Make stricter alignment requirements for the allocated header
This strict alignment allows reallocateUnaligned() to property account
for the padding occurring in cases when
alignof(QArrayData) < alignof(T) <= alignof(std::max_align_t), which
happens to be the case on e.g. 32-bit platforms with specific alignment
requirements.
This adds 4 bytes (the difference between alignof(std::max_align_t) and
sizeof(QArrayData)) of overhead for QString, QByteArray and certain QLists
on 32-bit systems.
Task-number: QTBUG-90359
Pick-to: 6.0
Change-Id: I8176a4cc79f100ee772b09425e88fe8ff3ae226a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
By not including ASM in the languages for which compiler flags are
adjusted, we can end up with inconsistent optimization and debug
settings between different languages. Make sure they are treated the
same.
Similarly, linker flags for MODULE target types should also have their
debugging and incremental linking flags adjusted, just like EXE and
SHARED targets. When building with MSVC, MODULE targets were having
the relevant flags stripped, but not then replaced with the desired
ones, leaving them empty in typical scenarios. This would primarily
affect plugins, which are built as MODULE rather than SHARED libraries.
Fixes: QTBUG-90237
Change-Id: I648ea74be1654d24cbecc592ce0ca4b59b2ae839
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
By default, use system PCRE2 for Bootstrap library, if not
cross-compiling.
Fixes: QTBUG-90556
Change-Id: I7291927565484073cadacec9a381b54b44ebeaec
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Make the document reflect the current status of the port.
Also link to https://doc.qt.io/qt-6/build-sources.html and
https://wiki.qt.io/Building_Qt_6_from_Git as official sources
on how to build Qt. This document should IMO rather
concentrate on documenting the build system internals.
Pick-to: 6.0
Change-Id: If62cb966b41b7452edb5b63725756916b66affac
Reviewed-by: Craig Scott <craig.scott@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Commit 370324f6e2 changed the call to 'xml' to 'xmlstarlet' because
some Linux distributions do not provide a binary called xml. Anyhow,
the official Windows packages of xmlstarlet only provide 'xml.exe'.
Therefore just check for both.
Pick-to: 5.15 6.0
Change-Id: I4381b256850e4101b760df21f2b0baffb4414620
Reviewed-by: Eric Lemanissier <eric.lemanissier@gmail.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This complements patch a148c7b5d71d244, where languageToCode(),
countryToCode() scriptToCode() methods were introduced, with matching
codeToLanguage(), codeToCountry(), and codeToScript() methods.
This allows us to remove the use of private Qt Core API in Qt Linguist.
[ChangeLog][QtCore][QLocale] Added static codeToLanguage(),
codeToCountry(), codeToScript() methods that convert ISO code strings
to the respective enum values.
Change-Id: If5c0843a718c006ade086a6f74ceb86ac6e0fce4
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Non-modal or window modal dialogs are shown at show(), via AppKit APIs
that are non-blocking. If we want to block execution at this point, we
need to spin our own event loop. The runModal API of NSSavePanel is not
meant to be used for blocking execution for already shown dialogs, but
is reserved for application modal dialogs.
This means we no longer trip over AppKit's understanding of what state
the dialog is in, which would result in the dialog not reporting back
any files. It also allows us to remove the guard for closing dialogs
twice.
We now also correctly close and end the application modal session if
the dialog is closed programmatically using Qt APIs.
Task-number: QTBUG-89959
Fixes: QTBUG-85547
Pick-to: 6.0 5.15
Change-Id: Ida3dc404417789d4823822ecfbf0935591c23878
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
Previously, different threads instantiating the same back-end could
collide in the register, with various unwelcome results.
Give Registry an atomic status flag to track whether it has been
populated or is being destroyed; and protect it with a mutex to ensure
distinct threads do not collide during registration or attempt to
register while the registry is being destroyed.
Document the correct way to instantiate custom backends, and that no
code other than the QCalendar implementation should instantiate the
built-in ones. Instantiators that follow these rules should be safe
from failed registrations, provided they don't pick a name that
conflicts with some other backend. They can also use the recent change
to semantics of registerAlias() to verify that registration *has*
succeeded.
Done-with: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Pick-to: 6.0 5.15
Task-number: QTBUG-88815
Task-number: QTBUG-85692
Fixes: QTBUG-84575
Change-Id: Ie78e700e71d610454152c05cafb38f6f713649ad
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The program to digest the public suffix list required the user to
initially grep away the comments and blank lines. Filtering those
lines out in the code to read the file is trivial, so save the user
one step in the process.
Change-Id: I08f2594fc4236a689c849d42b5446efa9ec2ef7a
Reviewed-by: Dimitrios Apostolou <jimis@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Added for macOS and X11 screens
Task-number: QTBUG-90535
Change-Id: Ifafe7a07ee2abc3c42cd12785db2d7329878375b
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Due to limitations of QByteArray it was not possible to store more than
2^31 bytes. This was fixed in Qt6 so throw away the casts to int in the
postgres plugin
Fixes: QTBUG-79059
Change-Id: I8ae7276a04d4936bcf5ba6c413e3412f6c342ff5
Pick-to: 6.0
Reviewed-by: Robert Szefner <robertsz27@interia.pl>
Reviewed-by: Andy Shaw <andy.shaw@qt.io>
For some use cases, vertical subpixel positioning may be
useful, as it allows you to vertically align text with
other painting primitives. This does impose an overhead,
so we make it opt-int with a render hint on the painter.
Note that this is only supported on Freetype currently.
It might be possible to support on older macOS versions,
prior to Mojave (which has disabled subpixel positioning
entirely), but since it would have limited usefulness
and Freetype is cross-platform anyway, I skipped that.
Note: This drive-by-fixes an issue with subpixel
positioning where glyphs would always be offset by 1/64,
because we added the aliasing offset *after* we had
determined the closest subpixel position. The idea of
this, as far as I can understand, is rather to snap to
nearest subpixel position upwards, not to add an offset
to all glyphs, so it should be added before finding the
correct position. It had a subtle visual effect when
animating the position. It might be that we could get
rid of it entirely, as I haven't been able to reproduce
any issues with that, but I have moved it instead, to
match what I believe the intention was.
[ChangeLog][QtGui][Text] Added render hint flag
QPainter::VerticalSubpixelPositioning which will position
text at subpixel positions vertically whenever supported.
In absence of this, text position will be rounded
vertically as before.
Fixes: QTBUG-35682
Change-Id: I8ce7a72a64e5a0924dac7c244e3e07c2938bfd09
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
This fixes INTEGRITY build error:
cannot access position 53 in array of 49 elements
constexpr const char *begin = func + prefix;
and many "expression must have a constant value" errors.
Change-Id: I0e043028fbcb479c6657b59ea144b4536cc65651
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The fusion style did not elide a very long text in a QComboBox as it's
done in other styles / non-popup mode.
Fixes: QTBUG-86632
Pick-to: 6.0 5.15
Change-Id: I356a61809220d41717a9039d2b33dd706d034941
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
When building a repository != qtbase, like qtactiveqt, QMAKE_PRL_LIBS
contained entries like Qt6Core and Qt6Widgets, meaning the linker line
would contain exactly these arguments.
When building an activeqt example with qmake, the link then failed with
"LNK1181: cannot open input file 'Qt6Widgets.obj'".
The correct entry would have been $$[QT_INSTALL_LIBS]/Qt6Widgets.obj.
Fix this by determining the full path to the import library in the first
step of the prl generation. This enables QtFinishPrlFile.cmake to
recognize Qt6Widgets as Qt module and generate the right QMAKE_PRL_LIBS
entry.
Pick-to: 6.0
Fixes: QTBUG-90520
Change-Id: Id0d9178da0e0dfc3ea4fadbbe8f5900d792ffc84
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
Qt is now always built with CMake.
The "cmake" keyword for QtTest blacklists remains for now. Removal is
tracked in QTBUG-90545.
Change-Id: I0011d56176a07c82698b2eb9aa330e77efa6cd34
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
+ Add Ubuntu-20.04 Github Actions runner
+ Use as many system libraries as possible on all
platforms except Windows
+ Checkout the git repository into a directory named "source",
so that it's possible to checkout more repositories in the future
+ Configure with install prefix set to a temporary path,
it will be used for uploading binary artifacts in the future
+ Configure with -pkg-config instead of -feature-pkg-config;
this was the traditional way and it works for cmake now
+ Split long lines using double quotes instead of backslashes,
makes the yaml file more readable and does not leak
backslashes into the command line
+ Adjust cache to work in multiple branches, provided that the
environment variables at the top of the file are set correctly
+ Do not let homebrew waste time cleaning up
+ Print logfiles in case of failure
+ Typos and other minor fixes.
Pick-to: 6.0
Change-Id: I1ac8afebbdbea50549a334c07b8b3117a823f0fe
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
ccache 4.0 is crashing consistently. Fix is merged upstream, but it's
not released yet:
https: //github.com/ccache/ccache/pull/736
Pick-to: 6.0
Change-Id: I27d154a1ff1ea3afe111bd8f931db58c6f518809
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
+ Specify the ccache cache directory to avoid the defaults, because
ccache >= 4.0 puts the cache directory in system-dependent paths
+ Adjust sloppiness to increase cache hit ratio
+ print compression stats but don't err if `-x` flag is missing
(only available only on ccache >= 4.0)
+ Print ccache config for debugging purposes
+ Enable ccache compression and limit size
Regarding the last point, ccache < v4.0 optionally compresses using the
zlib algorithm, but ccache >= v4.0 uses zstd and compression is enabled
by default. By forcing compression to true, all platforms will need
similar size cache. We have to limit it because the default (5GB) fills
up the cache artifacts size that Github Actions offers.
Pick-to: 6.0
Change-Id: Iea72d08b223c611ee4a4443625647a1f01809b08
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Initialize the offscreen platform plugin with no screens,
create QGuiApplication object.
Not much of the high-dpi related Qt API can be used
in this configuration, but at least Qt should not crash
on startup.
Task-number: QTBUG-71034
Change-Id: I6620843c3bd8b692c5c2419b1ba290e16175ba5b
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
... where it belongs.
To avoid the loop, introduce the drainAndStop() function, which allows
QWindowsPipeReader to flush the pipe itself. It determines the number of
bytes pending and blocks until the remainder of the process output is
received.
Note that the loop in drainOutputPipes() didn't actually have to
interleave the two pipes (because we're presuming that the operations
will finish instantly), so we don't do it now. Also, the code violated
the API contract: 'true' was returned when the 'wrong' channel received
data; this is now fixed as a side effect.
Change-Id: I38ed4861a238e39e793c3716e856e5bfdeed3d74
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
It's a purely command-line program, no need for gui.
Noticed on making a first (abortive) attempt to convert to CMake.
Change-Id: I767f59587d9fe902aec46cd1742576b6bfc84274
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Add a qCompare() overload for qsizetype, preventing it
falling back to the slow code path pre-formatting expected/actual.
toString() should use the correct format from the already present
int-types (see macro TO_STRING_IMPL).
Complements 94aa350621.
Task-number: QTBUG-38890
Change-Id: Ieb8cea7de086141a2c80f93b4c1be01572be96df
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Add '_WINDLL' definition for the MSVC compilers family.
Check the 'FEATURE_optimize_size' value before replacing compiler flags
in the qt_internal_add_optimize_full_flags function. This is required,
because Qt::Core and Qt::Gui modules lost their ability to shrink,
when selecting the appropriate build type or features.
Fixes: QTBUG-89952
Change-Id: I982c25ea84e4793b4006ead0ee516b3f3eb2a054
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Add QRhi APIs to retrieve and reload the contents of the "pipeline
cache".
The only API where there is a true pipeline cache is object is Vulkan
(VkPipelineCache). For OpenGL, the other backend where we support this,
it is simulated with program binaries. The Qt 5 style OpenGL program
binary disk cache continues to work like before, but one has now the
option to do things in a more modern, graphics API agnostic way, that
leads to generating a single blob instead of a large set of files in
some system location, allowing easier "pre-baking" of the cache content.
It is expected that Qt Quick exposes the two new functions in form
if QSG_RHI_ environment variables, thus allowing easy testing and
cache file generation.
As an example for the performance improvements this can give, consider
Vulkan, where we do not have any existing persistent caching mechanism
in place:
Running BenchmarkDemoQt6.exe --scene flythrough --mode demo creates 18
QRhiGraphicsPipeline objects from Qt Quick and Qt Quick 3D.
The total time spent in QRhiGraphicsPipeline::create() during application
startup for these 18 pipelines is 35-40 ms on a given Windows (NVIDIA)
system.
When exporting the pipeline cache contents to a file, and then, in a
subsequent run, reloading the cache contents, this is reduced to 5-7 ms
on the same system, meaning we get a 6-7x improvement.
The generated data is always specific to a given Qt version, RHI
backend, graphics device, and driver version. Much of the implementation
consists of adding and verifying the appropriate header to the blobs
retrieved from the driver, to allow gracefully ignoring data that was
generated with a device or driver that differs from the one used at
run time. This should provide robustness, even if the Vulkan or OpenGL
implementation is for some reason not prepared to identity and reject
incompatible cache/program blobs.
Fixes: QTBUG-90398
Change-Id: I67b197f393562434f372c7b7377f638abab85cb3
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
Commit d0071a4f87 broke static top-level builds:
CMake Error at .../qtbase/lib/cmake/Qt6Gui/Qt6GuiPlugins.cmake:17 (set_property):
set_property can not be used on an ALIAS target.
Fixed by moving the un-aliasing of the target before the set_property
call.
Pick-to: 6.0
Change-Id: I96c731353798a58ff9b3a664f3614ef03c897dcb
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
This is an abstraction for TLS backend and its factory, preparing to transition
to plugin-based design.
Task-number: QTBUG-65922
Change-Id: Ibe810e77fd1b715a6bea66cd3f44312b015ac274
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
This API gives the names of available backends and provides a basic
information about features/protocols supported by those backends.
Also, it has the 'loadBackend' functions which allow to select
a particular backend (which are becoming plugins).
At the moment, the implementation is still 'hardcoded', the
follow-up patch will allow to select different backends in runtime.
Task-number: QTBUG-65922
Change-Id: I05877de9c02857594e76b24d52e7578bdb01df69
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Also move the cmake_minimum_required call to the top of the file,
because the line before already requires some policy settings, which is
what cmake_minimum_required() is establishing.
For the standard Qt build, we make sure to get appropriate policy
settings by calling qt_internal_upgrade_cmake_policies().
Pick-to: 6.0
Change-Id: If97556a9dd00646e83957959d0f9f16916625160
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
Restore holding of a getID() call's return in a QJNIObjectPrivate
eliminated in commit 78cde1bfd94521bbe4972f31a79c959d0990ea77; this
keeps the Java internal object alive for its lifetime, where it might
otherwise be garbage-collected before we're done with it.
Fixes: QTBUG-88610
Pick-to: 5.15 6.0
Change-Id: Id65b059012f7bd3377a6562c1b647feb75a13679
Reviewed-by: BogDan Vatra <bogdan@kdab.com>
...when the platform does not support Vulkan. The version is left
at the default 0 then, so checking for >= 1.0 is wrong.
This allows the test to pass with the offscreen platform plugin.
Change-Id: I5afba8f1e703e4fa0ff41da91d18f5fabfb54868
Reviewed-by: Andreas Buhr <andreas.buhr@qt.io>
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
To optimize certain operations, it can be useful to know whether we are
currently evaluating a binding. For instance, we have properties whose
storage is only alloctaed on-demand when they are set. However, we would
also allocate them if they are used in a binding context, as we would
otherwise not properly track the dependency. Using
isAnyBindingEvaluating in the getter, we can detect this
situation, and avoid the allocation if it returns false.
This API is private for now, as it exposes some internals of the
property system and should be used with care. As it needs to access the
TLS variable, it also has a non-negligible cost.
Change-Id: I373aabee644fe7020b2ffba7d6a0ad9a1e1b4ec0
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
The timeout defaults to 0, give it 5s.
Pick-to: 6.0
Change-Id: I975810a1ecee8bb8b3a3f143f1379a9a09589a40
Reviewed-by: Alex Trotsenko <alex1973tr@gmail.com>
Projects may call find_package(Qt6 ...) multiple times. When enabling
examples and tests, this happens a lot. For a statically built Qt,
for modules that have plugins (e.g. Gui), every time the module's
config file was loaded, it was generating and adding another copy of
the import plugin sources to the module target. These accumulated and
created many duplicates, which in turn blew out generation time and
made the build very inefficient.
This change checks whether the import plugin sources have already been
processed for the module target and ensures they are only added once.
It records its status on the target itself so that both local and
global targets are supported.
Fixes: QTBUG-90465
Change-Id: I1f45b1ee771a933ee755d44f1e983d6d9113dad0
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
The macOS SDKs now follow the same major/minor version scheme as iOS,
where the major version is bumped for each OS release.
Pick-to: 6.0 5.15
Change-Id: I021421a9c58bb1ae64bc0c63b4215f14e4ada358
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Drawing a QPushButton with a menu results in a clipped label text and a
wrong aligned decoration and menu button.
The position of the menu button was calculated wrong (spacing has to be
subtracted since the start position is the right side). Also the text
label alignment was done before it was adjusted for the menu button
which results in an unaligned (in constrast to a pushbutton without a
menu) label and decoration.
Fixes: QTBUG-89619
Change-Id: I872d39ff2d8eb685fe6843e38c7f727868e1ee9a
Reviewed-by: Samuel Gaist <samuel.gaist@idiap.ch>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Funning <huangyub@uniontech.com>
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
iOS used to close the edit menu automatically when
the user tapped on the screen (even for menus shown
explicitly using the UIMenuController API). Apperently
this has now changed (probably as a part of
[UIMenuController setMenuVisible:] being deprecated in
iOS 13). So we now need to hide it explicitly.
Because of this, we no longer close the edit menu if
the user taps outside the input area. This patch will
fix this by updating the logic in QIOSTapRecognizer to
close the menu explicitly.
There are some more details around when to show or hide
the menu, which is commented on at the relevant places
directly in the patch.
Pick-to: 6.0 5.15
Fixes: QTBUG-90332
Change-Id: I336e3f4003817b39c7abf63f963fde1f3ef87c20
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
iOS used to close the edit menu automatically when
the user tapped on the screen (even for menus shown
explicitly using the UIMenuController API). Apperently
this has now changed (probably as a part of
[UIMenuController setMenuVisible:] being deprecated in
iOS 13). So we now need to hide it explicitly instead.
Because of this, the edit menu would be showing together
with the magnifier class while the user was dragging on
any of the handles. This patch will fix this, so that we
close the edit menu explicitly whenever the user starts
dragging on a handle.
Pick-to: 6.0 5.15
Fixes: QTBUG-80298
Change-Id: Iff2032d64db1b582fa7f048c6a1f37ec8a1528af
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
The benchmark simply calls QRegularExpression's public API methods, so
that we can assess how changes to the implication impact performance.
(Its addition is prompted by evaluation of whether adding a move
constructor saves more or less than the resulting need for
d-pointer null-checks costs.)
Task-number: QTBUG-86634
Change-Id: Idef775ef6cf9f9ded3ce7ba5b85e460571d12756
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
The plugins are meant to indicate what they do support, meaning users of
QNetworkInformation can choose to not care about which plugin is used
and rather just request what they want.
Task-number: QTBUG-86966
Change-Id: Ie130e1791250ec2a4470e3ba7081d982654af06c
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
