ebdab75d3fb462ba8b60561f003426502
Reviewed-by: Pasi Petäjäjärvi <pasi.petajajarvi@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
The latter didn't do much aside from call the former, which was
private with no other callers. This prepares the way for som tidy-up.
Change-Id: Iff741662365553e719a9c8b1253aff11de96455f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
By simply forwarding two QByteArray values (as views, since that
suffices) we can retain knowledge of their sizes and thus avoid the
need to null-terminate in order to forward to a hand-rolled strchr()
that depended on that. Using QBAV::contains() turns out to be quite
expensive, so use memchr() to implement a more efficient check.
In the process, turn a pointlessly repeated check into an assertion.
Change-Id: I316623206e1cf315a59d54cf574649d6acbb5c43
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QBA::toPercentEncoding() duplicated a condition to decide whether it
should include a percent-replacement character in its always-encode
list (which might allocate memory), where simply pretesting each
character against the percent achieves the same effect more cheaply.
Change-Id: Ic87a977713207d74b07ee6b811b0ead3b6945ef6
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
When a percent-replacement present in the exclude list would,
normally, not have been encoded, it was encoded regardless, as
otherwise round-tripping fails. However, when percent or its
replacement would normally have been encoded, its presence in the
exclude list lead to it not being encoded, which duly breaks
round-tripping. It also violates
https://datatracker.ietf.org/doc/html/rfc3986#section-2.4
[ChangeLog][Incompatible behavior change][QByteArray] When
converting to percent-encoding, it was possible to exclude the percent
character (or its replacement, unless this is one of the characters
not normally encoded) from conversion. Doing so would violate RFC
3986 Section 2.4 and break round-tripping of any string that contains
the percent character. The percent character, or its replacement, is
now always encoded, even if mentioned in the exclude list, as the
documentation has always claimed.
Change-Id: I885ce33e05dc6877f1b3700fb0870fa30556a5aa
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This was not previously tested.
Change-Id: Icd287b519f6bc5d450f4490990ac78b0d06774f6
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
WebGL is stricter than OpenGL ES and require that unused
vertex attribute arrays are disabled. This prevents
errors such as:
WebGL: INVALID_OPERATION: drawElements: no buffer
is bound to enabled attribute
Pick-to: 6.3
Change-Id: I68384a9f6954b6a1960ba6e8afd1fdbdfefe2335
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
Reviewed-by: Lorn Potter <lorn.potter@gmail.com>
Remove the member function from the API and re-add it as a hidden
friend on qfloat16, where it will be a complete type.
[ChangeLog][QtCore][Potentially Souce-Incompatible Changes] The
qfloat16 QDataStream operators are now hidden friends and only found
by argument-dependent lookup. Previously, these were member functions
on QDataStream.
Fixes: QTBUG-93499
Pick-to: 6.3
Change-Id: Ib3d4df7a3fe3a9d0938f3be8b70b50fef0416262
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
pthread_yield() is a non-posix extension and was deprecated for a long time.
It's been removed recently at least from Fedora 35. Use sched_yield() instead.
Change-Id: Iae47fa09cc89005aa466446149be87e1b673c074
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
for QInputDevice::primaryKeyboard() and
QPointingDevice::primaryPointingDevice().
This also reverts ae9fefe3c8.
Fixes: QTBUG-100790
Pick-to: 6.3
Change-Id: Id02f277db25f823eb29e939e25801325df8e4076
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Shawn Rutledge <shawn.rutledge@qt.io>
Found when retesting the testcase completer.zip from QTBUG-54642
Pick-to: 6.3 6.2 5.15
Change-Id: Id84eefeb3a33dc6d790cfa23755352381cc097a9
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
SDK headers that shipped with older versions of MinGW didn't declare
the IFileOperation COM interface, so the original implementation of
moveToTrash included an inferior fallback that used the Shell APIs.
That fallback was already removed in 6a51ff3f0d,
now remove the #ifdef'ery as well.
Change-Id: Iba44df8267f5c1e3e883368d848eb9be1bd02ac3
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Ievgenii Meshcheriakov <ievgenii.meshcheriakov@qt.io>
And remove their uses.
[ChangeLog][QtCore][Deprecation Notice] Deprecated QString::count()
and QByteArray::count() that take no parameters, to avoid confusion
with the algorithm overloads of the same name. They can be replaced
by size() or length() methods.
Change-Id: I6541e3235ab58cf750d89568d66d3b1d9bbd4a04
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This fixes the detection of font style names for the fonts "Bahnschrift"
and "Alef Bold" as described in Microsoft's documentation:
https://docs.microsoft.com/en-us/typography/opentype/spec/name
Fixes: QTBUG-101609
Fixes: QTBUG-101610
Pick-to: 6.3
Change-Id: I5bd2b72817c07195f1e98d1d924c6f673d9e24e0
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
This makes it easier to debug the font matching behavior.
Task-number: QTBUG-101436
Pick-to: 6.3
Change-Id: Id682384e7d9d669b9e736e00ef22cadfd0b4bfec
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
QPlatformTextureList holds a QRhiTexture instead of GLuint. A
QPlatformBackingStore now optionally can own a QRhi and a
QRhiSwapChain for the associated window. Non-GL rendering must use
this QRhi everywhere, whereas GL (QOpenGLWidget) can choose to still
rely on resource sharing between contexts. A widget tells that it
wants QRhi and the desired configuration in a new virtual function in
QWidgetPrivate returning a QPlatformBackingStoreRhiConfig. This is
evaluated (among a top-level's all children) upon create() before
creating the repaint manager and the QWidgetWindow.
In QOpenGLWidget what do request is obvious: it will request an
OpenGL-based QRhi. QQuickWidget (or a potential future QRhiWidget)
will be more interesting: it needs to honor the standard Qt Quick
env.vars. and QQuickWindow APIs (or, in whatever way the user
configured the QRhiWidget), and so will set up the config struct
accordingly.
In addition, the rhiconfig and surface type is (re)evaluated when
(re)parenting a widget to a new tlw. If needed, this will now trigger
a destroy - create on the tlw. This should be be safe to do in
setParent. When multiple child widgets report an enabled rhiconfig,
the first one (the first child encountered) wins. So e.g. attempting
to have a QOpenGLWidget and a Vulkan-based QQuickWidget in the same
top-level window will fail one of the widgets (it likely won't
render).
RasterGLSurface is no longer used by widgets. Rather, the appropriate
surface type is chosen.
The rhi support in the backingstore is usable without widgets as well.
To make rhiFlush() functional, one needs to call setRhiConfig() after
creating the QBackingStore. (like QWidget does to top-level windows)
Most of the QT_NO_OPENGL ifdefs are eliminated all over the place.
Everything with QRhi is unconditional code at compile time, except the
actual initialization.
Having to plumb the widget tlw's shareContext (or, now, the QRhi)
through QWindowPrivate is no longer needed. The old approach does not
scale: to implement composeAndFlush (now rhiFlush) we need more than
just a QRhi object, and this way we no longer pollute everything
starting from the widget level (QWidget's topextra -> QWidgetWindow ->
QWindowPrivate) just to send data around.
The BackingStoreOpenGLSupport interface and the QtGui - QtOpenGL split
is all gone. Instead, there is a QBackingStoreDefaultCompositor in
QtGui which is what the default implementations of composeAndFlush and
toTexture call. (overriding composeAndFlush and co. f.ex. in eglfs
should continue working mostly as-is, apart from adapting to the
texture list changes and getting the native OpenGL texture id out of
the QRhiTexture)
As QQuickWidget is way too complicated to just port as-is, an rhi
manual test (rhiwidget) is introduced as a first step, in ordewr to
exercise a simple, custom render-to-texture widget that does something
using a (not necessarily OpenGL-backed) QRhi and acts as fully
functional QWidget (modeled after QOpenGLWidget). This can also form
the foundation of a potential future QRhiWidget.
It is also possible to force the QRhi-based flushing always,
regardless of the presence of render-to-texture widgets. To exercise
this, set the env.var. QT_WIDGETS_RHI=1. This picks a
platform-specific default, and can be overridden with
QT_WIDGETS_RHI_BACKEND. (in sync with Qt Quick) This can eventually be
extended to query the platform plugin as well to check if the platform
plugin prefers to always do flushes with a 3D API.
QOpenGLWidget should work like before from the user's perspective, while
internally it has to do some things differently to play nice and prevent
regressions with the new rendering architecture. To exercise this
better, the qopenglwidget example gets a new tab-based view (that could
perhaps replace the example's main window later on?). The openglwidget
manual test is made compatible with Qt 6, and gets a counterpart in form
of the dockedopenglwidget manual test, which is a modified version of
the cube example that features dock widgets. This is relevant in
particular because render-to-texture widgets within a QDockWidget has
its own specific quirks, with logic taking this into account, hence
testing is essential.
For existing applications there are two important consequences with
this patch in place:
- Once the rhi-based composition is enabled, it stays active for the
lifetime of the top-level window.
- Dynamically creating and parenting the first render-to-texture
widget to an already created tlw will destroy and recreate the tlw
(and the underlying window). The visible effects of this depend on the
platform. (e.g. the window may disappear and reappear on some,
whereas with other windowing systems it is not noticeable at all -
this is not really different from similar situtions with reparenting
or when moving windows between screens, so should be acceptable in
practice)
- On iOS raster windows are flushed with Metal (and rhi) from now on
(previously this was through OpenGL by making flush() call
composeAndFlush().
Change-Id: Id05bd0f7a26fa845f8b7ad8eedda3b0e78ab7a4e
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
If a test crashes it might leave local sockets, causing test failures
on subsequent runs due to the socket already being "in use".
Pick-to: 6.3
Change-Id: Ie1107c414f4819026907071c7b8281b2e27b8541
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This made the screen names be non-unique (in my case
“28E850”), which causes the code in handleScreenChanges()
to think they are the same screen.
This reverts commit e9fd1c6aab.
Pick-to: 6.3 6.2
Change-Id: I97fb76aeb66857b4bf9b3c5b4bd6db6024446798
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Reviewed-by: Yuhang Zhao <2546789017@qq.com>
This introduces a new helper function,
qt_internal_add_repo_local_defines and makes use of it in
qt_internal_add_{module,test,executable,benchmark,plugin}. That function
checks whether QT_EXTRA_INTERNAL_TARGET_DEFINES is set. If it is, the
defines listed in there will be aded to all targets passed to the
functions mentioned above.
The intended usage is that QT_EXTRA_INTERNAL_TARGET_DEFINES gets set
in the repository local .cmake.conf. This allows e.g. opting in to
source incompatible changes in leaf modules (as long as those are
guarded by some define).
Pick-to: 6.2 6.3
Fixes: QTBUG-101640
Change-Id: I06c3693ee69f46e95a48de724621f0c97e7cc3a8
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
RFC2397 doesn't explicitly mention it, but references RFC2045, which,
in Section 2, states:
> All media type values, subtype values, and parameter names as
> defined are case-insensitive.
and goes on, in 6.1:
> mechanism := "7bit" / "8bit" / "binary" /
> "quoted-printable" / "base64" /
> ietf-token / x-token
>
> These values are not case sensitive
So regardless of whether "base64" is a parameter name, or a mechanism,
we need to treat it case-insensitively.
Use QLatin1String::endsWith() instead of QByteArray::endsWith(),
because the former takes Qt::CaseInsensitive while the latter would
need a toLower().
Add a test.
As a drive-by, use the same trick for the existing case-insensitive
comparison with "charset".
As a further drive-by, fix inappropriate uses of QLatin1String (=
where they don't prevent allocations).
[ChangeLog][QtCore][QUrl] Now recognizes the ";base64" marker in
"data:" URLs case-insensitively.
Pick-to: 6.3 6.2
Change-Id: Ife6ba771553aaad3b7c119c1fa631f41ffa8f590
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Destructors of polymorphic classes should be out-of-line to pin vtable
and type_info objects to a single TU.
Pick-to: 6.3
Task-number: QTBUG-45582
Change-Id: I541437a3ff53852527f6278355f7b549e62e17f7
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We want to mark the corresponding QColor ctor(s) explicit.
Use Qt::GlobalColor or the new QColor::fromString() instead.
Change-Id: I68bf75a094e6821b97682de5a0ffd975834d22d0
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
... from qstringalgorithms.h and qutf8stringview.h, in order to
centralize the declaration of the string classes, much like
qcontainerfwd.h does for Qt containers.
[ChangeLog][QtCore] Added header qstringfwd.h containing
forward-declarations of all Qt string classes.
Task-number: QTBUG-98434
Change-Id: I4cf1ede9fe6c286230f4c7e7abe379da28ce5d15
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Anton Kudryavtsev <antkudr@mail.ru>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Nothing serious, just was curious and played around a bit, so might as
well merge it.
There's much more here that could be done, but I lack the time to do
it.
Change-Id: Idd4ea09a7a413ed5911ffe16841f3d5c944ec1a5
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
... because we can.
Change-Id: I3a813fa9c33e47d20a00e6a331cbbcf6f85e63e7
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Use qOffsetStringArray() instead of an array of pairs of pointers.
This unearthed a problem with SIGNAL and SLOT calling qFlagLocation()
(meaning in debug mode, the array was runtime-initialized, not just
relocated), which we work around by using the new QT_STRINGIFY_
{SIGNAL,SLOT} macros now.
Saves 24 relocations, but, interestingly, saves only 4b in text size
on GCC 11.2 Linux AMD64 -O2 C++20 builds (TEXT and DATA combined)
while saving 760b on the equivalent Clang 10 libc++ build (expected,
in both cases: 24 * (sizeof(void*) - sizeof(quint16))):
Clang:
$ size qtbase/lib/libQt6Core.so.6.4.0
text data bss dec hex filename
5476156 90520 16185 5582861 55300d qtbase/lib/libQt6Core.so.6.4.0
$ ~/bin/relinfo.pl qtbase/lib/libQt6Core.so.6.4.0
qtbase/lib/libQt6Core.so.6.4.0: 6528 relocations, 3598 relative (55%), 2540 PLT entries, 2227 for local syms (87%), 0 users
$ ninja libQt6Core.so
[6/6] Creating library symlink qtbase/lib/libQt6Core.so.6 qtbase/lib/libQt6Core.so
$ size qtbase/lib/libQt6Core.so.6.4.0
text data bss dec hex filename
5475604 90312 16185 5582101 552d15 qtbase/lib/libQt6Core.so.6.4.0
$ ~/bin/relinfo.pl qtbase/lib/libQt6Core.so.6.4.0
qtbase/lib/libQt6Core.so.6.4.0: 6504 relocations, 3574 relative (54%), 2540 PLT entries, 2227 for local syms (87%), 0 users
GCC:
$ ~/bin/relinfo.pl qtbase/lib/libQt6Core.so.6.4.0
qtbase/lib/libQt6Core.so.6.4.0: 6300 relocations, 5343 relative (84%), 318 PLT entries, 1 for local syms (0%), 0 users
$ size qtbase/lib/libQt6Core.so.6.4.0
text data bss dec hex filename
6019871 75896 16952 6112719 5d45cf qtbase/lib/libQt6Core.so.6.4.0
$ ninja libQt6Core.so
[6/6] Creating library symlink qtbase/lib/libQt6Core.so.6 qtbase/lib/libQt6Core.so
$ ~/bin/relinfo.pl qtbase/lib/libQt6Core.so.6.4.0
qtbase/lib/libQt6Core.so.6.4.0: 6276 relocations, 5319 relative (84%), 318 PLT entries, 1 for local syms (0%), 0 users
$ size qtbase/lib/libQt6Core.so.6.4.0
text data bss dec hex filename
6020091 75672 16952 6112715 5d45cb qtbase/lib/libQt6Core.so.6.4.0
Pick-to: 6.3
Change-Id: I60749953f1a63d23d696a5a547cd924ec259ead3
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Factor common code into a separate macro to DRY.
The immediate paractical consequence is that we now have a way to
reliably get a const char[] from the SIGNAL or SLOT macros, e.g. for
storing in a QOffsetStringArray, without causing compile errors in
debug mode or running the danger of causing runtime initialization via
qFlagLocation, which is of limited value in this circumstance, anyway,
because qFlagLocation only stores the last two strings.
Pick-to: 6.3
Change-Id: I67401858e94eedc1200fdd08e695fd56d10f8560
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
QDoc interprets a sequence of special characters as something to auto-
link to; Escape the underscore character to prevent this.
Fixes: QTBUG-100562
Pick-to: 6.3
Change-Id: I61a10d7ea6878ef93197b0d329e16b8867d8ad30
Reviewed-by: Venugopal Shivashankar <Venugopal.Shivashankar@qt.io>
ANDROID_NATIVE_API_LEVEL is an alias for ANDROID_PLATFORM and the
Android's CMake docs [1] uses directly ANDROID_PLATFORM so let's use
that as well. Also, NDK r23b seems to have removed the part of code from
android.toolchain.cmake that handles ANDROID_NATIVE_API_LEVEL to set
the correct value to ANDROID_PLATFORM.
With this change, CMake will pass the value from the configure argument
-android-ndk-platform as -DANDROID_PLATFORM instead of
-DANDROID_NATIVE_API_LEVEL. Otherwise, if if -DANDROID_NATIVE_API_LEVEL
is passed directly to CMake, it should work as before.
[1] https://developer.android.com/ndk/guides/cmake#build-command
Pick-to: 6.3
Task-number: QTQAINFRA-4837
Change-Id: I5c21af53ac91e11a27c4b033313d22d1115c1abc
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
When a QDockWidget's dock areas are restricted by setAllowedAreas(...)
and a second QDockWidget is hovered over it, the first QDockWidget can
be docked to any dock area of the main window. Area restrictions will
be ignored.
That is due to the first QDockWidget being implicitely mutated into a
QDockWidgetGroupWindow upon hovering. By definition, the latter has no
docking restricitons.
This fix adds a check if a QDockWidgetGroupWindow has a single QDockWidget child.
In that case, the single child's area permissions will restrict docking.
Fixes: QTBUG-100670
Pick-to: 6.3 6.2 5.15
Change-Id: I903b074739953791634f482c9cf4b9a95a1d93d3
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Use "The default value is \c xxx" instead of "Default \c xxx".
Pick-to: 6.2 6.3
Change-Id: Id75b45499c7fc0e9d42e66fa2b8b3515c863c61c
Reviewed-by: Nicholas Bennett <nicholas.bennett@qt.io>
It broke drive-less (no C:\ prefix) paths passed to configure.
Invoking configure on Windows with the following args
qtbase/configure --
-DCMAKE_INSTALL_PREFIX=\Users\qt\work\install
called CMake with
-DCMAKE_INSTALL_PREFIX=\\Users\\qt\\work\\install
saying
Qt will be installed into '//Users/qt/work/install'
and while the build succeeded, installation would fail with
CMake Error at cmake_install.cmake:41 (file):
file cannot create directory:
//Users/qt/work/install/lib/cmake/Qt6BuildInternals. Maybe need
administrative privileges.
Note the double slash in the beginning is likely interpreted as a
Windows share URI / UNC path.
The same would happen when passing -prefix '\Users\qt\work\install'
As a reminder, we want to support drive-less prefix paths because
that's what Qt's Coin CI uses passes to ensure DESTDIR installation
works correctly.
Amends cb7f4030bc
Pick-to: 6.2 6.3
Fixes: QTBUG-94366
Change-Id: I9267b6f784babfbdaeafc65267ba96c75fa24112
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Previously we used sh $* for passing parameters of various calls like
cmake / configure. This did not properly preserve quoted arguments,
leading to potential issues.
For example when configuring qtbase with
cmake -DQT_QMAKE_DEVICE_OPTIONS="DISTRO_OPTS=hard-float boot2qt"
the 'boot2qt' option would be passed as a separate argument to cmake,
leading to the QT_QMAKE_DEVICE_OPTIONS cache var containing only
'DISTRO_OPTS=hard-float'.
Use "$@" to preserve the quotes.
Inspired by
https://wiki.bash-hackers.org/scripting/posparams#all_positional_parameters
Amends e9b8837c31
Pick-to: 6.2 6.3
Change-Id: Ib67d4bc138c65f7430457cb2a2baa55bff72e549
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
The -mno-direct-extern-access tells the compiler and linker that
references to symbols outside this ELF module mustn't be direct and must
instead always go through the GOT or PLT (the PLT can additionally be
disabled with -fno-plt). The ELF protected visibility tells the compiler
and linker that this symbol is present in the dynamic symbol table as an
export, but it cannot be interposed by another ELF module.
This option is required for user code to link properly to Qt, otherwise
they will get linker errors (assuming GNU binutils >= 2.39) or runtime
failures (glibc >= 2.35). Both versions of glibc and binutils are older
than GCC 12, so it's a safe assumption they are in use and downgrading
the toolchain or libc is not supported. Adding this option to the
compilation is assured for CMake and qmake-based projects.
For example, all accessess to QCoreApplication::self in QtCore, after
this change and with GCC 12 are relocation-free and direct:
000000000013ebf0 <QCoreApplicationPrivate::checkInstance(char const*)>:
13ebf0: cmpq $0x0,0x4f73d0(%rip) # 635fc8 <QCoreApplication::self>
13ebf8: setne %al
13ebfb: je a90fe <QCoreApplicationPrivate::checkInstance(char const*) [clone .cold]>
13ec01: ret
Meanwhile, accesses to the same variable in other modules are indirect
via the GOT:
66650: mov 0x876e1(%rip),%rax # edd38 <QCoreApplication::self@Qt_6>
66657: cmpq $0x0,(%rax)
This replaces the -Bsymbolic and -Bsymbolic-functions (broken)
functionality that Qt has been using or attempting to use since ~2006.
See https://gitlab.com/x86-psABIs/x86-64-ABI/-/issues/8#note_606975128
Change-Id: Iad4b0a3e5c06570b9f5f571b26ed564aa0811e47
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
