proEval(), proParser(): the loop was iterating over a temporary so it
couldn't have changed it. Hold the container in a const auto variable
and use ranged-for.
formatValue(): iterating over a const QList& parameter, and the container
isn't changed at the call sites during iterating, so Q_FOREACH wasn't
needed to beging with. Use ranged-for instead.
Task-number: QTBUG-115839
Change-Id: Idabe0bbd84b5bcc86cef275f80497651353a4d9e
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
systemEnvironment(): the loop was iterating over a temporary; hold the
temporary in a local const auto variable and use ranged-for.
runCommand(): the container is a const&, the loop doesn't change the
container and the containers the method is called on aren't changed
during iteration.
Task-number: QTBUG-115839
Change-Id: I6687e5ff64ff8c2fa26e34abf4044b98718e65d4
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
specifyMetaTagsFromCmdline(): the loop was iterating over a temporary so
it couldn't have modified it; hold it with a const auto variable and use
ranged-for.
relatedMetaObjectsNameConflict_data(): make the container const and port
to ranged-for.
Task-number: QTBUG-115839
Change-Id: I6a5afdf0e5a3dd47818da0025fbbeacd05335b39
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The loop was iterating over a temporary container, so it couldn't have
changed it. Use a const auto variable to hold the container and port to
ranged-for
Task-number: QTBUG-115839
Change-Id: I66e7cbdb811666ca352cdf064b1228caa346d876
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
badString(): const'ify the static QList, this is both faster as the
compiler doesn't need to check if it has been already initialized, and
it means we can use it directly in ranged-for as the method returns
const QList&.
Drive-by change: don't go the long way around to get a const char*:
qPrintable(QString("fail %1").arg(ba))
instead use:
QTest::addRow("fail %s", ba.constData())
(thanks to dfaure for pointing it out in review).
Task-number: QTBUG-115839
Change-Id: I6e8efa6df47ee94f1d71a63e22ab121647e6bf20
Reviewed-by: David Faure <david.faure@kdab.com>
The container is local to the function, but can't be made const due to
the way it's filled. The loop clearly doesn't modify the container so
use std::as_const and ranged-for.
Task-number: QTBUG-115839
Change-Id: Ia9f01dfaccfca3225fe0487aafd0a386605cf466
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
These are member containers of nested (in the test function) structs.
It's clear the container isn't modified in the loop body, so use
ranged-for and std::as_const.
Remove "#undef QT_NO_FOREACH".
Task-number: QTBUG-115839
Change-Id: I0588bf4b6520b42d6d8678d702192fb894956b05
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
This is a local container that isn't modified in the loop, so use a
ranged-for loop with std::as_const.
Task-number: QTBUG-115839
Change-Id: Ie9129e065f8ae9bd8c93cf95093a77529aef0803
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Use std::initializer_list/std::array for data known at compile time.
In files where Q_FOREACH isn't used any more, remove
"#undef QT_NO_FOREACH".
Drive-by change: de-duplicate some trivial code.
Task-number: QTBUG-115839
Change-Id: Ifb1a93579bd4ab8fd10f78665a28559cc61da7e2
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
These are local containers that are either:
- Already const and didn't need Q_FOREACH to begin with
- Can be simply made const, just by adding const keyword
In one case the unittest checked that the container's size is 1, so use
list.first() instead of a for-loop.
In files where Q_FOREACH isn't used any more, remove
"#undef QT_NO_FOREACH". Also remove those files from NO_PCH_SOURCES.
Drive-by changes:
- Remove parenthesis from one-line for-loops
- Make the for-loop variable a const& where a copy isn't needed
Task-number: QTBUG-115839
Change-Id: Ide34122b9cda798b80c4ca9d2d5af76024bc7a92
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
The loop was iterating over a temporary container, so it couldn't have
changed it. Store the container in a const auto variable and use
ranged-for.
In files where Q_FOREACH isn't used any more, remove
"#undef QT_NO_FOREACH".
Task-number: QTBUG-115839
Change-Id: I402df5fa48f4287f3cc989ddae1524da43999049
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
cleanupTestCase: the loops don't change the member containers, so use
std::as_const and ranged for.
iterateRelativeDirectory(): the for loop doesn't change the container,
so make it const to begin with, and use a ranged-for loop.
Drive-by change: make the for-loop variables const&, no need to create
unnecessary copies.
Task-number: QTBUG-115839
Change-Id: Ic2776459f695c9f334f83916b1c9bbe5646a3b9d
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
Each loop was iterating over temporary containers, so use a const auto
variable to hold it and use ranged-for
Drive-by change: make the for-loop variable a const& (QString,
QFileInfo).
Task-number: QTBUG-115839
Change-Id: Idffaedb8e2e8782a0f4f907995f62f3c0de44bba
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
filterLinks() is always called on a temporary QStringList, so make it
take by && (which proves it's always called on a temporary), and modify
the parameter directly.
Task-number: QTBUG-115839
Change-Id: I40611f40cc0096a58d5c9d8e68c5df06d43152e5
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
cleanupTestCase(): the loop doesn't modify the member container, so use
std::as_const and a ranged-for.
mounting(): the loop was iterating over a temporary QList, store it in a
local auto variable and use ranged-for.
Drive-by change: add braces to a multi-lined for block.
Task-number: QTBUG-115839
Change-Id: I0542cad4df3730d6a09b39e64a54a84fc0d57062
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Other recent commits have fixed the other Q_FOREACH uses in this file.
Task-number: QTBUG-115839
Change-Id: I03063f3e8f1e99c5a2aa2d9188260f3e79ca43bd
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
The density of Q_FOREACH uses in this and some other modules is still
extremely high, too high for anyone to tackle in a short amount of
time. Even if they're not concentrated in just a few TUs, we need to
make progress on a global QT_NO_FOREACH default, so grab the nettle
and stick to our strategy:
Mark the whole of Qt with QT_NO_FOREACH, to prevent new uses from
creeping in, and whitelist the affected TUs by #undef'ing
QT_NO_FOREACH locally, at the top of each file. For TUs that are part
of a larger executable, this requires these files to be compiled
separately, so add them to NO_PCH_SOURCES (which implies
NO_UNITY_BUILD_SOURCES, too).
In tst_qglobal.cpp and tst_qcollections.cpp change the comment on the
#undef QT_NO_FOREACH to indicate that these actually test the macro.
Task-number: QTBUG-115839
Change-Id: Iecc444eb7d43d7e4d037f6e155abe0e14a00a5d6
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Instead make "benchmarks" a member variable and call qDeleteAll() on it
in cleanupTestCase(). This doesn't make much difference since the
allocated resources would be freed when the whole test is destroyed
anyway, but still.
Change-Id: Iba66d32697fd3f2283185ee65a0a514176b4b258
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Previously name() has always used underscore and bcp47Name() dash; let
the user chose which one best fits their needs.
[ChangeLog][QtCore][QLocale] QLocale's name() and bcp47Name() now let
the caller chose what separator to use between the tags making up the
name, where there is more than one.
Change-Id: Ia689e6a3fb581b42905e7fb1ae7a7b688244d267
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
... and rename it to qbswap(), thus enabling the endian conversions
for Id128bytes via q{To,From}{Little,Big}Endian() functions.
Found during Qt 6.6 API Review.
Pick-to: 6.6
Change-Id: Ie320cee52ec2b9de0aaa112adec8febb7f5b68a2
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Accent color role has been renamed according to name rule of other
color roles in QPalette.
Fixes: QTBUG-116107
Pick-to: 6.6
Change-Id: I70ac98a1e97afbdc7ea5f8d79f808c307e170712
Reviewed-by: Mitch Curtis <mitch.curtis@qt.io>
Reduce test precision to account for rounding errors, and at the same
time increase the setup precession by premultiplying in rgba64 instead
of argb32, which makes the test randomness trigger more regularly.
Pick-to: 6.6
Task-number: QTBUG-115945
Change-Id: I3e95449ada26ff5bb0acc00412f345733603f4c0
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
The function was replacing the `>` character in generator expressions coming from `add_compile_definitions`. This was creating generator expression syntax errors. Discard generator expressions from character replacing.
Add tests for the three cases.
Fixes: QTBUG-111717
Change-Id: I694d2908738085fdf15112834f20183a9f393422
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
When changing the current index while the tab bar is not visible,
calculating the necessary scroll offset might result in wrong results if
the tab bar still has an old size. When the tab bar then gets shown and
resized, the scroll wouldn't be corrected, potentially leaving tabs
unnecessarily scrolled out.
We don't need to make the current index visible if the tab bar itself is
not visible, it's enough to flag the layout as dirty so that the next
show event (which either way makes the then current index visible)
triggers a laying out of the tab bar tabs.
Amends e851d4c06154bf02b23030ff1f7024a8b9edf874.
Fixes: QTBUG-115109
Task-number: QTBUG-113140
Pick-to: 6.6 6.5
Change-Id: I3d8633f9f8b907a36190123839a6104a17bfe138
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: David Faure <david.faure@kdab.com>
Reviewed-by: Santhosh Kumar <santhosh.kumar.selvaraj@qt.io>
Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
The test removes the file in memfs that IndexedDB settings use as the
backing store. This forces a new instance of IDB settings to read from
the actual IndexedDB, instead of the file.
Change-Id: I7c04a90ae80e47b7742bd133b2d9327ce0063fe2
Reviewed-by: Morten Johan Sørvig <morten.sorvig@qt.io>
It has always returned dash-joined forms of the locale names, and
callers who need an underscore-joined form have been obliged to
replace('-', '_') before using them. Given that everything it adds to
the list comes from QLocaleId methods that accept a separator, it's
trivial to let it offer the same choice to its callers and save them
this hassle.
Amended code in QTranslater and QMimeType to save them that hassle.
[ChangeLog][CoreLib][QLocale] QLocale::uiLanguages() now lets the
caller choose what separator to use between the tags that make up each
locale-identifier in the list returned.
Change-Id: I91fcd0b988d9a64e0e9ad9e851f6cb8c1be8ae50
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It's ... broken. Found and filed lots of bugs. Add #ifdef'ery and
QEXPECTED_FAIL() to document the state of affairs, hopefully reminding
us to fix these things come Qt 7.
Task-number: QTBUG-116064
Task-number: QTBUG-116076
Task-number: QTBUG-116077
Task-number: QTBUG-116079
Task-number: QTBUG-116080
Pick-to: 6.6 6.5
Change-Id: I29e89fdf995ddf60ef1e03c7af009e80980c9817
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
We were only ever testing with a 0 seed, even though the function was
called for all QFETCH_GLOBAL seeds.
Add the seed.
Amends 5e93361888.
Pick-to: 6.6 6.5 6.2 5.15
Change-Id: I3c78714ad6fb3f94233789dd2c8884d9b157fa76
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The loops don't change the m_connections container. The call chain is:
- registerObjectPeer() unittest constructs a MyServer, which connects
QDBusServer::newConnection to MyServer::handleConnection(), the
latter stores each new connection's name in m_connections
- An QTestEventLoop is entered, which triggers handleConnection(),
handleConnection() calls exitLoop() at the bottom (this is repeated
multiple times)
- server.unregisterObject() is called, iterating over m_connections
- server.registerObject() is called iterating over m_connections
- between the unregisterObject() call and the registerObject() calls
m_connections is not modified AFAICS
Thus no need for taking a copy of m_connections (not that it matters
much, it's a QStringList with size() == 3).
Change-Id: Idaea2ca4d3b27fc88d39f8434e3817a2a4098c72
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Amends f2f8820073.
I have no idea how this went through the CI, but assigning to a const
variable cannot possibly compile.
Reported-by: Axel Spoerl <axel.spoerl@qt.io>
Pick-to: 6.6 6.5
Task-nubmber: QTBUG-115839
Change-Id: I0f22dcd5ab691f92880ea3c6446aedca53df0721
Reviewed-by: Axel Spoerl <axel.spoerl@qt.io>
Extended linear Display P3 + FP16 is likely the thing to use
on platforms such as VisionOS and iOS (and optionally on macOS)
and perhaps iOS). Enable testing this on macOS with the hdr
manual test.
Change-Id: I67f0bdbadae8c7ebccae7de008f12fd8d9135529
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
The method calls show() on a dock widget group window, when the window
flags have changed. When all of its contained, tabbed dock widgets are
programmatically hidden or docked on the main window, an empty group
window is shown.
This patch implements bool hasVisibleDockWidgets(). It returns true, if
at least one of the group window's dockwidget children is not hidden.
It replaces show() by setVisible(), passing the return value of
hasVisibleChildren().
It adapts tst_QDockWidget::floatingTabs() to test the fix.
(Drive-by: remove dead code)
Fixes: QTBUG-115058
Pick-to: 6.6 6.5 6.2
Change-Id: Ifb8e2450e91a7c78decc06f592e160631ca2faf5
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
The method reads the next element in a loop, as long as valid elements
exist. Within the loop, it returns
- false if the end of an element has been reached
- true if a new element has started
When the document end has been reached, the loop continues, until
readNext() returns Invalid. Then, PrematureEndOfDocumentError is launched.
This is wrong, because reading beyond the document end is caused by a
missing return condition in the loop.
=> Treat document end like element end and return false without
reading beyond it.
=> Test correct behavior in tst_QXmlStream::readNextStartElement()
Fixes: QTBUG-25944
Pick-to: 6.6 6.5 6.2
Change-Id: I0160b65880756a2be541e9f55dc79557fcb1f09f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
By actually registering them.
Commit 850d850c5a changed from
qMetaTypeId<QDBusArgument>() to QMetaType::fromType<QDBusArgument>() and
in Qt 6, fromType() does not register the type with the database. That
means the lines became runtime no-ops at that time or during the
QMetaType updates since 6.0. All they did was instantiate the C++ inline
variable.
The testing also detected we didn't register QList<QDBusVariant> as an
alias for the "av" signature. I'm not entirely sure you're allowed to
use this because QtDBus does not like re-registration of the built-in
types, and "av" is already assigned to QVariantList. This is no trouble
for the parser, anyway.
Minor change to qdbuscpp2xml to allow reading from stdin, so we don't
have to create temporary files.
Pick-to: 6.5 6.6
Fixes: QTBUG-115964
Change-Id: I80612a7d275c41f1baf0fffd177a14925e7d23ac
Reviewed-by: Ievgenii Meshcheriakov <ievgenii.meshcheriakov@qt.io>
Strictly speaking, we don't need the *bus*, only libdbus-1, but some
machines in our CI appear to be misconfigured somehow. I don't
understand how they can both have and not have this library in the same
run.
Pick-to: 6.5 6.6
Change-Id: I80612a7d275c41f1baf0fffd177a66a04951948c
Reviewed-by: Ievgenii Meshcheriakov <ievgenii.meshcheriakov@qt.io>
We've known for a long time that this is producing worse code with GCC
because of how we implemented in Q_ASSUME_IMPL(). So bite the bullet and
actually deprecate the macro, replacing all extant Q_ASSUME() with
Q_ASSERT().
The replacement is in C++23. Backporting the support onto Q_ASSUME_IMPL
was previously rejected by reviewers.
[ChangeLog][Deprecation Notice] The Q_ASSUME() macro is deprecated. This
macro has different side-effects depending on the compiler used (GCC
compared to Clang and MSVC), and there are certain conditions under
which GCC is known to produce worse code than if the macro was absent.
To give a hint to the compiler for optimizations, use the C++23
[[assume]] attribute.
Change-Id: I80612a7d275c41f1baf0fffd177a3a4ad819fb2d
Reviewed-by: Volker Hilsheimer <volker.hilsheimer@qt.io>
This constructor matches way too many argument types (integral,
unscoped enums, FP types), so it's likely to cause mayhem, even if
left in as an explicit constructor.
We now have a named constructor for the same functionality, so just
drop the "unnamed" constructor.
"Unnamed" constructors are important when emplacement is more
efficient than construction + move, or when implicit conversion is
required. Neither is the case here: The named as well as the
"unnamed" constructors just copy ten bytes around, and the compiler
can optimize those extra copies away just fine.
Found in API review.
Pick-to: 6.6
Change-Id: I7faafd3ebf522fb2b0e450112fb95d643fece5ce
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
QSharedPointer is 'meh', see QTBUG-109570 and 18113e22e9.
This is just a textual replacement of
- QSharedPointer<(.+)>::create() → std::make_shared<\1>()
- QSharedPointer → std::shared_ptr
And it compiles and still passes. No non-standard APIs to fix up.
Task-number: QTBUG-109570
Change-Id: I827d4a9be0511780c3900bd53ffcbdcb6aacbc3b
Reviewed-by: Richard Moe Gustavsen <richard.gustavsen@qt.io>
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
All of these fall into the trivial category: loops over (readily made)
const local containers. As such, they cannot possibly depend on the
safety copy that Q_FOREACH performs, so are safe to port as-is to
ranged for loops.
There may be more where these came from, but these were the ones that
stood out as immediately obvious when scanning the 100s of uses in
qtbase, so I preferred to directly fix them over white-listing their
files with QT_NO_FOREACH (which still may be necessary for some files,
as this patch may not port all uses in that file).
Pick-to: 6.6 6.5
Task-nubmber: QTBUG-115839
Change-Id: I7b7893bec8254f902660dac24167113aca855029
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
The original code duplicated contained elements of a QStringList by
repeated appending it to itself, preventing the container from being
marked const.
Instead, keep a list of unique mime-types, and iterate over the list
eight times.
As a drive-by, port from QList to a C array ("never use a
dynamically-sized container for statically-sized data"), use u""_s
UDLs (since we're touching almost all lines of the function, anyway,
also in the unrelated mimeTypeForName() call).
This allows porting the Q_FOREACH loop (which anyway cannot deal with
C arrays) to a ranged for one (which can).
Pick-to: 6.6 6.5
Task-number: QTBUG-115839
Change-Id: I844ae38104bb2980ea194b85f9017a3e95791ea2
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
These are all trivial: all are over (already or newly-made) const
local variables.
As a drive-by, replace a QList legacy left-shift-based- with
initializer_list-construction.
Pick-to: 6.6 6.5
Task-number: QTBUG-115803
Change-Id: I453e24272c4c4b7dce5b91a0bd04481d833c50bb
Reviewed-by: Ivan Solovev <ivan.solovev@qt.io>
These are all trivial: all are over (already or newly-made) const
local variables.
As a drive-by, replace a QList with a C array
("never use a dynamically-sized container for statically-sized data").
Pick-to: 6.6 6.5
Task-number: QTBUG-115803
Change-Id: I1b1e8f093abf75093900631e6fe3cbc9e3019d34
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
These are all trivial: all are over (already or newly-made) const
local variables.
We don't have a mechanism to mark a subtree as Q_FOREACH-free, and
adding QT_NO_FOREACH to each executable is overkill, so we just have
to hope that no new uses are being introduced until we can mark the
whole QtBase module as Q_FOREACH-free.
Pick-to: 6.6 6.5
Task-number: QTBUG-115803
Change-Id: I13dc176756633674bab8c93a342ecdba6c5dd23e
Reviewed-by: Ahmad Samir <a.samirh78@gmail.com>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
When the placeholder text is changed after having been displayed, it
doesn't get updated on the screen any more, unless the entire viewport
is updated, e.g. because of a document change or a focus event.
This patch simplifies QPlainTextEditPrivate::updatePlaceHolderVisibility()
to update the visibility if the text document is empty.
It replaces the member QPlainTextEditorPrivate::placeholderVisible
by the function isPlaceHolderTextVisible(). It returns true, if the
document is empty and a placeholder text exists, and otherwise false.
It adapts and corrects tst_QPlainTextEdit::placeHolderVisibility():
- usage of new member function instead of data member.
- do not expect an empty placeholder to be visible.
Fixes: QTBUG-115831
Pick-to: 6.6 6.5 6.2
Change-Id: Ic4427ce7f7f1b8cde89957b9de0b978bd34ba923
Reviewed-by: Santhosh Kumar <santhosh.kumar.selvaraj@qt.io>
QAndroidPlatformInputContext::focusObjectStopComposing() sends an input
event for each character newly added by the Android virtual keyboard.
It then sends a second input event to notify that the cursor has
advanced to the position after the new character.
The implicit assumption is, that the receiver of the input event does
not change the text.
If e.g. QLineEdit::setText() is called in the QLineEdit::textEdited
slot, the text does change. If the change implies a cursor change,
QLineEdit notifies the platform input context about it.
However, by sending the second input event, QAndroidPlatformContent
returns the cursor back to the position after the last character added
by the virtual keyboard.
This patch joins the composed text and the cursor position into one
single input method event. A new cursor position, set by the receiver
of the input method event, is no longer overridden.
The patch adds test functionality to tst_QLineEdit::setText().
Fixes: QTBUG-115756
Pick-to: 6.6 6.5 6.2
Change-Id: I85ffac5d6bab93ccb144be0f5b8083258a270550
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Add more tests on WebAssembly platform for better tests coverage.
Change-Id: Iaaaa824ae6058a9ae5dba4c4038a7f687bfc17e0
Reviewed-by: Mikołaj Boc <Mikolaj.Boc@qt.io>
Take 2.
Re-land previously reverted commit, due to not handling resource names
that are not valid c++ identifiers. Now we sanitize the resource names
just like rcc does by replacing non-alphanumeric characters with
underscores.
Original commit message.
During the Qt 5 -> Qt 6 and qmake -> CMake porting time frame, it was
decided to keep resources in an object file (object library), rather
than putting them directly into a static library when doing a static
Qt build, so that the build system can take care of linking the
object file directly into the executable and thus not forcing
project developers to manually initialize resources with
the Q_INIT_RESOURCE() macro in project code.
This worked for most qmake and cmake projects, but it created
difficulties for other build systems, in the sense that these projects
would have to manually link to the resource object files, otherwise
they would get link time errors about undefined resource symbols,
assuming they kept the Q_INIT_RESOURCE() calls.
If the project code didn't contain Q_INIT_RESOURCE calls, the
situation would be even worse, the linker would not error out,
and the missing resources would only be discovered at runtime.
It's also an issue in CMake projects that try to link to the
library files directly instead of using the library target names,
which means the object files would not be automatically linked in.
Many projects try to do that because we don't yet offer a convenient
way to install libraries and reuse them in other projects (the SDK
case), so projects end up shipping only the libraries, without the
resource object files.
We can improve the situation by moving the resources back into their
associated static libraries, and only keeping a static initializer as
a separate object file / object library, which references the actual
resource initializer symbol, to ensure it does not get discarded
during linking.
This way, projects that link using targets get no behavior difference,
whereas projects linking to static libraries directly can still
successfully build as long as their sources have all the necessary
Q_INIT_RESOURCE calls.
To ensure the resource symbols do not get discarded, we use a few new
private macros. We declare the resource init symbols we want to keep as
extern symbols and then assign the symbol addresses to volatile
variables.
This prevents discarding the symbols with the compilers / linkers we
care about.
It comes at the cost of an additional static initializer per resource,
but we would get the same + a bigger performance hit if we just used
Q_INIT_RESOURCE twice (once in the object lib and once in project
code), which internally needs to traverse a linked list of all
resources to check if a resource was initialized or not.
For GHS / Integrity, we also need to use a GHS-specific pragma to keep
the symbols, which we currently use in qtdeclarative to ensure qml
plugin symbols are not discarded.
The same macros will be used in a qtdeclarative change to prevent
discarding of resources when linking to static qml plugins.
A cmake-based test case is added to verify that linking to static
libraries directly, without linking to the resource initializer
object libraries, works fine as long as the project code calls
Q_INIT_RESOURCE for the relevant resource.
This reverts commit bc88bb34ca.
Fixes: QTBUG-91448
Task-number: QTBUG-110243
Change-Id: Idce69db0cf79d3e32916750bfa61774ced977a7e
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
