'ld' only capable to resolve circular dependencies by wrapping the
suspected static libraries and objects using --start/end-group
arguments. We want to detect if linker is 'ld' at configure time to
decide how to link the resource objects if finalizers are not enabled.
The qt_config_compile_test function is extended with an extra argument
since it's required to pass custom cmake flags to the ld-related test.
Pick-to: 6.2
Change-Id: I484fcc99e2886952d8b0232f37e4e6a35d072931
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
In the future it might be useful to print the config summary entries
of a Qt module configure.cmake file even if the associated module
is not built and thus qt_feature_module_begin is not called.
The repo src/CMakeLists.txt could then use a combination of
qt_feature_evaluate_features and a conditional
qt_feature_record_summary_entries to ensure the that summary entries
are still shown.
Change-Id: I124efc82163ddae48d9e72c70a677ec4c6588fac
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Use the same approach we use for iOS, which is to set multiple
CMAKE_OSX_ARCHITECTURES values and let the clang front end
deal with lipo-ing the final libraries.
For now, Qt can be configured to build universal macOS libraries by
passing 2 architectures to CMake, either via:
-DCMAKE_OSX_ARCHITECTURES="x86_64;arm64"
or
-DCMAKE_OSX_ARCHITECTURES="arm64;x86_64"
Currently we recommend specifying the intel x86_64 arch as the first
one, to get an intel slice configuration that is comparable to a
non-universal intel build.
Specifying the arm64 slice first could pessimize optimizations and
reduce the feature set for the intel slice due to the limitation
that we run configure tests only once.
The first specified architecture is the one used to do all the
configure tests.
It 'mostly' defines the common feature set of both architecture
slices, with the excepion of some special handling for sse2 and
neon instructions.
In the future we might want to run at least the Qt architecture config
test for all specified architectures, so that we can extract all the
supported sub-arches and instruction sets in a reliable way.
For now, we use the same sse2 hack as for iOS simulator_and_device
builds, otherwise QtGui fails to link due to missing
qt_memfill32_sse2 and other symbols.
The hack is somewhat augmented to ensure that reconfiguration
still succeeds (same issue happened with iOS). Previously the sse2
feature condition was broken due to force setting the feature
to be ON. Now the condition also checks for a special
QT_FORCE_FEATURE_sse2 variable which we set internally.
Note that we shouldn't build for arm64e, because the binaries
get killed when running on AS with the following message:
kernel: exec_mach_imgact: not running binary built against
preview arm64e ABI.
Aslo, by default, we disable the arm64 slice for qt sql plugins,
mostly because the CI provisioned sql libraries that we depend on only
contain x86_64 slices, and trying to build the sql plugins for both
slices will fail with linker errors.
This behavior can be disabled for all targets marked by
qt_internal_force_macos_intel_arch, by setting the
QT_FORCE_MACOS_ALL_ARCHES CMake option to ON.
To disble it per-target one can set
QT_FORCE_MACOS_ALL_ARCHES_${target} to ON.
Task-number: QTBUG-85447
Change-Id: Iccb5dfcc1a21a8a8292bd3817df0ea46c3445f75
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@qt.io>
Features that are not emitted in the current
configuration (e.g. plugin-manifests on Linux) should not yield an
error. Instead, print a warning message like Qt5's configure did.
Set insignificant feature to OFF.
Remove the now unneeded emit_if parameter from
qt_feature_set_cache_value.
Pick-to: 6.1
Fixes: QTBUG-88305
Change-Id: I0f2ce152fca5f08417038c9bd2c07639ff6a3df4
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
When a feature is set in qtbase and is later set to another value,
an error occurs and an error message is given. This patch
changes the error message to contain both the preexisting and the
new value
Change-Id: Ifa9fc1f06bfde40e8fd5dfdf30165d4393abbd28
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
This property is dynamically determined and set in the
Qt6${module-name}PluginsConfig.cmake files now.
There is no need anymore to export this property.
Change-Id: I2d164864c4099034b88f9ad852eae8b9f9e55f8b
Reviewed-by: Alexey Edelev <alexey.edelev@qt.io>
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
Add an optional argument to store the report command while feature
evaluation-only stage. This is necessary, because according to the
actual feature evaluation process, evaluation happens only once and it's
impossible to record any command during the evaluation-only stage.
Fixes: QTBUG-90319
Change-Id: I215fbe0a28a2661cd2d97d253b2e2c787d295cbd
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Jason McDonald <macadder1@gmail.com>
Add output of feature condition error at the end of configuration
summary. Introduce qt_configure_add_report_error that may be used
to add error to end-point configure report without actual
configure interruption.
Fixes: QTBUG-88282
Change-Id: Idcf478da863ae6507438eb3370927d7d1f01e071
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Unset all QT_FEATURE_foo values for every build.
If any of FEATURE_foo is different of QT_FEATURE_foo, mark whole Qt
build as dirty. Reset FEATURE_foo for dirty builds to the calculated
value if it doesn't meet its condition.
Set Qt module as NOT FOUND if its target was not created during
configuration.
Main issue with this approach are generated files, that became trash
once the related features are disabled. This especially affects features
that enable/disable Qt modules. FooConfig.cmake files are created and
generate lots of warnings if feature was disabled. We may introduce a
module cleanup procedure at some point.
Fixes: QTBUG-85962
Pick-to: 6.0
Change-Id: Id71c1edb4027b24c6793063e40cc9d612c24ebce
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Add '_qt_internal_dump_expression_values' function that dumps all
values evaluated by 'qt_evaluate_config_expression'.
'_qt_internal_dump_expression_values' doesn't evaluate undefined
features, only collect actual values.
Fixes: QTBUG-88476
Pick-to: 6.0
Change-Id: I9d09ffbd9f9fa91bc4f36536c58e7f118b06f9b9
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Move 'INPUT_' variable detection logic out of
'qt_feature_set_cache_value' function. Use 'INPUT_'
variable to enable/disable related 'FEATURE_' in case if
'FEATURE_' is not defined.
Fixes: QTBUG-88501
Pick-to: 6.0 6.0.0
Change-Id: I1162ac0fb562036898a37b52c3f80229ec8a0970
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
INPUT_ variables enable FEATURE_ variable only in case if valid
"true"-like value is assigned from command line.
Amends dc43061e9a
Fixes: QTBUG-87755
Task-number: QTBUG-88142
Change-Id: I65e85c7548981fdec94366b531f6df6396be71b7
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
"configure" script translates feature-related parameters to INPUT_
variables instead of FEATURE_.
Both INPUT_ and FEATURE_ variables passed to cmake script are
equivalent. FEATURE_ has higher priority in case if both are defined.
Fixes: QTBUG-87755
Change-Id: If697a0d62ab839877a3196ea74e631582a570dda
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
CMake considers ON/OFF as booly string values regardless of the case.
Make the value comparison in QtFeature.cmake case-independent.
It's now possible to build Qt with '-DFEATURE_gui=off'.
Fixes: QTBUG-87948
Change-Id: I3d948e8219ad9728414803c8c4cd756034073b46
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Rolf Eike Beer <eb@emlix.com>
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 file provides common feature-functionality is supposed to be
included by QtFeature.cmake and QtProcessConfigureArgs.cmake.
Change-Id: Ifb483c2a9c4014d240c2a4d6ff07b5c0ceee9c3a
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
There was an inconsistency where configuring qtbase on Apple Silicon
with the following command line produced different results rom the
second variant.
$ cmake ../qtbase -DCMAKE_OSX_ARCHITECTURES=arm64
vs
$ CMAKE_OSX_ARCHITECTURES=arm64 cmake ../qtbase
That happened because the CMAKE_OSX_ARCHITECTURES variables was not
passed to project-based try_compile calls. This resulted in compile
tests like SIMD avx to succeeded on Apple silicon, which shouldn't.
To address that, always pass the architecture on macOS if it's
specified.
Change-Id: Ia12e86230cc6e5e11f387e3cbb273d90646ef2e3
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
qdevice.pri should embed info about the Android SDK, NDK, host,
platform api level and ABI.
The machine tuple test should not be run for uikit and Android
platforms.
Sysroot should also not be prepended for uikit and Android
platforms, otherwise it breaks Qt module include paths.
Task-number: QTBUG-85399
Task-number: QTBUG-82581
Change-Id: Ic48c88f6ab15d75c2ebc323c8d7a3b7e5596f3c1
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Show a status message when we run syncqt for a module (gives a bit
more insight into why the configuration seems to stall for a moment).
Remove some uninformative messages about generating config.h files
and Depends headers, etc.
Change-Id: I5ff2774b9cf5d92ddde564dc09f4197c2835ee4a
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
This adds the CMake equivalent to the classical -device-option key=value
configure argument:
-DQT_QMAKE_DEVICE_OPTIONS=key1=value1[;keyN=valueN]
The keys and values get dumped verbatim into qdevice.pri.
This patch also ports the machineTuple configure test. Its result is
written into qdevice.pri as value for the GCC_MACHINE_DUMP variable.
Change-Id: I29f2323fd87639fafaed99ec7446c7ee75504705
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
The plugin-manifests feature has the "negative" CONFIG value
"no_plugin_manifest". On negation, we're supposed to strip off the
leading "no_" instead of adding another one.
Change-Id: Id2c66da41f22881272d5b923f12b85d9fcc2c9d0
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Previously we checked the LINK_LIBRARIES property of the main module
target, but we should instead use the values of
INTERFACE_LINK_LIBRARIES set on the FooPrivate module.
Because both versionless targets and private targets are interface
libraries, we need to properly differentiate between them when
following versionless targets to their main associated target.
To do that, instead of using string comparison, export an additional
private _qt_is_versionless_target property, and query that.
Also make sure to set and export the _qt_config_module_name property
on the FooPrivate targets.
Also make sure to APPEND to EXPORT_PROPERTIES rather than override
(looking at you QtFeature.cmake).
Task-number: QTBUG-75666
Change-Id: Ia3261e218840e9f5217ab49755e8c876560e294d
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
Consider a negated feature config value like the following:
qt_feature_config("foo" QMAKE_PUBLIC_QT_CONFIG NEGATE)
If this feature was disabled, it would turn up in both,
enabled_features and disabled_features of module .pri files.
Also, QT_CONFIG would contain foo.
Expected however is that QT_CONFIG contains no-foo, and only
disabled_features contains foo.
Fix this by prepending a "no_" prefix to the value, similar to the
"no-" prefix in the qmake build. The qt_correct_config function was
adjusted to recognize "no_foo" and translate it to the qmakeish
"no-foo" config value.
Config values that start with "no_" but do not correspond to a feature
are left untouched. You can still have values like
"no_valley_too_deep" or "no_mountain_too_high".
Change-Id: I23d8b18c84e04ea6dfa25cc6ccd8f7e86211b144
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
If a find_package() in a try_compile project doesn't find a package,
and we then link against a non-existent target, the configuration
failure of the compile test also fails the configuration of the
project.
To avoid that, separate library targets from non-targets, and make sure
to only link against the targets if they exist.
pro2cmake now outputs modified compile test project code which iterates
over targets and non-target libraries, and links against them when
needed.
Change-Id: Ib0f4b5f07af13929c42d01a661df2cabdf9b926b
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
list(JOIN) removes a layer of escaping, which force doubling up
on the backslashes. Instead use string(REPLACE) thus making the
escaping a bit saner.
Change-Id: Ie3daf0112dd09fbcbaf8798552949470952611c9
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Some library entries in configure.json have a test entry.
An example is assimp in qtquick3d.
qmake tries to find the library via the sources section, and then tries
to compile the test found in config.tests/assimp/assimp.pro while
automagically passing it the include and link flags it found for assimp.
We didn't handle that in CMake, and now we kind of do.
configurejson2cmake will now create a corresponding
qt_config_compile_test call where it will pass a list of packages and
libraries to find and link against.
pro2cmake will in turn generate new code for the standalone
config.test project. This code will iterate over packages that need to
be found (like WrapAssimp) and then link against a list of passed-in
targets.
In this way the config.test/assimp/main.cpp file can successfully
use assimp code (due to propagated include headers).
qt_config_compile_test is augmented to take a new PACKAGES argument,
with an example as follows
PACKAGES PACKAGE Foo 6 COMPONENTS Bar
PACKAGE Baz REQUIRED
The arguments will be parsed and passed to the try_compile project,
to call find_package() on them.
We also need to pass the C/C++ standard values to the try_compile
project, as well as other try_compile specific flags, like the
toolchain, as given by qt_get_platform_try_compile_vars().
Change-Id: I4a3f76c75309c70c78e580b80114b33870b2cf79
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
QT_STATIC must be defined for static builds to have the right
import/export symbol macros defined. Originally, this macro is wrapped
in a condition. That's why we extend qt_feature_definition to be able
to write a prerequisite to qconfig.h.
Change-Id: I610e60acc7f5bdc031eff6d53a76e0b229a5c8c2
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Teaches configurejson2cmake about summaries / reports, so things
like enabled features, configure sections, notes, etc.
Add relevant CMake API for adding summary sections and entries,
as well as configure reports. The commands record the passed data,
and the data is later evaluated when the summary needs to be printed.
This is needed, to ensure that all features are evaluated by the
time the summary is printed.
Some report and summary entries are not generated if they mention a
feature that is explicitly exclduded by configurejson2cmake's feature
mapping dictionary. This is to prevent CMake from failing at configure
time when trying to evaluate an unknown feature. We should re-enable
these in the future.
A few custom report types are skipped by configurejson2cmake (like
values of qmake CONFIG or buildParts).
These will have to be addressed a case-by-case basis if still needed.
Change-Id: I95d74ce34734d347681905f15a781f64b5bd5edc
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
And in order to do this we must teach qt_config_compile_test a
COMPILE_OPTIONS argument.
Change-Id: I66fa45142b544e3a2fc599af1c1a4c69b442b318
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Introduce an internal qt_feature_evaluate_features() function
which takes a list of configure.cmake paths, and evaluates the
features declared in those files, thus setting a bunch of cache
variables.
This is required to implement the equivalent of what qtbase/src.pro
does, which includes the feature .pri files to decide whether
bundled 3rd party libraries need to be built.
Change-Id: I5552f488671c001eb3f204245b905ab981017a9f
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Need to pass additional -Zc:__cplusplus flag when using
MSVC, so that the __cplusplus define has correct values.
Additionally make the option be propagated to consumers of Qt
via the public Platform target, which QtCore links against.
Change-Id: Ie1283c25334b93f993529beb7fb32bdb001627f5
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Before we only exported features that had outputType PUBLIC or PRIVATE
on the various "QT_ENABLED_PUBLIC_FEATURES" target properties.
Now we also export features that have output type privateConfig,
publicConfig and publicQtConfig.
The new properties names are:
- QT_QMAKE_PUBLIC_CONFIG for outputType == publicConfig
- QT_QMAKE_PRIVATE_CONFIG for outputType == privateConfig
- QT_QMAKE_PUBLIC_QT_CONFIG for outputType == publicQtConfig
These need to be exported for 2 reasons:
- other modules that need to check the config values
- in preparation for generating proper qmake .prl and .pri
information for each module
Note that the config values are now considered actual features
when doing condition evaluation. So if there exists a feature "ssse3"
with outputType privateConfig, its enabled state can be checked via
QT_FEATURE_ssse3 in consuming modules (but not in the declaring
module).
These config values are also placed in the respective
QT_ENABLED_PUBLIC_FEATURES, QT_ENABLED_PRIVATE_FEATURES properties
when exporting a target, so the properties will now contain both
features and config values.
In order to make this work, feature name normalization has to happen
at CMake time, rather than done by the python script.
This means that features like "developer-build" need to retain the
dash in the qt_feature(), qt_feature_definition() and
qt_feature_config() calls, rather than generating "developer_build"
as the script did before.
The normalization is done at CMake time. Feature conditions,
CMake code, and -DFEATURE_foo=bar options passed on the command line
should still use the underscore version, but the original name is used
for the QT_QMAKE_PUBLIC_CONFIG properties.
Note that "c++11" like features are normalized to "cxx11".
Implementation wise, the configurejson2cmake script is adjusted to
parse these new output types.
Also QtBuild and QtFeature are adjusted to save the config values
in properties, and re-export them from GlobalConfig to Core.
Task-number: QTBUG-75666
Task-number: QTBUG-78178
Change-Id: Ibd4b152e372bdf2d09ed117644f2f2ac53ec5e75
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Tested locally with the following configurations:
- iOS device builds (arm64)
- iOS simulator builds (x86_64)
- iOS simulator_and_device builds (fat arm64 and x86_64 archives)
All iOS builds currently require a custom vcpkg fork which contains
fixes for building the required 3rd party libraries.
qtsvg, qtdeclarative, qtgraphicaleffects and qtquickcontrols2
have also been tested to build successfully.
simulator_and_device builds are also supported, but require an umerged
patch in upstream CMake as well as further patches to vcpkg.
Task-number: QTBUG-75576
Change-Id: Icd29913fbbd52a60e07ea5253fd9c7af7f8ce44c
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Inside qt_config_compile_test we use two different calls:
check_cxx_source_compiles when given code, and try_compile when
given a project.
The former caches its results, while the latter does not, which means
that the try_compile calls are re-executed on each reconfigure.
Do what check_cxx_source_compiles does internally, and don't rerun
the try_compile calls if the results were computed once.
Change-Id: I2691ff08b7bb46c3fa60bbf5ed6278731c9b8e21
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
We cannot use a generator expression in an if statement, it does not
work. Instead, we could inspect the CMAKE_C/CXX_COMPILE_FEATURES list,
but unfortunately that's not reliable. For example it detects that ICPC
supports C++17 when in fact that depends on the installed libstdc++.
Therefore this patch revives our own configure tests.
Change-Id: Ic3bc5762fbe81837722523e3881ac16e84628519
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Allow setting feature defines to ON, YES, TRUE, Y, OFF, NO, FALSE, N and
non-negative numbers like it's possible for other defines.
We'll not allow IGNORE and NOTFOUND, because they're a bit strange in
the context of features.
Change-Id: I7f6cd6dff4c68972921b499b28916c9ad789d510
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Reviewed-by: Qt CMake Build Bot
As well as SIMD tests.
Change-Id: I43318418f90a6b9736eb89f008459c7d6adac8f9
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Generate appropriate qt_config_compile_test() calls for config tests
that have CMake projects. These are protected by an if(EXISTS) check
so that configuration doesn't fail for repos where the config tests
have not been ported yet.
Adjust the qt_config_compile_test() function to use try_compile for
projects specified via new PROJECT_PATH argument.
Change-Id: I83c061e384f68688a654b782fd7a9bede282d1e3
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
This clarifies the weird pause when configuring, right after
the HAVE_reduce_relocations test.
Change-Id: I0edcbbd085286c20c4eaea8d8e15a509ec93d85f
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Some features check for module / target existence. Adapt conversion
script to handle that.
Reland after fixing it.
Change-Id: If4fb942c2e0d16e76a0b9b767bf478527851b0f7
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
In particular, Qt targets extended with other qt targets
(eg. Qt::VulkanSupport or Qt::LinuxAccessibilitySupport)
after the first add_qt_module were not taken into account when generating
Depends files.
Note that this patch updates the minimum required version
to CMake 3.15
Change-Id: I747deedd4d59e385876bc1a834ef9bdb6078911b
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
It breaks some conditions, and then the build fails not finding
some private header files in qpa, etc.
This reverts commit 35dc8f496d.
Change-Id: I1b51eac06fe9186181d3f0a7c78f22da7be534e2
Reviewed-by: Tobias Hunger <tobias.hunger@qt.io>
Some features check for module existence. Adapt conversion script
and QtBuild feature condition parser to handle that.
Change-Id: I063e49a6fe9f8e9cf3aec985fd78ed4430398586
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
This commit introduces infrastructure work to allow static builds of Qt
to handle importing of plug-ins.
Change-Id: Ife0ca3ca7276ea8ec96fe0eb6adf934fad7620ec
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Current lates CMake has a limitation that it does not allow exporting
custom properties from INTERFACE libraries. GlobalConfig is such a
library, which means that so far all the global features were not
actually exported.
Copy the feature property values from GlobalConfig to Core. Because
Core is an actual shared library, it keeps the custom properties
when exported, and thus Core feature properties will contain the sum
of Core and GlobalConfig feature values.
Change-Id: Idde305cbaf9ab85ecfbe29522dcbac1c44022b17
Reviewed-by: Tobias Hunger <tobias.hunger@qt.io>
